THE EMPIRICAL TURN IN FAMILY LAW

THE EMPIRICAL TURN IN FAMILY LAW

Historically, the legal system justified family law’s rules and policies through morality, common sense, and prevailing cultural norms. In a sharp departure, and consistent with a broader trend across the legal system, empirical evidence increasingly dominates the regulation of families.
There is much to celebrate in this empirical turn. Properly used, empirical evidence in family law can help the state act more effectively and efficiently, unmask prejudice, and depoliticize contentious battles. But the empirical turn also presents substantial concerns. Beyond perennial issues of the quality of empirical evidence and the ability of legal actors to use it, there are more fundamental problems: Using empirical evidence focuses attention on the outcomes of legal rules, discouraging a debate about contested and competing values. Reliance on empirical evidence overlays a veneer of neutrality on normative judgments. And uncritically adopting evidence about present conditions without interrogating the role of historical discrimination that continues to disadvantage some families can replicate that discrimination.
Given the promise and peril of the empirical turn in family law, this Essay proposes a framework to guide the use of this evidence. The framework preserves space for debating multiple values and advises decisionmakers when to use empirical evidence, with particular attention to the dangers for nondominant families. The framework also recommends strengthening evidentiary gatekeeping and elevating the potential for legal scholarship to serve as a bridge from the broader research base to the courts. With this guidance in place, empirical evidence can take its rightful place as a useful but cabined tool in the legal regulation of families.

INTRODUCTION

  1. EMPIRICAL EVIDENCE COMES TO FAMILY LAW
    1. The Historical Baseline
    2. Empirical Evidence Across the Institutions of Family Law
      1. Litigation
        1. Marriage Equality
        2. Abortion
        3. Intimate Partner Violence and the Child Welfare System
        4. Juvenile Sentencing
        5. Child Custody
      2. Legislation
      3. Administration
    3. Patterns in the Empirical Turn
  2. THE PROMISE AND PERIL OF AN EMPIRICAL FAMILY LAW
    1. The Benefits of Evidence-Based Family Law
    2. The Empirical Turn in Critical Perspective
      1. Reliability and Translation.
      2. Facts and Values
        1. Skewing Debates
        2. Providing Political Cover
        3. Replicating Discrimination
  3. GUIDING THE USE OF EMPIRICAL EVIDENCE IN FAMILY LAW
    1. The Proper Role of Empirical Evidence
    2. Practical Tools
      1. Gatekeeping
      2. Attention to Intersecting Identities
      3. A Translation Role for Legal Scholars

CONCLUSION

Introduction

At the crux of the fight over marriage equality, a court in northern California conducted a remarkable twelve-day trial. 1 Perry v. Schwarzenegger, 704 F. Supp. 2d 921, 929 (N.D. Cal. 2010), aff’d sub nom. Perry v. Brown, 671 F.3d 1052 (9th Cir. 2012), vacated and remanded sub nom. Hollingsworth v. Perry, 133 S. Ct. 2652 (2013); see also Kenji Yoshino, Speak Now: Marriage Equality on Trial 91–228 (2015) (describing the trial in detail). Faced with a challenge to the State’s marriage restriction, the court heard evidence on a range of social facts relating to family structure and child outcomes, the physical and economic benefits of marriage, the nature of sexual orientation, and the increased risk of physical and mental harm from discrimination and stigma. 2 See Perry, 704 F. Supp. at 932–38. This empirical evidence was pivotal to the court’s decision striking down the marriage restriction. A crucial part of the decision was the finding that children of same-sex parents have simi­lar outcomes to children raised by different-sex couples, undermining California’s rationale for differentiating couples based on sexual orientation. 3 See id. at 935, 950, 963–73, 981, 994–1003.

This example of relying on empirical evidence—defined broadly as research and data gathered through both quantitative and qualitative methods 4 This Essay loosely contrasts this empirical information with values. There exists a longstanding, if contested, distinction between the two. See David Hume, A Treatise of Human Nature 458–70 (L.A. Selby-Bigge & P.H. Nidditch eds., Oxford Univ. Press 2d ed. 1978) (1739–1740) (discussing the distinction between moral judgments and empirical facts). As explored throughout this Essay, the line between empirical evidence and values is blurred, and one often informs the other. Empirical evidence can influence values in numerous ways identified in this Essay, but values can also influence empirical evidence, partly because knowledge is inherently situated in culture. This is particularly true in the social sciences, but it can also be true in the hard sciences. The political and cultural valence of juvenile crime, for example, accounts at least in part for the research agenda of neuroscientists interested in adolescent brain development. See generally Hilary Putnam, The Collapse of the Fact/Value Dichotomy and Other Essays 28–45 (2002) (demonstrating how the distinction between facts and values breaks down in numerous ways); Ruth Anna Putnam, Creating Facts and Values, 60 Phil. 187, 190–204 (1985) (arguing facts are value-laden and values are fact-laden). —to resolve fundamental questions of family law is hardly unique. Consistent with an increasingly widespread reliance on empirical evidence across sectors of the economy, academic disciplines, and within the law, 5 See infra text accompanying notes 34–43. family law decisionmakers regularly draw on sociology, psychology, neuroscience, data analytics, and related social and hard sciences to make critical choices about the legal regulation of families. 6 See infra section I.B. In addition to using empirical evidence to decide constitutional cases, courts turn to psychological research about parental alienation syndrome to decide custody suits. 7 See infra section I.B.1.e. Lawmakers draw on studies about the harms of foster care to drastically revamp the child welfare system. 8 See infra section I.B.2. And agency officials mine data about risk factors for child abuse and neglect to construct predictive analytics for family intervention. 9 See infra section I.B.3.

Despite this stark shift in family law, there is limited, albeit growing, scholarly attention to the subject. Some family law scholars have explored the use of empirical evidence in specific contexts, 10 See, e.g., Libby Adler, Just the Facts: The Perils of Expert Testimony and Findings of Fact in Gay Rights Litigation, 7 Unbound: Harv. J. Legal Left 1, 19–21 (2011) (noting the multiple downsides of the empirical battle over marriage equality, including that the facts found are not stably pro-gay and instead can be used by anti-gay advocates); Susan Frelich Appleton, Gender and Parentage: Family Law’s Equality Project in Our Empirical Age, in What Is Parenthood? Contemporary Debates About the Family 237, 245–48 (Linda C. McClain & Daniel Cere eds., 2013) (critiquing the use of empirical studies to determine the proper place of gender in parentage laws); Elizabeth Bartholet, Creating a Child-Friendly Child Welfare System: The Use and Misuse of Research, 13 Whittier J. Child & Fam. Advoc. 1, 10–14 (2014) (arguing that inaccurate statistical reports in a national study of child maltreatment—which found similar rates of maltreatment across races—fed claims of bias in the child welfare system and supported family-preservation efforts); Joan S. Meier, Johnson’s Differentiation Theory: Is It Really Empirically Supported? 12 J. Child Custody 4 (2015) [hereinafter Meier, Johnson’s Differentiation Theory] (analyzing the empirical basis for a purported typology of intimate partner violence and finding that the empirical evidence does not support the claimed differentiation). Helpfully, the Family Court Review published a “Researchers’ Roundtable,” addressing the problem of translating social science about families into legal rules. See Researchers’ Roundtable, 54 Fam. Ct. Rev. 134, 134–66 (2016). but few interrogate or analyze the larger trend. 11 There are two notable exceptions. Professor Margaret Brinig, one of the leading advocates and producers of empirical work in family law, has identified several reasons why research on families may not be reliable and may not translate well into legal rules and policies, including the following: population heterogeneity, the confidentiality or unavailability of data, bias by principal investigators, and the absence of control groups. See Margaret F. Brinig, Empirical Work in Family Law, 2002 U. Ill. L. Rev. 1083, 1084–94 [hereinafter Brinig, Empirical Work] (describing challenges in conducting reliable empirical work on family law and arguing that scholars and legislatures should neither respond too quickly to any single study nor overstate the likelihood that a law will change behavior). Brinig thus addresses questions about data reliability and the translation of empirical evidence by legal actors, both touched upon in sections II.B.1 and II.B.2. Brinig does not, however, address the other concerns of this Essay: the tendency of empirical evidence to obscure the importance of a range of values, conceal normative judgments, and normalize discrimination. See infra section II.B.2. Additionally, Professor Peggy Cooper Davis has explored an important aspect of an empirically based family law: the process by which judges absorb and use social science evidence in family law cases as they relate to legislative facts. See Peggy C. Davis, “There Is a Book Out . . .”: An Analysis of Judicial Absorption of Legislative Facts, 100 Harv. L. Rev. 1539, 1542 (1987) [hereinafter Davis, There Is a Book Out] (noting the article was “undertaken to identify the ways in which judges find and use legislative facts; to discover whether there are patterns of misuse; and to document the effect of legislative facts upon the development of law”). Davis does not, however, engage more broadly with the use of empirical evidence across the institutions of family law and many of the questions explored in this Essay. Perhaps most importantly, Davis does not explore the central concern of this Essay: that empirical evidence discourages a debate about competing values. See infra sections II.B.2, III.A. For an argument that family law sorely lacks an empirical basis and that family law scholars should do more to produce and engage with empirical work about families, see Carl E. Schneider & Lee E. Teitelbaum, Life’s Golden Tree: Empirical Scholarship and American Law, 2006 Utah L. Rev. 53, 78–91.

There is much to celebrate about this empirical turn in family law. Traditionally, decisionmakers in family law have drawn on a combination of moral judgments, prevailing cultural norms, and perceived common sense. 12 See infra section I.A. An empirical grounding for family law has considerable advantages over this historical approach. 13 See infra section II.A. A detailed understanding of family life and the legal system is essential to the development of effective rules. Rigorous consideration of empirical evidence can guide state investments, promoting the efficient and effective use of scarce resources, and it can give decisionmakers a clearer sense of areas in which legal inputs might yield particular social outcomes. It can help family law be more inclusive and move beyond narrow dominant norms. And it holds the potential to help depoliticize battles over family recognition and support, or at least to separate political arguments and social beliefs from the empirical evidence. 14 Indeed, with the new presidential Administration skeptical about well-settled science, evidence may be more important than ever. Perhaps the best example of President Trump’s willingness to ignore overwhelming scientific evidence is climate change. Compare Juliet Eilperin, Trump Says “Nobody Really Knows” If Climate Change Is Real, Wash. Post (Dec. 11, 2016), http://www.washingtonpost.com/news/energy-environment/wp/2016/12/11/trump-says-nobody-really-knows-if-climate-change-is-real/?utm_term=.9a2c5009c609 [http://perma.cc/9SV7-FHPC] (reporting then-President-elect Trump as saying about climate change, “I’m still open-minded. Nobody really knows”), with Scientific Consensus: Earth’s Climate Is Warming, Global Climate Change, NASA, http://climate.nasa.gov/scientific-consensus/ [http://perma.cc/YF6S-FUZQ] (last visited Sept. 11, 2017) (“[Ninety-seven] percent or more of actively publishing climate scientists agree: Climate-warming trends over the past century are extremely likely due to human activities.”). President Trump’s skepticism about the well-settled evidence on childhood vaccines, however, is a good example within family law. See Michael D. Shear et al., Anti-Vaccine Activist Says Trump Wants Him to Lead Panel on Immunization Safety, N.Y. Times (Jan. 10, 2017), http://www.nytimes.com/2017/01/10/us/politics/anti-vaccine-activist-trump-immunizations.html (on file with the Columbia Law Review) (describing how then-President-elect Trump met with Robert F. Kennedy, Jr., a vaccine skeptic, and asked him to lead a commission on vaccine safety). For a discussion of empirical evidence and the legal debate about mandatory vaccines, see infra text accompanying notes 237–244. It is not surprising, then, that Professor Kenji Yoshino would say, of legal contests that turn on legislative facts, “[l]et there be a trial.” 15 Yoshino, supra note 1, at 280 (emphasis omitted); see also Linda Greenhouse & Reva B. Siegel, The Difference a Whole Woman Makes: Protection for the Abortion Right After Whole Woman’s Health, 126 Yale L.J. Forum 149, 156–61 (2016), http://www.yalelawjournal.org/pdf/11.Greenhouse-SiegelFinalforPDF_io54a7ck.pdf [http://
perma.cc/4U7V-DMPL] (arguing a rigorous application of the undue burden test that draws on empirical evidence is a promising new front in the fight to protect reproductive rights).

But not so fast. Despite its considerable benefits, the empirical turn in family law also presents substantial concerns. As a threshold matter, there are well-rehearsed issues with the quality of the research and the capacity of legal actors to use empirical evidence in a nuanced manner. 16 See infra section II.B.1. These concerns take on a particular hue in the context of family law, in which research about families addresses complex questions such as the relationship between family structure and child outcomes. 17 See infra sections II.B.1–.2. Moreover, even the most sophisticated research can leave out variables that are difficult to quantify and yet are central to family life—love and a sense of belonging, distrust and a sense of dislocation. 18 See infra section II.B.1. In these ways, empirical evidence tells us something, but not everything, about family life.

More fundamentally, empirical evidence exerts a gravitational influ­ence on decisionmaking in a number of deeply troubling ways. To begin, the empirical turn focuses attention on the outcomes of legal rules. Many of these outcomes, notably child well-being and the reduction of family violence, embody important values, and family law is rightly focused on these concerns. But there are competing, and often contested, values also at play in family law, including equality, autonomy, pluralism, and inclusion, to mention but a few. 19 See infra section II.B.3. Simply naming the values at play in family law is a fraught endeavor, both because they are so numerous and because they often conflict. By mentioning a few in the text, and omitting others clearly at play, such as the privatization of dependency, the goal is not to elevate some values over others but rather to underscore that family law is heavily and inevitably value-laden. For a discussion of the changing values in family law, see Naomi R. Cahn, The Moral Complexities of Family Law, 50 Stan. L. Rev. 225, 227–29, 236–48 (1997) (reviewing Nancy E. Dowd, In Defense of Single-Parent Families (1997) & Barbara Dafoe Whitehead, The Divorce Culture (1997)) (arguing that, with the challenge to traditional morality as a justification for family law, there is a competing vision for family law—the “new family morality”—that embraces values including gender equality, caregiving, and commitment). The focus on the outcomes of legal rules discourages a forthright debate about these competing values and the tradeoffs inherent in any legal regulation. Moreover, even if decisionmakers address the full range of values at issue, empirical evidence does not tell decisionmakers how to weight the competing values. It can clarify the stakes in a debate and show how different policy options further different values, but empirical evidence does not help decisionmakers prioritize competing values and thus should not play an outsized role.

Compounding the problem, decisionmaking based on empirical evidence appears neutral, allowing legal actors to sidestep difficult and contentious debates, such as which families deserve legal recognition and support or which families should receive coercive intervention because of concerns about child abuse and neglect. 20 See infra section II.B.2.b. In the child welfare system, for example, the adoption of predictive analytics to triage suspected cases of child abuse and neglect allows agency officials to throw up their hands and claim they are simply following the algorithm, thus avoiding questions about whether the system improperly intervenes in the lives of low-income families of color. 21 See infra section III.B.2.

Finally, relying on empirical evidence poses particular dangers to nondominant families. The use of empirical evidence risks describing present conditions without interrogating the role of historical discrimination that continues to disadvantage some families. 22 See infra section II.B.2.c. Advocates challenging the heightened protections for removing Native American children from their homes in cases of abuse and neglect, for example, contend that this policy harms child well-being. 23 For a discussion of this example, see infra text accompanying notes 313–316 (explaining that the plaintiffs are challenging both the heightened removal standards and the preference for placement with Native American families). But evidence on child outcomes obscures the role of historical discrimination—indeed, genocide—against Native American families. Government policies are a direct cause of instability in Native American families, 24 See infra text accompanying notes 345–347. and uncritically adopting the evidence on outcomes replicates discrimination against this marginalized population.

To understand these concerns with empirical evidence, consider debates over custody and visitation rights for unmarried fathers. 25 See infra text accompanying notes 317–320. Unlike divorced and married fathers, there is not clear evidence that maintaining a relationship between unmarried fathers and their children improves child outcomes. 26 See infra text accompanying note 318. As courts and legislatures decide the rules for unmarried fathers, the absence of empirical evidence showing positive child outcomes is likely to be influential. But focusing on child well-being ignores the other value-based reasons for protecting father–child relationships, including gender equality and fathers’ liberty interests in the relationship. Foregrounding child outcomes provides a seemingly neutral rationale for the choice not to protect the parental rights of a socially marginalized group. And taking the empirical evidence at face value disregards government policies, such as mass incarceration, 27 See Michelle Alexander, The New Jim Crow: Mass Incarceration in the Age of Colorblindness 4 (rev. ed. 2012) (describing how mass incarceration operates “as a stunningly comprehensive and well-disguised system of racialized social control that functions in a manner strikingly similar to Jim Crow”). that make it harder for some unmarried fathers, particularly fathers of color, to play a beneficial social and economic role in their children’s lives. 28 See Donald Braman, Doing Time on the Outside: Incarceration and Family Life in Urban America 37–96, 177–87 (2007) (describing the impact of incarceration on prisoners’ families); Kimani Paul-Emile, Beyond Title VII: Rethinking Race, Ex-Offender Status, and Employment Discrimination in the Information Age, 100 Va. L. Rev. 893, 894–98 (2014) (describing the employment impact of a criminal record and noting that this disproportionately affects men of color). In short, there are reasons to be highly cautious about the use of empirical evidence in some contexts.

To be clear, this Essay does not condemn empirical evidence writ large. As noted throughout, empirical evidence does and should play a vital role in answering many family law questions and guiding family policies. But it is essential to have a nuanced understanding of what the empirical turn means for family law and to be ready to bend the arc to­ward its most promising trajectory.

How, then, should family law use empirical evidence? This Essay proposes a framework for taking advantage of the benefits of empirical evidence while also guarding against the significant concerns raised by the empirical turn. Not all questions are amenable to resolution through empirical analysis, and decisionmakers must know when and why to use this evidence. Decisionmakers should generally rely on empirical evidence when seeking to achieve a particular, agreed-upon outcome, such as reducing family violence, when the valence of the choice is relatively uncontested and when there is a general agreement about how to balance competing values. In this context, empirical evidence can guide choices among rules and policies, highlighting effective and effi­cient means for reducing violence. But contested and competing values inhere in family law, and it is critical to preserve space for debating these values. Empirical evidence can play a limited role in debates about values, but decisionmakers should not use it to avoid a debate about con­tested values and norms, nor should decisionmakers prioritize only those values that are more amenable to measurement or are more compelling because of evidence.

Even when empirical evidence is relevant, decisionmakers must be cautious about how they use it. To guide this nuance, this Essay’s framework calls for more effective gatekeeping mechanisms across the institutions of family law. It warns decisionmakers to be attentive to the potential for empirical evidence to reflect and refract the legal salience of intersecting identities, including race, gender, and class. And the framework encour­ages a robust role for legal scholars to make empirical evidence accessible and comprehensible for those crafting legal rules and policies.

This Essay focuses on the empirical turn in family law, but empirical evidence is now an entrenched feature of the legal system. 29 See infra text accompanying notes 38–43. By exploring the benefits and dangers of the empirical turn in one context, this Essay contributes to the broader debate about the use of empirical evidence trans-substantively in the law. 30 To give one example, Professor Jeffrey Rachlinski has argued that empirical evidence may make sense of that large debate in domains in which singularly focused outcomes can be reasonable organizing principles—as in medicine, with the goal of treating patients effectively, or business, with the goal of profit maximization. See Jeffrey J. Rachlinski, Evidence-Based Law, 96 Cornell L. Rev. 901, 917–19 (2011). In law, however, the endeavor is more fraught because law is inherently political, with often contested and competing goals. See id. Although some of the concerns identified in this Essay are family law specific, such as the general disregard for the Daubert test in family court, 31 See infra text accompanying notes 166–169 (describing the failure of most family court judges to use the Daubert test to screen out unreliable expert testimony). most of the issues are generalizable, and this Essay thus holds lessons for other areas of the law. 32 Thus, this Essay resists family law exceptionalism, see Janet Halley, What Is Family Law?: A Genealogy Part I, 23 Yale J.L. & Human. 1, 3–6 (2011) (describing and challenging family law exceptionalism), and instead uses family law as an example of the benefits, but also potential dangers, of the broader empirical turn.

The Essay proceeds as follows. Part I describes the increasingly widespread use of empirical evidence across many family law contexts. Part II unpacks the substantial benefits of, and significant concerns raised by, this empirical turn. Part III draws on this descriptive and analytical foundation to chart a path for the use of empirical evidence in family law.

I. Empirical Evidence Comes to Family Law

Empirical evidence is so thoroughly integrated into modern life that it is easy to overlook its existence. 33 Cf. Ludwig Wittgenstein, Philosophical Investigations § 129 (P. M. S. Hacker & Joachim Schulte eds., G. E. M. Anscombe et al. trans., 4th ed. 2009) (1953) (“The aspects of things that are most important for us are hidden because of their simplicity and familiarity.”). Decisionmakers in the public, 34 In the public sector, Mayor Michael Bloomberg was a leader in basing public policies on quantitative data, using them in multiple areas, including public health, building and transportation safety, climate change, and poverty. See Michael Flowers, Beyond Open Data: The Data-Driven City, in Beyond Transparency: Open Data and the Future of Civic Innovation 185, 187–95 (Brett Goldstein & Lauren Dyson eds., 2013) (describing New York City’s experience beginning to use data to drive decisions on multiple fronts including city infrastructure and building safety); Alan Feuer, The Mayor’s Geek Squad, N.Y. Times (Mar. 23, 2013), http://www.nytimes.com/
2013/03/24/nyregion/mayor-bloombergs-geek-squad.html (on file with the Columbia Law Review) (describing the use of data-driven decisionmaking in multiple policy areas under Mayor Bloomberg). But see Floyd v. City of New York, 959 F. Supp. 2d 540, 658–67 (S.D.N.Y. 2013) (finding the city’s data-driven stop-and-frisk policy unconstitutional). For further discussion of evidence-based policymaking in the public sector, see Ian Ayres, Super Crunchers: Why Thinking-by-Numbers Is the New Way to Be Smart 69–87 (2007) (describing the trend toward evidence-based policymaking in the United States and elsewhere). For a skeptical view on the ability of governments to synthesize and then translate data into social policy, see Peter Schuck, Why Government Fails So Often: And How It Can Do Better 161–67 (2014) (describing the challenges with this process, including the unavailability of data, the competing interpretations of data, and the uncertainty of how regulated entities will react to social policies).
private, 35 See Ayres, supra note 34, at 29–31 (describing this trend in business and using the example of Wal-Mart, which deploys data to make virtually every decision); Kristina McElheran & Erik Brynjolfsson, The Rise of Data-Driven Decision Making Is Real but Uneven, Harv. Bus. Rev. (Feb. 3, 2016), http://hbr.org/2016/02/the-rise-of-data-driven-decision-making-is-real-but-uneven [http://perma.cc/U9Z7-YR9B] (describing data-driven decisionmaking in the private sector, using manufacturing as a case study). and nonprofit sectors 36 See, e.g., Annual Letter 2013: Measuring Progress, Bill & Melinda Gates Found. (Jan. 2013), http://www.gatesfoundation.org/Who-We-Are/Resources-and-Media/Annual-Letters-List/Annual-Letter-2013 [http://perma.cc/MQW6-R6U8] (describing a core objective of the Gates Foundation’s approach to philanthropy as developing effective measurement tools to ensure that programs are furthering identified goals); Evidence-Based Decision Making, Laura & John Arnold Found., http://www.arnoldfoundation.org/initiative/evidence-based-policy-innovation/evidence-based-decision-making/ [http://perma.cc/5DLP-VYNY] (last visited Sept. 27, 2017) (“[M]any of the[] well-intentioned efforts [to address issues such as hunger, homelessness, and unemployment] have failed to produce adequate improvements. If we are to solve these problems, we must dramatically accelerate the pace at which we learn what works and insist on services that deliver measurable results.”). regularly rely on empirical evidence. 37 One of the first sectors of the economy to embrace empirical evidence was medicine. See Shirley Reynolds, The Anatomy of Evidence-Based Practice: Principles and Methods, in Evidence-Based Practice: A Critical Appraisal 17, 17–22 (Liz Trinder ed., 2000) (describing the history of evidence-based medicine). An acknowledgement of the widespread use of empirical evidence entered popular culture through Michael Lewis’s book about the professional baseball team, the Oakland A’s. See generally Michael Lewis, Moneyball: The Art of Winning an Unfair Game (2003). Similarly, the legal system has long integrated empirical evidence. 38 For a discussion of the distinction between evidence used to establish adjudicative versus legislative facts, see infra text accompanying notes 208–216. Since at least the era of the Brandeis brief, legal actors have drawn on data from the social and hard sciences. 39 See Brief for Defendant in Error, Muller v. Oregon, 208 U.S. 412 (1908) (No. 107) (available in full on HeinOnline). But see Noga Morag-Levine, Facts, Formalism, and the Brandeis Brief: The Origins of a Myth, 2013 U. Ill. L. Rev. 59, 61–62 (arguing the use of social science evidence in the courts predated the Brandeis brief). The use of social science data and methodologies is one of the distinguishing features of the Legal Realist movement. See Karl N. Llewellyn, Some Realism About Realism—Responding to Dean Pound, 44 Harv. L. Rev. 1222, 1247 (1931) (describing the difficulty of evaluating lower courts’ actions and noting “the techniques of the social sciences are being drawn upon and modified to make the work possible”). Economic analysis is thoroughly embedded in the law. 40 See United States v. Carroll Towing Co., 159 F.2d 169, 173 (2d Cir. 1947) (using cost-benefit analysis to determine negligence); Guido Calabresi, The Decision for Accidents: An Approach to Nonfault Allocation of Costs, 78 Harv. L. Rev. 713, 722–42 (1965) (using economic analysis in the context of accident law). The judicial system has developed methods for identifying and adjudicating the reliability of expert testimony. 41 See Fed. R. Evid. 702 (setting forth the standard for admission of expert testimony); Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579, 589–90 (1993) (setting forth a test for admitting expert testimony that requires evidence to be grounded in reliable scientific methodology and reasoning and must be relevant to the facts of the case); Alice B. Lustre, Annotation, Post-Daubert Standards for Admissibility of Scientific and Other Expert Evidence in State Courts, 90 A.L.R. 5th 453 §§ 1–2 (2001) (describing the adoption of the Daubert standard and alternatives in state courts). And legal scholars have embraced empirical work across multiple fields, 42 See Daniel E. Ho & Larry Kramer, Introduction: The Empirical Revolution in Law, 65 Stan. L. Rev. 1195, 1195–97 (2013) (introducing a special issue dedicated to identifying and exploring the many ways legal scholars and the legal system integrate empirical evidence across multiple fields). with an entire field of empirical legal studies dedicated to producing, not simply using, empirical evidence. 43 See Society for Empirical Legal Studies, Cornell Univ. Law Sch., http://
www.lawschool.cornell.edu/sels/ [http://perma.cc/MH4L-GJMH] (last visited Sept. 11, 2017) (describing the society’s purpose, sponsored events, scholarship, and Journal of Empirical Legal Studies).

As this Part shows, family law increasingly embraces empirical evidence as well. Beginning in the second half of the twentieth century, and accelerating over the past several decades, family law has regularly drawn on empirical evidence, profoundly changing the process of judg­ing, legislating, and administering the law. 44 See infra section I.B. But as the first section shows, this was not always the case.

A. The Historical Baseline

Family autonomy is one of the animating principles at the heart of family law. When the Supreme Court first held in the early twentieth cen­tury that there are limits on the state’s power to interfere with parental decisionmaking, the Court relied only on the language of rights, not em­pirical evidence. In a pair of cases, the Court held that, under the Due Process Clause of the Fourteenth Amendment, the state cannot unduly burden the “liberty of parents and guardians to direct the upbringing and education of children under their control.” 45 Pierce v. Soc’y of Sisters, 268 U.S. 510, 534–35 (1925); see also Meyer v. Nebraska, 262 U.S. 390, 400 (1923) (establishing a liberty interest in parental control of children’s education). The Court found that “[t]he child is not the mere creature of the State; those who nurture him and direct his destiny have the right, coupled with the high duty, to recognize and prepare him for additional obligations.” 46 Pierce, 268 U.S. at 535.

Later in the century, when the Court qualified parental rights by acknowledging children’s interests, again, it did not cite evidence and instead relied on perceived common sense. In Parham v. J.R., the Court held that the law could presume that parents make medical decisions to further their children’s welfare, and thus children are not constitutionally entitled to formal adversarial proceedings when parents seek to commit them to psychiatric hospitals. 47 442 U.S. 584, 602–03 (1979). The Court justified this presumption by claiming that “natural bonds of affection lead parents to act in the best interests of their children” 48 Id. at 602. and that “pages of human experience . . . teach that parents generally do act in the child’s best interests.” 49 Id. at 602–03. The notion that love between parents and children will lead parents to care for their children has deep roots. See 1 William Blackstone, Commentaries *434–35 (“[P]rovidence has done it more effectually than any laws, by implanting in the breast of every parent . . . [an] insuperable degree of affection, which not even the deformity of person or mind, not even the wickedness, ingratitude, and rebellion of children, can totally suppress or extinguish.”); 2 James Kent, Commentaries on American Law 160 (1826) (“The obligation of parental duty is so well secured by the strength of natural affection, that it seldom requires to be enforced by human laws.”); see also id. at 159 (“The wants and weaknesses of children render it necessary that some person maintain them, and the voice of nature has pointed out the parent as the most fit and proper person.”).

More broadly, decisionmakers in family law established and justified legal rules by relying on traditional morality and norms, such as the need to police sexuality outside marriage and reinforce gender roles within marriage. 50 See Carl E. Schneider, Moral Discourse and the Transformation of American Family Law, 83 Mich. L. Rev. 1803, 1808–19 (1985) (describing moral justification as a traditional basis for family law). These values were widely accepted, and decisionmakers did not generally invoke empirical evidence to support laws furthering these values and norms. 51 Cf. id. at 1807–19 (describing the historical basis for family regulation—conventional morality—and the shift away from this grounding since the 1960s).

Even when the Court was presented with empirical evidence in family law cases, it often did not rely on it, turning instead to basic values. In Loving v. Virginia, the state argued the Court should defer to the legis­lature because there was “conflicting scientific opinion [about] the effects of interracial marriage, and the desirability of preventing such alliances, from the physical, biological, genetic, anthropological, cultural, psychological and sociological point of view.” 52 Brief and Appendix on Behalf of Appellee at 7, Loving v. Virginia, 388 U.S. 1 (1967) (No. 395), 1967 WL 93641. For a description of this aspect of the litigation, see Linda C. McClain, Prejudice, Moral Progress, and Not Being “On the Wrong Side of History”: Debating the Legacy of Loving for the Right to Marry 6–15 (Mar. 17, 2017) (unpublished manuscript) (on file with the Columbia Law Review). The Court did not engage with this evidence and instead found that Virginia’s law, which restricted intermarriage with whites and not along other racial lines, could be understood only as an expression of white supremacy and therefore was invidious racial discrimination. 53 Loving, 388 U.S. at 11. Similarly, in Palmore v. Sidoti—a custody battle between two white parents, with the father chal­lenging the mother’s custody because her new husband was Black—the Court was uninterested in whether a child raised by an interracial couple might suffer stigma and harm; instead, the Court focused on the importance of race-neutral decisionmaking. 54 466 U.S. 429, 433 (1984) (“The question, however, is whether the reality of private biases and the possible injury they might inflict are permissible considerations for removal of an infant child from the custody of its natural mother. We have little difficulty concluding that they are not.”).

Outside of constitutional law, courts also relied on common sense and traditional norms, not empirical evidence. For instance, the spousal immunity privilege permits a spouse to refuse to provide adverse testi­mony in a criminal trial of the other spouse, 55 See Trammel v. United States, 445 U.S. 40, 53 (1980) (“[T]he witness-spouse alone has a privilege to refuse to testify adversely; the witness may be neither compelled to testify nor foreclosed from testifying.”). There are numerous nuances not relevant here, such as the differences between federal and state law. For details, see Milton C. Regan, Jr., Spousal Privilege and the Meanings of Marriage, 81 Va. L. Rev. 2045, 2052–55 (1995). and the marital communications privilege protects confidential communications between spouses. 56 See Blau v. United States, 340 U.S. 332, 334 (1951) (holding a defendant’s refusal to reveal his wife’s location, which she may have confidentially shared with him, was lawful); Wolfle v. United States, 291 U.S. 7, 14 (1934) (noting the “basis of the immunity given to communications between husband and wife is the protection of marital confidences”). For a discussion of both spousal privileges, see Dan Markel et al., Criminal Justice and the Challenge of Family Ties, 2007 U. Ill. L. Rev. 1147, 1168–69. Courts justified these privileges by claiming that they promote marital harmony and solidarity. 57 See Wolfle, 291 U.S. at 14 (stating the marital communications privilege was “regarded as so essential to the preservation of the marriage relationship as to outweigh the disadvantages to the administration of justice”); United States v. Armstrong, 476 F.2d 313, 315 (5th Cir. 1973) (noting the spousal immunity privilege “preserve[s] family peace by preventing husband and wife from becoming adversaries in a criminal proceeding”). But courts did not cite any evidence to support the contention that testifying against each other or breaching marital confidences would introduce strife and distrust into marriage. 58 See, e.g., Wolfle, 291 U.S. at 14 (citing the reasoning for the privilege but providing no supporting evidence to suggest that it is necessary to preserve marital harmony). Instead, courts relied on common sense and traditional notions of marriage. 59 In Trammel, the Court reasoned that “[w]hen one spouse is willing to testify against the other in a criminal proceeding . . . their relationship is almost certainly in disrepair; there is probably little in the way of marital harmony for the privilege to preserve.” Trammel, 445 U.S. at 52.

When legal actors did invoke empirical evidence, it was generally pseudoscience. Consider the odious history of sterilization programs. From 1929 to 1975, North Carolina sterilized approximately 7,600 people, 60 Daniel J. Kevles, In the Name of Eugenics: Genetics and the Uses of Human Heredity 92–94 (1985); Adam Owens, N.C. Dedicates Marker to Eugenics Program, WRAL (June 22, 2009), http://www.wral.com/news/local/story/5406081/ [http://perma.cc/
HP3U-9PYQ]; Victims of State Sterilization Tell Their Stories, WRAL (June 30, 2011), http://www.wral.com/news/video/9755940/#/vid9755940 [http://perma.cc/YV4Z-8YAX].
targeting low-income women of color and people with low IQ test scores or low levels of education. 61 Clarence J. Gamble, Eugenic Sterilization in North Carolina, 12 N.C. Med. J. 550, 550–51 (1951). Programs like this were justified by the “science” of phrenology and the like, purportedly proving the inferiority of people of color and low-income populations. 62 See Robert V. Guthrie, Even the Rat Was White: A Historical View of Psychology 92–104 (1976); Kevles, supra note 60, at 92–94.

In sum, the traditional mode of analysis and the justification for legal rules was not empirical, at least as we understand the term today. As the next section describes, this absence of empirical evidence did not last.

B. Empirical Evidence Across the Institutions of Family Law

In stark contrast to the historical baseline, empirical analysis in family law is now widespread. This trend is consistent with the empiricization of law generally, but it also responds to a particular demand in family law. In the last part of the twentieth century, the Supreme Court largely rejected traditional morality and dominant norms as acceptable justifications for family law. 63 For two, among many, such decisions, see Clark v. Jeter, 486 U.S. 456, 461–65 (1988) (adopting intermediate scrutiny for statutory distinctions based on the marital status of parents and discussing earlier cases suggesting heightened scrutiny was warranted for classifications based on illegitimacy); Orr v. Orr, 440 U.S. 268, 281–83 (1979) (striking down a state statute authorizing the award of alimony for only women, not men). It was thus necessary to find new justifications for the regulation of families, creating an opening for empirical evidence. Additionally, as family norms rapidly changed during the same period—including an increase in divorce, a rise in cohabitation, and more childbearing outside of marriage 64 See Clare Huntington, Failure to Flourish: How Law Undermines Family Relationships 28–31 (2014) [hereinafter Huntington, Failure to Flourish] (describing these changes). —societal consensus about family values began to wane. 65 See Cahn, supra note 19, at 227–29, 236–49. Empirical evidence thus appealed as a seemingly neutral basis for decisionmaking. Responding to these changes, courts and legislatures embraced empirical evidence, with psychological theories about parents and children fundamentally shaping child custody laws 66 See Painter v. Bannister, 140 N.W.2d 152, 156–58 (Iowa 1966) (reviewing the social science literature and drawing on the then-dominant theory of psychological parenthood to determine that the child’s best interests would be served by remaining in the care of his grandparents, with whom the child had lived for nearly three years, rather than the biological father); Davis, There Is a Book Out, supra note 11, at 1542–47 (describing how social science research influenced the best-interests standard beginning in the 1960s). and research on child abuse leading to mandatory reporting laws and the modern child welfare system. 67 C. Henry Kempe, et al., The Battered-Child Syndrome, 181 J. Am. Med. Ass’n 17, 23 (1962) (describing the results of a national study of injured children and defining a new diagnosis—Battered Child Syndrome); Lois A. Weithorn, Protecting Children from Exposure to Domestic Violence: The Use and Abuse of Child Maltreatment Statutes, 53 Hastings L.J. 1, 55–60 (2001) (describing the Battered Child Syndrome study and its pivotal role in the creation of the modern child welfare system). Another pioneer was Daniel Patrick Moynihan, who was an early advocate for using social science to inform government policy. See James T. Patterson, Freedom Is Not Enough: The Moynihan Report and America’s Struggle over Black Family Life—From LBJ to Obama 14 (2010) (describing Moynihan’s efforts to “put into practice his faith in the capacity of expert social scientists to fashion public policy”). Some examples of influential empirical evidence in family law over the succeeding decades include the work of Professors Martha Fineman and Robert Mnookin. See Martha Albertson Fineman, The Illusion of Equality: The Rhetoric and Reality of Divorce Reform 18–20, 32–33, 55, 61–73 (1991) (describing the negative effect of no-fault divorce on women); Robert H. Mnookin, Child-Custody Adjudication: Judicial Functions in the Face of Indeterminacy, Law & Contemp. Probs., Summer 1975, at 226, 229–30 (documenting the inconsistent, unpredictable, and undesirable consequences of the best-interests standard in child welfare and child custody cases).

This section describes the empirical turn in three contexts: litigation, legislation, and administration. As this section shows, both the type and quality of empirical evidence vary. Decisionmakers use evidence from the hard sciences, demographic statistics about changes in family form, and social science studies about the relationship between those changes and child outcomes. Some empirical evidence satisfies scientific standards for reliability, but other evidence decidedly does not. Although the empirical turn reaches across all of family law, this section describes it in detail in the context of litigation, with a particular focus on the marriage equality cases. This fine-grained description of the marriage equality litigation lays the groundwork for the critique of empirical evidence in Part II and the proposed framework in Part III.

1. Litigation

a. Marriage Equality. — In the early days of the marriage equality movement, 68 See Douglas NeJaime, Before Marriage: The Unexplored History of Nonmarital Recognition and Its Relationship to Marriage, 102 Calif. L. Rev. 87, 117–46 (2014) (arguing that even before advocates began making explicit and sustained claims for marriage in the 1990s, marriage still shaped the battle for relationship recognition—domestic partnerships and civil unions—with advocates patterning these models of relationship recognition on marriage).
This Essay focuses on the marriage equality litigation as an example of the importance of empirical evidence in LGBT family rights because the issue had such widespread social salience. But the evidentiary battle played out in related cases as well, such as litigation over the adoption rights of LGBT adults. These cases had many of the same hallmarks. For example, in a federal case challenging Florida’s ban on adoption by “homosexuals,” Fla. Stat. § 63.042(3) (2014) (amended 2015), the Eleventh Circuit upheld the state law, finding that the State Legislature could rationally conclude that the ban was necessary because there was no conclusive evidence that children do not benefit from growing up with two different-sex, married parents. Lofton v. Sec’y of the Dep’t of Children & Family Servs., 358 F.3d 804, 825 (11th Cir. 2004). The court stated:
“[W]e must ask not whether the latest in social science research and professional opinion support the decision of the Florida legislature, but whether that evidence is so well established and so far beyond dispute that it would be irrational for the Florida legislature to believe that the interests of its children are best served by not permitting homosexual adoption.”
Id. The court also noted the absence of longitudinal studies following subjects into adulthood, concluding that the Legislature could find the relevant research insufficiently developed to rely upon, see id. at 826, and that “the question of the effects of homosexual parenting . . . is one on which even experts of good faith reasonably disagree,” id. The court determined that the Legislature could rationally rely on the “unprovable assumption” that a married man and woman is the “optimal social structure” for childrearing, and thus, the State could prefer this family structure to others in the context of adoption. Id. at 819–20, 826.
Six years later, the Florida District Court of Appeal struck down the adoption ban, relying heavily on empirical evidence. See Fla. Dep’t of Children & Families v. Adoption of X.X.G., 45 So. 3d 79, 91–92 (Fla. Dist. Ct. App. 2010). The court referred to the trial court’s finding that rigorous research, including longitudinal studies that followed participants for up to fourteen years, demonstrated no difference between parenting by LGBT and heterosexual parents and no difference in the adjustment of children raised by same-sex and different-sex parents. See id. at 86–87. The trial court had closely examined the evidence, rejecting one expert’s analysis of social science studies because other experts testified that the analysis had fundamental errors and most of the scientific community disagreed with the analysis. See id. at 88. The trial court also rejected expert testimony that LGBT adults have a higher lifetime prevalence of certain mood and substance disorders, concluding that if every demographic group with elevated rates of these disorders was excluded from adopting, then only Asian American men would be allowed to adopt. See id. at 89.
the debate did not focus on children. 69 See NeJaime, supra note 68, at 117–21, 151 (describing the arguments made for and against relationship recognition, which centered primarily on the intimate bond between partners and their economic interdependency, and further noting that social conservatives were worried about the impact of domestic partnerships on the traditional definition of marriage). In the 1980s, however, a significant number of lesbians had begun to conceive children and raise them with a partner, 70 See George Chauncey, Why Marriage? The History Shaping Today’s Debate over Gay Equality 105 (2004) (“[T]he lesbian baby boom of the 1980s represented something new: a generation of women who lived openly as lesbians and no longer felt obliged to marry a man in order to have a child.”). and advocates began to argue that same-sex relationships should be recognized to help protect the children. 71 Id. Many LGBT parents were already raising children, but they were generally raising children conceived in previous different-sex relationships. Id. The change in the 1980s was that lesbians began conceiving children within same-sex relationships. Id.

With LGBT parenting as a central component of the marriage equality movement, advocates on both sides began to make arguments about the quality of that parenting. 72 This began with the litigation in Hawaii. Addressing a challenge to the constitutionality of the State’s law, Haw. Rev. Stat. Ann. § 572-1 (LexisNexis 2015), the Hawaii Supreme Court found the law discriminated on the basis of sex and therefore was subject to heightened scrutiny. See Baehr v. Lewin, 852 P.2d 44, 65–66 (Haw. 1993). The court remanded the case to allow the State to introduce evidence to satisfy this standard. In a bench trial, the plaintiffs presented expert witnesses to testify about the effects of same-sex parenting on child development. See Baehr v. Miike, No. 91-1394, 1996 WL 694235, at *10–16 (Haw. Cir. Ct. Dec. 3, 1996), rev’d mem., 994 P.2d 566 (Haw. 1999). The trial court found that the testimony presented by two experts for the plaintiffs, a sociologist and a psychologist, was “especially credible,” id. at *10, and that the State had not produced sufficient evidence to establish adverse public consequences from allowing same-sex couples to marry or that traditional marriage needed to be protected, see id. at *16–17. After noting that the evidence indicated that the most important factor in child development was the quality of the parent–child relationship, the court concluded that sexual orientation is not an indicator of parental fitness. See id. at *17. Advocates had a ready source of social scientific evidence—studies on LGBT parenting that had been conducted in response to custody battles in the 1970s. 73 Marie-Amélie George, The Custody Crucible: The Development of Scientific Authority About Gay and Lesbian Parents, 34 Law & Hist. Rev. 487, 493–99 (2016) (describing cases in which courts were deciding whether to award custody to the heterosexual parent or the parent who had begun a relationship with a same-sex partner). At first, opponents of marriage equality argued that LGBT parents harmed their children. 74 Edward Stein, The “Accidental Procreation” Argument for Withholding Legal Recognition for Same-Sex Relationships, 84 Chi.-Kent L. Rev. 403, 408 (2009) (“This argument, which I call the ‘gays make bad parents’ argument, was embraced in some form by all three appellate courts that heard challenges to prohibitions against same-sex marriages in the 1970s.”). Over time, this argument morphed into a claim that even if children were not actively harmed, the optimal childrearing environment for a child was with two different-sex, married parents. 75 See, e.g., Hernandez v. Robles, 855 N.E.2d 1, 7 (N.Y. 2006) (crediting the argument that “it is better, other things being equal, for children to grow up with both a mother and a father”).

After the Supreme Court held in 2003 that the State could not reflexively draw on traditional values to regulate lesbians and gay men, at least in criminal law, 76 Lawrence v. Texas, 539 U.S. 558, 571 (2003) (noting that “for centuries there have been powerful voices to condemn homosexual conduct as immoral” but that “[t]he issue is whether the majority may use the power of the State to enforce these views on the whole society through operation of the criminal law”). the focus on empirical evidence became all the more important. The state needed to show a reason, beyond moral judgments, for keeping same-sex couples out of marriage. Ultimately, opponents of marriage equality abandoned the comparative argument about same-sex parents and different-sex parents entirely and instead cited a new outcomes-based justification for the restriction: that the state had an interest in channeling procreative sex into marriage to ensure a child had two parents, and thus the state could privilege different-sex marriage as a way of inducing these couples to marry. 77 See Baskin v. Bogan, 766 F.3d 648, 660 (7th Cir. 2014) (describing the prevention of nonmarital childbearing as the sole argument advanced by Indiana); Perry v. Brown, 671 F.3d 1052, 1086 (9th Cir. 2012), vacated and remanded sub nom. Hollingsworth v. Perry, 133 S. Ct. 2652 (2013) (“The primary rationale Proponents offer for Proposition 8 is that it advances California’s interest in responsible procreation and childrearing.”).

Advocates of marriage equality made their own empirically grounded claims. Citing demographic evidence, they demonstrated that same-sex couples were raising children in increasing numbers. 78 See infra note 202. Drawing on social science evidence, they argued that children raised by same-sex couples have similar outcomes to children raised by different-sex couples. 79 See infra text accompanying notes 97, 112. For a summary of the underlying studies, see Carlos A. Ball, Social Science Studies and the Children of Lesbians and Gay Men: The Rational Basis Perspective, 21 Wm. & Mary Bill Rts. J. 691, 702–15 (2013). Relying on economic and social science evidence, they con­tended that marriage provides economic and emotional stability to children. 80 See infra text accompanying note 97. And, finally, looking to social science research, advocates posited that children raised by same-sex couples would benefit from their parents’ access to marriage. 81 See infra text accompanying notes 97–98, 112.

Some courts were skeptical about the relevance of this empirical evidence, especially early in the movement. In the state court litigation over New York’s marriage restriction, 82 See Hernandez v. Robles, 855 N.E.2d 1, 5 (N.Y. 2006) (describing the history of this litigation and noting that the trial courts in four cases granted summary judgment, one in favor of the plaintiffs challenging the New York law and three in favor of the State). the Court of Appeals applied rational basis review and found the different-sex requirement did not violate the state constitution. 83 See id. at 9–12. In a plurality opinion, the court invoked family law’s traditional methodology, relying on “the undisputed assumption that marriage is important to the welfare of children.” 84 Id. at 7. On this basis, the court held that the Legislature could rationally decide that different-sex couples are far more likely than same-sex couples to procreate and could seek to stabilize these families to channel procreation into marriage. 85 Id. The court likewise held that the Legislature could rationally conclude that it is better for a child to grow up with a man and a woman as the two parents. 86 Id. The court cited no evidence, instead stating that “[i]ntuition and experience suggest that a child benefits from having before his or her eyes, every day, living models of what both a man and a woman are like.” 87 Id. In response to the plaintiffs’ argument that the optimal childrearing proposition was factually untrue, the court dismissed any social science evidence as inconclusive, and thus “the Legislature could rationally think otherwise.” 88 Id. at 7–8 (“[T]he studies . . . do not establish beyond doubt that children fare equally well in same-sex and opposite-sex households . . . . In the absence of conclusive scientific evidence, the Legislature could rationally proceed on the commonsense premise that children will do best with a mother and father in the home.”).

In many other cases, however, the empirical evidence played a critical role. 89 See, e.g., Varnum v. Brien, 763 N.W.2d 862, 896, 899–901 (Iowa 2009) (applying intermediate scrutiny to Iowa’s marriage restriction and concluding that the governmental justification that different-sex parents provide children with the optimal childrearing environment did not pass). In one of the final lower court decisions striking down different-sex marriage requirements, Judge Posner eviscerated arguments made by Indiana and Wisconsin in support of their marriage restrictions: He concluded that “more than unsupported conjecture that same-sex marriage will harm heterosexual marriage or children or any other valid and important interest of a state is necessary to justify discrimination on the basis of sexual orientation.” Baskin v. Bogan, 766 F.3d 648, 671 (7th Cir. 2014). Indeed, he found that “the grounds advanced by Indiana and Wisconsin for their discriminatory policies are not only conjectural; they are totally implausible.” Id. Often courts considered this evidence at summary judg­ment, 90 See Varnum, 763 N.W.2d at 899 (citing the “abundance of evidence and research . . . supporting the proposition that the interests of children are served equally by same-sex parents and opposite-sex parents” and noting that the “opinions that dual-gender parenting is the optimal environment for children” are “largely unsupported by reliable scientific studies”). but in two of the three cases that made it to the U.S. Supreme Court, the trial court conducted lengthy trials, developing a rich factual record that largely turned on empirical evidence. 91 This section discusses the trials in Perry v. Schwarzenegger, 704 F. Supp. 2d 921, 929 (N.D. Cal. 2010), aff’d sub nom. Perry v. Brown, 671 F.3d 1052 (9th Cir. 2012), vacated and remanded sub nom. Hollingsworth v. Perry, 133 S. Ct. 2652 (2013), and DeBoer v. Snyder, 973 F. Supp. 2d 757, 761–68 (E.D. Mich.), rev’d, 772 F.3d 388 (6th Cir. 2014), rev’d sub nom. Obergefell v. Hodges, 135 S. Ct. 2584 (2015). The trial court in Windsor v. United States did not conduct a trial and instead granted summary judgment to the plaintiff. See 833 F. Supp. 2d 394, 396 (S.D.N.Y.), aff’d, 699 F.3d 169 (2d Cir. 2012), aff’d, 133 S. Ct. 2675 (2013). In the first such case, Perry v. Schwarzenegger, the federal challenge to California’s constitutional amendment limiting marriage to different-sex couples, 92 Cal. Const. art. I, § 7.5 (“Only marriage between a man and a woman is valid or recognized in California.”), ruled unconstitutional in Perry, 704 F. Supp. 2d 921. the trial court held a twelve-day bench proceeding. 93 See 704 F. Supp. 2d at 929. The effect on child outcomes of being raised by two married, different-sex parents was a central issue in the litigation. 94 In proposing the amendment to the state constitution, the ballot initiative contained this explanatory language in favor of the amendment: “[T]he best situation for a child is to be raised by a married mother and father.” Ron Prentice et al., Argument in Favor of Proposition 8, in California General Election: Official Voter Information Guide, General Election Ballot 56 (2008), http://vig.cdn.sos.ca.gov/2008/general/pdf-guide/vig-nov-2008-principal.pdf [http://perma.cc/H9PQ-HXYE]. At trial, the defendants continued this line of argument, contending that limiting marriage to different-sex couples “[p]romotes ‘statistically optimal’ child-rearing households; that is, households in which children are raised by a man and a woman married to each other.” Perry, 704 F. Supp. 2d at 931 (quoting defendants’ written submissions).

The trial was largely a battle of experts. The plaintiffs produced nine experts, and the defendants produced two. 95 Perry, 704 F. Supp. 2d at 932. The plaintiffs’ experts included two historians; one testified about the historical meaning of marriage, and the other put California’s constitutional amendment in the historical context of discrimination against LGBT people. 96 See id. at 933–37. Three psychologists were called to testify: one about the evidence on LGBT parenting, one about the physical and economic benefits of marriage, and one about the nature of sexual orientation. 97 See id. at 934–36. Two economists weighed in on the economic benefits to a state flowing from marriage as compared with domestic partnerships and the economic benefits of marriage to the couple. 98 See id. at 934–36, 938. A social epidemiologist testified about the increased risk of physical and mental harms for gays and lesbians as a result of the constitutional amendment. 99 See id. at 935–36. And a political scientist testified about the extent to which homophobia infects the political process. 100 See id. at 937. In response, proponents 101 The plaintiffs had sued the Governor and Attorney General of California as well as several other government officials, but none of the defendants was willing to defend the constitutional amendment. Id. at 928. Thus, five proponents of the constitutional amendment argued in favor of the amendment. See id. at 928, 954. of the ballot initiative that led to the constitutional amendment introduced the founder and president of the Institute for American Values; opining as an expert, he testified about marriage, fatherhood, and family structure, contending that children do best when raised by married, biological parents. 102 Id. at 945–50. The proponents also introduced a political science expert, who testified about the political power of LGBT people in California. See id. at 950–52.

The trial court engaged in a lengthy analysis of this empirical evidence. The court dissected the methodological and substantive components of the underlying social science, focusing on sample size, replicability, and so on. 103 The court noted, for example, that the studies finding that married different-sex parents provide the optimal childrearing environment did not explore the outcomes for children raised in a stable same-sex household, id. at 935, and thus these studies “do not inform conclusions about outcomes for children raised by same-sex parents in stable, long-term relationships,” id. at 981. The court readily accepted the plaintiffs’ ex­pert testimony, 104 See id. at 940–44. but it found that the opinions of the proponents’ experts were “not supported by reliable evidence or methodology” and therefore were “entitled to essentially no weight.” 105 Id. at 950; see also id. at 947, 952.

This evidentiary battle was dispositive in the resulting decision, with the court concluding that “[t]he trial evidence provides no basis for establishing that California has an interest in refusing to recognize marriage between two people because of their sex” 106 Id. at 934. and that “the evidence presented at trial fatally undermines the premises underlying proponents’ proffered rationales for” the constitutional amendment. 107 Id. at 938. The trial court required only some evidence to support the constitutional amendment: “An initiative measure adopted by the voters deserves great respect. The considered views and opinions of even the most highly qualified scholars and experts seldom outweigh the determinations of the voters. When challenged, however, the voters’ determinations must find at least some support in evidence.” Id. Speaking directly to the question of LGBT parenting, the court found that “[c]hildren raised by gay or lesbian parents are as likely as children raised by heterosexual parents to be healthy, successful and well-adjusted. The research supporting this conclusion is accepted beyond serious debate in the field of developmental psychology.” 108 Id. at 980. The trial court made eighty findings of fact, including that children raised by same-sex parents benefit economically and psychologically when their parents are able to marry, see id. at 973, that the gender and sexual orientation of the parent do not affect child outcomes, see id. at 980, and that having two different-sex parents does not increase the likelihood a child will have positive life outcomes, see id. at 981. The court thus concluded that California’s different-sex limitation on marriage violated both the Due Process Clause and the Equal Protection Clause of the U.S. Constitution. 109 See id. at 991–1003.

The nine-day bench trial challenging Michigan’s constitutional amendment, 110 Mich. Const. art. I, § 25 (“To secure and preserve the benefits of marriage for our society and for future generations of children, the union of one man and one woman in marriage shall be the only agreement recognized as a marriage or similar union for any purpose.”); DeBoer v. Snyder, 772 F.3d 388, 397 (6th Cir. 2014) (stating that the trial court “held a nine-day trial on the issue”), rev’d sub nom. Obergefell v. Hodges, 135 S. Ct. 2584 (2015). DeBoer v. Snyder, was similarly replete with expert testi­mony and debates about methodology, sample sizes, and correlation versus causation. 111 973 F. Supp. 2d 757 (E.D. Mich. 2014), rev’d, 772 F.3d 388, rev’d sub nom. Obergefell, 135 S. Ct. 2584. As in California, the plaintiffs in the Michigan litigation contended that the limitation violated the federal Due Process Clause and Equal Protection Clause. See id. at 760. Michigan defended the provision by arguing that it served four legitimate purposes: giving children “‘biologically connected’ role models of both genders that are necessary to foster healthy psychological development,” “avoiding the unintended consequences that might result from redefining marriage,” “upholding tradition and morality,” and channeling procreation into stable relationships. See id. The court assumed the standard of review was rational basis. See id. at 760–61. The plaintiffs introduced empirical evidence about families headed by same-sex parents, academic achievement for children raised by same-sex couples, an ongoing study about relationship stability, and so on. 112 See id. at 761–64. Defendants introduced competing empirical evidence on the same subjects. 113 See id. at 765–67. As in Perry, the trial court meticulously scrutinized this empirical evidence. The court acknowledged that the plaintiffs’ studies used convenience samples—small, self-selected populations rather than large, representative samples—but noted that this was the standard methodology in the relevant fields. 114 See id. at 761–62. By contrast, the court criticized the defendants’ evidence because it compared children who had not experienced a family breakup with those who had; the court noted that the comparison should be between children raised in stable homes with different-sex parents and children raised in stable homes with same-sex parents. 115 See id. at 765. The court further found that one of the central studies relied upon by the defendants had been funded by a party who was certain the study would show the value of different-sex marriage, which undermined the credibility of the study. 116 See id. at 766. For these reasons, the court found the testimony of the defendants’ experts “entirely unbelievable and not worthy of serious consideration.” 117 Id. at 766–68. The trial court found the testimony of all three experts “clearly represent[s] a fringe viewpoint that is rejected by the vast majority of their colleagues across a variety of social science fields.” Id. at 768.

This empirical evidence was critical to the court’s ruling, with the court concluding that the Michigan constitutional amendment could not even pass the rational basis test under the Equal Protection Clause. 118 See id. at 768–69. In light of the holding on equal protection grounds, the court declined to address the due process argument. See id. at 768. The court rejected all of the State’s rationales, 119 See id. at 770. including the argument that married different-sex parents provide the optimal environment for raising children. The court found that there was no evidence that children benefit from being raised by married different-sex parents and that even if it were true, carrying this argument to its logical conclusion would lead the State to restrict marriage for demographic groups that are correlated with poor outcomes for children, such as minority and low-income families, a proposition the court called an “absurdity.” 120 Id. at 770–72.

By the time marriage equality came before the U.S. Supreme Court, the empirical decisionmaking in the lower courts had largely settled the question about same-sex parenting and child outcomes, and there was no reason to rehash the debate. 121 Indeed, the defenders of the different-sex requirement did not make evidence on same-sex parenting a central issue. See, e.g., Brief of Petitioners at 31–48, Hollingsworth v. Perry, 133 S. Ct. 2652 (2013) (No. 12-144), 2013 WL 457384 [hereinafter Perry Petitioners Brief]. Instead, they relied on other arguments, notably that the state has a particular interest in channeling procreative sex into marriage, that caution counsels in favor of a go-slow approach, and that the issue should be left to the democratic process. See, e.g., Brief for Respondent at 11–35, Obergefell v. Hodges, 135 S. Ct. 2584 (2015) (No. 14-556), 2015 WL 1384100; Perry Petitioners Brief, supra, at 31–61.
In Perry, at least some of the Justices wanted to revisit the empirical question. At oral argument, for example, Justice Scalia noted that “there’s considerable disagreement among . . . sociologists as to what the consequences [are] of raising a child in a . . . single-sex family, [and] whether that is harmful to the child or not,” see Transcript of Oral Argument at 19, Perry, 133 S. Ct. 2652 (No. 12-144), 2013 WL 6908183, and “that there’s no scientific answer to that question at this point in time,” see id. at 20. Justice Kennedy acknowledged that the “sociological information is new” but that the Court should focus on the injury to the children of same-sex parents. See id. at 21. The case was ultimately decided on jurisdictional grounds, see Perry, 133 S. Ct. at 2668, so the opinion did not mention the evidence.
In Obergefell v. Hodges, although some amici addressed the empirical evidence directly, see, e.g., Brief of Amicus Curiae American Sociological Association in Support of Petitioners at 5–27, Obergefell, 135 S. Ct. 2584 (No. 14-556), 2015 WL 1048442 (discussing the empirical evidence at length); Brief of Amici Curiae the Ruth Institute and Dr. Jennifer Roback Morse, PhD, in Support of Respondents and in Opposition to Reversal at 21–29, Obergefell, 135 S. Ct. 2584 (No. 14-556), 2015 WL 1501656 (same), neither the parties nor the Court addressed it in any detail, and it was mentioned only in passing at oral argument. See Transcript of Oral Argument at 35, Obergefell, 135 S. Ct. 2584 (No. 14-556), 2015 WL 2399419 (documenting an exchange between Justice Scalia and Solicitor General Donald Verilli about whether “all of the evidence shows there is no problem” with same-sex couples raising children).
It is unsurprising, then, that the Court’s opinion in Obergefell did not mention the underlying evidence on same-sex parenting and instead waxed poetic about the importance of marriage. 122 See Obergefell, 135 S. Ct. at 2601 (describing marriage as “a keystone of our social order”); see also id. at 2594, 2608 (arguing marriage “embodies the highest ideals of . . . family” and “is essential to our most profound hopes and aspirations”). Even though it was not at the forefront of the decision, however, it is likely that the empirical evidence played a role. It is hard to imagine the Court, and Justice Kennedy in particular, approving marriage equality if there were evidence that this family form harmed children.

b. Abortion. — Empirical evidence has also played a key role in abortion jurisprudence. The basic liberty right of a woman to decide whether to continue or terminate a pregnancy rests on notions of privacy and individual liberty, 123 See Planned Parenthood of Se. Pa. v. Casey, 505 U.S. 833, 847 (1992) (plurality opinion) (“It is a promise of the Constitution that there is a realm of personal liberty which the government may not enter.”); Roe v. Wade, 410 U.S. 113, 153 (1973) (“This right of privacy . . . is broad enough to encompass a woman’s decision whether or not to terminate her pregnancy.”). but empirical evidence has long been pivotal in decisions recognizing and effectuating this right. 124 In Roe v. Wade, the Court compared data on mortality rates from early-pregnancy abortions and childbirth to establish that maternal mortality rates for abortions performed before the end of the first trimester are “as low as or lower than the rates for normal childbirth,” 410 U.S. at 149, and the viability test established in the case rests on a scientific understanding of fetal development, see id. at 163 (explaining the state has a compelling interest in protecting the life of the fetus at the point of the viability “because the fetus then presumably has the capability of meaningful life outside the mother’s womb”). For another example, see Planned Parenthood of Cent. Mo. v. Danforth, 428 U.S. 52, 76–79 (1976) (invalidating a state ban on a procedure using saline amniocentesis by relying on data comparing the use of this procedure and alternative procedures). Empirical evidence has taken on even greater importance since 1992, when the test to determine the constitutionality of an abortion regulation became whether the restriction imposes an “undue burden” on the right to reproductive freedom. 125 See Casey, 505 U.S. at 876–77. This is fundamentally an empirical inquiry, centered on whether the purpose or effect of the restriction creates a substantial obstacle to exercising the right. 126 See id. (“[A] statute which, while furthering . . . [a] valid state interest, has the effect of placing a substantial obstacle in the path of a woman’s choice cannot be considered a permissible means of serving its legitimate ends.”); id. at 878 (“Unnecessary health regulations that have the purpose or effect of presenting a substantial obstacle to a woman seeking an abortion impose an undue burden on the right.”).

In Whole Woman’s Health v. Hellerstedt, the Court pored over evidence about the state restrictions, notably the requirement that doctors performing abortions have admitting privileges in nearby hospitals. 127 See 136 S. Ct. 2292, 2310–11 (2016); see also Tex. Health & Safety Code Ann. § 171.0031(a) (West 2017) (“A physician performing or inducing an abortion . . . must, on the date the abortion is performed or induced, have active admitting privileges at a hospital that . . . is located not further than 30 miles from the location at which the abortion is performed or induced.”). The Court described the trial evidence—both expert testimony and peer-reviewed medical studies—finding that abortions are a safe procedure and that hospital admissions are rare. 128 See Whole Woman’s Health, 136 S. Ct. at 2310–11. The Court also considered the extensive evidence introduced at trial that the requirement had led to the closure of half of the abortion facilities in Texas. 129 See id. at 2312. Based on this evidence, the Court concluded that the requirement imposed a substantial obstacle to the exercise of the right. 130 See id. at 2311–12. The other state restriction at issue in the case required facilities providing abortions to meet the standards for ambulatory surgical centers. In finding that the requirement did not promote women’s health, was unnecessary, and placed a substantial obstacle to exercising the right to reproductive choice, the Court credited expert testimony at trial predicting the closure of clinics, finding that although the prediction was not ultimately borne out, it had relied on the “scientific method” of making a hypothesis and then attempting to verify the hypothesis with further studies. See id. at 2314–17.

In another abortion case, the Supreme Court also relied on empirical evidence—albeit much more questionable evidence—to draw conclusions about the mental health consequences of an abortion. In Gonzales v. Carhart, the Court held that it was permissible for a state to prohibit a medical procedure, in part because there was evidence that after an abortion “[s]evere depression and loss of esteem can follow.” 131 550 U.S. 124, 159 (2007) (citing Brief for Sandra Cano, the Former “Mary Doe” of Doe v. Bolton, et al. as Amici Curiae in Support of Petitioner at 22–24, Carhart, 550 U.S. 124 (No. 05-380), 2006 WL 1436684 [hereinafter Sandra Cano Brief]). Another reason for prohibiting the procedure was that a woman might later regret her decision to terminate a pregnancy once she found out the details of the procedure. See id. (“[S]ome women come to regret their choice to abort the infant life they once created and sustained.”); see also id. at 159–60 (“It is self-evident that a mother who comes to regret her choice to abort must struggle with grief more anguished and sorrow more profound when she learns . . . that she allowed a doctor to pierce the skull and vacuum the fast-developing brain of her unborn child . . . .”). The Court acknowledged that there was no evidence to support this regret rationale. See id. at 159 (“While we find no reliable data to measure the phenomenon, it seems unexceptionable to conclude some women come to regret their choice to abort the infant life they once created and sustained.”). Writing for the majority, Justice Kennedy cited an amicus brief representing the views of 181 women who had had an abortion and who felt that the procedure created “adverse emotional and psychological health effects.” 132 Sandra Cano Brief, supra note 131, at 1. This brief also cited to a South Dakota task force that purportedly established the detrimental mental health effects of an abortion. 133 See id. at 16–21. For a discussion of the task force, see infra text accompanying notes 291–294.

As this last example demonstrates, judicial reliance on empirical evidence is not inevitably a neutral or thorough process. Indeed, in Carhart the parties had not extensively litigated the mental health effects of abortions in the two-week bench trial in the case. 134 See Carhart v. Ashcroft, 331 F. Supp. 2d 805, 852–1002 (D. Neb. 2004) (summarizing the medical evidence at trial but noting the psychological effects of abortion only in passing), aff’d sub nom. Carhart v. Gonzales, 413 F.3d 791 (8th Cir. 2005), rev’d, 550 U.S. 124 (2007). If they had, the challengers likely would have introduced the abundant evidence finding that abortions are not correlated with mental health problems, 135 See Brenda Major et al., Am. Psychol. Ass’n, Report of the APA Task Force on Mental Health and Abortion 4–5 (2008), http://www.apa.org/pi/women/programs/abortion/mental-health.pdf [http://perma.cc/LN5J-H76C] (“[A]mong adult women who have an unplanned pregnancy the relative risk of mental health problems is no greater if they have a single elective first-trimester abortion than if they deliver that pregnancy. The evidence regarding the relative mental health risks associated with multiple abortions is more equivocal.”). and the Supreme Court would have had to grapple with this more complete evidentiary record. The discussion below returns to the issue of selective use of empirical evidence. 136 See infra section II.B.2.

c. Intimate Partner Violence and the Child Welfare System. — Empirical evidence was also a decisive factor in a significant rights-based case that significantly affected the child welfare system—the litigation over New York City’s practice of responding to intimate partner violence by removing children from their homes and placing them in foster care. 137 See Nicholson v. Williams, 203 F. Supp. 2d 153, 208–10, 228–29 (E.D.N.Y. 2002) (describing the city’s practice of finding that a victim of intimate partner violence was neglecting the child). The city persisted in this practice even when the child had not witnessed the violence firsthand, when the child was not the direct victim of the abuse, and when the mother was otherwise adequately caring for the child. See id. at 169–72, 228. For a firsthand account of the litigation from the perspective of the plaintiffs’ attorney, see generally Jill M. Zuccardy, Nicholson v. Williams: The Case, 82 Denv. U. L. Rev. 655 (2005). In a class action challenging the practice, the district court held a twenty-four-day bench trial, with twelve expert witnesses. 138 Nicholson, 203 F. Supp. 2d at 165; Zuccardy, supra note 137, at 662 (noting that the trial included twelve expert witnesses). These witnesses addressed the research on children and intimate partner violence, focusing on whether witnessing intimate partner violence produced adverse effects for children and whether a home with intimate partner violence was more likely to be a home with child abuse. 139 See Nicholson, 203 F. Supp. 2d at 197–98. Drawing on extensive studies and their own work, five experts offered the following opinions: Children experience a range of negative effects, from minor psychological disturbance to post-traumatic stress syndrome; numerous factors influence a child’s reaction including the severity of the abuse, the child’s proximity to the abuse, and the parent’s ability to support the child; witnessing domestic violence is correlated with a higher risk of substance abuse and violence as an adult, but the vast majority of children exposed to intimate partner violence do not experience these problems as an adult; and even when children are exposed to severe intimate partner violence, if they are then in a safe place and the violence ends, significant psychological problems disappear completely for the majority of children. See id. Plaintiffs’ experts testified that although intimate partner violence and child abuse often occur together, it is almost always the same adult inflicting both kinds of abuse—it is not typically a situation in which the father hits a mother who then hits the child. See id. at 198. The experts also described the research on the detrimental effects of removal on children, especially in cases of intimate partner violence, 140 Id. at 198–99. The experts noted that removing a child from the home, and thus disrupting the parent–child relationship, can have extreme consequences for the child’s sense of security and safety and that when the child has been removed because of intimate partner violence, the child’s sense of danger is often heightened because the child is concerned about the parent left behind. Id. Further, the experts testified that children often blame themselves for the removal, leading to psychological problems, and placement in foster care introduces a new set of risks, including abuse and neglect, inadequate medical care, and disruption of the child’s contacts with family, school, and community. Id. at 199. and the best practices for addressing intimate partner violence in the child welfare system. 141 Id. at 200–05 (describing extensive expert testimony and a report by the National Council of Juvenile and Family Court Judges). This evidence suggested to the court that mothers should not be accused of neglect merely for being victims of domestic violence, perpetrators should be held accountable, children should be protected by offering services to the mother, removal should be used only as a last resort, and child welfare workers should be adequately trained on the dynamics of these cases. See id.

Relying extensively on the expert testimony, the district court ruled that the city’s practice was unconstitutional. 142 See id. at 233–60 (finding a likelihood of success on the merits for the plaintiffs’ Fourth, Ninth, Thirteenth, Fourteenth, and Nineteenth Amendment claims and thus issuing a preliminary injunction). Although the decision was subsequently narrowed on appeal, 143 The case had a lengthy subsequent history not relevant to the issue of how courts use empirical evidence. For a discussion of this history, see Zuccardy, supra note 137, at 669. the decision continues to resonate across the child welfare system. 144 See Kathleen A. Copps, The Good, the Bad, and the Future of Nicholson v. Scoppetta: An Analysis of the Effects and Suggestions for Further Improvements, 72 Alb. L. Rev. 497, 510–12, 523–25 (2009) (describing the changes in New York City as a result of the litigation and cataloguing the decision’s influence in other jurisdictions). Moreover, there is no question that that empirical evidence was critical to the outcome of the case. 145 See Nicholson, 203 F. Supp. 2d at 198–99, 250 (reviewing expert testimony about the harm of removing children from their homes even when there is domestic violence in the home and concluding that “[t]he evidence demonstrates that the compelling state interest in protecting children” is hindered by “policies of prosecuting abused mothers and removing their children”).

d. Juvenile Sentencing. — Empirical evidence has had a profound effect on Eighth Amendment jurisprudence. In a series of cases addressing the constitutionality of sentences for crimes committed by juveniles, the Supreme Court relied heavily on research in the fields of neuroscience, psychology, and sociology. 146 For a small sample of this literature, see, e.g., Alison S. Burke, Under Construction: Brain Formation, Culpability, and the Criminal Justice System, 34 Int’l J.L. & Psychiatry 381, 382–83 (2011) (presenting neuroscience research indicating “the brain is still growing and maturing during adolescence” and arguing that charging children as adults is overly punitive); Eveline A. Crone & Maurits W. van der Molen, Developmental Changes in Real Life Decision Making: Performance on a Gambling Task Previously Shown to Depend on the Ventromedial Prefrontal Cortex, 25 Developmental Neuropsychology 251, 252 (2004) (presenting research indicating juvenile brains do not fully appreciate future consequences); Jay N. Giedd, Structural Magnetic Resonance Imaging of the Adolescent Brain, 1021 Annals N.Y. Acad. Sci. 77, 83 (2004) (noting research indicating important brain functions that “inhibit impulses, weigh consequences of decisions, prioritize, and strategize” are “still under construction for a decade after the throes of puberty and may therefore be related to some of the behavioral manifestations of the teen years”); Nitin Gogtay et al., Dynamic Mapping of Human Cortical Development During Childhood Through Early Adulthood, 101 Proc. Nat’l Acad. Sci. 8174, 8174, 8178 (2004) (mapping brain development over time and noting that regions associated with more complex and integrative tasks developed last). The underlying research shows that, although brain structure is primarily in place by age five or six, the brain continues to develop through early adulthood. Neuro-scientists have focused in particular on the prefrontal cortex, finding that adolescents have less forethought and impulse control than fully matured adults. 147 See Burke, supra note 146, at 382–83 (“Research shows that youths also have a less than fully developed brain and this difference can account for many behavioral discrepancies between adolescents and adults. Because the brain is still forming and changing during the teenage years, the culpability of adolescent behavior may be diminished.”); Crone & van der Molen, supra note 146, at 274 (noting that study participants, with advancing age, “made increasingly more advantageous choices”); Giedd, supra note 146, at 83 (concluding that “brain structure goes through explosive changes during the teen years”). Additionally, adolescents are still developing their characters and personalities, and there are many opportunities for change and growth. 148 See Laurence Steinberg, Age of Opportunity: Lessons from the New Science of Adolescence 18–64 (2014) (arguing that adolescence is a “remarkable period of brain reorganization and plasticity”).

Working with scholars in other disciplines, legal scholars played a pivotal role in translating this research into legal rules and principles. 149 See Barry C. Feld, Adolescent Criminal Responsibility, Proportionality, and Sentencing Policy: Roper, Graham, Miller/Jackson, and the Youth Discount, 31 Law & Ineq. 263, 277–83 (2013) (reviewing the legal literature, which heavily draws on the underlying interdisciplinary research). One of the most productive collaborations was between legal scholar Elizabeth Scott and psychologist Laurence Steinberg. In a highly influential article, Scott and Steinberg laid out a framework for a devel­opmentally sensitive approach to juvenile justice. 150 Laurence Steinberg & Elizabeth S. Scott, Less Guilty by Reason of Adolescence: Developmental Immaturity, Diminished Responsibility, and the Juvenile Death Penalty, 58 Am. Psychologist 1009, 1014 (2003). They contended that developmental insights should inform the approach to juvenile crime: The immaturity of adolescents means they are not as morally culpable, their vulnerability to peer pressure makes it difficult for adolescents to leave a situation in which a crime may be committed, and their still-developing characters means there is an opportunity for rehabilitation. 151 See id.

Starting with Roper v. Simmons in 2005, 152 543 U.S. 551, 569–70 (2005). and through the most recent pronouncement in Montgomery v. Louisiana in 2016, 153 136 S. Ct. 718, 733 (2016). the Court drew heavily on the underlying research. The Court embraced the developmentally sensitive framework proposed by Scott and Steinberg, 154 Roper, 543 U.S. at 569–70. holding that the Eighth Amendment places substantial constraints on sentences for crimes committed by minors. 155 See id. at 578 (holding it unconstitutional to impose the death penalty upon a seventeen-year-old minor who committed first-degree murder); see also Montgomery, 136 S. Ct. at 736 (holding Miller v. Alabama applies retroactively); Miller v. Alabama, 132 S. Ct. 2455, 2471 (2012) (partially extending Graham v. Florida to minors guilty of homicide but clarifying that the Eighth Amendment “mandates only that a sentencer follow a certain process—considering an offender’s youth and attendant characteristics—before imposing [life without the possibility of parole on a minor convicted of homicide]”); Graham v. Florida, 560 U.S. 48, 82 (2010) (“The Constitution prohibits [imposing] a life without parole sentence on a juvenile offender who did not commit homicide. A State need not guarantee the offender eventual release, but if it imposes a sentence of life it must provide him or her with some realistic opportunity to obtain release . . . .”).

e. Child Custody. — Finally, empirical evidence is playing an influential role in contemporary custody decisions. In custody disputes between parents, every state uses some variant of the best-interests-of-the-child standard, 156 All states have some variant on the best-interests standard, but there is a preference for continued contact with both parents. See Theresa Glennon, Still Partners? Examining the Consequences of Post-Dissolution Parenting, 41 Fam. L.Q. 105, 114–17 (2007). which itself was shaped by empirical evidence. 157 See Davis, There Is a Book Out, supra note 11, at 1542–49. This test gives nearly boundless discretion to the court. 158 See Mnookin, supra note 67, at 226, 255–62. As scholars have shown, 159 See Elizabeth S. Scott & Robert E. Emery, Gender Politics and Child Custody: The Puzzling Persistence of the Best-Interests Standard, 77 Law & Contemp. Probs., no. 1, 2014, at 69, 72–75 (arguing that courts often cannot obtain verifiable information about parenting because the qualitative proxies to determine best interests, such as closeness of the relationship between a parent and child, are highly complex and difficult to assess, and the standard gives no guidance on weighing multiple factors). courts are ill-equipped to implement this standard, and thus courts look to more definite criteria, such as each parent’s willingness and ability to foster a relationship between the child and the other parent. 160 See id. at 95–100 (discussing how courts rely on expert opinions that assess family violence or “parental alienation” to evaluate custody disputes).

To determine a parent’s openness to the child’s ongoing relationship with the other parent, some courts have relied on so-called parental alienation syndrome. 161 For a summary of the history and ongoing use of parental alienation syndrome, see Joan S. Meier, A Historical Perspective on Parental Alienation Syndrome and Parental Alienation, 6 J. Child Custody 232, 235–50 (2009) [hereinafter Meier, A Historical Perspective]; Joan S. Meier & Sean Dickson, Mapping Gender: Shedding Empirical Light on Family Courts’ Treatment of Cases Involving Abuse and Alienation, 35 Law & Ineq. 311, 316–19 (2017). Developed by a single psychologist based on interviews with only his clients and self-published without the benefit of peer review, 162 See Richard A. Gardner, Child Custody Litigation: A Guide for Parents and Mental Health Professionals 76–104 (1986); Carol S. Bruch, Parental Alienation Syndrome and Alienated Children—Getting It Wrong in Child Custody Cases, 14 Child & Fam. L.Q. 381, 386–87 (2002) (noting Gardner self-published most of his work). Gardner conceded that most of the studies supporting his theory were based on direct patient observation, rather than experiments and statistical analysis. See Richard A. Gardner, Commentary on Kelly and Johnston’s “The Alienated Child: A Reformulation of Parental Alienation Syndrome,” 42 Fam. Ct. Rev. 611, 617–18 (2004). parental alienation syndrome is not recognized in the Diagnostic and Statistical Manual of Mental Disorders (DSM), 163 See Am. Psychiatric Ass’n, Diagnostic and Statistical Manual of Mental Disorders (5th ed. 2013). and it has been uniformly discredited by psychologists. 164 See, e.g., Robert E. Emery, Parental Alienation Syndrome: Proponents Bear the Burden of Proof, 43 Fam. Ct. Rev. 8, 8–9 (2005) (noting that “there have been no independent, objective, or public replications of Gardner’s assertions”); Janet R. Johnston & Joan B. Kelly, Rejoinder to Gardner’s “Commentary on Kelly and Johnston’s ‘The Alienated Child: A Reformulation of Parental Alienation Syndrome,’” 42 Fam. Ct. Rev. 622, 622 (2004) (rejecting calling parental alienation syndrome “a syndrome and grant[ing] it status as a DSM diagnostic category . . . outright”); Joan B. Kelly & Janet R. Johnston, The Alienated Child: A Reformulation of Parental Alienation Syndrome, 39 Fam. Ct. Rev. 249, 249–50 (2001) (“[T]here is a relative absence of any empirical or research support for the reliable identification of [parent alienation syndrome], other than Gardner’s (and other proponents’) clinical experience and ‘expert testimony.’”). For a summary of the research finding that parental alienation syndrome has no scientific basis, see Bruch, supra note 162, at 383–89; Meier, A Historical Perspective, supra note 161, at 235–40. Despite this lack of scientific basis, many courts and mental health professionals, who provide highly influential custody evaluations to courts, have invoked parental alienation syndrome. 165 See Bruch, supra note 162, at 387–88 (discussing cases in which courts allowed parental alienation syndrome testimony but noting that when Richard Gardner, the psychologist who identified parental alienation syndrome, attempted to testify, most courts disallowed the testimony, either because it went to the ultimate determination of custody or because the court found the syndrome unsupported); Meier, A Historical Perspective, supra note 161, at 240 (describing the ubiquity of parental alienation syndrome in family court); Scott & Emery, supra note 159, at 99–100 & n.164 (citing cases and discussing the continuing widespread reliance on parental alienation syndrome).

In light of the looser evidentiary rules used in family court, 166 See Meier, A Historical Perspective, supra note 161, at 240; Meier & Dickson, supra note 161, at 319; Scott & Emery, supra note 159, at 99–100 (explaining that family courts often do not screen custody opinions from mental health professionals, may believe the professionals are neutral and therefore do not need the scrutiny, and may believe the court appointment itself suffices as a validation of the professional’s scientific credibility). the Daubert standard, developed for the purpose of distinguishing reliable scientific evidence from unreliable scientific evidence, 167 Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579, 589–90 (1993) (holding that before admitting expert testimony, courts must determine that the evidence is based on reliable scientific methodology and reasoning); see also John Conley & Jane Moriarty, Scientific and Expert Evidence 82 (2d ed. 2011) (explaining that Daubert applies in federal courts but that most states have adopted the standard or a similar one). is not an effective tool for combatting parental alienation syndrome. Family courts rarely use the test, either to screen expert witnesses in court or when drawing on reports from mental health professionals evaluating custody. 168 See Meier, A Historical Perspective, supra note 161, at 240–41 (“To a troubling degree, family courts and even courts of appeal are increasingly accepting the application of [parental alienation syndrome] . . . while sidestepping the admissibility question . . . .”); Scott & Emery, supra note 159, at 99–100 (explaining how “few jurisdictions require systematic scrutiny” of mental health professionals’ opinions). As a result, highly unreliable and unscientific evidence continues to dominate in family court. 169 See Meier, A Historical Perspective, supra note 161, at 240–41; Scott & Emery, supra note 159, at 99–100. But the tide may be beginning to turn. See Maxine Eichner, Bad Medicine: Parents, the State, and the Charge of “Medical Child Abuse,” 50 U.C. Davis L. Rev. 205, 271 & n.294 (2016) (listing cases in which courts, including family courts, rejected parental alienation syndrome as unscientific and unreliable). As Joan Meier explains, even though courts are less likely to rely on parental alienation syndrome, they now—and still problematically—invoke the related notion of parental alienation, which is not characterized as a syndrome, per se, but rather as a behavior that weighs against awarding custody to the alienating parent. See Meier, A Historical Perspective, supra note 161, at 245–50.

2. Legislation. — Empirical evidence also plays an important role in lawmaking. Legislatures regularly use empirical evidence to identify problems and determine appropriate solutions. When Congress passed the Adoption and Safe Families Act (ASFA) of 1997, 170 Adoption and Safe Families Act of 1997, Pub. L. No. 105-89, 111 Stat. 2115 (codified as amended in scattered sections of 42 U.S.C. (2012)). it relied on a wealth of empirical evidence about problems plaguing the child welfare system. 171 See Robert M. Gordon, Drifting Through Byzantium: The Promise and Failure of the Adoption and Safe Families Act of 1997, 83 Minn. L. Rev. 637, 646–50 (1999). Congress held hearings and found that the child welfare system was not serving the interests of children because family-preservation efforts were keeping some children in dangerous homes, the problem of “foster care drift” (the term used to describe both long stays in foster care and placement in multiple homes) was getting worse, and children would be better served by promoting adoption rather than family preservation. 172 See Promotion of Adoption, Safety, and Support for Abused and Neglected Children: Hearing Before the S. Comm. on Fin., 105th Cong. 1–9 (1997) (discussing the “demands on the [child welfare] system” and exploring the “pressure points at which reform might be aimed”); Improving the Well-Being of Abused and Neglected Children: Hearing Before the S. Comm. on Labor & Human Res., 104th Cong. 9–10 (1996) (statement of Richard J. Gelles, Director, Family Violence Research Program) (recounting the results of studies finding that current family-preservation efforts were ineffective); Barriers to Adoption: Hearing Before the Subcomm. on Human Res. of the H. Comm. on Ways & Means, 104th Cong. 4–64 (1996) (considering whether federal intervention was necessary to promote adoption); Federal Adoption Policy: Hearing Before the Subcomm. on Human Res. of the H. Comm. on Ways & Means, 104th Cong. 4–9 (1995) (exploring how to “simultaneously achieve the goals of both family preservation and timely adoption”); Child Welfare Programs: Hearing Before the Subcomm. on Oversight of the H. Comm. on Ways & Means, 104th Cong. 4–8 (1995) (examining lessons learned since the passage of the Adoption Assistance and Child Welfare Act of 1980, which required the Department of Health and Human Services to review the foster care caseload of every state); see also Gordon, supra note 171, at 646–50 (describing this legislative history, including the role of the Clinton Administration in pushing Congress to reform the child welfare system). Congress responded to these problems by adopting a standard that set a time limit on family-reunification efforts, 173 As a condition of receiving federal funds, states had to commence proceedings to terminate parental rights for children who had been in foster care for fifteen of the most recent twenty-two months. See 42 U.S.C. § 675(5)(E). thus moving children to permanent homes more quickly and making child safety and permanency—rather than family preservation—the para­mount concerns of the child welfare system. 174 ASFA conditioned federal funds on states developing a foster-care and adoption-assistance plan in which “the child’s health and safety shall be the paramount concern.” Id. § 671(a)(15)(A). In enacting ASFA, however, Congress selectively relied on empirical evidence. It did not focus on empirical evidence about competing concerns, such as the risks a child faces in foster care or the developmental harm of separating a child from a caregiver, especially during early childhood. 175 See Gordon, supra note 171, at 646–50 (describing the evidence presented in support of ASFA, which did not include material on these issues).

Legislatures have also used empirical evidence to develop responses to intimate partner violence. Since the 1970s, scholars in multiple disciplines have generated significant research about the extent of intimate partner violence, the harms it causes, and effective responses. 176 Intimate partner violence grew into a distinct academic discipline in the 1970s and 1980s, with scholars from different disciplines, including law, providing a theoretical framework for reform and legitimizing the study of the problem. See Leigh Goodmark, A Troubled Marriage: Domestic Violence and the Legal System 9–28 (2012) [hereinafter Goodmark, Troubled Marriage] (describing the legal and scholarly history of the intimate partner violence movement beginning in the 1970s). Since then, intimate partner violence has become a robust area of inquiry, studied by scholars in multiple disciplines, including sociology, see, e.g., Richard J. Gelles & Murray A. Straus, Intimate Violence (1988), psychology, see, e.g., Paula Nicolson, Domestic Violence and Psychology: A Critical Perspective (2010); Katherine M. Kitzmann et al., Child Witnesses to Domestic Violence: A Meta-Analytic Review, 71 J. Consulting & Clinical Psychol. 339 (2003), medicine, see, e.g., Empowering Survivors of Abuse: Health Care for Battered Women and Their Children (Jacquelyn C. Campbell ed., 1998), and economics, see, e.g., Anna Aizer, The Gender Wage Gap and Domestic Violence, 100 Am. Econ. Rev. 1847 (2010). Both the federal and state legislatures have invoked this evidence to enact laws to protect victims and punish perpetrators. Evidence drove many aspects of the federal Violence Against Women Act, 177 Violence Against Women Act of 1994, Pub. L. No. 103-322, 108 Stat. 1902 (codified as amended in scattered sections of 16, 28, and 42 U.S.C. (2012)). federal immigration law provisions that address the incentive for a noncitizen to stay with a violent partner, 178 See, e.g., 8 U.S.C. § 1186a(c)(4)(C) (2012) (allowing a victim of intimate partner violence to petition separately from a spouse or former spouse for the removal of conditional residency); id. § 1229b(b)(2)–(4) (authorizing a victim of intimate partner violence to self-petition for cancellation of removal rather than rely on a family member). and state laws that create civil and criminal protection orders. 179 See Jeannie Suk, Criminal Law Comes Home, 116 Yale L.J. 2, 13–17 (2006) (describing these orders). The empirical evidence does not always support the various legislative mandates, 180 See infra text accompanying notes 277–282 (discussing the conflicting evidence about mandatory-arrest policies and recidivism). but there is no question that legislatures regularly use empirical evidence to develop legal responses to intimate partner violence.

3. Administration. — Under the banner of “Bringing Business Intelligence to Child Welfare,” 181 Innovations in Action, Eckerd Kids, http://www.eckerd.org/about-eckerd-kids/what-were-doing/innovations-in-action/ [http://perma.cc/8B72-7PJ4] (last visited Sept. 11, 2017). administrative agencies around the country are beginning to use predictive analytics—which employs statistics and modeling to forecast future events 182 See Eric Siegel, Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die 15 (rev. & updated ed. 2016) (“Predictive analytics (PA)—Technology that learns from experience (data) to predict the future behavior of individuals in order to drive better decisions.”); Charles Nyce, Am. Inst. for CPCU & Ins. Inst. of Am., Predictive Analytics White Paper 1 (2007), http://www.the-digital-insurer.com/wp-content/uploads/2013/12/78-Predictive-Modeling-White-Paper.pdf [http://perma.cc/4QUQ-Y9A4] (providing an overview of predictive analytics). Predictive analytics is used in myriad settings, from health care to retail marketing, and by both private and public actors. See Viktor Mayer-Schönberger & Kenneth Cukier, Big Data: A Revolution that Will Transform How We Live, Work, and Think 123–49 (2013) (describing the use of predictive analytics in the private sector); Siegel, supra, at 23–45 (describing the use of predictive analytics more broadly); Robert Brauneis & Ellen P. Goodman, Algorithmic Transparency for the Smart City, 20 Yale. J.L. & Tech. (forthcoming 2018) (manuscript at 6–11) (on file with the Columbia Law Review) (describing the use of data analytics at the state and particularly local level as part of the move toward “smart cities”). —in their child welfare systems. 183 See, e.g., Gladys Carrión, Comm’r, N.Y.C. Admin. for Children’s Servs., Testimony to the Commission to Eliminate Child Abuse and Neglect Fatalities 10 (Aug. 6, 2015), http://www1.nyc.gov/assets/acs/pdf/testimony/2015/Commission_to_Eliminate_Child_Abuse_and_Neglect_Fatalities_8_5_15.pdf [http://perma.cc/3JAR-BGXA] (describing New York City’s early use of predictive analytics); U.S. Dep’t of Health & Human Servs., Child Maltreatment 2015, at 133–34, 199–201 (2017) [hereinafter Child Maltreatment], http://
www.acf.hhs.gov/sites/default/files/cb/cm2015.pdf [http://perma.cc/5MGJ-N6MG] (describing the use of predictive analytics in Connecticut and Oklahoma); Marquis Cabrera, Florida Leverages Predictive Analytics to Prevent Child Fatalities—Other States Follow, Huffington Post (Dec. 21, 2015), http://www.huffingtonpost.com/
marquis-cabrera/florida-leverages-predictive_b_8586712.html [http://perma.cc/ZV6T-S3CK] (describing a nonprofit’s contracts to develop and use predictive analytics in Alaska, Connecticut, Florida, Illinois, Maine, and Oklahoma); Dina Gusovsky, Can Life as a Data Point Save America’s At-Risk Children?, CNBC News (Jan. 14, 2016), http://
www.cnbc.com/2016/01/14/an-80-billion-annual-tax-bill-thats-failing-our-children.html [http://perma.cc/UG73-KRK9] (describing a pilot program using predictive analytics in Los Angeles County); Laura Santhanam, Can Big Data Save These Children?, PBS NewsHour (Mar. 22, 2016), http://www.pbs.org/newshour/updates/can-big-data-save-these-children/ [http://perma.cc/66G7-XWU2] (describing the use of predictive analytics in Allegheny County, which includes Pittsburgh); Eckerd Kids, supra note 181 (describing the use of predictive analytics in Florida).
Without this technology, social workers in the child welfare system investigate cases and make pivotal decisions about cases using their experience, intuition, and rudimentary risk assessment tools, only some of which are empirically validated. 184 In a typical case, the child welfare agency receives a report of abuse or neglect, and a caseworker investigates the claim and speaks with family members, the school, and other individuals and institutions in the child’s life. The caseworker then decides whether to substantiate the allegation of abuse or neglect and thus begin the process of state intervention in the family. See Child Maltreatment, supra note 183, at 6 (explaining the first step is determining whether the report should be screened in for an investigation and noting reports are screened out for a variety of reasons, including inadequate information in the report). For a description of the different risk assessment tools currently in use, see Richard J. Gelles, Out of Harm’s Way: Creating an Effective Child Welfare System 104–08 (2017). Triaging cases is a daunting prospect, with an estimated four million reports of abuse and neglect annually, 185 See Child Maltreatment, supra note 183, at 6 (noting that the number of these reports has increased by almost sixteen percent in four years). limited resources, and child safety and family integrity hanging in the balance. 186 See Clare Huntington, Rights Myopia in Child Welfare, 53 UCLA L. Rev. 637, 642–52 (2006) [hereinafter Huntington, Rights Myopia] (describing the competing demands and the stakes in the child welfare system).

Predictive analytics brings a data-driven approach to this process. 187 For an excellent overview of the use of predictive analytics in child welfare as well as a discussion of the many concerns associated with this practice, see Christopher E. Church & Amanda J. Fairchild, In Search of a Silver Bullet: Child Welfare’s Embrace of Predictive Analytics, 68 Juv. & Fam. Ct. J. 67, 68–78 (2017). Agencies are using it in somewhat different ways, but the model—first developed in New Zealand in 2012—reviews reports of abuse and neglect and determines which cases are most serious and thus deserving of inten­sive follow-up and intervention. 188 See Rhema Vaithianathan et al., Vulnerable Children: Can Administrative Data Be Used to Identify Children at Risk Of Adverse Outcomes? 6–9 (2012), http://
www.msd.govt.nz/documents/about-msd-and-our-work/publications-resources/research/
vulnerable-children/auckland-university-can-administrative-data-be-used-to-identify-children-at-risk-of-adverse-outcome.pdf [http://perma.cc/2YKU-WPZZ]. The model was also designed to work at a much earlier stage, identifying children at risk of abuse and neglect at the time a family member applies for a public benefit. See id. at 6. No child welfare agency in the United States is using predictive analytics at this stage. Id.
After an agency receives a call about a particular family, the model mines the databases of several government systems—education, criminal justice, health, public benefits, and so on—to obtain information about the family, 189 See Gusovsky, supra note 183 (describing this process in Los Angeles County). combines this with demo­graphic data, including age of the parents and children, education levels, and family structure, and then adds details about the family’s past involvement with the child welfare system, including whether the parent spent any time in foster care. 190 See Vaithianathan et al., supra note 188, at 10–11 (describing the family and child demographic characteristics that are included); Gusovsky, supra note 183 (noting the Los Angeles model combines information from the databases with information about the family because “experts say that whoever is living with the child has a big, if not the greatest, influence on his or her well-being”). The model runs these data points through a proprietary algorithm, producing a risk score for the child. 191 See Vaithianathan et al., supra note 188, at 7–11 (describing how the algorithm uses the variables to generate a risk score, or “the chance that the child who has started the spell will have an adverse outcome by some given age”); Gusovsky, supra note 183 (detailing how the algorithm “provides a total risk score for each child based on numerous factors, as well as a map of that child’s social network and data points related to those connections, such as criminal history”).

In the United States, Eckerd Kids is a nonprofit organization championing the use of predictive analytics and contracting with agen-cies around the country. 192 See Cabrera, supra note 183 (describing Eckerd’s contracts with Alaska, Connecticut, Florida, Illinois, Maine, and Oklahoma). It has developed its own predictive analytics model that combs datasets and looks at risk factors, including the age of the child, the presence of a “paramour” in the home, a history of substance abuse and intimate partner violence, and the parent’s experience in the child welfare system as a child. 193 Eckerd Kids, supra note 181. Identified cases are slated for intensive follow-up, with the technological tool also prompting the caseworker to follow recommended steps. 194 See Cabrera, supra note 183 (describing the implementation of Eckerd’s Rapid Safety Feedback program, which uses predictive analytics to reduce child fatality). When tested against past cases, predictive analytics has been relatively accurate in identifying the cases that resulted in child fatalities or severe injuries, 195 See Vaithianathan et al., supra note 188, at 15–25 (describing the accuracy of the predictions and noting that the algorithm had a seventy-six percent area under the Receiver Operator Characteristic curve—a model with 100% area under the curve is considered to have a perfect fit in terms of predictive power); Santhanam, supra note 183 (describing how Allegheny County found the model highly predictive of abuse and noting that “[a]mong children with the highest risk score, 40 percent were removed from their homes less than a year later” and “[a]mong those with the lowest risk score, the likelihood of entering foster care was . . . 0.3 percent”). See generally Emily Putnam-Hornstein et al., Preventing Severe and Fatal Maltreatment: Making the Case for the Expanded Use and Integration of Data, 92 Child Welfare 59, 64–70 (2013) (describing the model and the benefits of drawing on multiple sources of data through an automated system rather than the current approach). but it has a high rate of false positives. 196 See Church & Fairchild, supra note 187, at 71–72 (discussing the high rate of false positives); Vaithianathan et al., supra note 188, at 18 (noting if services are offered to the 3,284 children in the two groups with the highest risk scores, “1,211 children will have a maltreatment finding before age 5 and 2,073 [will] not”); Daniel Heimpel, Uncharted Waters: Data Analytics and Child Protection in Los Angeles, Chron. Soc. Change (July 20, 2015), http://chronicleofsocialchange.org/featured/uncharted-waters-data-analytics-and-child-protection-in-los-angeles/10867 [http://perma.cc/H4VA-C6WR] (noting that when applied to Los Angeles data, predictive analytics correctly identified seventy-six percent of the cases that resulted in death, near death, or severe injury but that the model also led to a false positive rate of more than ninety-five percent). A different concern is the lack of transparency in the algorithm. See Brauneis & Goodman, supra note 182, at 11–22.

Administrative agencies also use empirical evidence to develop programs to support families. The abundant research establishing the importance of early childhood development has been particularly influential. 197 For a summary of this research, see generally Jack P. Shonkoff & Deborah A. Phillips, Nat’l Acad. Sci., From Neurons to Neighborhoods: The Science of Early Childhood Development (2000). In cities and states around the country, administrative agencies are adopting programs to promote language development and other skills in the first few years of life. In Providence, Rhode Island, the Mayor’s Office implemented Providence Talks, a program to boost language skills during early childhood with biweekly coaching sessions and a “word pedometer” to help parents track their children’s language exposure. 198 About, Providence Talks, http://www.providencetalks.org/about [http://perma.cc/
3AEY-B7HY] (last visited Sept. 11, 2017).
In Oklahoma, the Department of Education offers a range of early childhood programs and services designed to prepare children for kindergarten. 199 Early Childhood and Family Education, Okla. State Dep’t of Educ., http://
sde.ok.gov/sde/early-childhood-and-family-education [http://perma.cc/55CB-KV3Z] (last updated Aug. 17, 2017).
And in multiple cities and states, administrative agencies provide voluntary home-visiting programs to promote child health and learning in the first years of life. 200 See, e.g., Home Visiting, Mo. Dep’t of Health & Senior Servs., http://
health.mo.gov/living/families/homevisiting/ [http://perma.cc/K7KY-K9Z7] (last visited Sept. 11, 2017).
These agencies regularly invoke the research on early childhood development to support their programs. 201 See, e.g., Research, Providence Talks, http://www.providencetalks.org/research/ [http://perma.cc/6X69-9YGZ] (last visited Sept. 11, 2017).

C. Patterns in the Empirical Turn

As the above description illustrates, there are many kinds of empirical evidence in family law. Sometimes empirical evidence reflects relatively uncontested statistics, such as the number of children being raised by same-sex parents. 202 See, e.g., DeBoer v. Snyder, 973 F. Supp. 2d 757, 763 (E.D. Mich.) (relying on testimony from demographer Gary Gates that 5,300 children in Michigan were being raised by same-sex couples), rev’d, 772 F.3d 388 (6th Cir. 2014), rev’d sub nom. Obergefell v. Hodges, 135 S. Ct. 2584 (2015); Perry v. Schwarzenegger, 704 F. Supp. 2d 921, 968 (N.D. Cal. 2010) (citing a study by the Williams Institute for the proposition that eighteen percent of same-sex couples in California are raising children), aff’d sub nom. Perry v. Brown, 671 F.3d 1052 (9th Cir. 2012), vacated and remanded sub nom. Hollingsworth v. Perry, 133 S. Ct. 2652 (2013); see also Obergefell, 135 S. Ct. at 2601 (relying on demographic data in an amicus brief showing that married same-sex couples experience economic advantages that unmarried same-sex couples do not).
Statistics can, of course, be subject to contest. For further discussion of counting families headed by same-sex parents, see Gary J. Gates & Michael D. Steinberger, Same-Sex Unmarried Partner Couples in the American Community Survey: The Role of Misreporting, Miscoding and Misallocation 13–21 (May 2010) (unpublished manuscript), http://economics-files.pomona.edu/steinberger/research/Gates_Steinberger_ACS_
Miscode_May2010.pdf [http://perma.cc/3LY6-ZGGP] (demonstrating that the U.S. Census both undercounts and overcounts children raised by same-sex parents). For an acknowledgement of the difference in counting same-sex couples raising children versus families headed by an LGBT parent, who may be single, see Same-Sex Couple and LGBT Demographic Data Interactive, Williams Inst. (May 2016), http://
williamsinstitute.law.ucla.edu/visualization/lgbt-stats/?topic=LGBT#demographic [http://
perma.cc/NE7G-MFLU].
Sometimes empirical evidence is embedded in a data-driven metric, such as predictive analytics. And often, empirical evidence reflects hotly contested correlations, such as the relationship between race and child maltreatment 203 See Annie E. Casey Found., Disparities and Disproportionality in Child Welfare: Analysis of the Research 7 (2011), http://www.aecf.org/m/resourcedoc/AECF-DisparitiesAndDisproportionalityInChildWelfare-2011.pdf [http://perma.cc/G9VV-EUC7] (conducting a meta-analysis examining the correlation between race and child maltreatment);  Andrea J. Sedlak et al., U.S. Dept. of Health & Human Servs., Supplementary Analyses of Race Differences in Child Maltreatment Rates in the NIS-4, at 4 (2010), http://www.acf.hhs.gov/sites/default/files/opre/nis4_supp_analysis_race_diff_mar2010.pdf [http://perma.cc/QJJ2-H9XE] (offering potential explanations for the NIS-4’s finding of a correlation between race and child maltreatment). or family struc-ture and child outcomes. 204 See supra text accompanying notes 112–117.

Similarly, the quality of empirical evidence used in family law ranges broadly. Sometimes evidentiary standards sufficiently screen for quality, such as the trial courts’ rejection of some of the evidence introduced by supporters of different-sex marriage requirements. 205 See supra text accompanying note 105. But sometimes courts, especially family courts, use evidence that falls far short of scientific standards, such as the invocation of parental alienation syndrome. 206 See supra text accompanying notes 161–169. And both courts and legislatures can use evidence that is politically motivated and highly selective, such as the finding that women who terminate a pregnancy experience depression and other mental health side effects. 207 See supra text accompanying notes 131–135; infra text accompanying notes 291–294.

Beyond these differences, decisionmakers use empirical evidence in a variety of ways. In litigation, judges typically use empirical evidence to establish what Professor Kenneth Culp Davis famously, if somewhat confusingly, called legislative facts. 208 See Kenneth Culp Davis, An Approach to Problems of Evidence in the Administrative Process, 55 Harv. L. Rev. 364, 402–03 (1942) [hereinafter Davis, Problems of Evidence] (distinguishing legislative and adjudicative facts); Laurens Walker & John Monahan, Social Frameworks: A New Use of Social Science in Law, 73 Va. L. Rev. 559, 561–71 (1987) (describing a “social framework” role for empirical evidence: using it to evaluate a contested issue in a case, such as the reliability of the testimony of a specific eyewitness, by providing a broader context for understanding the reliability of eyewitness testimony generally). These are not facts found by the legislature (or facts about legislatures) but rather background social facts about the world used to decide broad questions of law and policy. 209 See Davis, Problems of Evidence, supra note 208, at 402–03. Courts can consider evidence outside the record to establish legislative facts, but it is often the plaintiffs who introduce the evidence. Advocates deploy empirical evidence to establish legislative facts, such as introducing evidence about the economic benefits of marriage, and to establish adjudicative facts, such as introducing evidence about Shaken Baby Syndrome to show that a particular child was injured by a parent. See supra text accompanying note 97; infra text accompanying notes 283–287. Advocates also use empirical evidence to undermine an opponent’s case, such as the plaintiffs in the marriage equality cases presenting empirical evidence about child outcomes to show that bias, not a concern about child well-being, animated the different-sex marriage requirement. See supra text accompanying notes 96–100. The use of empirical evidence to adjudicate legislative facts is not new. From the Brandeis brief to the doll study in Brown v. Board of Education, 210 See 347 U.S. 483, 494 & n.11 (1954) (“Whatever may have been the extent of psychological knowledge at the time of Plessy v. Ferguson, [the finding that children of color are harmed educationally and psychologically by segregation] is amply supported by modern authority.”); Richard Kluger, Simple Justice: The History of Brown v. Board of Education and Black America’s Struggle for Equality 315–40 (1975) (describing the efforts of the plaintiffs to build their social science case in the trial court and the difficulty of showing that psychological harm stemmed from legal segregation and not prejudice more broadly). courts have looked to empirical evidence for this purpose. 211 See Kenneth Culp Davis, Facts in Lawmaking, 80 Colum. L. Rev. 931, 933–40 (1980) (arguing empirical evidence—as it relates to legislative facts—influences constitutional reasoning). Examples include the use of evidence about group size and decisionmaking to determine the constitutionality of a five-person jury, evidence on pornography and illegal behavior to uphold a zoning regulation, and evidence about the availability of contraceptives and early sexual activity to uphold a law restricting access to birth control for minors. Id.

In family law, legislative facts deeply influence judicial deter-minations of rights. When a court finds, based on empirical evidence, that children raised by same-sex parents have similar outcomes as children raised by different-sex parents, this legislative fact informs the court’s judgment about the state’s purported rationale in limiting marriage to different-sex couples. 212 See supra section I.B.1.a. When a court finds, based on empirical evidence, that witnessing intimate partner violence can be harmful to children but that foster care presents its own harm, these legislative facts help the court assess the constitutionality of policies removing children from homes with intimate partner violence. 213 See supra section I.B.1.c. In a variety of contexts, then, social science and hard science inform judicial understandings of the implications of legal rules or the underlying conditions giving rise to familial conflicts.

Legislative facts differ from adjudicative facts, which are case-specific facts about the parties before the court. 214 See Davis, Problems of Evidence, supra note 208, at 402–03. Courts regularly use empirical evidence to establish adjudicative facts as well. Thus, when the Court applied the undue burden test in Whole Woman’s Health, it dissected the evidence about the actual effects of the restrictions at issue on the availability of abortion in Texas. 215 See supra text accompanying notes 127–130.

In legislatures, lawmakers also establish facts, although these are generally called legislative findings, not facts. Lawmakers use empirical evidence to establish legislative priorities, understand the contours of a given social problem, and divine possible links between policy tools and preferred outcomes. 216 There is considerable literature about the comparative institutional competence of legislatures, as compared with courts, to gather facts. See, e.g., Cass R. Sunstein & Adrian Vermeule, Interpretation and Institutions, 101 Mich. L. Rev. 885, 886–87, 920–25 (2003) (comparing the institutional capacities of courts and legislatures).

Administrative agencies use empirical evidence to develop and implement policies. When an agency decides on a priority, or im-plements a state mandate, it turns to empirical evidence to guide the policy choices. The Mayor’s Office in Providence used research on effect-ive early childhood interventions to design the program elements of Providence Talks. 217 See supra text accompanying note 198. Administrative agencies also use empirical evidence to distribute resources. When deciding how to allocate caseworker time in the child welfare system, predictive analytics directs attention to the highest-risk cases. 218 See supra text accompanying note 188. And, finally, administrative agencies use empirical evidence to evaluate the effectiveness of their efforts. Intervention programs, for instance, now regularly incorporate evaluation mech-anisms. 219 See Flowers, supra note 34, at 192–95 (discussing various means of evaluating collected data and of implementing insights).

This description of the empirical turn is not to claim that legal actors are using only empirical evidence across family law’s institutions. Moreover, when legal actors use empirical evidence, they can do so for a variety of reasons, as elaborated below. 220 See infra section II.B.2.b. But there is no doubt that family law has embraced empirical evidence, a turn that has considerable upsides and downsides, as the next Part explores.

II. The Promise and Peril of an Empirical Family Law

Family law can and should draw on the wealth of research on families. There are numerous advantages to an empirically based family law, and this Part briefly outlines these benefits. But there are also substantial bases for concern. Beyond the common problems of reliability and translation by legal actors, there are fundamental concerns about the multiple ways empirical evidence skews decisionmaking. As this Part argues in detail, empirical evidence focuses attention on the outcomes of legal rules, not competing values. It provides political cover for the value judgments that are made. And it risks replicating historical discrimination against nondominant families.

A. The Benefits of Evidence-Based Family Law

At the most basic level, family law and policies should draw on a well-informed understanding of family life. A periodic, congressionally mandated study of the incidence of child maltreatment, for example, illuminates the risk factors for child abuse and neglect. That study found that children in families with low socioeconomic status experience seven times the rate of neglect as children in families with higher socioeconomic status. 221 See Andrea J. Sedlak et al., U.S. Dep’t. of Health & Human Servs., Fourth National Incidence Study of Child Abuse and Neglect (NIS-4): Report to Congress 12 (2010), http://www.acf.hhs.gov/sites/default/files/opre/nis4_report_exec_summ_pdf_jan2010.pdf [http://perma.cc/EXU5-83KW]. This is a critical starting point for addressing and attempting to reduce the incidence of child neglect.

More specifically, a rigorous, research-based approach to family law helps the government be more effective in its efforts, giving legal actors a clearer sense of where legal inputs might yield particular social outcomes. A recent study, for example, found that providing legal counsel to victims of intimate partner violence had substantial benefits: Over time, women reported substantially less physical violence in their lives, improved psychological well-being, and increased income. 222 See Carolyn Copps Hartley & Lynette M. Renner, The Longer-Term Influence of Civil Legal Services on Battered Women 7–8, 52–62 (2016), http://www.ncjrs.gov/pdffiles1/nij/grants/249879.pdf [http://perma.cc/3PFU-HZTY]. As explored in section II.B, there are often questions about a study’s methodology and thus its relevance to legal debates. In this study, for example, the researchers could not, for both ethical and methodological reasons, use a control group. See id. at 30 (explaining that it would be unethical to assign some victims to a no-treatment group when the victims are facing imminent and significant danger). The researchers thus used a panel-study method, comparing the same group over time. See id.
A related benefit is that empirical evidence informs both academic and policy debates. An empirical study of the impact of burdens of proof in child protection cases, for example, demonstrated that increasing the standard of proof decreased the number of substantiated reports, primarily affecting cases that were difficult to prove. See Nicholas E. Kahn, Josh Gupta-Kagan & Mary Eschelbach Hansen, The Standard of Proof in the Substantiation of Child Abuse and Neglect, 14 J. Empirical Legal Stud. 333, 356–57 (2017). Similarly, the sociological research identifying a typology of intimate partner violence, see Michael P. Johnson, A Typology of Domestic Violence: Intimate Terrorism, Violent Resistance, and Situational Couple Violence (2008), has led to a sustained debate about both the typology itself, see Meier, Johnson’s Differentiation Theory, supra note 10, at 4, 6, 12–16 (arguing Johnson’s data do not support his typology or his claim that intimate terrorism is a rare phenomenon and identifying flaws in the research and conclusions), and the typology’s legal consequences, see Nancy Ver Steegh, Differentiating Types of Domestic Violence: Implications for Child Custody, 65 La. L. Rev. 1379, 1384–414 (2005) (contemplating the effect of Johnson’s research on state intervention, particularly on allocating resources and determining child custody). In short, empirical work provides an important grounding for law-reform debates.
Another example is groundbreaking work by economists Raj Chetty and Emmanuel Saez on economic mobility. 223 How Can We Improve Economic Opportunities for Our Children? We Use Big Data to Identify New Pathways to Upward Mobility, Equal. of Opportunity Project, http://
www.equality-of-opportunity.org/ [http://perma.cc/M6H9-BTET] (last visited Sept. 11, 2017).
Their work has unearthed specific factors—particularly racial segregation, concentrated poverty, and a lack of transportation infrastructure—that deeply influence eco­nomic mobility. 224 See Raj Chetty et al., Where is the Land of Opportunity? The Geography of Intergenerational Mobility in the United States, 129 Q.J. Econ. 1553, 1554–62, 1586–620 (2014). The research shows the positive benefits of some government efforts, such as the U.S. Department of Housing and Urban Development’s Moving to Opportunity program, which helps families move from areas of high poverty to more mixed-income, higher-opportunity neighborhoods. 225 See Raj Chetty et al., The Effects of Exposure to Better Neighborhoods on Children: New Evidence from the Moving to Opportunity Experiment, 106 Am. Econ. Rev. 855, 870–76 (2016).

Relatedly, empirical evidence can guide state investments, promoting the more efficient and effective use of scarce resources. Economist James Heckman has shown that investing in early childhood is far more cost effective for producing desirable long-term outcomes, such as high school graduation rates and adult earnings, than investments later in childhood and in adult training programs. 226 See James J. Heckman, Giving Kids a Fair Chance 5–6, 13–41 (2013). At the federal level, President Obama was a leader in using empirical evidence to evaluate governmental programs, determining which were supported by evidence and which were not and, thus, should be changed or defunded. 227 See Ron Haskins & Greg Margolis, Show Me the Evidence: Obama’s Fight for Rigor and Results in Social Policy 2–30 (2015) (explaining the Obama Administration’s policy of evaluating social intervention programs on the basis of rigorous evidence). Obama ran into political resistance as well as bureaucratic inertia when he tried to drop programs, 228 See Ron Haskins & Jon Baron, The Obama Administration’s Evidence-Based Social Policy Initiatives: An Overview, in Evidence for Social Policy and Practice 28, 28–29 (2011), http://www.nesta.org.uk/sites/default/files/evidence_for_social_policy_and_practice.pdf [http://perma.cc/QHX4-ZC4A] (observing that, even when an empirical evaluation finds a program to be ineffective, “it does not follow that the Administration or Congress will take action”). but he was successful in supporting new programs with a strong evidence bases, such as teen pregnancy prevention efforts. 229 See Haskins & Margolis, supra note 227, at 67–101 (explaining the Obama Administration’s approach to reducing teen pregnancy). As this example demonstrates, empirical evidence does not necessarily overcome entrenched interests and political preferences, but it can structure the debate about state policies and investments in families and children.

Empirical evidence can also justify state intervention in families, overcoming the basic rule of family autonomy. The child welfare system, for example, is predicated on the empirically grounded understanding that child abuse and neglect are harmful to children. 230 See Weithorn, supra note 67, at 55–60 (describing the work of Henry Kempe establishing Battered Child Syndrome and the role of this research in spurring the creation of the modern child welfare system). Similarly, legal rules and policies around intimate partner violence are based on the knowledge that intimate partner violence has significant and far-reaching negative consequences both for individuals and society more broadly. 231 S. Rep. No. 103-138, pt. 3, at 41–42 (1993) (citing evidence about the rate of intimate partner violence and the ensuing cost and noting that “we spend $5 to $10 billion a year on health care, criminal justice, and other social costs of domestic violence”).

Additionally, drawing on empirical evidence can help unmask prejudice and help dislodge stereotypes. A persistent cultural image is the dysfunctional Black family and particularly the absent Black father. 232 See Office of Policy Planning & Research, U.S. Dep’t of Labor, The Negro Family: The Case for National Action 5, 47 (1965), http://web.stanford.edu/~mrosenfe/
Moynihan%27s%20The%20Negro%20Family.pdf [http://perma.cc/2SMA-6M6R] (arguing measures such as the Civil Rights Act of 1964 were insufficient to assure African Americans full participation in society and partly blaming “the deterioration of the Negro family” on its “tangle of pathology . . . capable of perpetuating itself without assistance from the white world although also noting the large context of “three centuries of injustice”).
Recent empirical evidence, both qualitative and quantitative, challenges this stereotype. As compared with white and Latino men, Black men who do not live with their children are more likely to maintain better co-parenting relationships with the mothers of their children and more likely to be involved with their children. 233 See Kathryn Edin & Timothy J. Nelson, Doing the Best I Can: Fatherhood in the Inner City 215 (2013) (“[O]ur black fathers are more involved than the white fathers are with their children, especially when the kids are younger.”); Marcia J. Carlson et al., Coparenting and Nonresident Fathers’ Involvement with Young Children After a Nonmarital Birth, 45 Demography 461, 473 (2008) (finding, among nonresident fathers, Black men were more likely than white or Hispanic men to have maintained contact with their children); Robert I. Lerman, Capabilities and Contributions of Unwed Fathers, Future Child., Fall 2010, at 63, 64, 75 (“Black fathers are more likely than white and Hispanic fathers to maintain close contact with their children, especially in cases when the father neither marries nor cohabits with the mother.”). Another example is the de-biasing evidence produced in the marriage equality cases. In Perry, one of the proponents of California’s constitutional amendment was the secretary of the America Return to God Prayer Movement, which operated a website containing statements urging people to vote for the amendment because “homosexuals are twelve times more likely to molest children.” 234 Perry v. Schwarzenegger, 704 F. Supp. 2d 921, 937 (N.D. Cal. 2010) (citing trial transcript pages 1919–22), aff’d sub nom. Perry v. Brown, 671 F.3d 1052 (9th Cir. 2012), vacated and remanded sub nom. Hollingsworth v. Perry, 133 S. Ct. 2652 (2013); see also id. at 982–83 (using evidence to refute the stereotypes “that gays and lesbians are affluent, self-absorbed and incapable of forming long-term intimate relationships” and that they are “disease vectors or . . . child molesters who recruit young children into homosexuality”). The trial court used evidence introduced at trial to rebut this and other stereotypes. 235 See id. at 982–85 (finding no evidence to support this stereotype). Similarly, trial courts used the Daubert test to dismiss expert testimony that was rooted in prejudice rather than scientifically grounded empirical evidence. 236 See id. at 948 (rejecting expert testimony of think tank founder David Blankenhorn because “nothing in the record other than the ‘bald assurance’ of Blankenhorn suggests that Blankenhorn’s investigation into marriage has been conducted to the ‘same level of intellectual rigor’ characterizing the practice of anthropologists, sociologists or psychologists” (first quoting Daubert v. Merrell Dow Pharm., 43 F.3d 1311, 1316 (9th Cir. 1995) (on remand); then quoting Kumho Tire Co. v. Carmichael, 526 U.S. 137, 152 (1999))).

Finally, an empirically based family law helps provide a counter-weight to politically fraught battles over family regulation. Overwhelming evidence has shown that childhood vaccines promote individual as well as communal health. 237 Ctrs. for Disease Control & Prevention, Why Are Childhood Vaccines So Important?, Vaccines & Immunizations, http://www.cdc.gov/vaccines/vac-gen/howvpd.htm [http://perma.cc/928E-7EQZ] (last updated Aug. 18, 2017) (explaining that vaccines have prevented countless cases of diseases and saved millions of lives). There are risks for individual children, 238 Ctrs. for Disease Control & Prevention, Who Should NOT Get Vaccinated with These Vaccines?, Vaccines & Preventable Diseases, http://www.cdc.gov/vaccines/vpd/should-not-vacc.html [http://perma.cc/3S5Q-DRY2] (last updated May 8, 2017) (specifying circumstances in which certain children should not be vaccinated, often because of allergies or underlying medical conditions). but there is no credible evidence that vaccines during early childhood cause widespread harm among children or that vaccines contribute to conditions such as autism. 239 Ctrs. for Disease Control & Prevention, Vaccines Do Not Cause Autism, Vaccine Safety, http://www.cdc.gov/vaccinesafety/concerns/autism.html [http://perma.cc/9YKE-R9ZG] (last updated Nov. 23, 2015). Based on this evidence, professional groups, such as the American Academy of Pediatricians, have stated their unqualified support for vaccinating young children. 240 Karen Remley, American Academy of Pediatrics Reiterates Safety and Importance of Vaccines, Am. Acad. of Pediatrics (Sept. 17, 2015), http://www.aap.org/
en-us/about-the-aap/aap-press-room/pages/American-Academy-of-Pediatrics-Reiterates-Safety-and-Importance-of-Vaccines.aspx [http://perma.cc/R7KJ-ZAWQ].
Empirical evidence does not easily combat motivated cognition, 241 See Emily Balcetis & David Dunning, See What You Want to See: Motivational Influences on Visual Perception, 91 J. Personality & Soc. Psychol. 612, 612–13 (2006) (describing the literature finding that people perceive the world around them, including factual information, in a manner that is consistent with their beliefs); Cultural Cognition Project at Yale Law School, http://www.culturalcognition.net/ [http://perma.cc/NP8A-5U8Y] (last visited Sept. 11, 2017) (“Cultural cognition refers to the tendency of individuals to conform their beliefs about disputed matters of fact (e.g., whether humans are causing global warming; whether the death penalty deters murder; whether gun control makes society more safe or less) to values that define their cultural identities.”). For an example in family law, see The Cultural Cognition Project at Yale Law School, Cultural Cognition of Gay and Lesbian Parenting: Summary of First Round Data Collection 16 [hereinafter Cultural Cognition Project Study], http://static1.1.sqspcdn.com/static/f/386437/4705742/1264041920357/
Stage+1+Report.pdf?token=s1fuz1P6cvBYibVBP2pLhKH7oBQ%3D [http://perma.cc/5A5D-RSZT] (last visited Sept. 11, 2017) (“A majority of Americans say that their position on gay and lesbian adoption is centered on the welfare of children. However, few say they would change their minds if shown convincing contrary evidence.”).
and thus, it has not necessarily changed the minds of those vehemently opposed to childhood vaccines. 242 See, e.g., Autism and the Vaccine Debate, Focus for Health, http://
www.focusforhealth.org/autism-and-vaccine-debate/ [http://perma.cc/5C45-ZUXC] (last visited Sept. 11, 2017) (“The debate over whether or not vaccines contribute to the development of autism is far from settled.”).
Nor does evidence always overcome political preferences. 243 See Shear et al., supra note 14 (describing how then-President-Elect Trump met with Robert F. Kennedy, Jr., a vaccine skeptic, and asked him to lead a commission on vaccine safety). But the empirical evidence has helped sway public policy, with states such as California tightening grounds for legal exemptions. 244 See S.B. 277, 2015 Leg., Reg. Sess. (Cal. 2015) (describing the new law as “eliminat[ing] the exemption from existing specified immunization requirements based upon personal beliefs”). More broadly, the widespread availability of data and work in behavioral economics and other fields showing that human decisionmaking is prone to multiple biases has encouraged the use of empirical evidence and data-driven decisionmaking. 245 See Daniel Kahneman, Thinking, Fast and Slow 3–15 (2011) (showing how human decisionmaking reflects multiple, predictable biases); Mayer-Schönberger & Cukier, supra note 182, at 6–18 (discussing the role of big data in “humankind’s quest to quantify and understand the world”). The reliance in the public, private, and nonprofit sectors on data, algorithms, and so on is an attempt to correct for these biases and imperfections. In short, empirical evidence holds the potential—even if it does not always deliver—to help depoliticize debates and focus attention on workable solutions. 246 For another example, see Clare Huntington, Postmarital Family Law: A Legal Structure for Nonmarital Families, 67 Stan. L. Rev. 167, 184–91 (2015) [hereinafter Huntington, Postmarital Family Law] (describing the debate about child outcomes and family structure and analyzing the growing evidence that family structure itself, and not just poverty and other characteristics that often accompany nonmarital childbearing, contribute to worse child outcomes).

B. The Empirical Turn in Critical Perspective

Despite the many benefits of an empirically based family law, there are also significant concerns. This section draws on the larger literature about evidence-based and data-driven decisionmaking to identify and explore what is concerning and fraught about the empirical turn in family law. 247 See, e.g., Joel Best, Damned Lies and Statistics: Untangling Numbers from the Media, Politicians, and Activists 32 (2001) (explaining many statistics are produced inaccurately, by guessing or by relying on flawed sampling, measurements, or definitions); Cathy O’Neil, Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy 7–8, 10, 12 (2016) (arguing “weapons of math destruction” are self-perpetuating models that work in bulk, are relied upon despite the fact that they are prone to error, and affect a range of decisions, including advertising and prison sentencing). For one of the foundational critiques of empiricism within law, see Lee Epstein & Gary King, The Rules of Inference, 69 U. Chi. L. Rev. 1, 6–7, 23, 87, 125–27, 131, 211–12 (2002) (arguing legal scholars often do not follow the same rules of inference as are used in the social and natural sciences, give “over-confident” conclusions, fail to address selection biases, exclude reliability analyses, lack blind peer-review and documentation of data sources, and rely on improper random sampling). This section identifies threshold concerns about the quality of research and the ability of legal actors to use empirical evidence. These concerns are not particularly new, and this section thus describes these problems expeditiously. The section then turns to the heart of the Essay, arguing that even in a world of perfect information, with legal actors well trained in the use of evidence, the empirical turn can influence decisionmaking in troubling ways.

1. Reliability and Translation. — Much empirical evidence on the family is less reliable than a casual observer might conclude. The most fundamental question is whether empirical evidence on families—especially social scientific evidence—satisfies basic scientific norms. Methodological concerns are rife. As noted in the marriage equality deci­sions, many studies on the family use small convenience samples and not large cross-sections of the population. 248 See supra text accompanying note 114. The Fragile Families and Child Wellbeing Study is a notable exception. See infra text accompanying notes 272–274. Further, because of ethical concerns, subjects are not randomly assigned to control and intervention groups. It would be unethical, for example, to remove some children from homes that are perceived to pose a threat to the child’s physical safety and leave another group of children in homes with the same perceived threat level. Hence, when researchers compare outcomes for children placed in foster care and children left at home, they are not comparing similar groups. 249 For a discussion of these challenges, see Joseph J. Doyle, Jr., Causal Effects of Foster Care: An Instrumental-Variables Approach, 35 Child. & Youth Servs. Rev. 1143, 1143–44 (2013). Researchers try to account for this in a variety of ways. See, e.g., id. (describing the possibility of testing the causal effects of removal by using the natural experiment of varying removal recommendations among caseworkers and finding that, for marginal cases, foster care placement is associated with higher juvenile delinquency rates later in life and no corresponding increase in child safety).

By contrast, when the state offers a limited benefit—say, spaces in a Head Start program—it can randomly assign participants. But even then, and again for ethical reasons, the state will often recommend that families in the control group receive another kind of intervention. This is indeed what happened with the Head Start studies and may be one reason why these studies show a modest impact for Head Start—the comparison was not between Head Start and no preschool but rather between Head Start and a different program. 250 See U.S. Dep’t Health & Hum. Servs., Head Start Impact Study Final Report, at iii–v (2010), http://www.acf.hhs.gov/sites/default/files/opre/executive_summary_final.pdf [http://perma.cc/YY6C-7G5X] (noting that sixty percent of the children in the control group were enrolled in some kind of group program and describing the impact of Head Start but also noting that the effects largely did not persist into the school years). Indeed, when researchers study the impact of early childhood programs as a whole, there are marked benefits, both in the short and long term. See Lynn A. Karoly, M. Rebecca Kilburn & Jill S. Cannon, Early Childhood Interventions: Proven Results, Future Promise 55–78, 128–29 (2005). There are numerous other reasons family research may be less reliable than some other areas of research. 251 See Robert E. Emery et al., “Bending” Evidence for a Cause: Scholar-Advocacy Bias in Family Law, 54 Fam. Ct. Rev. 134, 135, 141–44 (2016) [hereinafter Emery et al., “Bending” Evidence] (noting there are relatively few family researchers and thus relatively few studies overall and discussing the debate about overnight visits for very young children, which turned on only four studies); Irwin Sandler et al., Convenient and Inconvenient Truths in Family Law: Preventing Scholar-Advocacy Bias in the Use of Social Science Research for Public Policy, 54 Fam. Ct. Rev. 150, 151 (2016) (“[A] major limitation on the use of research to shape policy and practice in the family law field is the paucity of research evidence that is sufficiently replicated and based on valid methodology to have clear and unambiguous implications for practice and policy . . . .”).

Another problem potentially compromising reliability is bias. Empirical evidence carries a mantle of objectivity, but despite the availability of clear research standards, 252 See Emery et al., “Bending” Evidence, supra note 251, at 136 (noting that in social science scholarship “the scholar’s purpose is to be, insofar as possible, self-aware and critical about prior assumptions, personal values, and biases, willing to subject hypotheses to rigorous inquiry and falsifiable tests, and prepared to consider alternative interpretations of the data”). studies are not always neutral explorations of the world of families. 253 In addition to the problems identified in the text, data can be flawed. See Solon Barocas & Andrew D. Selbst, Big Data’s Disparate Impact, 104 Calif. L. Rev. 671, 684 (2016) (describing the ways “institutions might maintain systematically less accurate, precise, timely, and complete records for certain classes of people” and noting that “[e]ven a dataset with individual records of consistently high quality can suffer from statistical biases that fail to represent different groups in accurate proportions”); see also Michael Mattioli, Disclosing Big Data, 99 Minn. L. Rev. 535, 546 (2014) (“Data is often deeply infused with the subjective judgments of those who collect and organize it.”). This bias can come from multiple sources. When research is funded by an entity with a stake in the answer, this vested interest casts a shadow over the research. 254 This was an issue in the marriage equality context, with the trial court in DeBoer discounting one of the central studies relied upon by the defendants because it had been funded by a party seeking to show children benefited when raised by different-sex parents. See DeBoer v. Snyder, 973 F. Supp. 2d 757, 766 (E.D. Mich.), rev’d, 772 F.3d 388 (6th Cir. 2014), rev’d sub nom. Obergefell v. Hodges, 135 S. Ct. 2584 (2015). Further, a researcher can influence a study if the researcher has an ideological commitment—sometimes acknowledged, sometimes unacknowledged—that seeps into the study design and the interpretation of the results. 255 See Emery et al., “Bending” Evidence, supra note 251, at 134, 141–44 (coining the term scholar-advocacy bias, defined “as the intentional or unintentional use of the language, methods, and approaches of social science research, as well as one’s status as an expert, for the purpose and/or outcome of legitimizing advocacy claims at the cost of misrepresenting research findings”). For an example of this critique, see Ummni Khan, Antiprostitution Feminism and the Surveillance of Sex Industry Clients, in Feminist Surveillance Studies 189, 193–202 (Rachel E. Dubrofsky & Shoshana Amielle Magnet eds., 2015) (describing this problem in research on sex workers conducted by those with an antiprostitution bias). A controversial example is research on family preservation efforts for children at risk of foster care placement. Some scholars contend this research has an ideological bias in favor of family preservation, which, in turn, influences the result. 256 See Elizabeth Bartholet, Thoughts on the Liberal Dilemma in Child Welfare Reform, 24 Wm. & Mary Bill Rts. J. 725, 726–27 (2016) (critiquing the “corrupt policy-research merger” and contending that some researchers choose programs, such as family preservation, that reflect the researchers’ values and then set out “not to test, but instead to prove the programs’ efficacy” in an effort to persuade policymakers to adopt the preferred program).

Researchers also bring cultural biases around race and class to their research, which can influence the focus of inquiry, study design, and the interpretation of data. Research on parenting, for example, usually follows a taxonomy of parenting styles: authoritarian, authoritative, permissive, and disengaged. 257 See Diana Baumrind, The Development of Instrumental Competence Through Socialization, in 7 Minnesota Symposia on Child Psychology 3, 13–14 (Anne D. Pick ed., 1973) (describing the taxonomy of parenting styles); see also Jeanne Brooks-Gunn & Lisa B. Markman, The Contribution of Parenting to Ethnic and Racial Gaps in School Readiness, Future Child., Spring 2005, at 139, 148 (applying Baumrind’s scheme to Black mothers). This taxonomy, however, is based on white, middle-class families and does not incorporate a style of parenting more associated with Black women—strict but nurturing. 258 See Brooks-Gunn & Markman, supra note 257, at 148 (noting that studies have found that Black mothers more frequently exhibit “tough love” parenting styles than do white mothers). Once identified, researchers could determine that this style of parenting is correlated with positive outcomes for children, but before it was acknowledged, this style of parenting was subsumed under the more pathologized model of authoritarian parenting. 259 See id. In this way, researchers can interpret some behavior as pathological rather than merely adaptive or a different way of flourishing.

This cultural myopia can profoundly affect research. The influential Adverse Childhood Experiences (ACE) Study, which largely studied white, middle-class individuals, found a correlation between family-based trauma and abuse and long-term health. 260 Vincent J. Felitti et al., Relationship of Childhood Abuse and Household Dysfunction to Many of the Leading Causes of Death in Adults: The Adverse Childhood Experiences (ACE) Study, 14 Am. J. Preventive Med. 245, 251 (1998). But when researchers conducted follow-up studies with a more diverse population, they found that other childhood traumas, including experiencing discrimination, living in an unsafe neighborhood, and witnessing community violence, were also highly predictive of long-term health. 261 See Peter F. Cronholm et al., Adverse Childhood Experiences: Expanding the Concept of Adversity, 49 Am. J. Preventive Med. 354, 355 (2015) (noting other community-level factors associated with long-term health); David Finkelhor et al., Improving the Adverse Childhood Experiences Study Scale, 167 J. Am. Med. Ass’n Pediatrics 70, 70–71 (2013) (“Among the predictors missing from the ACE Study model are peer rejection, exposure to violence outside the family, low socioeconomic status, and poor academic performance.”); see also Nancy E. Dowd, Straight Out of Compton: Developmental Equality and a Critique of the Compton School Litigation, 45 Cap. U. L. Rev. 199, 235–45 (2017) (exploring this research and its implications for law). The original study did not account for these consequences of structural racism and inequality, which affect low-income youth and youth of color, and thus did not fully capture the adverse circumstances of some children’s lives. 262 More broadly, some of the foundational research on child development was conducted using only white children. See Guthrie, supra note 62, at 50–52; Nancy E. Dowd, Black Boys Matter: Developmental Equality, 45 Hofstra L. Rev. 47, 59–61 (2016) (describing this research and its built-in biases).

Relatedly, researchers—intentionally or not—can frame research to confirm preexisting beliefs rather than to challenge them. For example, a persistent question in welfare policy is whether providing low-income families with material support will encourage these families to have additional children. 263 This concern is reflected in family cap provisions, which limit welfare benefits to existing children. At least nineteen states have such provisions. Welfare Reform: Family Cap Policies, Nat’l Conference of State Legislatures (Jan. 31, 2011), http://www.ncsl.org/issues-research/human-services/welfare-reform-family-cap-policies.aspx [http://perma.cc/9UWU-UMTC]. For example, in California until January 2017, if a child was born to a woman who had been receiving assistance for the ten previous months, the woman could not receive additional support for the child unless the pregnancy was the result of rape, incest, or “conceived as a result of contraceptive failure if the parent was using an intrauterine device, a Norplant, or the sterilization of either parent.” Cal. Welf. & Inst. Code § 11450.04(b)(3) (repealed 2017). Researchers typically do not, however, ask middle-income families whether the availability of public education or the larger houses enabled by the home mortgage-interest deduction encourage them to have more children. It might be that these middle-class parents are indeed responding to such incentives. This would show that families across the income spectrum respond to governmental incentives and subsidies, but the failure to pose the question across class lines skews public policy. Only low-income families are assumed to possess path-ologies that need to be tamed.

A different element of reliability is whether the data are complete. It is important to ask which groups are not included in a study. Family scholars in multiple disciplines routinely rely on data from the American Time Use Survey (ATUS). 264 See, e.g., Suzanne M. Bianchi et al., Changing Rhythms of American Family Life 174, 223 (2006) (indicating that the study considered “time diaries” that were collected as part of the ATUS). Conducted by the U.S. Department of Labor’s Bureau of Labor Statistics, the ATUS measures the amount of time participants spend doing specified activities, including paid labor, childcare, housework, and leisure activities. 265 Frequently Asked Questions (FAQs), U.S. Dep’t of Labor, Bureau of Labor Statistics, http://www.bls.gov/tus/atusfaqs.htm#1 [http://perma.cc/SC6B-23PN] (last visited Sept. 11, 2017). The data are collected through phone interviews. 266 Id. The Bureau of Labor Statistics tries to include families without telephones by sending letters to their addresses and providing a toll-free number for individuals to call, 267 Id. but it is not surprising that the survey radically underrepresents people on the margins of society. 268 See Katharine G. Abraham et al., Nonresponse in the American Time Use Survey: Who Is Missing from the Data and How Much Does It Matter?, 70 Pub. Opinion Q. 676, 678, 697–98 (2006) (analyzing respondents and finding that people who are weakly integrated into the community are less likely to participate, largely because they are not contacted). ATUS thus gives an incomplete picture of American families.

There is a limit to what is known and, more fundamentally, what is knowable. 269 See Emery et al., “Bending” Evidence, supra note 251, at 135 (“Given the broad reach of family law, the rare use of random assignment studies (the ‘gold standard’ of scientific research), and the relatively small number of studies (and researchers) in the field, the ultimate empirical truth regarding many family law controversies often is ‘more research is needed.’”). Much social science research, especially on the family, cannot account for all the variables affecting outcomes. There is over­whelming evidence that children raised by married parents have better outcomes than children raised in any other family structure. 270 See Huntington, Failure to Flourish, supra note 64, at 31–34 (discussing child outcomes and parental marital status). Once factors that are correlated with family structure are taken into account, particularly income and parental education, the differences are far less pronounced. 271 See id. at 37. But even then, there remains a gap. The lead researchers of the Fragile Families and Child Wellbeing Study hypothesize that family structure plays a causal role in child outcomes because nonmarital families experience higher levels of relationship instability and multi-partner fertility; these factors contribute to worse outcomes because the relationship stress associated with changing partners negatively affects parenting. 272 See Sara S. McLanahan & Irwin Garfinkel, Fragile Families: Debates, Facts, and Solutions, in Marriage at the Crossroads: Law, Policy, and the Brave New World of Twenty-First-Century Families 141, 151–53 (Marsha Garrison & Elizabeth S. Scott eds., 2012). But even accounting for this causal role of family structure, the question about the full impact of family structure on child outcomes is partly unanswerable. After controlling for observable characteristics, there may be other, nonobservable characteristics that affect both family structure and outcomes. A person with strong interpersonal skills might choose to get married and stay married, and this kind of person might also be a more effective parent. This separate characteristic would drive the family structure and the child outcome, but it is difficult for an outside researcher to identify this characteristic. 273 See Jane Waldfogel et al., Fragile Families and Child Wellbeing, Future Child., Fall 2010, at 87, 92–93 (“A common challenge in research in this area is that parents who are single or cohabiting may have attributes . . . that differ from those of married parents and that also foster adverse child and adolescent outcomes.”). Researchers try to account for this selection bias in a number of different ways, but there is no easy way around the problem. 274 The longitudinal nature of the Fragile Families Study is an attempt to account for selection bias by identifying events early in a child’s life, such as a high-conflict parental relationship, that predate a family breakup and might separately influence the child’s outcomes. See Wendy Sigle-Rushton & Sara McLanahan, Father Absence and Child Well-Being: A Critical Review, in The Future of the Family 116, 127 (Daniel P. Moynihan et al. eds., 2004).

Finally, empirical evidence captures only those variables amenable to measurement. So much of family law and family life, however, is un-quantifiable. Studies about child outcomes tend to focus on metrics such as educational progress, adult earnings, and mental and physical health. 275 See supra text accompanying notes 97–102 (describing the evidence about child well-being introduced in the marriage equality litigation). These are important aspects of a child’s life, but so too are the intangibles: a child’s sense of belonging, trust, and feeling loved, not to mention the converse of dislocation, distrust, and feeling unwanted and unloved. 276 See, e.g., Erik H. Erikson, Growth and Crises of the Healthy Personality, in 1 Identity and the Life Cycle 50, 51–99 (1959) (contending that individuals develop through eight psychosocial stages and that each stage involves the acquisition of a virtue, including hope, will, purpose, competence, fidelity, love, care, and wisdom). Family law must account for these considerations. But by measuring some but not all salient aspects of family life, empirical evidence presents an incomplete picture of the relevant factors.

Compounding quality and reliability concerns is the reality that legal actors and advocates are not necessarily able to make nuanced judgments based on empirical evidence. This is partially about capacity—whether lawyers, judges, legislators, and executive branch officials have sufficient training to understand empirical research, including the limitations of most studies. As noted, research findings are often more tentative than lay readers might think. Consumers of these studies, however, are not always aware of their contingent nature.

Legal actors sometimes adopt policies based on incomplete or preliminary evidence. Early research on mandatory arrest and recidivism rates for perpetrators of intimate partner violence, for example, showed that recidivism rates decreased when police were required to arrest a perpetrator. 277 See Brinig, Empirical Work, supra note 11, at 1095–96 (describing this research). Based on this research, many jurisdictions adopted a mandatory-arrest policy. 278 See id. (“Based on Sherman’s study, the single most frequent reform these days is mandatory arrest.”). Subsequently, numerous researchers could not replicate the results or found only a modest correlation. 279 See id. at 1096 (citing the subsequent research). For an excellent analysis of the original study and the replication efforts, see generally Christopher D. Maxwell et al., Nat’l Inst. of Justice, U.S. Dep’t of Justice, The Effects of Arrest on Intimate Partner Violence: New Evidence from the Spouse Assault Replication Program (2001), http://www.ncjrs.gov/pdffiles1/nij/188199.pdf [http://perma.cc/S7JE-GBR5]. At least part of the reason was that the original study focused on a particular population—men who were employed and integrated into their communities and thus had a reputation to protect. 280 Brinig, Empirical Work, supra note 11, at 1095. The subjects of the subsequent studies were not employed and had less of a stake in the community and thus, arguably, less of an incentive to avoid arrest. 281 See id. at 1095–96 (“When the domestic abuse offender has less stake in community or employer reputation, mandatory arrest may cause more rather than less recidivism.”). Even with the new evidence about the non-generalizability of the original research, jurisdictions have been slow to abandon mandatory-arrest policies, notwithstanding considerable criticism, particularly from communities of color, about the detrimental impact of increased police involvement. 282 See, e.g., Aya Gruber, The Feminist War on Crime, 92 Iowa L. Rev. 741, 797–98 (2007) (explaining how “tough-on-crime advocates could rally around domestic violence reform while continuing to internalize and perpetuate racial characterizations of victims and criminals”).

Even when studies seem conclusive, interpretations and under-standing of data can change over time. 283 See Thomas S. Kuhn, The Structure of Scientific Revolutions 10–22, 164–72 (1962) (arguing science does not develop in cumulative, measured steps and instead reigning paradigms—the accepted wisdom on a particular topic—ultimately become unstable when they are repeatedly challenged). Shaken Baby Syndrome is a notorious example. Many parents and caregivers have been convicted of homicide based on seemingly settled evidence that three factors—retinal bleeding, bleeding in the protective layer of the brain, and brain swelling—are evidence that the caregiver shook the child violently and caused the child’s death. 284 See Deborah Tuerkheimer, Flawed Convictions: “Shaken Baby Syndrome” and the Inertia of Injustice 1–66 (2014); see also Eichner, supra note 169, at 273–78 (discussing the role of pediatricians in supporting the research and legal reliance on the triad of symptoms). More recent research, however, casts serious doubt on the importance of these three clinical findings, instead suggesting that most child deaths may well have been the result of other causes, particularly underlying conditions or older brain injuries. 285 See Tuerkheimer, supra note 284, at 17–30. The legal system is only just beginning to take note. A federal court recently vacated the murder conviction of a caregiver on actual innocence grounds, concluding that the evidence to support Shaken Baby Syndrome is “more an article of faith than a proposition of science.” 286 Del Prete v. Thompson, 10 F. Supp. 3d 907, 957 n.10 (N.D. Ill. 2014). But many parents and caregivers remain in prison, convicted based on what may now be understood as spurious findings. 287 See Tuerkheimer, supra note 284, at 173–93.

Perhaps in recognition of their own limited ability to understand empirical evidence, legal actors can be overly deferential to researchers and thus often do not act as competent gatekeepers. Some “science” is pure junk, but it can still affect the legal system—the “garbage in, garbage out” problem. Parental alienation syndrome, discussed above, is a good example. By labeling something a “syndrome,” the research takes on the veneer of science. 288 Professor Maxine Eichner has documented another example of this phenomenon. See Eichner, supra note 169, at 213 (criticizing the diagnosis of “medical child abuse”). As Eichner demonstrates, doctors have developed a new diagnosis of “medical child abuse,” defined as parents seeking supposedly unnecessary medical treatment, even when other doctors have ordered the treatment. Id. at 210. Based on this diagnosis, the child welfare system initiates child protective proceedings, which can lead to the removal of a child from the home, and the state sometimes brings criminal charges against parents. Id. at 211. Courts have accepted a diagnosis of medical child abuse as a distinct form of child abuse. Id. at 219. In addition to arguing that the diagnosis of medical child abuse violates the constitutional rights of parents to make medical decisions for their children, and that it circumvents the need for proving actual child abuse, Eichner contends that the diagnosis of medical child abuse is based on “flawed science and flawed medical practice.” Id. at 213. But the underlying work does not begin to meet the basic standards for scientific research and has been roundly rejected by researchers in multiple fields. 289 See supra note 164. Even though the court system theoretically has a gatekeeping mechanism to check the reliability of this kind of research—the Daubert standard and its corollaries in state courts—judges do not always use this standard, especially in family court. 290 See supra text accompanying notes 166–169.

The gatekeeping problem is arguably worse in the legislative branch, in which there is no accepted practice for gathering and examining empirical evidence in a manner that promotes neutrality and reliability. Compounding the problem, courts generally defer to legislative judgments, thus extending the reach of such evidence. This insulation and lack of accountability make the legislative process particularly susceptible to co-optation by political forces or incompetence. When the South Dakota Legislature created a task force to study abortion in the state, 291 See H.B. 1233, 2005 Leg., 80th Sess. (S.D. 2005). the process was tilted heavily in favor of finding problems with abortion, including the idea that abortion hurts women’s physical and mental health. 292 See Reva Siegel & Sarah Blustain, Mommy Dearest?, Am. Prospect, Oct. 2006, at 22, 22–23. The findings of the task force provided the basis for South Dakota’s restrictive laws on abortion, 293 See id. and the Supreme Court relied on them indirectly in Gonzales v. Carhart. 294 See 550 U.S. 124, 159 (2007) (citing Sandra Cano Brief, supra note 131, at 22–24).

There is somewhat less concern in the administrative context because courts have more latitude to act as checks on administrative decisionmaking. If agencies do not consider all aspects of a problem, then courts may strike down agency action as arbitrary and capricious or, in formal rulemaking and formal adjudication contexts, lacking substantial evidence. 295 See Administrative Procedure Act, 5 U.S.C. § 706(2)(A) (2012) (stating that a court may set aside an agency’s action if it is “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law”). An agency acts arbitrarily and capriciously when it:
“[H]as relied on factors which Congress has not intended it to consider, entirely failed to consider an important aspect of the problem, offered an explanation for its decision that runs counter to the evidence before the agency, or is so implausible that it could not be ascribed to a difference in view or the product of agency expertise.”
Motor Vehicle Mfrs. Ass’n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983).
These are the prevailing federal standards, and state standards are similar, 296 See Arthur Earl Bonfield, The Federal APA and State Administrative Law, 72 Va. L. Rev. 297, 297–302 (1986) (finding states modeled their administrative laws after the Model State Administrative Procedure Act, which drew upon the earlier federal Administrative Procedure Act). but this oversight might play out differently at the local level because these agencies typically do not have the same resources as state and federal governments. Local government agencies are deeply involved in the operations of the child welfare system and other aspects of family law yet may be less likely to explore all aspects of a problem or have the capacity to handle empirical evidence in a careful and sophisticated manner. 297 See Nestor M. Davidson, Localist Administrative Law, 126 Yale L.J. 564, 618 (2017) (arguing that “many local agencies have limited resources and lack the independent ability to develop sophisticated technical knowledge” but bring a different kind of expertise in their “mediating and information-collecting function”).

A different concern is the difficulty of translating social science into legal rules. It is not always clear how to do so because of the evolving nature of the underlying research; because legal rules require definite lines, even if expressed in a standard; and because the research may be more equivocal, not lending itself to this kind of concrete guidance. 298 For a discussion of these challenges, see Emily Buss, What the Law Should (and Should Not) Learn from Child Development Research, 38 Hofstra L. Rev. 13, 13–14 (2009) (identifying concerns with the law determining children’s capacities based on scientific research into child development); Emery et al., “Bending” Evidence, supra note 251, at 135–41 (highlighting a “scholar-advocacy bias” in the family law context); Terry A. Maroney, The False Promise of Adolescent Brain Science in Juvenile Justice, 85 Notre Dame L. Rev. 89, 145–60 (2009) (describing the limitations of using neuroscience in the context of juvenile justice); Joan S. Meier, Dangerous Liaisons: Social Science and Law in Domestic Violence Cases 1–3 (Feb. 2, 2017) (unpublished manuscript) (on file with the Columbia Law Review) (arguing law seeks definitive answers by examining specific facts and discrete events whereas social science engages in ongoing inquiries to answer amorphous questions about human nature). There is the further problem that empirical work does not necessarily point in one policy direction in light of the judgment calls and trade-offs inherent in lawmaking. 299 See Rachlinski, supra note 30, at 917–19 (arguing evidence-based decision-making is an easier fit in medicine and business, which have clear goals, such as treating a patient effectively or maximizing profit, than in law, which is inherently political, with often contested and competing goals). For example, research on overnight visits for very young children has not established a clear harm to children if they move back and forth between parents’ homes, but it does suggest some reason to be concerned. 300 See Emery et al., “Bending” Evidence, supra note 251, at 141–44 (summarizing this research). This preliminary evidence could be used to argue in favor of primary custody in one parent. But it could just as easily be used to argue that concerns about child attachment are not so grave as to require sole physical custody in one parent and that, instead, other values, including the involvement of both parents, should shape the resulting doctrine. 301 See id. (making this point); Clare Huntington, Early Childhood Development and the Law, 90 S. Cal. L. Rev. 755, 772–82, 799–807 (2017) [hereinafter Huntington, Early Childhood Development] (arguing states should at least consider the underlying research but should also balance a young child’s need for continuity against other interests, including the interests of both the child and the noncustodial parent in having a relationship). In short, empirical evidence often raises as many questions as it might answer.

2. Facts and Values. — Beyond reliability and translation issues, there are three fundamental concerns: Using empirical evidence focuses attention on the outcomes of legal rules, avoiding debates about contested values; empirical evidence allows decisionmakers to cloak value-based judgments in seemingly neutral garb; and empirical evidence puts families to the test, potentially replicating historical discrimination against nondominant families.

a. Skewing Debates. — Family law almost always involves a consideration of both outcomes and underlying values. Who should control a frozen embryo? 302 See, e.g., McQueen v. Gadberry, 507 S.W.3d 127, 149, 158 (Mo. Ct. App. 2016) (affirming the trial court’s decision that the divorcing couple’s frozen embryos were marital property and awarding the embryos to the couple jointly, not to be used unless both parties agreed to implantation). Should a mentally disabled parent be allowed to care for a child? 303 See, e.g., Lisa Miller, Who Knows Best, N.Y. Mag. (Jan. 25, 2016), www.thecut.com/2016/01/how-intelligent-to-be-a-parent.html [http://perma.cc/M47C-JSSV] (describing a mother’s experience with the child welfare system removing her child because of the mother’s low cognitive functioning). How much leeway should parents have to make decisions for their children? 304 See infra note 306 (discussing legal regulation of corporal punishment). And so on. Deciding how the law should answer these questions is partly about outcomes. Which legal rule about embryo disposition will lead to the most desirable bargaining regime between would-be parents? How does a child fare when raised by a mentally disabled parent? Does parental discretion further child well-being? But each question is also a debate about what society currently values. How should the law balance one person’s desire to procreate against another person’s desire not to? How much should the law protect a parent’s right to care for a child when the parenting may be substandard but not imminently dangerous? How important is pluralism in parenting? Moreover, many outcomes are values. When family law seeks to maximize child well-being, this is both an outcome and a value. When family law seeks to reduce family violence, this is both an outcome and a value. In short, values and outcomes often blend in practice, with most family law rules and policies evincing an inexorable mix of both.

Additionally, most family law questions involve competing values. Consider one of the above examples: parental discretion in childrearing matters. The law gives parents considerable leeway to use reasonable corporal punishment, 305 See, e.g., Mich. Comp. Laws § 750.136b(9) (2015) (“This section does not prohibit a parent or guardian, or other person permitted by law or authorized by the parent or guardian, from taking steps to reasonably discipline a child, including the use of reasonable force.”); Miss. Code Ann. § 97-5-39(2)(g) (2015) (“Nothing . . . [in] this subsection shall preclude a parent or guardian from disciplining a child of that parent or guardian, or shall preclude a person in loco parentis to a child from disciplining that child, if done in a reasonable manner . . . .”); S.C. Code Ann. § 16-3-95(D) (2014) (“This section may not be construed to prohibit corporal punishment or physical discipline which is administered by a parent or person in loco parentis in a manner which does not cause great bodily injury upon a child.”); Wyo. Stat. Ann. § 6-2-503(b)(i) (2015) (“Physical injury . . . exclud[es] reasonable corporal punishment.”). and courts distinguish permissible corporal punishment from impermissible child abuse. 306 See, e.g., State v. Matavale, 166 P.3d 322, 341–42 (Haw. 2007) (finding no criminal liability for striking a fourteen-year-old with various objects for lying); Willis v. State, 888 N.E.2d 177, 179–80 (Ind. 2008) (finding no criminal liability for a single mother who hit her eleven-year-old child with history of lying and stealing); State v. Wilder, 748 A.2d 444, 456–57 (Me. 2000) (finding that the father’s actions did not exceed the standard of permissible corporal punishment and thus the criminal conviction could not stand); Commonwealth v. Dorvil, 32 N.E.3d 861, 870 (Mass. 2015) (exempting parents and guardians from liability when the amount of force is reasonable, the use of force reasonably relates to the welfare of the child, and it does not cause, nor create, a substantial risk of physical or mental harm); Cobble v. Comm’r of Dep’t of Soc. Servs., 719 N.E.2d 500, 508 (Mass. 1999) (finding the conduct of a father did not fall within regulatory definition of child abuse when the father spanked his nine-year-old son with a belt for misbehaving in school); N.J. Div. of Youth & Family Servs. v. P.W.R., 11 A.3d 844, 855 (N.J. 2011) (finding slapping a sixteen-year-old on the face a few times was not excessive corporal punishment when the slaps left no marks and further noting that use of “excessive” in the statute “plainly recognizes the need for some parental autonomy in the child-rearing dynamic”). At issue in this line drawing is a concern about child outcomes (whether and what kinds of corporal punishment harm a child), the underlying value of child well-being, and other values, including family autonomy from the state, parental discretion, societal pluralism, and children’s dignitary interests. Deciding where to draw the line between conduct that is protected corporal punishment and conduct that is the basis for state intervention implicates all of these values.

A central problem with empirical evidence is that it focuses attention on the outcomes of legal rules and the values underlying those outcomes. As detailed in Part I, the debate about marriage equality was centered on the effect of same-sex parenting on children. Child well-being is an important outcome and value, but there were other values at stake too, notably equality, inclusion, and pluralism. The availability of abundant evidence on same-sex parenting and child well-being meant that this value took center stage, distracting from the other values. It was fortunate for advocates of marriage equality that the evidence on children’s outcomes aligned with the values of equality, inclusion, and pluralism. But if the empirical evidence had shown that children of same-sex couples somehow had worse outcomes, such as lower high school graduation rates, it would have been considerably harder to argue for marriage equality on the basis of these other values. To be sure, this partly reflects the institutional context of judicial review. If the state had evidence of worse child outcomes, then there might have been a legitimate state reason for restricting marriage to different-sex couples, and a court might have upheld the restriction, depending on the level of scrutiny applied. But this tendency to focus on outcomes and the under­lying values—a tendency encouraged and exacerbated by the availability of empirical evidence—is also a risk at the legislative level, with lawmakers choosing to focus on child outcomes rather than consider a range of values at stake in family law rules.

Moreover, the widespread availability of empirical evidence encourages decisionmakers to focus on those values that have been measured and are more amenable to measurement, giving shorter shrift to the values that have not been measured or are harder to measure. 307 As compared with child well-being and levels of family violence, it is harder, but not impossible, to measure whether a rule advances other values, such as pluralism. The question is whether such evidence exists. Child well-being and family violence, in particular, are susceptible to measurement and often are measured in social science studies of families. 308 See, e.g., Jennifer S. Barber et al., The Relationship Context of Young Pregnancies, 35 Law & Ineq. 175, 175–76, 192–96 (2017) (reporting the results of a study of eighteen- and nineteen-year-old women that focused on the correlation between pregnancy and intimate partner violence); Fragile Families & Child Wellbeing Study, Data and Documentation, Princeton Univ., http://fragilefamilies.princeton.edu/documentation [http://perma.cc/P5A6-LL7E] (last visited Sept. 11, 2017) (listing the types of data collected in a landmark longitudinal study on families, including “information on attitudes, relationships, parenting behavior, demographic characteristics, health (mental and physical), economic and employment status, neighborhood characteristics, and program participation”). Empirical evidence about these two values ensures they garner attention in a debate, potentially overshadowing other values. Moreover, these values are relatively uncontested and present a compelling case: It is hard to justify a legal rule based on competing values when the rule might be at odds with child well-being or a reduction in family violence. 309 This is not to suggest that, in practice, family law decisionmakers necessarily prioritize these factors. See Meier & Dickson, supra note 161, at 329 (reporting the results of a study of custody cases and finding that in cases in which the father alleged the mother was alienating the children, courts switched custody from mother to father at approximately the same rate regardless of whether the mother alleged the father was abusing the children).

Further, even if there were empirical evidence about a variety of competing values, this would not necessarily tell decisionmakers how to strike a balance among the values. Empirical evidence can clarify the stakes in a debate—showing how much different policy options advance or compromise each value 310 See Michael Simkovic, Young Scholar Medal Recipient’s Address at the 93rd Annual Meeting of the American Law Institute: What Can We Learn from Credit Markets? 12–13 (May 18, 2016) (on file with the Columbia Law Review) (“My role is to help policymakers understand the parameters of the tradeoffs that they’re facing. If there are tradeoffs between economic growth and equality or various values that we care about, they need information to understand the nature of those tradeoffs.”). —but empirical evidence does not tell us the importance of the values. 311 Some scholars believe values can be quantified and thus compared. See W. Kip Viscusi, Fatal Tradeoffs: Public and Private Responsibilities for Risk 19–21 (1992) (describing the willingness-to-pay tool as a means to quantify the value of life); Simkovic, supra note 310, at 13 (“As soon as you have two absolute values, then you need to start making tradeoffs because they sometimes come into conflict. Which means you need some method of measuring those values against each other, and the measurement that economists use is money.”). This Essay takes the position that this is a dubious proposition, particularly with inherently noneconomic values such as protection of the parent–child relationship and family autonomy. Cf. Frank Ackerman & Lisa Heinzerling, Pricing the Priceless: Cost-Benefit Analysis of Environmental Protection, 150 U. Pa. L. Rev. 1553, 1562–78 (2002) (critiquing the economic approach to measuring values in the environmental context). Empirical evidence can demonstrate, for example, that heightened evidentiary standards for abuse and neglect lead to fewer substantiated cases and thus less state intervention, 312 See supra note 222 (describing such empirical evidence). but the empirical evidence does not tell us how to weigh family autonomy against child safety. For this reason, empirical evidence can play only a limited role in helping decisionmakers strike a balance among competing values.

Consider the Indian Child Welfare Act (ICWA). This law and its state equivalents provide both procedural and substantive protections for Native American children facing removal from their homes, with an emphasis on keeping children at home and placing them with a Native American family if they are removed; the law also gives tribes, rather than state family courts, jurisdiction over cases involving Native American children. 313 See 25 U.S.C. §§ 1903, 1911–1913, 1915 (2012). For a description of state equivalents, see State Statutes Related to the Indian Child Welfare Act, Nat’l Conference of State Legislatures (July 19, 2017), http://www.ncsl.org/research/human-services/
state-statutes-related-to-indian-child-welfare.aspx [http://perma.cc/6AC6-CJNK]. Congress enacted ICWA for numerous reasons, including the need to redress historical discrimination against Native American families and respect the sovereignty of tribes. See 25 U.S.C. §§ 1901(4)–(5) (setting forth congressional findings, including “that an alarmingly high percentage of Indian families are broken up by the removal, often unwarranted, of their children from them by nontribal public and private agencies and . . . [many of those children] are placed in non-Indian foster and adoptive homes and institutions”). Historically, “the States, exercising their recognized jurisdiction over Indian child custody proceedings through administrative and judicial bodies, have often failed to recognize the essential tribal relations of Indian people and the cultural and social standards prevailing in Indian communities and families.” Id. § 1901(5); see also Adoptive Couple v. Baby Girl, 133 S. Ct. 2552, 2557 (2013) (noting that Congress enacted ICWA to address the “consequences to Indian children, Indian families, and Indian tribes of abusive child welfare practices that resulted in the separation of large numbers of Indian children from their families and tribes through adoption or foster care placement, usually in non-Indian homes” (internal quotation marks omitted) (quoting Miss. Band of Choctaw Indians v. Holyfield, 490 U.S. 30, 32 (1989))).
ICWA embraces a broad understanding of child well-being that includes a child’s ties to a tribe; the law also recognizes a tribe’s interest in sovereignty and continued vitality. 314 See 25 U.S.C. § 1901 (setting forth congressional findings, including that “there is no resource that is more vital to the continued existence and integrity of Indian tribes than their children”). For an excellent summary of ICWA, the current challenge to the laws, and the constitutional basis for the law, see Sarah Krakoff, They Were Here First: American Indian Tribes, Race, and the Constitutional Minimum, 69 Stan. L. Rev. 491, 506–17 (2017).

In a series of challenges to ICWA, plaintiffs are contesting the law on multiple grounds, including constitutional arguments that ICWA is con­trary to the interests of children because it prioritizes “blood” ties over children’s welfare, in essence contending that the law’s emphasis on preserving tribal ties and respecting tribal sovereignty compromises children’s outcomes. 315 See, e.g., Civil Rights Class Action Complaint for Declaratory and Injunctive Relief at 2, Carter ex rel. A.D. v. Washburn, No. CV-1259, 2017 WL 1019685 (D. Ariz. Mar. 16, 2017) [hereinafter Carter ex rel. A.D. Complaint], appeal docketed, No. 17-15839 (9th Cir. Apr. 26, 2017). One case was dismissed for lack of standing, see Nat’l Council for Adoption v. Jewell, 156 F. Supp. 3d 727, 733 (N.D. Va. 2015), and another for mootness, although the court noted that the “case presents significant constitutional questions” that “merit careful consideration” in the right case, Doe v. Piper, No. 15-2639 (JRT/DTS), 2017 WL 3381820, at *5 (D. Minn. Aug. 4, 2017). Another case is in progress. See Carter ex rel. A.D. v. Washburn, No. CV-15-01259-PHX-NVW, 2016 WL 5464582, at *5 (D. Ariz. Sept. 29, 2016) (granting motions to intervene in a case challenging the constitutionality of ICWA). For a description of the challenges, including the institutional supporters, see Casey Tolan, A Series of New Lawsuits Is Challenging How Native American Kids Are Adopted, Splinter (July 17, 2015), http://splinternews.com/a-series-of-new-lawsuits-is-challenging-how-native-amer-1793849248 (on file with the Columbia Law Review). The litigation is still in its early stages, but as courts and legislatures consider the law, the debate about ICWA will likely turn to evidence about outcomes for children raised in Native American families as compared with non–Native American families. 316 In one case, the plaintiffs alleged that removing children “from caring, loving [pre-adoptive] homes and forc[ing them] into placements . . . sometimes leads to abuse, psychological harm, or even physical trauma and death.” Carter ex rel. A.D. Complaint, supra note 315, at 7. The plaintiffs further alleged that “[i]n many instances, children are left in abusive or neglectful Indian families where they are subjected to grave physical or psychological harm as a result of ICWA” and that “[s]ubjecting these children and families to ICWA creates delay and uncertainty in the journey to permanent family status, and the prospect and reality of displacement from stable, loving families causes great harm to children.” Id. For existing research on ICWA and child well-being, see Gordon E. Limb et al., An Empirical Examination of the Indian Child Welfare Act and Its Impact on Cultural and Familial Preservation for American Indian Children, 28 Child Abuse & Neglect 1279, 1285 (2004) (finding eighty-three percent of Indian children who were studied and placed with foster or pre-adoptive families were placed according to ICWA guidelines and that seventy-one percent of cases involving foster care placement and eighty-nine percent of cases involving involuntary termination of parental rights included qualified expert witnesses). The Limb study further found that the state took tribal culture into account in eighty-four percent of cases, made efforts to prevent breaking up families in ninety-four percent of cases, and attempted to reunify the family in fifty-six percent of cases. Id. Inadequate recordkeeping and incomplete data, however, make it difficult to assess ICWA. See U.S. Gov’t Accountability Office, GAO-05-290, Indian Child Welfare Act: Existing Information on Implementation Issues Could Be Used to Target Guidance and Assistance to States 1, 2–5 (2005) (discussing the limitations in states’ ICWA recordkeeping).

There are two concerns with this use of empirical evidence. First, it focuses the debate on those values that have been measured. Decisionmakers will consider empirical evidence about child well-being, especially as defined along traditional metrics, such as physical health and high school graduation rates. Empirical evidence could document ICWA’s progress toward other, competing values—such as whether ICWA preserves a child’s ties to the tribe, the tribe’s tie to the child, and tribal sovereignty—but if there is no evidence about these factors, then it is too easy to ignore them in the debate. In short, when there is empirical evidence about some but not all values, it can bias the debate in favor of those values that are measured.

Second, additional empirical evidence about a range of competing values would still beg the question about how to balance these values. Empirical evidence about whether ICWA advances tribal sovereignty would not tell us how much to weight this value—what sovereignty means to the tribe or to society more broadly. Similarly, showing whether Native American children who are kept home or placed with Native American families have different outcomes from children removed and placed with non–Native American families would clarify the stakes of the debate, but the empirical evidence does not tell us how to balance high school graduation rates against tribal ties. Instead, it is still necessary to have an independent debate about how to balance these competing values.

The gravitational pull of empirical evidence toward measurable values and away from an explicit debate about contested and competing values is present in numerous legal contexts. Consider the example of unmarried fathers. A number of laws about parental rights, custody, and visitation favor unmarried mothers over unmarried fathers and thus are ripe for constitutional challenge. 317 Under marital-presumption laws, married fathers are automatically considered legal fathers at birth, whereas unmarried fathers must take affirmative steps to establish their legal rights; further, fifteen states have laws that automatically grant an unmarried mother sole custody at the time of birth. For a discussion of these laws, including a listing of the relevant statutes, see Huntington, Postmarital Family Law, supra note 246, at 202–05. There are active battles over custody and visitation rules in many state legislatures. See Scott & Emery, supra note 159, at 76–83. And courts are considering challenges as well. See, e.g., In re Adoption of J.S., 358 P.3d 1009, 1033 (Utah 2015) (rejecting an unmarried father’s challenge to his son’s adoption on the grounds that Utah could properly require unmarried fathers, but not unmarried mothers, to complete an affidavit to perfect parental rights). Most unmarried children do not live with their fathers. See Carlson et al., supra note 233, at 472 (2008) (“[A]mong the large (and growing) fraction of all children born outside of marriage today, more than three-fifths will be living apart from their biological father by age 5.”). Advocates for fathers seeking to challenge these laws, however, face a problem because there is not a clear empirical basis for finding that child contact with unmarried fathers improves child outcomes. 318 This is not to argue that children are unharmed by father absence. Indeed, there is considerable evidence showing such harm. See Colter Mitchell et al., Father Loss and Child Telomere Length, Pediatrics, Aug. 2017, at 2, 6–7 (reviewing the literature and showing the results of a new study showing the harm, at a cellular level, of father absence from death, incarceration, or divorce or separation). The precise question, however, is whether involvement of never-married nonresidential fathers—who are much more likely than divorced fathers to be low income and have low levels of educational attainment, and who have high levels of criminal justice involvement, see Huntington, Postmarital Family Law, supra note 246, at 186–88—improves the outcomes of their children. On this point, there is not yet compelling evidence. More broadly, there is evidence showing a correlation between improved child outcomes and involvement by nonresidential fathers, see Kari Adamsons & Sara K. Johnson, An Updated and Expanded Meta-Analysis of Nonresident Fathering and Child Well-Being, 27 J. Fam. Psychol. 589, 595–98 (2013) (reporting the findings of a meta-analysis of fifty-two studies of involvement by nonresidential fathers), but these studies do not disaggregate fathers by marital status and thus are likely skewed by the inclusion of divorced fathers, who tend to be much more involved in the lives of their children than never-married fathers, see Huntington, Postmarital Family Law, supra note 246, at 189–90 (contrasting the experience of children with nonresidential fathers with “children of divorced parents, who see their fathers more frequently”). Relatively new evidence is showing that the parenting practices of social fathers, regardless of marital status, are equal to and in some cases superior to the parenting practices of biological fathers, as measured by engagement with the nonbiological child, shared responsibility with the biological mother, cooperation with the biological mother, and trust by the biological mother. See Lawrence M. Berger et al., Parenting Practices of Resident Fathers: The Role of Marital and Biological Ties, 70 J. Marriage & Fam. 625, 629–36 (2008) (comparing social and biological fathers along these variables). Moreover, there is evidence that unmarried mothers and fathers experience high levels of intimate partner violence. 319 See Barber et al., supra note 308, at 176, 192–95 (reporting the results of a study of eighteen- and nineteen-year-old women, which found that “pregnancy relationships included more than twice the amount [of] disrespect as non-pregnancy relationships, more than triple the rate of threats, and four times the rate of physical assault”).

This empirical background makes it seem like the right rule for child well-being is to give more custody rights to unmarried mothers, putting children with the parent who will better care for them and decreasing interaction between parents who may have a history of violence. 320 See, e.g., Leslie Joan Harris, Family Policy After the Fragile Families and Relationship Dynamics Studies, 35 Law & Ineq. 223, 225–35 (2017) (drawing on the evidence about increased intimate partner violence to argue against a default joint custody rule for unmarried parents). Basing a rule on this evidence furthers some (albeit very important) values—child well-being and a decrease in family violence. But it obscures competing values, especially equality between parents and protection of the father–child relationship because neither clearly furthers the desired outcome of child well-being and both implicate family violence. In this way, empirical evidence about some values can eclipse a consideration of other values. And as with debates about ICWA, the evidence does not tell decisionmakers how to balance the competing values.

Despite these concerns, empirical evidence can be relevant to the debate about values, and it is important not to draw a stark divide between the two. 321 See supra note 4 (discussing the conflation of facts and values). Values inform empirical evidence: To the extent society values something—say, child well-being—researchers are more likely to study it. 322 See supra text accompanying notes 25–26, 73, 146–151 (discussing empirical data related to father–child relationships, child custody research, and the neuroscience of juvenile brain development). But empirical evidence can inform values, too. As a foundational matter, many values in family law contain embedded factual assumptions. One reason that society continues to support traditional families is the assumption that, on average, this family form has benefits for child well-being. 323 See Hernandez v. Robles, 855 N.E.2d 1, 7–12 (N.Y. 2006) (noting the “undisputed assumption that marriage is important to the welfare of children”). And indeed there is evidence to support this assumption. 324 See Huntington, Failure to Flourish, supra note 64, at 31–44 (describing the research establishing that children of married couples have better outcomes than other family forms but also noting the degree to which the outcomes are correlative, not causal). But if there were evidence that this was not true, this evidence might slowly change the underlying value.

Empirical evidence informs values in a more subtle way as well. In the marriage equality cases, the courts focused on outcomes for children. 325 See supra text accompanying notes 94, 108, 119–120. But the cases were also a proxy for the larger value-laden debate about whether a family headed by a same-sex couple is socially acceptable. The two are related: To the extent same-sex parents do not harm children, this can help make the family form more acceptable. 326 The process of social acceptance is complex. See Suzanne B. Goldberg, Multidimensional Advocacy as Applied: Marriage Equality and Reproductive Rights, 29 Colum. J. Gender & L. 1, 3–18, 33–40 (2015) (describing the importance of advocating for social change on multiple fronts). Simply providing more information about a family form does not necessarily change views, see Cultural Cognition Project Study, supra note 241, at 16 (finding that “[a] majority of Americans say that their position on gay and lesbian adoption is centered on the welfare of the children” but that “few say they would change their minds if shown convincing contrary evidence”). For a discussion of the relationship between social acceptance of same-sex couples and the marriage equality litigation, see Elizabeth S. Scott & Robert E. Scott, From Contract to Status: Collaboration and the Evolution of Novel Family Relationships, 115 Colum. L. Rev. 293, 313–14 (2015) (arguing that novel families can gain social and legal acceptance if they satisfy social-welfare criteria and that LGBT families and advocates were able to make this showing through what the authors identify as collaborative processes). And to the extent the family form is more acceptable, this may help improve outcomes for the children. 327 Courts have acknowledged this connection, finding that legal recognition would help children by solidifying the legal relationships in the family and giving them important social capital. See, e.g., DeBoer v. Snyder, 973 F. Supp. 2d 757, 771 (E.D. Mich.), rev’d, 772 F.3d 388 (6th Cir. 2014), rev’d sub nom. Obergefell v. Hodges, 135 S. Ct. 2584 (2015).

The feedback loop can also work in the other direction. There is considerable evidence that children raised in plural marriage families have worse outcomes than children raised in other family structures. 328 In a case before the Supreme Court of British Columbia, the court relied on the trial testimony of an expert witness in the field of economics and a review of the social science literature to find that the practice of plural marriage fosters institutional control of women by men through “early and arranged marriages, the payment of brideprice, easy divorce and the devaluing of romantic love”; that women in polygynous relationships suffer from higher rates of domestic violence and death in childbirth, have shorter lifespans than women in monogamous marriages, suffer from marital dissatisfaction and low self-esteem, and have more economic difficulties than women in monogamous marriages because of a lack of, or inequitable division of, resources; that children of polygamous marriages, as compared with their counterparts in monogamous families, have higher infant mortality, elevated risks of abuse and neglect, more emotional, behavioral, and physical problems, and lower educational attainment. See Reference re: Section 293 of the Criminal Code of Canada, 2011 BCSC 1588, paras. 780–86, 789–90 (Can. B.C.). There is some evidence that this stems from the practice of plural marriage itself, 329 See Adrienne D. Davis, Regulating Polygamy: Intimacy, Default Rules, and Bargaining for Equality, 110 Colum. L. Rev. 1955, 1976 (2010) (summarizing the evidence on polygamy and finding that the natural sex ratio does not fit the needs of polygamous communities). In polygamous communities, more females need to be recruited, some men need to be excluded, or younger females need to be married; in polyandrous societies, female infanticide is known to occur. Id. at 1976–77. Professor Adrienne Davis further notes that polygamy tends to lead to statutory rape, incest, and low levels of education, which contribute to problems with individual well-being. Id. at 1977. influencing the view that plural marriage is socially unacceptable. But it is also plausible that the poor outcomes are at least partly due to the state’s hostility toward these families. All states prohibit plural marriage as a matter of civil law, and many states also make it a crime. 330 See id. at 1968. Thus, families based on plural marriage tend to be highly isolated. 331 Utah Office of the Attorney Gen., The Primer: A Guidebook for Law Enforcement and Human Services Agencies Who Offer Assistance to Fundamentalist Mormon Families 8 (2011), http://site.demo.utah.gov/attorneygeneral/wp-content/
uploads/sites/9/2013/08/The_Primer.pdf [http://perma.cc/5DFD-B2NG] (“The fun-damentalists adapted to a secret, underground lifestyle to avoid prosecution and what they perceived as persecution from the ‘world.’”).
Moreover, the illegality of polygamy may well create a selection effect, with social outliers choosing this family form. If polygamy were legal, a wider range of people might choose it, including adults with a broader range of social and economic resources. It is hard to tell, then, whether the different outcomes stem from growing up in a family with plural marriage, growing up in extreme social isolation, growing up with under-resourced parents, or some combination of these factors. Regardless of the particular cause and effect, the point here is that empirical evidence and values are interrelated: The social value of monogamy can influence the empirical analysis of plural marriage, which in turn reinforces the social value of monogamy.

In sum, a fundamental concern with empirical evidence is that it focuses attention on the outcomes of legal rules, emphasizing those values that have been and are susceptible to measurement and distracting attention from competing values. A debate so conceived is thus incomplete, with decisionmakers considering some but not all aspects of the legal question. Further, empirical evidence cannot tell decisionmakers how to balance competing values. It clarifies the stakes and provides information about the extent to which rules advance or compromise different values, but empirical evidence does not tell decisionmakers which values to consider and how to prioritize them. In both ways, then, empirical evidence discourages a forthright debate about competing values.

b. Providing Political Cover. — A second problem is that empirical evidence can provide political cover for the value judgments that are made. 332 See Dan M. Kahan, The Supreme Court, 2010 Term—Foreword: Neutral Principles, Motivated Cognition, and Some Problems for Constitutional Law, 125 Harv. L. Rev. 1, 35 (2011) (“It’s also likely . . . that the Court sometimes consciously resorts to empirical factfinding for strategic reasons. The Justices might well believe that their decision . . . will provoke less conflict . . . if framed in the seemingly neutral idiom of fact as opposed to the morally evocative idiom of constitutional principle.”). Professor Libby Adler has made a related argument, about how the claim of expertise, including by scientists, is itself ideological. See Adler, supra note 10, at 35 (“[E]xpertise has a history of serving a depoliticizing function. It cloaks political ideology in neutral garb for purposes of gaining legitimacy in a discourse in which bare political desire stands counterpoised to legal correctness.”). Prioritizing some family relationships—and some families—over others is an inevitable aspect of state regulation, 333 See Huntington, Failure to Flourish, supra note 64, at 55–80. The state recognizes some but not all economic and affective ties between individuals. See Laura A. Rosenbury, Friends with Benefits?, 106 Mich. L. Rev. 189, 200 (2007) (describing the close relationships, notably including friendship, that family law does not regulate). The state provides support to some family groupings but not others. The Earned Income Tax Credit, for example, supports certain extended family relationships, such as an aunt or uncle caring for a niece or nephew, but not other relationships, such as an adult caring for a minor cousin. See 26 U.S.C. §§ 152(c)–(d) (2012). And the state provides different kinds of support to different families. See Huntington, Failure to Flourish, supra note 64, at 55–80 (describing how state support for families differs by income group, noting that state scrutiny and suspicion tends to accompany support for low-income families, and using tax deductions for middle-income families and food stamps and housing subsidies for low-income families to illustrate this point). but it is also a deeply contested political and normative endeavor. Similarly, choosing how to regulate families necessarily involves a host of trade-offs. Allowing different-sex, but not same-sex, couples to marry reflected a state determination about desirable family forms. And reforming the child welfare system to prioritize child permanency over family integrity, as Congress did in ASFA, 334 See supra text accompanying notes 170–175. reflected a judgment about the lesser importance of families of origin. In these examples, the state relied, at least in part, on empirical evidence. In supporting different-sex marriage, states cited evidence purporting to show the benefits to children of this family structure. 335 See supra text accompanying notes 101–102 (describing states’ arguments in the marriage equality cases). And in its radical reform of the child welfare system in 1997, Congress relied on evidence about children lingering in foster care and the poor outcomes for these children. 336 See supra note 171.

This reliance on evidence may reflect well-intentioned efforts to inform policy, but there is a more cynical explanation as well. Empirical evidence obfuscates the normative and political nature of the judgments. By citing evidence supporting the effects of different-sex marriage on child outcomes, states sidestepped (at least temporarily) the politically fraught battle over families headed by same-sex parents. And by citing evidence about poor outcomes for children in foster care, Congress side­stepped the racially charged debate about breaking up families of color and instead contended that it was promoting child well-being. In short, rather than acknowledge that a rule or policy reflects values or elevates one value over another, decisionmakers can claim they are simply following the evidence.

Professor Suzanne Goldberg has written about a version of this phenomenon in constitutional law, arguing that courts play an inevitable role in choosing among contested normative judgments about a group’s capacities. 337 See Suzanne B. Goldberg, Constitutional Tipping Points: Civil Rights, Social Change, and Fact-Based Adjudication, 106 Colum. L. Rev. 1955, 1965–76, 1980–88 (2006). Goldberg contends that courts are uncomfortable with this role and thus conceal their normative choice with facts. 338 See id. at 1965–76. She explains that courts at times rely on thin facts—which are empirical, uncontested facts, such as women, not men, give birth—courts mostly rely on thick facts. 339 See id. at 1965–70. Thick facts are conclusions that appear to be based on empirical evidence but actually reflect deeply normative judgments, for example, that giving birth creates a connection between mother and child. 340 See id. As Goldberg explains, thick facts “contain both description (group X has a particular characteristic) and evaluation (the characteristic limits the status or capacity of group X). Yet courts regularly ignore the contestable evaluation . . . .” Id. at 1965 (footnote omitted). These thick facts are not based on actual empirical evidence but instead on intuition and assumptions. 341 See id. at 1965–70.

Goldberg argues that empirical evidence can be an antidote to this tendency, contending that, when properly used, 342 Goldberg recognizes the danger of courts cherry-picking studies that support their preferred norm without interrogating the methodology or assumptions of the study. See id. at 1989. this evidence can lead courts to challenge underlying assumptions and intuitions. 343 See id. at 1989–92. The marriage equality trials arguably vindicated this argument. When advocates presented evidence about same-sex parenting, they were able to move beyond the notion that gay and lesbian parents would harm children. 344 See supra text accompanying notes 234–235.

But empirical evidence can conceal normative judgments in a different way. Even if decisionmakers are citing actual studies and doing so rigorously, this reliance on evidence obscures the underlying value judgments. A legislature can claim it is prohibiting plural marriage because of the harm to children. A child welfare agency can state that it is prioritizing child well-being by promoting adoption because of the harm of staying in foster care. And a court can contend it is promoting child development by upholding a law giving greater custody rights to unmarried mothers because of the high rate of intimate partner violence among unmarried couples. These rationales may be backed by evidence, but they also obscure underlying value-based choices. Resorting to empirical evidence makes it seem like the decisionmaker is not taking a political position and instead is simply following the evidence.

c. Replicating Discrimination. — Finally, using empirical evidence can be particularly problematic for nondominant families because it risks replicating historical discrimination. Evidence on unequal outcomes for different demographic groups may be accurate, but the underlying conditions that create the family situations are often the result of systemic discrimination and inequality. Thus, using empirical evidence about poor outcomes to justify legal rules inflicts a second act of discrimination.

Return to the legal challenges to ICWA. Apart from failing to consider multiple, competing values, a focus on evidence about child outcomes in Native American homes as compared with other homes obscures centuries of aggressively hostile state policies toward Native American families. 345 See generally Charles Wilkinson, Blood Struggle: The Rise of Modern Indian Nations (2005) (presenting a comprehensive history of the modern tribal sovereignty movement). Indeed, the explicit, long-standing state policy was to remove children from Native American homes and assimilate the children into the dominant culture. 346 See Lorie Graham, Reparations and the Indian Child Welfare Act, 25 Legal Stud. F. 619, 624–31 (2001) (describing this history). Before ICWA, one study found that up to thirty-five percent of Native American children were removed from their homes, and eighty-five percent of these children were placed in non–Native American families. See Steven Unger, The Destruction of American Indian Families 1–2 (1977). In light of this and numerous other historical efforts to destabilize tribes, it is understandable that Native American families might not produce the same outcomes, such as educational achievement, as non–Native American families, as measured by traditional metrics. 347 See Education, Nat’l Cong. of Am. Indians, http://www.ncai.org/policy-issues/education-health-human-services/education [http://perma.cc/Q29F-BS4U] (last visited Sept. 12, 2017) (“Native students’ academic achievement and educational attainment lags far behind that of their white peers. Over the past 10 years, Native students have been the only population to have not improved in reading or math . . . . Native youth face some of the lowest high school graduation rates . . . .” (footnote omitted)). But uncritically adopting this evidence on outcomes would perpetuate and replicate discrimination against this marginalized population.

The same dynamic would likely play out in any litigation over unmarried fathers. A decision by policymakers to base custody rules on the absence of evidence that unmarried fathers contribute to better outcomes for their children would fail to account for the reasons why these fathers struggle to provide for their children economically and socially. 348 See generally Edin & Nelson, supra note 233 (examining the challenges of fatherhood among low-income, urban men). Unmarried fathers are overwhelmingly low income, with low levels of education and high levels of criminal justice involvement. 349 See McLanahan & Garfinkel, supra note 272, at 147 tbl.8.1 (noting that forty-five percent of single, unmarried fathers do not have a high school diploma, as compared with nineteen percent of married fathers; and that thirty-nine percent have been incarcerated, as compared with seven percent of married fathers). But for at least some of these men—particularly African Americans—the state played an active role in shaping their disadvantage. Historically, the state created and sanctioned pervasive systems of labor exploitation and racial discrimination, from slavery to sharecropping to segregation, all of which limited the economic potential of Black men. 350 See generally, e.g., William Julius Wilson, The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy 10–12, 109–24 (2d ed. 2012) (arguing that a “racial division of labor has been created due to . . . centuries[] of discrimination and prejudice”). Even when these systems were no longer lawful, the state adopted numerous other policies that compromised the economic capacities of Black men, 351 See generally Richard Rothstein, The Color of Law: A Forgotten History of How Our Government Segregated America 153–76 (2017) (discussing government policies that prevented African Americans from climbing the economic ladder). from redlining residential neighborhoods, resulting in racially concentrated poverty, to maintaining a criminal justice system that stops and frisks young men of color at highly disproportionate rates 352 See Floyd v. City of New York, 959 F. Supp. 2d 540, 658–67 (S.D.N.Y. 2013) (describing the disproportionate impact of New York City’s stop-and-frisk policy on men of color). and treats drug violations by Black men much more harshly than similar violations by white men, despite similar rates of drug use and selling. 353 See Alexander, supra note 27, at 98–100. Indeed, given this history, it is all the more noteworthy that unmarried Black fathers are more likely than unmarried white or Latino fathers to play an active role in their children’s lives. See supra text accompanying notes 232–233. A legal rule based on evidence about father involvement and child outcomes would fail to hold the state accountable for the present conditions of many of these fathers and would risk replicating past discrimination. 354 The argument is somewhat different if the state, in support of custody rules, cites evidence about higher rates of intimate partner violence between unmarried couples. See supra text accompanying note 319. In that context, the central concern is that focusing on family violence overshadows other values, notably gender equality and the father–child relationship. See supra text following note 320.

The problem persists in the policy context, in which empirical evidence about child outcomes in different families raises the question of whether the state will support all families. As Professor Mary Anne Case has argued, when parties make a claim on state resources, such as childcare subsidies, it is not unreasonable for the state to ask for good outcomes in return for that support. 355 Mary Anne Case, How High the Apple Pie? A Few Troubling Questions About Where, Why, and How the Burden of Care for Children Should Be Shifted, 76 Chi.-Kent L. Rev. 1753, 1772–76 (2001) (making this argument in the context of public support for childcare in commenting on Linda C. McClain, Care as a Public Value: Linking Responsibility, Resources, and Republicanism, 76 Chi.-Kent L. Rev. 1673 (2001)). This triggers an inquiry into which families are capable of creating these outcomes. 356 See id. Requiring improved outcomes might lead to less, not more, investment in struggling families to the extent there is evidence that disadvantaged families are not well positioned to foster good outcomes for their children. Compounding the problem, increasingly, studies indicate that it is hard to improve parenting in low-functioning families—there is a dearth of evidence demonstrating the effectiveness of parenting programs for such families. 357 See Shonkoff & Phillips, supra note 197, at 226 (“[E]fforts to change the course of [children’s] development by strengthening parenting have met with mixed success. Shifting parental behavior in ways that shift the odds of favorable outcomes for children is often remarkably difficult.”).

In sum, putting families to an empirical test about their ability to produce good outcomes for their children can lead decisionmakers to replicate discrimination. Focusing on empirical evidence about outcomes does not require interrogating the reasons why a group may not be able to show a positive contribution. It can also direct attention away from questions about what families need as a matter of dignity and equal citizenship and instead toward a determination of which families can perform as the state wishes.

III. Guiding the Use of Empirical Evidence in Family Law

If family law’s empirical turn brings both considerable benefits but also substantial reasons for concern, it is essential to develop a framework to guide the use of this evidence. Decisionmakers must know when and why to use it, because not all questions can be resolved through empirical evidence. This Part proposes a framework to guide this process—encouraging the use of empirical evidence when appropriate while also preserving space for a wide-ranging debate about competing values. Further, when empirical evidence is relevant, decisionmakers must be cautious about how they rely on it. This Part provides tools to guide the use of this evidence, calling for more effective gatekeeping across all of family law’s institutions, increased attention to the dangers of perpetuating discrimination along intersecting identities, and a robust role for legal scholars in translating empirical evidence into legal rules and policies.

A. The Proper Role of Empirical Evidence

Nearly all family law questions reflect contested and competing values, but some questions do so more than others. Gaining traction on the issue of when to use empirical evidence requires an appreciation of this aspect of family law—some questions involve a relative consensus about a value or which value to prioritize, and some questions involve a relative lack of consensus about a value or how to prioritize competing values. Empirical evidence is more relevant to the former kinds of questions and less relevant to the latter.

As a threshold matter, then, decisionmakers should determine whether there is consensus about a value or the priority of values. 358 This raises the questions of how to determine whether there is a consensus on the underlying judgment and whether there must be consensus that there is consensus. As with any complex issue, the poles are relatively easy to identify. Compare Baker v. State, 744 A.2d 864, 886 (Vt. 1999) (recognizing—as the first state supreme court to do so—a constitutional claim to relationship recognition for same-sex couples), with Baskin v. Bogan, 766 F.3d 648, 672 (7th Cir. 2014) (affirming “[t]he district court judgments invalidating and enjoining . . . prohibitions of same-sex marriage” in one of the last lower court decisions before Obergefell). With the framework proposed in this section, no more may be needed: The goal of the threshold question is determining whether the question falls at one end or the other of a consensus continuum. If there is a relative consensus on these issues, then empirical evidence can help identify policy choices that further the consensus value or effectuate the decided balance of values. In this context, society and the legal system have largely resolved the relevant value-based judgments, and empirical evidence helpfully elucidates policy choices. It is widely agreed in modern society, for example, that intimate partners should not physically assault each other, a judgment that, as hard as it is to recall, was once much more unsettled. 359 See, e.g., Goodmark, Troubled Marriage, supra note 176, at 9. Today, the issue facing decisionmakers is not whether but instead how to achieve this goal, and the question for family law is one of implementation. Empirical evidence helps decisionmakers compare alternative rules and policies to determine which will best reduce intimate partner violence. To be sure, no values are entirely uncontested, and even furthering seemingly incontrovertible outcomes requires some consideration of competing values, such as individual autonomy and family privacy. 360 See Leigh Goodmark, Should Domestic Violence Be Decriminalized?, 40 Harv. J.L. & Gender 53, 57–69 (2017) (describing the debate about whether intimate partner violence should be decriminalized in favor of alternative approaches and identifying the various values at play in the debate, including the autonomy of victims and the desire for retribution). If there is a residual debate about competing values, such as family privacy versus a reduction in family violence, empirical evidence can play a limited role in elucidating the debate. Understanding, for example, whether more intrusive measures, such as drop-by police visits for families with a history of violence, would reduce family violence might inform the debate about how to trade off these competing values. But when there is a clear goal and a consensus about the underlying value, as well as some sense about how to prioritize competing values, high-quality empirical evidence can be immensely useful in the development of policy.

If the answer to the threshold question is that there is no consensus about a value or there is considerable disagreement about how to balance competing values, then empirical evidence is less relevant, and decisionmakers must be particularly careful in their use of this evidence. The evidence should not dominate the underlying normative question, and nonquantified or nonquantifiable values must be given due credit. In these contexts, it is important to look to sources of authority other than empirical evidence—empathy, a capacity to reason, a sense of justice. 361 See John Rawls, Political Liberalism 28 (2d ed. 1993) (summarizing his theory of reflective equilibrium); Brian Leiter, Constitutional Law, Moral Judgment, and the Supreme Court as Super-Legislature, 66 Hastings L.J. 1601, 1602 (2015) (arguing the Supreme Court acts as a super-legislature in that most constitutional questions before the Court are not dictated by law and thus the Justices regularly, and properly, consult their own moral and political values). This can be a messy and difficult process, and first principles may not be amenable to compromise. Indeed, it is challenging to find agreement around values in a rapidly changing, pluralistic society, but the debate must be resolved on its own terms, as one of values and norms. 362 This raises a host of questions beyond the scope of this Essay, such as whether courts are well equipped to make moral judgments. At the very least it is essential to recognize that the debate implicates these contested and competing values and that empirical evidence will not fully resolve the issue.

The legal challenges to ICWA are a good example of the limited relevance of empirical evidence. The lawsuits raise several value-laden debates: how to balance a child’s well-being as defined by traditional metrics against a child’s well-being as measured by a broader set of interests, including tribal ties; how to weight child well-being versus tribal well-being; and the value to be attributed to tribal sovereignty. 363 See supra notes 313–316 and accompanying text. For all of these questions, empirical evidence can elucidate some of the stakes of the debate, showing, for example, whether children truly have worse outcomes when kept in Native American homes with some level of abuse or neglect. 364 If decisionmakers do consult empirical evidence for this more limited question, they must do so aware of the dangers, particularly the temptation to justify a political choice with data and the potential for empirical evidence to replicate historical discrimination. But empirical evidence fundamentally cannot answer the question of whether and how to balance these competing values. 365 Even when empirical evidence might help answer a factual question that informs values, such as the immutability of sexual orientation, we might not want decisionmakers to consider this evidence. Professor Edward Stein has argued that the existence of a genetic basis of sexual orientation, for example, should be irrelevant to questions about sexual-orientation discrimination. See Edward Stein, Sexual Orientations, Rights, and the Body: Immutability, Essentialism, and Nativism, 78 Soc. Res. 633, 638–54 (2011) [hereinafter Stein, Sexual Orientation]; see also Janet E. Halley, Sexual Orientation and the Politics of Biology: A Critique of the Argument from Immutability, 46 Stan. L. Rev. 503, 506 (1994) (arguing immutability is not a requirement for suspect-class status). Even if a genetic tie could be shown, and he is dubious that it can, Stein is concerned that consulting evidence about the genetic nature of sexual orientation improperly turns an ethical judgment about fairness and inclusion into a scientific question. See Stein, Sexual Orientation, supra, at 634. He believes society should decide questions about the acceptability of sexual orientation based on values, not science. See id. at 654–55.

Most family law questions fall somewhere between these poles, with an incomplete agreement about the underlying values and an incomplete agreement about how to balance competing values. 366 Part of this depends on the level of specificity with which the question is posed. See Cass R. Sunstein, Incompletely Theorized Agreements, 108 Harv. L. Rev. 1733, 1735–36 (1995) (describing “incompletely theorized agreements,” whereby judges “agree on the result and on relatively narrow or low-level explanations for it,” as “an important source of social stability and an important way for diverse people to demonstrate mutual respect, in law especially but also in liberal democracy as a whole”). In these contexts, it is important to make empirical evidence only one part of the debate and to be explicit about the range of values at stake.

Consider predictive analytics in the child welfare system. This tool appears to be a cost-effective means for reducing severe injuries and death, but it also has a high rate of false positives. 367 See supra text accompanying notes 187–196. The legal and policy question is whether this is an acceptable method for triaging child welfare cases. As a threshold matter, the question falls between the two poles because, although there is a consensus that society should protect children from abuse and neglect, there is an ongoing debate about how to balance this goal with other concerns, including family autonomy, family integrity, and disproportionate intervention in families of color. 368 See Huntington, Rights Myopia, supra note 186, at 643–55 (describing these competing values).

Empirical evidence is useful to the inquiry, but it should not dominate. When an administrative agency is determining how to reduce child abuse, it should consider multiple approaches to the problem. Empirical evidence properly guides this process of comparing policies, shedding light on effectiveness and efficiency. Based on this evidence, an agency might choose predictive analytics. But because the tool sweeps in many children not at high risk for death and severe injury, it is essential to ask whether the trade-off is acceptable. Empirical analysis can shed some light on an acceptable rate of false positives by demonstrating the harm to a child of unnecessary involvement with the child welfare system and the harm to a child from unaddressed abuse or neglect. But tolerance for false positives is also a value-based judgment, and, therefore, considerations other than empirical evidence must be addressed. 369 Cf. 4 Blackstone, supra note 49, at *352 (“[B]etter that ten guilty persons escape, than that one innocent suffer.”). For further discussion of values and the use of predictive analytics in other contexts, see Brendan O’Flaherty, Assessment and Prediction in Homelessness Services and Elsewhere, 50 Austl. Econ. Rev. 229, 229–30, 232, 234 (2017) (arguing that the use of predictive analytics to allocate homelessness services can work against notions of fairness, which may be easier to achieve than optimality, and can fail to identify those truly in need of services). Again, empirical evidence can tell us the risk that accompanies false positives—how many families are improperly swept into the child welfare system. And empirical evidence can tell us whether other policy options further different values—for example, whether caseworker judgment, as compared with predictive analytics, leads to more or less intervention for families of color. But empirical evidence cannot completely determine the acceptability of false positives or how to weight these and other concerns. This must be debated as a question of competing values. In short, empirical evidence is useful but not dispositive.

Turning to a more complex question, consider plural marriage. Parties encouraged by Obergefell are beginning to challenge the two-person limit on marriage—sometimes called the next frontier in marriage equality 370 See, e.g., Douglas E. Abrams et al., Contemporary Family Law 127–29 (4th ed. 2012). —both whether it should be criminalized and, more positively, whether states must allow multiple parties to marry one another. 371 See Collier v. Fox, CV 15-83-BLG-SPW-CSO, 2015 WL 12804521, at *1–4 (D. Mont. Dec. 8, 2015) (dismissing a case brought by a man and two women challenging Montana’s criminal law prohibiting bigamy because the parties lacked standing since they had not engaged in plural marriage and instead were seeking a pre-enforcement injunction against the State without any evidence that the State would enforce the provision); Brown v. Buhman, 947 F. Supp. 2d 1170, 1176 (D. Utah 2013) (finding Utah’s criminal prohibition on bigamy unconstitutional as applied to cohabitation), vacated as moot, 822 F.3d 1151 (10th Cir. 2016) (finding the case moot because the State adopted a policy of nonprosecution for cohabitation and thus vacating the decision below), cert. denied, 137 S. Ct. 828 (2017). Taking the affirmative question of a right to marry, the framework proposed above guides the use of empirical evidence. As a threshold matter, there is a prevailing, but not uniform, view that plural marriage is unacceptable. 372 For a discussion of the relationship between plural marriage and social acceptance, see Scott & Scott, supra note 326, at 313–14, 364–69 (arguing that the state will recognize those novel families that satisfy social-welfare criteria, particularly interdependence and long-term mutual care, and that polygyny is so strongly associated with harms, notably sexual abuse and extreme gender inequality, that it will be particularly challenging for these families to gain social recognition). Some commentators argue that the practice is morally repugnant and incompatible with notions of justice and basic democratic values, 373 See Stephen Macedo, Just Married: Same-Sex Couples, Monogamy, and the Future of Marriage 146–48, 164–65, 184–90, 203 (2015) (arguing against legal recognition of plural marriage in part because of its associated harms); see also John Witte, Jr., The Western Case for Monogamy over Polygamy 457–65 (2015) (describing the low social acceptability of polygamy and arguing that “any change in traditional polygamy laws must come from below . . . by gradual democratic adjustments”). and others contend that it is possible to have an ethical vision of polyamory. 374 See Elizabeth F. Emens, Monogamy’s Law: Compulsory Monogamy and Polyamorous Existence, 29 N.Y.U. Rev. L. & Soc. Change 277, 283 (2004) (“Polyamory is a lifestyle embraced by a minority of individuals who . . . articulate an ethical vision that I understand to encompass five main principles: self-knowledge, radical honesty, consent, self-possession, and privileging love and sex over other emotions and activities such as jealousy.”); see also Gregg Strauss, The Positive Right to Marry, 102 Va. L. Rev. 1691, 1762–65 (2016) (arguing that “the law cannot reconcile the inequalities of liberty created by traditional polygamy” but noting that other forms of plural marriage, apart from one man and multiple wives, might be reconcilable with equality). In Brown v. Buhman, the Utah district court assessed the context of plural marriage and was skeptical about any harm associated with it. 947 F. Supp. 2d at 1188. The court noted that the historical perception of the harm was not about actual harm to children but instead the harm from “introducing a practice perceived to be characteristic of non-European people—or non-white races—into white American society.” Id.; see also id. at 1191 (noting the State had not introduced evidence supporting a finding of social harm and thus there was no material issue of fact). That said, the court did not decide the substantive due process challenge by relying on empirical evidence and instead rejected it because of the long tradition in Anglo-American law of prohibiting plural marriage. See id. at 1194–97. Further, there are clearly competing values, with a question about how to balance inclusion and pluralism on the one hand and child well-being and gender equality on the other. 375 See Gregg Strauss, Is Polygamy Inherently Unequal?, 122 Ethics 516, 516–17, 524–44 (2012) (arguing that the traditional structure of polygyny (one spouse marrying multiple partners) necessarily creates inequalities but that other forms of plural marriage—poly-fidelity (each spouse is married to every other spouse) and molecular polygamy (any spouse can marry outside the original family)—do not, theoretically, present such structural inequalities). Compare Brown, 947 F. Supp. 2d at 1188, 1191, 1194–97 (expressing skepticism about any harm associated with plural marriage but also acknowledging the long tradition of prohibiting plural marriage), with Scott & Scott, supra note 326, at 342 (expressing concern about the relationship between plural marriage and harm to children and women). With this question falling between the two poles, empirical evidence may be helpful, but it is important to calibrate its role.

If a legislature were to take up the issue, lawmakers should debate whether, as a matter of values, being married to more than one person at the same time is the kind of decision that consenting adults should be allowed to make. Lawmakers will need to balance values such as autonomy, pluralism, and inclusion against the values of child well-being and gender equality. Although there is considerable evidence on child outcomes and harm to women, 376 See supra notes 328–329 and accompanying text. it is important for decisionmakers to cabin that evidence and not let it completely decide the issue. Moreover, legislators must explicitly embrace the normative nature of the judgment and not simply hide behind the evidence of poor child outcomes.

When courts consider a legal challenge to the two-person restriction on marriage, judges should approach the question somewhat differently from lawmakers. 377 The framework proposed in this section directs decisionmakers to limit the use of empirical evidence, but I recognize the argument that, in constitutional cases, citizens are debating values and typically do so in a language that can reach other citizens who hold different values. See Robert Post & Reva Siegel, Roe Rage: Democratic Constitutionalism and Backlash, 42 Harv. C.R.-C.L. L. Rev. 373, 379–85 (2007). Empirical evidence, with its pretense to be value neutral, can be a powerful language, and thus there is a temptation to bring it into normative debates. For further discussion of the role of empirical evidence in constitutional decisionmaking, see Kahan, supra note 332, at 31–41 (describing the debate about the use of empirical evidence in constitutional decisionmaking); Timothy Zick, Constitutional Empiricism: Quasi-Neutral Principles and Constitutional Truths, 82 N.C. L. Rev. 115, 179–202 (2003) (analyzing courts’ increased reliance on empirical evidence in constitutional cases). As a matter of constitutional doctrine, the relevance of empirical evidence depends on the type of constitutional challenge to a ban on plural marriage. In a substantive due process claim, the question is whether the right to marry more than one person is a fundamental right, which is a normative, values-based question. 378 As the Court stated in Obergefell, substantive due process “requires courts to exercise reasoned judgment in identifying interests of the person so fundamental that the State must accord them its respect.” Obergefell v. Hodges, 135 S. Ct. 2584, 2598 (2015). This will entail an analysis of the values that the Court identified as inherent in the right to marry: that personal choice in marriage reflects individual autonomy, that marriage is the ultimate form of relationship recognition, that marriage is also about children and families, which are separately protected in the Constitution, and that marriage is a foundational social institution. See id. at 2599–602. There may be some room within this analysis for a consideration of empirical evidence, particularly with respect to the consideration of children and the foundational nature of marriage. But this is at heart an inquiry about values and should be treated as such.

If a court does recognize plural marriage as a fundamental right, then the court would proceed to the second prong of the analysis—whether “the infringement is narrowly tailored to serve a compelling state interest.” 379 Reno v. Flores, 507 U.S. 292, 302 (1993). This analysis is about means-end fit, asking whether the state’s interest in limiting marriage to two people is sufficient to overcome the infringement on the now-recognized fundamental right to marry more than one person at a time. In contrast to identifying the existence of a fundamental right, assessing the state’s interest does call for empirical evidence. This inquiry would focus on the connection between plural marriage and harm to children and vulnerable family members, asking whether prohibiting plural marriage reduces the risk of these harms. But even in this prong of the analysis, there are non-empirical questions at stake—for instance, what risk of harm that may result from allowing plural marriage is sufficient to justify the prohibition? This, too, needs to be answered as a values-based question. In short, a substantive due process analysis allows only a small role for empirical evidence. 380 The analysis is somewhat different in an equal protection challenge. In an equal protection challenge to a ban on plural marriage, empirical evidence is relevant to whether the identified category is a protected class. See City of Cleburne v. Cleburne Living Ctr., 473 U.S. 432, 440–42 (1985) (describing the four characteristics of a protected class: whether the class has been the subject of historical discrimination, has a defining characteristic that bears a relationship to an ability to contribute to society, has obvious or immutable characteristics, and is a minority or politically powerless). If the category is a protected class, then empirical evidence is also relevant to the tailoring between the classification and the goal of the legislation. Empirical evidence would be relevant to the state’s articulation of its goal, presumably protecting children and vulnerable family members from an arguably exploitative family form. If the category is not a protected class, then under rational basis review, the legislature does not need to show empirical evidence to support its speculation that the law advances its goal. See FCC v. Beach Commc’ns, Inc., 508 U.S. 307, 315 (1993) (explaining that under rational basis review “legislative choice is not subject to courtroom factfinding and may be based on rational speculation unsupported by evidence or empirical data”).

Both legislators and judges must recognize the interrelationship of the values and empirical evidence surrounding plural marriage. As noted above, one informs the other, and neither can be neatly disentangled. 381 See supra notes 328–331 and accompanying text. The widespread value of monogamy likely influences the empirical evidence, and this evidence likely influences the value. There is no easy way around this entwining, but acknowledging the relationship will help decisionmakers be more wary of both the empirical evidence and the values influenced by the data, not placing too much weight on either and understanding that shifting one may influence the other.

In sum, contested and competing values pervade family law. It is critical to preserve space to debate these issues explicitly, as difficult as it often is. Similarly, it is essential for decisionmakers to guard against the tendency to use empirical evidence to cover political choices about preferred values and norms. The framework proposed in this section accepts the relevance of empirical evidence but guards against its domination. Properly implemented, the framework thus addresses the risk that the empirical turn will allow decisionmakers to avoid and obscure value-based judgments. Finally, the approach recognizes that values can be informed by empirical evidence, and thus it allows for evidence to have its own, perhaps modest, place in the conversation about values.

B. Practical Tools

Once decisionmakers determine that empirical evidence is relevant to the question at hand, this evidence must be used in a manner that addresses the concerns in Part II. This section identifies three tools to achieve these ends: (1) improved gatekeeping across family law’s institutions, (2) greater attention to the ways empirical evidence reflects but also refracts intersecting identities, and (3) a robust role for legal scholars in the translation of empirical evidence into legal rules and policies. The goal of this section is to offer initial thoughts on these tools, intended to spark a larger debate about how to use empirical evidence in a manner that takes advantage of its benefits while guarding against its dangers.

1. Gatekeeping. — Legal actors need to be both more sophisticated and more skeptical in their consumption and use of empirical evidence. A core issue is developing more effective gatekeeping mechanisms. As noted above, federal courts generally follow basic evidentiary rules about the admission of expert testimony, 382 For further guidance to courts on how to use empirical evidence, see John Monahan & Laurens Walker, A Judges’ Guide to Using Social Science, 43 Ct. Rev. 156, 162 (2007) (advising courts how to obtain, evaluate, and use empirical evidence to establish legislative facts, adjudicative facts, and social frameworks). but, too often, family courts do not and thus admit unreliable evidence. 383 See supra text accompanying notes 166–169. For further discussion of how courts can better evaluate social science research, see John Monahan & Laurens Walker, Social Authority: Obtaining, Evaluating, and Establishing Social Science in Law, 134 U. Pa. L. Rev. 477, 498–508 (1986) (proposing that courts evaluate scientific research analogously to how they evaluate precedent and prioritize research that “(1) has survived the critical review of the scientific community; (2) has employed valid research methods; (3) is generalizable to the case at issue; and (4) is supported by a body of other research”); see also Davis, There Is a Book Out, supra note 11, at 1594–602 (describing how courts can be more rigorous in their use of social science, expert testimony, and extra-record literature to determine legislative facts). Although generally desirable, it may be unrealistic to reinvigorate Daubert in family court in light of overwhelming caseloads, minimal support, and generally loose evidentiary standards. 384 See supra text accompanying notes 166–169. To address this structural problem, one approach is to create a review process separated from the individual courtroom, thus removing the burden from family court judges. A centralized body of experts could vet the empirical evidence related to the issues regularly addressed in family court: adolescent development, domestic violence, coparenting, and the like. This is more than a call for judicial training, which would help prepare judges for the empirical world. It is a centralized method for reviewing the voluminous research and determining which pieces meet the Daubert standard, thus improving the quality of the research family court judges use. This approach would present its own challenges, such as ensuring the neutrality and sophistication of the members of the review board. But if this panel operated as an additional resource, supplementing but not automatically supplanting individual review, it would improve the quality of empirical evidence used in family courts.

Even a strict application of the Daubert standard, however, does not account for the ways in which empirical evidence fails to reflect historical, government-supported discrimination. Courts will need to address this separately. In the context of plural marriage, for example, when a court is considering evidence of any harms associated with plural marriage, the court must also ask whether these harms inevitably flow from the family structure or whether they are at least partly the product of historical discrimination. 385 See supra text accompanying notes 328–331. To the extent there is evidence that children of plural marriage have worse outcomes than children in other family structures, courts must question—assuming the underlying evidence does not—whether these outcomes are due, at least in part, to state hostility to plural marriage.

Beyond courts, legislative bodies should develop better gatekeeping mechanisms, as they do not currently have a standard practice for ensuring the reliability and relevance of empirical evidence. This is admittedly an enormous topic, and this section cannot identify all of the current failures and possible solutions, but the basic idea of the mechanism is to encourage legislatures to consider multiple sides of an issue. When Congress enacted a series of reforms to the child support system in the 1970s, it had abundant evidence that child support orders were inadequately enforced, 386 See Joanna L. Grossman & Lawrence M. Friedman, Inside the Castle: Law and the Family in 20th Century America 224–31 (2011) (describing the history of federal child support laws and their inadequacies and shortfalls). but it did not have evidence that a robust child support system would ameliorate child poverty. 387 This evidence would have been hard to find because low-income, nonresidential parents generally do not have an income to support their children. See Office of Child Support Enf’t, U.S. Dep’t of Health & Human Servs., Family-Centered Innovations Improve Child Support Outcomes 1 (2011) [hereinafter Family-Centered Innovations], http://www.acf.hhs.gov/sites/default/files/ocse/family_centered_innovations.pdf [http://
perma.cc/SUL6-UTYA] (“One quarter of all custodial and noncustodial parents are poor and nearly two-thirds of custodial families in the child support program have incomes below 200 percent of the poverty threshold.”). Fathers in these families are often unable to pay even small amounts of money to support their children. See Laurie S. Kohn, Engaging Men as Fathers: The Courts, the Law, and Father-Absence in Low-Income Families, 35 Cardozo L. Rev. 511, 531–44 (2013) (“Research suggests that the majority of low-income fathers’ failure to meet their obligations is not because of their unwillingness to support their children, but because they do not earn enough to satisfy their obligations.”).
Instead of considering alternative means for improving child poverty, such as a guaranteed child-allowance provision, common in many European countries for more than half a century, 388 See Duncan Lindsey, The Welfare of Children 319–20 (2d ed. 2004) (describing universal children’s allowance programs in Europe and elsewhere). Congress reinforced the cultural norm of economically independent families. 389 At common law, parents were required to provide economic support to their children. See Blackstone, supra note 49, at *435 (seeing this duty as “a principle of natural law,” arising from “nature herself” and the act of “bringing [a child] into the world” because it “would be in the highest manner injurious to their issue, if they only gave the children life, that they might afterwards see them perish”); Kent, supra note 49, at 161 (“The father is bound to support his minor children, if he be of ability, even though they have property of their own; but this obligation in such a case does not extend to the mother.”). If Congress had had additional evidence about various alternatives, at least it would have had to explain why it was not choosing the child allowance—a more effective, 390 If Congress had compared alternative means, it would have had to address the research showing that guaranteed child incomes is a more effective means of supporting low-income families than child support. See Lindsey, supra note 388, at 313–38. albeit less politically popular, policy.

With judicial review of their actions and, for some agencies, internal norms of evidence-gathering, administrative agencies have some check on the quality of their decisionmaking, but these agencies still have tremendous leeway in their use of empirical evidence. It is essential to develop a system to ensure the empirical evidence agencies use is of high quality and addresses the concerns about data reliability, the proper use of the evidence, and so on. Administrative agencies’ capacity to address these issues is exacerbated at the local government level because budget constraints and the limits of institutional capacity will mean that agencies do not generally have access to sophisticated tools for evaluating empirical evidence. One solution is for agencies to partner with universities. These research institutions can play a useful role in helping agencies both identify and critique relevant empirical evidence. The University of Florida, for example, sponsored a two-day working conference on early-childhood development, bringing together academics, childcare providers, and government officials. 391 Early Childhood National Summit, Univ. of Fla. Anita Zucker Ctr. for Excellence in Early Childhood Studies, http://ceecs.education.ufl.edu/national-summit/ [http://perma.cc/Z8AA-QE2D] (last visited Sept. 12, 2017). The fruitful exchange was an opportunity to translate the existing knowledge about early-childhood development into actionable steps at the national, state, and local level.

2. Attention to Intersecting Identities. — As Professor Robin Lenhardt has argued, often the law appears to be race neutral but, in operation, perpetuates racial inequality. 392 See R.A. Lenhardt, The Color of Kinship, 102 Iowa L. Rev. 2071, 2074–77 (2017). She contends that when the state underinvests in schools serving mostly students of color, fails to combat racial segregation in housing, and does not protect the safety of the water supply, families, and especially low-income families of color, suffer the consequences. 393 See id. at 2088. Indeed, there are numerous ways family law obscures the importance of intersecting identities, including race, class, and gender. Courts often cast constitutional claims by unmarried fathers as a contest between married and unmarried fathers, not between unmarried mothers and unmarried fathers, thus avoiding a debate about sex discrimination and gender roles in the family. 394 See Serena Mayeri, Foundling Fathers: (Non-)Marriage and Parental Rights in the Age of Equality, 125 Yale L.J. 2292, 2292, 2334–42 (2016) (arguing that in foundational cases establishing the limited rights of unmarried fathers, the Supreme Court framed the conflict as a clash between husbands and unmarried fathers, not between mothers and fathers, and thus avoided questions of sex neutrality in custody law). Debates about unmarried fathers risk glossing over the race and class dimensions of marital status, ignoring the reality that most unmarried fathers are low income and disproportionately of color. 395 See McLanahan & Garfinkel, supra note 272, at 147 tbl.8.1 (reporting characteristics of participants in the Fragile Families and Child Wellbeing Study). At the time of the focal child’s birth, 29.5% of the married fathers had earned at least a college degree, as compared with 3.8% of unmarried nonresidential fathers; 12.5% of the unmarried fathers were white, and 58.4% were Black. Id. Many scholars do recognize this aspect of the debate. See, e.g., June Carbone & Naomi Cahn, Nonmarriage, 76 Md. L. Rev. 55, 79, 106, 118 (2016) [hereinafter Carbone & Cahn, Nonmarriage] (stating that those who do not marry differ from those who do marry in ways that affect children and that nonmarried parents tend to have fewer resources than married parents). Thus obscured, the debate appears to be between men who willingly assume responsibilities of fatherhood by marrying the mother of their children and men who shirk these responsibilities. This framing, however, ignores some of the reasons for low marriage rates in poor communities, notably government-sponsored discrimination that has compromised the economic potential of some men, and especially Black men, making it harder for these men to provide for their children. 396 See supra text accompanying notes 350–353. It also obscures the role of the state in using marriage as a tool to subjugate nondominant populations—policies that have made marriage far less appealing to such groups. 397 See generally Katherine Franke, Wedlocked: The Perils of Marriage Equality (2015) (highlighting the potential for undesirable state-enforced legal rules that can accompany civil marriage).

Following Lenhardt’s invitation to uncover the law’s role in perpetuating inequality, it is essential for decisionmakers to see how empirical evidence can direct attention to inequality but also obscure and compound it. Beginning with the potential for empirical evidence to help uncover inequality, it can show how families of color are disproportionately represented in the child welfare system, 398 Children’s Bureau, U.S. Dep’t of Health & Human Servs., The AFCARS Report: Preliminary FY 2015 Estimates as of June 2016, at 2 (2016), http://www.acf.hhs.gov/sites/
default/files/cb/afcarsreport23.pdf [http://perma.cc/4N5F-SUFR].
arguably from a built-in bias toward seeing pathology in such families. 399 See Emily Putnam-Hornstein et al., Race and Ethnic Disparities: A Population-Based Examination of Risk Factors for Involvement with Child Protective Services, 37 Child Abuse & Neglect 33, 34, 42–44 (2013) (discussing the literature making this claim as well as the evidence in a study of a California birth cohort); see also supra note 203 (citing sources discussing possible explanations for the correlations between race and child maltreatment). And in the child support system, empirical evidence can show that the majority of men who do not pay child support lack the economic means to do so. 400 Family-Centered Innovations, supra note 387, at 2. By documenting these differences, decisionmakers can use this empirical evidence to better understand the ongoing salience of race, class, and gender and begin to pay attention to the problem. 401 See, e.g., Office of Child Support Enf’t, U.S. Dep’t of Health & Human Servs., Annual Report to Congress FY 2015, at iii, 9–11 (2016), http://www.acf.hhs.gov/sites/
default/files/programs/css/fy2015_part_01.pdf [http://perma.cc/KP25-5QP9] (describing the reorientation of the child support system in response to the challenges facing low-income men).
As noted above, empirical evidence can also be an antidote to discrimination and demeaning stereotypes, such as the inaccurate image of the uninvolved Black father. 402 See supra text accompanying note 233.

On the other side of the ledger, however, empirical evidence can obscure inequality. When decisionmakers take this evidence at face value, without interrogating the underlying causes of differential outcomes along identity lines, primarily race and class, it masks the source of the inequality. 403 See supra section II.B.1. Even more troubling, empirical evidence can compound the salience of intersecting identities. In predictive analytics, a repeated criticism is the algorithm’s reliance on a parent’s experience in the child welfare system as a child. 404 See supra text accompanying note 190. Even if this is a reliable predictor, including this data point risks reinscribing discrimination against low-income families of color. For Native American families, in particular, this risk factor would expose exceptionally high numbers of parents to greater scrutiny. Native American children face the highest cumulative risk of foster care placement, even with the protections of ICWA in place. 405 See Christopher Wildeman & Natalia Emanuel, Cumulative Risks of Foster Care Placement by Age 18 for U.S. Children, 2000-2011, 9 PLOS ONE 1, 5 (2014) (finding a 5.91% cumulative risk of placement in foster care for all children in the United States before age eighteen, but with sharp differentials by race and ethnicity). Asian children face a 2.14% risk, white children a 4.86% risk, Latino children a 5.35% risk, Black children a 10.99% risk, and Native American children a 15.44% risk of placement in foster care. Id. To the extent the prior foster care placements reflect some form of bias and not only higher rates of maltreatment, including the foster care placement history of a Native American parent as a factor in the algorithm recreates this discrimination.

In sum, when using empirical evidence to make decisions, it is essential for decisionmakers—legislators, courts, and administrative officials—to be attuned to the potential for empirical evidence not only to uncover inequality but also to compound it. Decisionmakers must ask whether the underlying evidence is a reflection of historical discrimination. And then they must ask whether the use of empirical evidence will perpetuate that discrimination.

3. A Translation Role for Legal Scholars. — Finally, there is a pivotal role for legal scholars in guiding the use of empirical evidence. 406 Family law scholars can also produce empirical research, as some are already doing. See, e.g., Douglas W. Allen & Margaret Brinig, Do Joint Parenting Laws Make Any Difference?, 8 J. Empirical Legal Stud. 304, 304 (2011) (studying the effect of a joint cus­tody presumption on custody determinations); Margaret F. Brinig & Douglas W. Allen, “These Boots Are Made for Walking”: Why Most Divorce Filers Are Women, 2 Am. L. & Econ. Rev. 126, 126–30 (2000) (studying which parent files for divorce most often and examining the relationship between filing and custody disposition); Meier & Dickson, supra note 161, at 331–34 (reporting the results of a study finding a negative correlation between a mother alleging abuse and the award of child custody). The legal system always needs better research and more reliable data, but more fundamentally it needs sophisticated translation of social and hard science into legal rules and policies. Researchers outside of the legal academy are not well positioned to turn their findings into legal rules and policies, and indeed trying to do so risks compromising the underlying research. 407 See Emery et al., “Bending” Evidence, supra note 251, at 135. But legal scholars are well suited to this work. 408 See Huntington, Early Childhood Development, supra note 301, at 792–801, 806–10 (arguing legal scholars should play a pivotal role in translating the empirical evidence on early childhood into legal rules and policies). Policy analysis is one of the archetypes of legal scholarship, with scholars identifying problems, objectively comparing evidence about alternatives, and recom­mending solutions. 409 Martha Minow, Archetypal Legal Scholarship: A Field Guide, 63 J. Legal Educ. 65, 66 (2013); see also id. at 67 (describing another model of legal scholarship as “[t]est[ing] a proposition about society or the economy or about human beings that is used by lawyers or assumed in legal sources” and then either conducting or analyzing empirical work about this assumption before “digest[ing] the findings for legal audiences”). Moreover, this translation work is more than simply reconciling divergent studies. It requires the integration of other bases for legal regulation, including a consideration of values, morality, and so on. 410 Cf. Sandler et al., supra note 251, at 151 (describing the role of advocacy, as compared with research, and noting “[s]uccessful advocacy involves multiple factors beyond the translation of empirical research findings into action, including attending to the morality, ethics, related laws and legal procedures, civil rights, social values and mores, feasibility, and economic costs of action proposals”). Legal scholars can help put empirical work in context, noting instances in which the research advances a particular goal but the legal system may want to do something other than simply advance that goal.

Family law scholars are on the right path, regularly drawing on empirical work from a range of disciplines, including sociology, 411 See, e.g., June Carbone & Naomi Cahn, The Triple System of Family Law, 2013 Mich. St. L. Rev. 1185, 1218–19 (relying on the Fragile Families and Child Wellbeing Study). economics, 412 See, e.g., Margaret F. Brinig, The Effect of Transactions Costs on the Market for Babies, 18 Seton Hall Legis. J. 553, 553–60 (1994) (relying on economic principles and research to analyze adoption law). psychology, 413 See, e.g., Solangel Maldonado, Cultivating Forgiveness: Reducing Hostility and Conflict After Divorce, 43 Wake Forest L. Rev. 441, 442–48 (2008) (relying on psychological research to argue for a model of forgiveness in family law). and neuroscience. 414 See, e.g., Maroney, supra note 298, at 145–60 (exploring the use of neuroscientific evidence in the juvenile justice system). Family law scholars integrate the wealth of information available about families, from in-depth ethnographic studies about unmarried parents 415 See, e.g., Linda C. McClain, The Other Marriage Equality Problem, 93 B.U. L. Rev. 921, 937 (2013) (relying on the ethnographic work of Kathryn Edin, among others). to detailed statistical portraits of marriage rates. 416 See, e.g., June Carbone & Naomi Cahn, Marriage Markets: How Inequality Is Remaking the American Family 11–20 (2014). Often, the empirical evidence does not answer questions as much as fuel additional debate, with schol­ars using empirical evidence to support different arguments. 417 Compare Carbone & Cahn, Nonmarriage, supra note 395, at 111 (citing empirical evidence about the ability of unmarried parents to negotiate financial support and time with the children without the legal system, and thus arguing that the legal system should not change the current treatment of unmarried parents), with Huntington, Postmarital Family Law, supra note 246, at 227–29 (drawing on research about nonmarital families to argue for reforms to family law, particularly that unmarried fathers should have similar rights to unmarried mothers). For other scholarship using empirical evidence to make arguments about whether and how to regulate unmarried couples, see Cynthia Grant Bowman, Unmarried Couples, Law, and Public Policy 169–70 (2010) (arguing the differences between married and cohabitating couples—such as different levels of stability, domestic violence, and economic interdependence—combined with the fact that many cohabitating couples are economically disadvantaged and have children, should encourage governmental recognition of cohabitational unions); Marsha Garrison, Is Consent Necessary? An Evaluation of the Emerging Law of Cohabitant Obligation, 52 UCLA L. Rev. 815, 839–41, 861–63 (2005) (drawing on empirical evidence indicating that married and cohabitating couples view their relationships differently and display different behavior to argue that children benefit from having married parents and that the law should encourage marriage by treating married couples differently from cohabitating couples). But this scholarship does advance the conversation. The point here is to encourage and guide this continued engagement with empirical work.

As family law scholars continue this translation project, however, it is important to be attentive to the concerns about empirical evidence iden­tified in this Essay. A starting point is recognizing our own ideological commitments, which can seep into the selection, consideration, and translation of empirical evidence. Beyond this basic point, family law scholars should consider the guidelines set forth in this Part—using empirical evidence to advance specified outcomes while remaining mindful of the limited role for empirical evidence in debates about con­tested and competing values. And scholars should pay particular atten­tion to the role of empirical evidence and intersecting identities. This involves the consideration of factors such as the socially constructed hierarchies of families along various identities; the role of historical discrimination in creating current family conditions, particularly the stratification along race and class lines; the role of race in constructing all families, including white families; the acceptance of a range of family forms without using the nuclear family as the default norm; and so much more. 418 Lenhardt, supra note 392, at 2101–03 (proposing that these and other factors should be considered in understanding the role of race in family law).

Conclusion

Family law’s reliance on empirical evidence is not going away. Nor should it. Using empirical evidence to inform difficult policy choices and evaluate different legal rules is and should be an integral component of good governance. But the empirical turn also presents considerable cause for concern. Beyond the oft-cited issues of data reliability and the ability of legal actors to use empirical evidence appropriately, there are fundamental concerns that empirical evidence focuses attention on outcomes rather than on competing values, cloaks normative judgments, and risks replicating historical, state-sponsored discrimination.

To address these issues while also capturing the benefits of empirical evidence, this Essay has proposed a framework to guide the use of empirical evidence. This evidence has utility, but decisionmakers in family law must preserve space for debating values, avoid using evidence to claim neutrality, and be wary about the potential for reinscribing dis­crimination. When empirical evidence is relevant, more effective gatekeeping mechanisms across the institutions of family law are critical, as is closer attention to the ways empirical evidence can compound the legal salience of intersecting identities. Finally, family law scholars should continue to play an active role in the translation of empirical evidence into legal rules and policies.

By identifying the trend toward increased use of empirical analysis, and discussing its potential benefits and drawbacks, it is possible to chart a better course for the use of empirical evidence in family law. Using empirical evidence when appropriate, but cabining it when necessary, creates a more effective but also a more balanced and inclusive family law, benefiting individuals, families, and society as a whole.