Federal Agencies Archives - Columbia Law Review

Introduction Partisan balance requirements (PBRs) are a common—and controversial—feature of federal agency design. These provisions generally mandate that only a bare majority of commissioners of a multimember agency can hail from the same political party. More than half of all multimember agencies within the federal government are now subject to these party-balance rules, ranging from little-known entities like the African Development Foundation to high-profile agencies such as the Federal Communications Commission (FCC) and the Securities and Exchange Commission (SEC).[footnote]See Kirti Datla & Richard L. Revesz, Deconstructing Independent Agencies (and Executive Agencies), 98 Cornell L. Rev. 769, 797 tbl.4 (2013).[/footnote] Yet, despite the prevalence of PBRs, scholars of administrative law lack even a basic understanding of how these mandates affect the ideological composition and policy outputs of multimember agencies. This Article sheds light on PBRs and their consequences. A number of commentators have doubted whether PBRs materially affect the ideological composition of multimember agencies. They point out that Presidents can comply with the letter of the law simply by selecting an appointee who is nominally registered as a member of the opposite party or as an independent, even if that individual is otherwise the President’s ideological ally. This view was perhaps most clearly stated in a 1976 report commissioned by the Senate Commerce Committee, which averred that neither Republican nor Democratic Presidents had chosen “‘bona fide, honest-to-God’ members of the other party” to fill cross-party seats.[footnote]Staff of the S. Comm. on Commerce, 94th Cong., Appointments to the Regulatory Agencies: The Federal Communications Commission and the Federal Trade Commission (1949–1974), at 386 (Comm. Print 1976).[/footnote] Until scholars can determine empirically whether PBRs meaningfully constrain a President’s ability to appoint ideologically sympathetic commissioners to multimember agencies, it will be difficult to say much about the positive or normative implications of these statutory requirements. And yet, in the last half century, not a single academic article has systematically examined whether and how PBRs affect the ideological composition of the agencies to which they apply. One study has sought to measure the effect of partisan balance on voting patterns at the FCC specifically,[footnote]Daniel E. Ho, Congressional Agency Control: The Impact of Statutory Partisan Requirements on Regulation 1–3 (Feb. 12, 2007) (unpublished manuscript), http://perma.cc/F8MW-6WN8 (“[T]he effect of commissioner ideology on voting is profound. Commissioner partisan affiliation exhibits robust and large predictive power over votes, even holding constant the party of the appointing president.”); see also Wenmouth Williams, Jr., Impact of Commissioner Background on FCC Decisions: 1962–1975, 20 J. Broadcasting 239, 241 (1976) (providing an earlier look at FCC voting patterns based on political affiliation).[/footnote] and a handful of scholars have offered anecdotal observations that speak to this question.[footnote]See infra notes 53–58 and accompanying text.[/footnote] But for the last systematic, cross-agency empirical analysis of PBRs’ effect on agencies’ ideological composition, one has to look all the way back to a 1964 study by political scientists Stuart Nagel and Martin Lubin.[footnote]Stuart Nagel & Martin Lubin, Regulatory Commissioners and Party Politics, 17 Admin. L. Rev. 39 (1964) (studying the influence of party affiliation on the decisionmaking of seven regulatory agencies).[/footnote] A lot has changed since 1964. Back then, both major political parties were ideological “big tents”: The Democratic Party included a staunchly conservative southern base, which found itself to the right of “Rockefeller Republicans” from the North on several major issues.[footnote]See Sean Farhang & Ira Katznelson, The Southern Imposition: Congress and Labor in the New Deal and Fair Deal, 19 Stud. Am. Pol. Dev. 1, 1–2 (2005) (describing how southern congressional Democrats, who had previously aligned with northern Democrats, came to align with Republicans on labor issues in the 1940s); Eric Schickler, Kathryn Pearson & Brian D. Feinstein, Congressional Parties and Civil Rights Politics from 1933 to 1972, 72 J. Pol. 672, 672–73 (2010) (describing how northern Democrats took the lead on advancing civil rights legislation beginning in the 1940s, with Republicans in the middle—offering tepid support and occasional behind-the-scenes opposition—and with southern Democrats firmly opposed).[/footnote] By the turn of the twenty-first century, however, partisan identity had become a much more reliable predictor of ideology—a phenomenon known as “partisan sort.”[footnote]See Matthew Levendusky, The Partisan Sort: How Liberals Became Democrats and Conservatives Became Republicans 2–11 (2009) (defining the term “partisan sort”).[/footnote] Meanwhile, developments in campaign finance law, information technology, and empirical social science have made individual ideology much easier to observe. The Federal Election Campaign Act of 1971 for the first time required candidates and political parties to disclose the identities of their contributors as well as the amounts donated.[footnote]Pub. L. No. 92-225, §§ 304, 308(a)(4), 86 Stat. 3, 14–15, 17 (1972).[/footnote] Those records have been open to the public in a computer-searchable form since 1980,[footnote]Telephone Interview with Senior Pub. Affairs Specialist, Pub. Records Branch, FEC (May 30, 2017) (call notes on file with the Columbia Law Review ); see also 49 Fed. Reg. 22,335, 22,336 (May 29, 1984) (setting fees for Freedom of Information Act (FOIA) requests concerning “computer search[es] of an entire individual contributor file for contributions made by a particular individual or individuals”).[/footnote] and they have been posted on the Internet in a searchable format since 1998.[footnote]Press Release, FEC, FEC Launches New Web Info (July 21, 1998), http://classic.fec.gov/press/press1998/websrch.htm [http://perma.cc/RA8G-EAAW]. The FEC first made “comprehensive computer files containing campaign finance information” available online in 1996 and introduced a search function two years later. Id.[/footnote] While an individual’s campaign contributions are not a perfect proxy for ideology, recent political science research suggests that donations closely track ideological preferences.[footnote]See Adam Bonica, Mapping the Ideological Marketplace, 58 Am. J. Pol. Sci. 367 (2014) [hereinafter Bonica, Ideological Marketplace].[/footnote] By leveraging information on campaign contributions made by individuals across several election cycles, researchers can distinguish among liberals, moderates, and conservatives within the same political party—both those who hold government office and those who don’t—in a much more reliable way than past generations of scholars possibly could. All this means that a comprehensive, cross-agency empirical analysis of the effect of PBRs on the ideological composition of multimember boards and commissions is long overdue. This Article fills that void. Before delving into the analysis, this Article considers two explanations as to why limitations on the partisan composition of agencies might affect the ideological composition of multimember agencies: a supply-side theory and a demand-side theory. (It also considers alternative theories that focus on search costs and identity signaling.) The supply-side theory posits that there is only a small pool of liberal Republicans and conservative Democrats who are potentially qualified to serve as commissioners, and so Presidents looking for competent cross-party appointees whose ideological preferences track their own will often find themselves out of luck. Thus, Presidents who are required by PBRs to make cross-party appointments to multimember agencies will be compelled by supply-side constraints to choose a genuine cross-partisan. Whereas the supply-side theory focuses on the first stage of the appointment process—the selection of a nominee—the demand-side theory focuses on the end stage: confirmation. The demand-side theory emphasizes that nominees to federal agencies ultimately must be “sold” to the Senate, since a majority vote of the Senate is necessary for confirmation. On this view, consumers in the confirmation market (senators) demand cross-party appointees who are bona fide members of that party and thus force the President to choose commissioners from the other side of the ideological spectrum. The supply-side and demand-side theories both suggest that PBRs will affect the ideological composition of multimember agencies, but they generate different predictions as to when and why. If the supply-side theory is correct, then we would expect to see the effect of PBRs growing over time in tandem with partisan sort. If the demand-side theory is correct, then we would expect to see the effect of PBRs most strongly in periods of divided government, when the opposition party in the Senate has leverage to force the President’s hand. To assess the effect of PBRs and test these and other theories, we gather data on the identities of 578 appointees to twenty-three agencies over the course of six presidencies and thirty-six years (1979 through 2014). We then use data on those individuals’ campaign contributions to estimate their ideological preferences. We compare the (estimated) ideological preferences of appointees to those of their appointing Presidents, and we compare co-party appointees (i.e., a Democrat appointed by a Democratic President or a Republican appointed by a Republican President) to cross-party appointees (i.e., a Republican appointed by a Democratic President or vice versa). We show how the effects of PBRs on the ideological composition of federal agencies vary over time and with changes in control of Congress. Our main findings are as follows. First , we find that partisan balance requirements do “bite.” When Presidents are required by law to make cross-party appointments, they generally choose “bona fide, honest-to-God” members of the opposite party—or, at least, individuals whose observed ideologies are significantly different from the President’s own. Looking at the thirty-six-year period overall, Republican Presidents appoint Democrats who are more liberal than they are and more liberal than their Republican co-party appointees are. Likewise, Democratic Presidents appoint Republicans who are more conservative than they and their Democratic co-party appointees are. Second , the effect of PBRs on the ideological composition of multimember agencies is much more pronounced today than it was at the beginning of our study period. This finding is consistent with the supply-side theory and our expectations based on partisan sort. President Jimmy Carter often named relatively moderate Republicans as cross-party appointees; Presidents Ronald Reagan and George H.W. Bush often appointed centrist or conservative Democrats. By the time President Bill Clinton took office, however, this phenomenon had subsided. Republicans appointed by Presidents Clinton and Obama tended to be quite conservative, and Democrats appointed by President George W. Bush tended to be quite liberal. Third , we find no apparent relationship between the presence or absence of divided government and the ideological distance between the President and cross-party appointees. This finding is at odds with the demand-side theory, which would lead us to expect the effect to be most pronounced when the White House and the Senate are controlled by different parties. While we do not dismiss the demand-side explanation entirely, our analysis suggests that supply-side factors do more to mediate the effect of PBRs on the ideological composition of multimember agencies. This Article concludes by considering the implications of our empirical analysis for the scholarly understanding of PBRs. It identifies three main accounts of partisan balance requirements in the administrative law literature. One account, which pulls insights from political science, posits that PBRs reduce the costs borne by Congress in monitoring agency actions (the monitoring account ). On this view, PBRs ensure that agencies include members whose policy preferences diverge from the sitting President’s, and these minority members will alert lawmakers if members from the President’s party stray from legislators’ preferences.[footnote]See infra section III.A.[/footnote] A second account, drawing on social psychology literature, suggests that PBRs can improve agencies’ deliberative processes by offsetting tendencies toward group polarization (the deliberation account ).[footnote]See Cass R. Sunstein, Deliberative Trouble? Why Groups Go to Extremes, 110 Yale L.J. 71, 74 (2000) (defining group polarization as involving “members of a deliberating group predictably mov[ing] toward a more extreme point in the direction indicated by the members’ predeliberation tendencies”).[/footnote] According to this view, PBRs ensure that multimember agencies include a diverse set of perspectives, thereby counterbalancing tendencies toward group polarization. Apart from any effect on the cost of congressional monitoring, this account suggests that PBRs can improve federal administration by raising the quality of agency decisionmaking.[footnote]See infra section III.B.[/footnote] A third account, rooted in constitutional law scholarship on the “unitary executive,” posits that PBRs obstruct the President in the execution of her constitutional responsibilities (the obstruction account ).[footnote]See Ronald J. Krotoszynski, Jr., Johnjerica Hodge & Wesley W. Wintermyer, Partisan Balance Requirements in the Age of New Formalism, 90 Notre Dame L. Rev. 941, 991–99 (2015) (“Although statutory partisan balance requirements may appear innocuous or even beneficial at first glance, these provisions have far-reaching ramifications on the President’s ability to properly influence and oversee federal agencies.”).[/footnote] According to this view, PBRs “quite literally force Presidents to rely on political enemies to carry out their executive duties,”[footnote]Id. at 991.[/footnote] and thus potentially prevent Presidents from implementing their policy agendas.[footnote]Id. at 995.[/footnote] These three accounts are not mutually exclusive: In theory, PBRs could reduce the cost to Congress of monitoring multimember agencies, counterbalance tendencies toward group polarization within agencies, and force Presidents to rely on their political enemies. Moreover, these three perspectives on PBRs all rest on the same empirical premise: that PBRs actually operate as meaningful constraints on the President’s ability to appoint commissioners who share her ideology and policy preferences. Our findings shore up that empirical premise, though they fall short of confirming the more ambitious claims that each account makes. Our analysis also indicates that the effects of PBRs on the operation of multimember agencies are potentially quite different today than they were three decades ago, when PBRs had a less dramatic effect on ideological composition. This finding suggests that conclusions regarding PBRs drawn from the period before partisan sort should be reassessed anew. Finally, and perhaps counterintuitively, this Article suggests that, despite concerns that PBRs weaken presidential control over multimember agencies, the fact that PBRs are relatively effective actually makes them less threatening to the sitting President. The rationale for this counterintuitive conclusion is explained at greater length below.[footnote]In brief, PBRs ensure that Presidents inherit commissions with only a bare majority of members from the opposition party. For a commission with fixed terms for appointees and a PBR, this means that in expectation, it will take less time for a new President to “flip” the commission to include a majority of her party’s members than it would for an otherwise identical multimember commission with a fixed-term provision and no PBR. See infra section III.C.[/footnote] For the optimist, our results might suggest that there is an unexpected upside to partisan sort: As the parties become less ideologically diverse, multimember agencies become more so. While the phenomenon of partisan sort has fueled concerns about group polarization in government institutions,[footnote]See David A. Jones, The Polarizing Effect of a Partisan Workplace, 46 PS: Pol. Sci. & Pol. 67, 69–71 & figs.3 & 4 (2013) (reporting evidence of group polarization at federal and other inside-the-Beltway workplaces); see also James B. Stewart, Case Study in Chaos: How Management Experts Grade a Trump White House, N.Y. Times (Feb. 2, 2017), http://www.nytimes.com/2017/02/02/business/donald-trump-management-style.html (on file with the Columbia Law Review ) (reporting a management expert’s concerns regarding “group think” in the Trump Administration).[/footnote] our results suggest that the combination of partisan sort and partisan balance requirements improves the prospects for ideological diversity within multimember agencies. We caution, though, that sweeping normative conclusions are premature: Our study addresses the effect of PBRs on the ideological diversity of multimember agencies, not the effect of ideological diversity on agency functioning. We cannot yet claim that PBRs are desirable because ideologically diverse agencies function better than agencies with monolithic memberships. What we can say is that for such a claim to be true, it must first be true that PBRs actually lead to more ideologically diverse agencies, and our results strongly affirm that first premise. While our study is primarily backward-looking, our findings have clear implications going forward for the evaluation of cross-party appointments under President Trump and subsequent administrations. Without knowing whether past Presidents have named genuine cross-partisans to PBR agencies, it is difficult to determine whether a new White House occupant is acting consistent with established norms. At the same time, a President can potentially skew the composition of PBR agencies by quickly making co-party appointments while stalling on cross-party nominations. However, by documenting past practices regarding appointments to PBR agencies, this study establishes a baseline that can facilitate future comparative evaluations. This Article proceeds in three parts. Part I provides background on PBRs, reviews the existing literature on these requirements, and considers causal pathways by which these requirements might affect the ideological composition of multimember agencies. Part II describes our data and methodology and presents our main empirical results. Part III discusses the implications of our results and identifies areas for future research. I. Partisan Balance Requirements: History, Theory, and Existing Evidence This first Part provides an overview of partisan balance requirements in agency organic statutes and considers reasons why these requirements might (or might not) affect the ideological makeup of boards and commissions. Section I.A briefly summarizes the history of PBRs. Section I.B surveys the literature on the efficacy of PBRs. Section I.C considers causal pathways through which PBRs might operate and explains how these different causal stories can be disentangled. A. A Short History of Partisan Balance Requirements Partisan balance requirements for federal agencies date at least as far back as the Edmunds Anti-Polygamy Act of 1882.[footnote]Act of Mar. 22, 1882, ch. 47, 22 Stat. 30; see also Krotoszynski et al., supra note 15, at 964.[/footnote] The Act prohibited polygamists from voting or running in Utah elections. A new Utah Commission was tasked with enforcing the anti-polygamy requirements. Section 9 of the Act stated that the Board should consist of “five persons, to be appointed by the President, by and with the advice and consent of the Senate, not more than three of whom shall be members of one political party . . . . ”[footnote]§ 9, 22 Stat. at 32 (emphasis added).[/footnote] Similar PBRs appeared in several more nineteenth-century statutes, including the Pendleton Act of 1883, which created the Civil Service Commission; [footnote]Act of Jan. 16, 1883, ch. 27, 22 Stat. 403; see also Krotoszynski et al., supra note 15, at 966.[/footnote] the Interstate Commerce Act of 1887, which created the Interstate Commerce Commission; [footnote]Act of Feb. 4, 1887, ch. 104, 24 Stat. 379.[/footnote] and an 1890 law establishing a nine-member board of general appraisers that would ascertain the value of merchandise for customs purposes.[footnote]Act of June 10, 1890, ch. 407, § 12, 26 Stat. 131, 136.[/footnote] Congress also included PBRs in a number of statutes creating commissions to negotiate treaties with Native American tribes at the century’s end.[footnote]See Act of June 10, 1896, ch. 398, 29 Stat. 321, 342; Act of July 13, 1892, ch. 164, 27 Stat. 120, 138–39; Act of Aug. 19, 1890, ch. 807, 26 Stat. 336, 354; Act of Mar. 2, 1889, ch. 412, § 14, 25 Stat. 980, 1005. Justice Brandeis documents the early history of PBRs in his dissent in Myers v. United States, 272 U.S. 52, 269–71 & n.51 (1926) (Brandeis, J., dissenting).[/footnote] Perhaps not coincidentally, the 1880s also saw a national movement for civil service reform, and these first PBRs limited the President’s ability to adopt a “to the victor go the spoils” posture in appointments.[footnote]See Jerry L. Mashaw, Federal Administration and Administrative Law in the Gilded Age, 119 Yale L.J. 1362, 1389 (2010).[/footnote] Yet these first few PBRs do not appear to have elicited much discussion in the House or in the Senate, and—speculation aside—the initial reason for their insertion remains obscure.[footnote]See Krotoszynski et al., supra note 15, at 964–65, 967.[/footnote] The presidency of Woodrow Wilson brought with it the next golden age of PBRs. Wilson’s first term saw the creation of five major multimember agencies with PBRs in their organic statutes: the Federal Trade Commission (FTC),[footnote]Act of Sept. 26, 1914, ch. 311, § 1, 38 Stat. 717, 718 (providing that “[n]ot more than three of the [appointed] commissioners shall be members of the same political party”).[/footnote] the Federal Farm Loan Board,[footnote]Act of July 17, 1916, ch. 245, § 3, 39 Stat. 360, 360 (“Of the four members to be appointed by the President, not more than two shall be appointed from one political party.”).[/footnote] the U.S. Shipping Board,[footnote]Act of Sept. 7, 1916, ch. 451, § 3, 39 Stat. 728, 729 (“Not more than three of the commissioners shall be appointed from the same political party.”).[/footnote] the U.S. Employees’ Compensation Commission,[footnote]Act of Sept. 7, 1916, ch. 458, § 28, 39 Stat. 742, 748 (“No more than two of [the appointed] commissioners shall be members of the same political party.”).[/footnote] and the U.S. Tariff Commission.[footnote]Act of Sept. 8, 1916, ch. 463, § 700, 39 Stat. 756, 795 (providing that there will be “six members, who shall be appointed by the President, by and with the advice and consent of Senate, not more than three of whom shall be members of the same political party”).[/footnote] At the time, senators suggested that PBRs would make agencies less prone to partisan bias and more likely to act on the basis of expertise—though precisely how PBRs might produce these results remained unspecified.[footnote]See Krotoszynski et al., supra note 15, at 969–70 (discussing Congress’s rationale for creating PBRs).[/footnote] The creation of new agencies with PBRs in their organic statutes has continued ever since. In the last century, Congress has established at least forty bodies subject to PBRs.[footnote]See id. at 1009–15 tbl.1, 1016 tbl.2, 1017 tbl.3.[/footnote] Most of these statutes provide for an odd number of members, no more than a bare majority of whom may be from the same political party.[footnote]Id. (noting the partisan membership requirements for independent agencies listed in the tables).[/footnote] The organic statutes of a small handful of agencies (most significantly, the Federal Election Commission (FEC)) provide for an even number of members, with no more than half from the same party.[footnote]See Federal Election Campaign Act Amendments of 1976, Pub. L. No. 94-283, sec. 1, § 101, 90 Stat. 475, 475 (codified as amended at 2 U.S.C. § 437c(a)(1) (2012)). Other extant agencies with an even number of members and a requirement that no more than half hail from the same party include the International Trade Commission, 19 U.S.C. § 1330(a) (2012), and the Election Assistance Commission, 52 U.S.C. § 20923(b)(2) (Supp. II 2015).[/footnote] Significantly, the PBR provisions in agency organic statutes do not technically require a Democratic President to appoint Republicans or a Republican President to appoint Democrats. For example, the Equal Employment Opportunity Commission’s (EEOC) organic statute states that “not more than three [of the five commissioners] shall be members of the same political party.”[footnote]See 42 U.S.C. § 2000e-4(a) (2012).[/footnote] Thus, a Republican President could appoint three Republican commissioners and two independents (or, for that matter, two libertarians), rather than naming two Democrats. In practice, though, we see very few cases of Presidents naming independents to cross-party posts on PBR agencies. Indeed, in our review of 216 cross-party appointments to twenty-three PBR agencies between 1979 and 2014, we have identified only seven cases in which a President named an independent at a time when a PBR prohibited him from appointing a member of his own party to an agency.[footnote]See infra notes 39–42 and accompanying text.[/footnote] These seven cases, moreover, do not suggest a pattern of Presidents attempting to manipulate PBRs by filling cross-party seats with like-minded independents. Three of the seven cases involved a Democratic President filling a cross-party seat with an appointee who, though registered as an independent, had previously worked as an aide to a Republican senator.[footnote]The three are: Don Zimmerman, appointed by President Carter to the NLRB in 1980; Reginald Jones, appointed by President Clinton to the EEOC in 1996; and Jackie Clegg, appointed by Clinton to the Export-Import Bank in 1997. See James A. Gross, Broken Promise: The Subversion of U.S. Labor Relations Policy, 1947–1994, at 245–46 (1995) (noting that Zimmerman had served as an adviser to Republican Senator Jacob Javits of New York); Lee Davidson, Utah Nominee Gets High Marks, Deseret News (Salt Lake City) (May 23, 1997), http://www.deseretnews.com/article/562187/Utah-nominee-gets-high-marks.html [http://perma.cc/DUE5-UMHA] (noting that Clegg had served as an aide to Republican Senator Jake Garn of Utah); Press Release, White House Office of the Press Sec’y, President Clinton Names Reginald E. Jones to the Equal Employment Opportunity Commission (June 5, 1996), http://clintonwhitehouse6.archives.gov/ 1996/06/1996-06-06-jones-on-equal-employment-opportunity-commission.html [http://perma.cc/87C9-QDYT] (noting that Jones, at the time of his appointment, was senior legislative counsel to then-Republican Senator James Jeffords of Vermont).[/footnote] Two more involved Republican Presidents filling cross-party seats with independents who would later be chosen by Democratic Presidents for positions to which the Democratic President could have named a Democrat—an indication that the individual, though registered as an independent, was generally aligned with the Democratic Party.[footnote]The two are: Mary Schapiro, appointed by President Reagan to the SEC in 1988 and reappointed by President George H.W. Bush in 1989, and Thomas Curry, appointed to the Federal Deposit Insurance Corporation (FDIC) by President George W. Bush in 2003. See Binyamin Appelbaum, Official from F.D.I.C. Picked to Lead Banking Regulator, N.Y. Times (July 1, 2011), http://www.nytimes.com/2011/07/02/business/ 02currency.html (on file with the Columbia Law Review ) (noting President Obama’s nomination of Curry to be Comptroller of the Currency, a position not subject to a PBR); Hannah Bergman, FDIC Board Is Factor in OCC Choice, Am. Banker (Jan. 12, 2005), http://www.americanbanker.com/news/fdic-board-is-factor-in-occ-choice (on file with the Columbia Law Review ) (noting that Curry was an independent named by President George W. Bush to a cross-party seat); Robert D. Hershey, Jr., Clinton Expected to Name S.E.C. Member to Head C.F.T.C., N.Y. Times (May 3, 1994), http://www.nytimes.com/1994/ 05/03/business/clinton-expected-to-name-sec-member-to-head-cftc.html (on file with the Columbia Law Review ) (reporting that President Clinton would name Schapiro to chair the Commodity Futures Trading Commission (CFTC) at a time when the commission already had two Republican members, meaning Clinton could have named a fellow Democrat); Kathleen Pender, Mary Schapiro’s Mixed Marks as SEC Head, S.F. Chron. (Nov. 26, 2012), http://www.sfgate.com/business/networth/article/Mary-Schapiro-s-mixed-marks-as-SEC-head-4068463.php [http://perma.cc/9LBY-75RP] (noting that Schapiro, named by President Obama in 2009 to chair the SEC, served with two Republicans, meaning that Obama could have named a third registered Democrat instead of Schapiro); Vicky Stamas, President to Nominate Reum to Fill Position Remaining at the SEC, Bond Buyer, Aug. 21, 1992, at 5 (on file with the Columbia Law Review ) (noting that Schapiro was an independent in a cross-party seat).[/footnote] One case involved President Clinton appointing to the National Transportation Safety Board (NTSB) an independent who had been recommended by the Board’s outgoing Republican member.[footnote]President Clinton appointed John Goglia to the NTSB in 1995 and reappointed him in 1999. See Matthew Brelis, US Transportation Appointee Brings Hands-On View; 29 Years as Air Mechanic Won Spot for Saugus Man, Bos. Globe, Oct. 10, 1995, at 18 (on file with the Columbia Law Review ); The Former Board of the NTSB, Nat’l Transp. Safety Bd., http://www.ntsb.gov/about/board/Pages/former_members.aspx [http://perma.cc/2FPY-ZG4H] (last visited Oct. 2, 2017) (noting that Goglia served on the NTSB from 1995 to 2004).[/footnote] Indeed, in only one case did we find any record of complaints that the President had appointed an independent rather than a member of the opposite party to fill a cross-party seat.[footnote]President Reagan named Mary Azcuenaga to the FTC in 1984. Her appointment generated grumbling from one Democratic senator who “wondered why Mr. Reagan, if he wants to appoint independents, cannot pick them for Republican seats.” Robert D. Hershey, Jr., Washington Watch; No Democrats Seen for F.T.C., N.Y. Times (Oct. 28, 1985), http://www.nytimes.com/1985/10/28/business/washington-watch-no-democrats-seen-for-ftc.html (on file with the Columbia Law Review ) (paraphrasing Senator Wendell Ford, a Democrat from Kentucky). However, the outgoing Democratic commissioner was reported to favorably view Azcuenaga, who replaced him. See Thomas Ferraro, Reagan Expected to Name Woman to Replace FTC’s Pertschuk, UPI (Aug. 14, 1984), http://www.upi.com/Archives/1984/08/14/Reagan-expected-to-name-woman-to-replace-FTCs-Pertschuk/9862461304000 [http://perma.cc/A7SA-GSHN] (quoting an aide to the outgoing Democratic commissioner who said that the commissioner “respects” Azcuenaga and was “relatively pleased” that she would be his successor).[/footnote] The only agency at which Presidents have repeatedly sought to fill cross-party seats with ideologically sympathetic independents is—perhaps surprisingly—an agency with no regulatory authority: the U.S. Commission on Civil Rights.[footnote]See Charlie Savage, Maneuver Gave Bush a Conservative Rights Panel, Bos. Globe (Nov. 6, 2007), http://archive.boston.com/news/nation/articles/2007/11/06/ maneuver_gave_bush_a_conservative_rights_panel/?page=full [http://perma.cc/57GP-FE7N] (“Especially since the 1980s, presidents and lawmakers have tried to tilt the panel by appointing independents who shared their party’s views on civil rights.”).[/footnote] (We exclude the Commission on Civil Rights from our dataset because of its lack of authority to bind others through regulation, adjudication, or enforcement, as well as the fact that half of its members are chosen by Congress rather than the President and its members generally serve part-time.[footnote]Of the Commission’s eight members, four are appointed by the President, two are appointed by the President pro tempore of the Senate on the recommendation of the Senate Majority and Minority Leaders, and two are appointed by the Speaker of the House on the recommendation of the House Majority and Minority Leaders. 42 U.S.C. § 1975(b) (2012).[/footnote]) The Commission’s organic statute states that “[n]ot more than 4 of the [8] members shall at any one time be of the same political party.”[footnote]Id.[/footnote] The meaning of this provision was tested in 2003 and 2004, when two Republican members of the Commission changed their registration to independent.[footnote]See Savage, supra note 43 (“Critics say Bush in effect installed a fifth and sixth Republican on the panel in December 2004, after two commissioners, both Republicans when appointed, reregistered as independents.”).[/footnote] Their switches allowed President George W. Bush to name two additional Republicans to the commission, bringing the number of Republican or recently Republican members of the panel to six.[footnote]Id.[/footnote] At the time, the Office of Legal Counsel in the Justice Department issued an opinion stating that, for purposes of the statutory PBR, “the relevant consideration is the party affiliation of the other members at the time the new member is appointed.”[footnote]Political Balance Requirement for the Civil Rights Comm’n, 28 Op. O.L.C. 295, 295 (2004).[/footnote] In the Office of Legal Counsel’s view, application of the statutory PBR did not require looking beyond party registration to determine the partisan identity of a prospective or current commissioner. Aside from the Office of Legal Counsel’s opinion, there is very little legal precedent regarding the interpretation or application of partisan balance requirements in agency organic statutes. The constitutionality of PBRs has been challenged in two cases, but the courts did not reach the merits of the constitutional issue in either case. In FEC v. NRA Political Victory Fund , the National Rifle Association’s political action committee argued that the provision in the six-member FEC’s organic statute capping the number of commissioners from any political party at three violated the President’s appointment power.[footnote]6 F.3d 821, 824 (D.C. Cir. 1993).[/footnote] The D.C. Circuit rejected the argument on standing grounds, reasoning that “it is impossible to determine in this case whether the statute actually limited the President’s appointment power” because it was not clear that the President otherwise would have sought to appoint more than three commissioners from his own party.[footnote]Id.[/footnote] In National Committee of the Reform Party v. Democratic National Committee , the Ninth Circuit rejected a challenge to the FEC’s partisan balance requirement on similar grounds, emphasizing that “without the statute the President could have appointed exactly the same members.”[footnote]168 F.3d 360, 365 (9th Cir. 1999) (internal quotation marks omitted) (quoting NRA Political Victory Fund , 6 F.3d at 825).[/footnote] We know of no case in which a litigant has sought judicial enforcement of a PBR—that is, has argued for the invalidation of an agency action on the grounds that a commission had too many members of the same party. At the same time, we have found no reported instances of outright PBR violations—no case, that is, in which a President has sought to exceed the statutory cap on commissioners from a single party. The letter, if not the spirit, of PBRs is followed uniformly. But to say that PBRs are followed as a technical matter does not explain what substantive effects these provisions have. The following section turns to that question and surveys the scant scholarly literature that exists to da B. Partisan Balance Requirements Bite? Our central aim is to test the hypothesis that partisan balance requirements affect the ideological composition of multimember agencies. A contrary view posits that cross-party commissioners “come in sheep’s clothing” and do not differ from a President’s co-party appointees in any way other than the superficial fact of party registration.[footnote]See Ho, supra note 3, at 2–3 (collecting sources that make this claim).[/footnote] One scholar has noted the example of Ray Wakefield, a nominally Republican but questionably conservative Roosevelt nominee to the FCC.[footnote]See Lawrence W. Lichty, Members of the Federal Radio Commission and Federal Communications Commission 1927–1961, 6 J. Broadcasting 23, 25 (1961); accord Williams, supra note 3, at 241 (concluding that “party affiliation has not been a reliable predictor of voting” on the FCC).[/footnote] Timothy Nokken and Brian Sala cite examples from the Eisenhower, Kennedy, Nixon, and Carter Administrations.[footnote]Timothy P. Nokken & Brian R. Sala, Confirmation Dynamics: A Model of Presidential Appointments to Independent Agencies, 12 J. Theoretical Pol. 91, 95 (2000). In particular, Nokken and Sala note that President Eisenhower appointed two nominally Democratic FTC commissioners whose ideological loyalties were suspect: Robert Secrest, a conservative Democratic congressman from Ohio, and William Kern, previously an FTC staff member “whose principal Democratic credentials were that his father had been William Jennings Bryan’s running mate in 1908.” Id. (quoting Staff of the S. Comm. on Commerce, 94th Cong., Appointments to the Regulatory Agencies: The Federal Communications Commission and the Federal Trade Commission (1949–1974), at 385 (Comm. Print 1976)). Nokken and Sala add that “Kennedy and Nixon adopted similar appointment strategies by appointing liberal ‘Lindsay Republicans’ and conservative ‘Connally Democrats,’ respectively.” Id. And they further mention the case of Frank Reiche, one of President Carter’s cross-party appointees to the FEC, whom some Republican senators opposed because he was not sufficiently conservative. Id. at 95 & n.6; see also Princeton Attorney Confirmed to Post on Election Panel, Asbury Park Press, July 26, 1979, at 62 (on file with the Columbia Law Review ) (noting that twenty-one out of forty-one Republican senators voted against Reiche, with some conservative Republicans saying that he did not represent their views).[/footnote] An especially notorious example from the Nixon Administration is that of James Quello, a nominally cross-party appointee to the FCC who—despite his Democratic affiliation—donated $1,100 to Nixon’s reelection effort in 1972.[footnote]See Staff of the S. Comm. on Commerce, 94th Cong., Appointments to the Regulatory Agencies: The Federal Communications Commission and the Federal Trade Commission (1949–1974), at 357 (Comm. Print 1976).[/footnote] A report commissioned by the Senate Commerce Committee and completed at the end of the Ford presidency concluded that cross-party appointees “typically” support the President who selected them.[footnote]Id. at 366.[/footnote] Alan Morrison observed at the end of the Reagan years that the outgoing President had “largely succeeded” in “placing in all agencies, independent and otherwise, those who believe firmly in the Reagan view of government—Democrats as well as Republicans.”[footnote]Alan B. Morrison, How Independent Are Independent Regulatory Agencies?, 1988 Duke L.J. 252, 253–54.[/footnote] And Senator Bob Dole once accused President Clinton of appointing “Clinton Republicans” to cross-party seats.[footnote]Al Kamen, Who’s Republican? And Who’s to Decide?, Wash. Post (Sept. 6, 1993), http://www.washingtonpost.com/archive/politics/1993/09/06/whos-republican-and-whos-to-decide/8a85544f-f1a1-4558-ac1f-5c498e5609e6/?utm_term=.a10e33cb8cfe (on file with the Columbia Law Review ).[/footnote] We know of only one study other than our own that undertakes a quantitative cross-commission analysis of the effect of PBRs on the ideological composition of multimember agencies: the 1964 study by Stuart Nagel and Martin Lubin referenced above. Nagel and Lubin focus on seven commissions: the now-defunct Civil Aeronautics Board and Interstate Commerce Commission (ICC), the FCC, the Federal Power Commission (now the Federal Energy Regulatory Commission (FERC)), the FTC, the National Labor Relations Board (NLRB), and the SEC.[footnote]Nagel & Lubin, supra note 5, at 39.[/footnote] They draw data from three years each spaced a decade apart: 1936, 1946, and 1956.[footnote]Id.[/footnote] They consider a decision to be “liberal” if it goes “in favor of the consumer, shipper, or investor (rather than the seller, producer, transporter, or broker), in favor of labor (rather than management), in favor of a small business or increased competition (rather than a larger firm or decreased competition).”[footnote]Id. at 40.[/footnote] They compare each commissioner’s liberalism score with her agency’s average liberalism score. On this dimension, Democrats appointed by Republican Presidents are most likely to be above their agency’s average liberalism score (64%), followed by Democrats appointed by Democratic Presidents (54%), followed by Republicans appointed by Democratic Presidents (46%), followed by Republicans appointed by Republican Presidents (33%).[footnote]Id. at 42 tbl.1.[/footnote] The findings are reported in a way that makes it impossible to determine just how much PBRs affect voting behavior, though the results seem to suggest that PBRs lead Republican Presidents to choose somewhat more liberal commissioners and lead Democratic Presidents to select somewhat more conservative commissioners. While these results are suggestive, Nagel and Lubin’s method of categorizing decisions as “liberal” and “conservative” is certainly vulnerable to critique. First, the terms “liberal” and “conservative” are not easy to operationalize—and any attempt to do so inevitably imports biases. (Why do we think liberals are more likely to favor “increased competition,” and how should we code cases in which the competitive consequences are hotly disputed?) Second, the types of cases that will reach an adjudicatory commission may vary with the commission’s ideological composition—the administrative law version of the Priest-Klein effect familiar from the civil procedure literature.[footnote]See George L. Priest & Benjamin Klein, The Selection of Disputes for Litigation, 13 J. Legal Stud. 1, 14–18 (1984) (suggesting the disputes that reach litigation are ones in which each party has close to a 50% probability of victory).[/footnote] Individual workers and labor unions may be more likely to bring ambitious cases to the NLRB when they know that the Board has a liberal majority; likewise for employers when they know that the board tilts toward the right. Professor Daniel Ho’s methodologically rigorous study of voting patterns among FCC commissioners addresses the first of these two aforementioned imperfections of the Nagel and Lubin study. Ho pioneers an approach—a Bayesian multilevel ideal point model of mixed ordinal votes—to map votes onto a unidimensional spectrum without requiring researchers to decide which positions are “liberal” and which are “conservative.”[footnote]Ho, supra note 3, at 14–16.[/footnote] As Ho summarizes, his analysis of the FCC suggests that most cross-party appointees are “genuine”: With a small number of exceptions, “[c]ross-party appointees don’t don sheep’s clothing.”[footnote]Id. at 19, 24.[/footnote] In other words, Democratic commissioners appointed by Republican Presidents to satisfy the statutory PBR appear to be relatively liberal, and Republican commissioners appointed by Democratic Presidents to satisfy the statutory PBR appear to be relatively conservative. Somewhat surprisingly, the most conservative commissioner according to Ho’s estimates is a Republican appointed by Democratic President Clinton, and the three most liberal commissioners are Democrats appointed by Republican Presidents Reagan, George H.W. Bush, and George W. Bush.[footnote]Id. at 21 fig.4.[/footnote] Ho’s study of the FCC provides us with extraordinarily useful insights regarding voting behavior at a single agency. But while Ho’s single-agency analysis suggests that partisan balance requirements do “bite,” it is not clear whether his FCC-specific results are generalizable to other multimember agencies.[footnote]The FCC is, by many accounts, a particularly politicized agency. See Benjamin Kapnik, Affirming the Status Quo?: The FCC, ALJs, and Agency Adjudications, 80 Geo. Wash. L. Rev. 1527, 1528 (2012) (“The FCC has long been considered a politicized agency . . . .”); Philip J. Weiser, Institutional Design, FCC Reform, and the Hidden Side of the Administrative State, 61 Admin. L. Rev. 675, 689 (2009) (noting the “super-politicized” environment at the FCC); cf. Kimberly A. Zarkin & Michael J. Zarkin, The Federal Communications Commission: Front Line in the Culture and Regulation Wars 49 (2006) (“[T]he FCC is an organization that is particularly open to interference by outside forces.”). While these characterizations are difficult to verify, they cast doubt on the external validity of FCC-specific findings .[/footnote] Unfortunately, Ho’s innovative method cannot be applied broadly beyond the FCC: His method relies on having a large set of nonunanimous commission votes to analyze, but at many multimember agencies—including the FEC, FTC, Nuclear Regulatory Commission (NRC), and SEC—dissents and concurrences are rare.[footnote]See Ho, supra note 3, at 8 tbl.1 (showing that the number of dissents and concurrences per commissioner is 172 for the FCC, compared to four for the FEC, twelve for the FTC, three for the NRC, and six for the SEC).[/footnote] For this reason, we look to other data sources and methodological approaches in order to assess the cross-agency effects of PBRs. Ho’s study also does not seek to determine why PBRs might be effective. The next section turns to that question and sets forth two theories with testable implications that might explain the efficacy of these provisions. Part II goes on to consider whether PBRs do in fact affect the ideological composition of multimember agencies and assess the evidence in support of the two separate theories. C. Why Might Partisan Balance Requirements Bite? We focus here on two causal mechanisms that might explain why PBRs affect the ideological composition of multimember agencies: a supply-side theory and a demand-side theory . In the “market” for commissioners, the President draws from a limited supply of individuals with the experience and expertise to serve competently on multimember federal agencies. The President then must “sell” the nominee to sometimes-skeptical senators—the consumers in our market metaphor. If PBRs do indeed affect the ideological composition of multimember agencies, it may be due to supply-side constraints, or demand-side constraints, or both. 1. The Supply-Side Theory . — The supply-side theory holds that the President selects from a pool of potential cross-party appointees that is dominated by individuals whose ideological preferences differ from the President’s own. Consider the challenge facing a Democratic President who is filling a seat on the five-member FCC at a time when the FCC already has three Democratic members (its maximum number of same-party commissioners). Presumably the President cares at least somewhat about the competence of commissioners and so desires an appointee with relevant experience in communications law or policy. The pool of individuals who have such experience and who are not members of the Democratic Party will be weighted toward moderates and conservatives. While it might not be impossible for the Democratic President to find a liberal non-Democrat who also has relevant qualifications for the commissionership, it is—at the very least—easier for a Democratic President to draw liberals from within the President’s own party than from without. The plausibility of the supply-side theory depends on the strength of the relationship between partisanship and ideology. If the Republican Party includes large pools of liberals as well as conservatives, then a Democratic President making a cross-party appointment can satisfy the PBR by selecting a relatively liberal Republican. Partisan sort, however, undermines the ability of Democratic Presidents to find ideologically sympathetic Republicans to fill posts on multimember agencies (and likewise for Republican Presidents seeking out ideologically sympathetic Democrats). The political science literature on partisan sort distinguishes between “elite polarization” and “mass polarization.” By “elites” (or alternatively, the “political class”), we refer to public officials, interest group leaders, activists, substantial donors, and political commentators.[footnote]See Morris P. Fiorina & Matthew S. Levendusky, Disconnected: The Political Class Versus the People, in 1 Red and Blue Nation?: Characteristics and Causes of America’s Polarized Parties 49, 50 (Pietro S. Nivola & David W. Brady eds., 2006) (describing various constituencies in the “political class” and the existence of significant differences in political views within this group).[/footnote] While the extent to which mass polarization has occurred is contested,[footnote]Compare Alan I. Abramowitz & Kyle L. Saunders, Ideological Realignment in the U.S. Electorate, 60 J. Pol. 634, 649 (1998) (finding that “increasing ideological polarization of the Democratic and Republican Parties in the Reagan and post-Reagan eras made it easier for voters to recognize the differences between the parties’ policy stands” and choose “their party identification” based on policy preferences rather than strict party allegiance), with Morris P. Fiorina, Samuel A. Abrams & Jeremy C. Pope, Polarization in the American Public: Misconceptions and Misreadings, 70 J. Pol. 556, 558 (2008) (stating that “party sorting has proceeded much less in the general public than among party elites”).[/footnote] there is close to a consensus that elites have become more polarized over the course of the past several decades.[footnote]See Fiorina et al., supra note 70, at 557 (“There is general agreement that party elites have become significantly more distinct over the course of the past several decades . . . .”).[/footnote] Because appointees to multimember agencies are generally drawn from the “political class,”[footnote]See infra note 108 (looking at the propensity of PBR appointees to contribute to political candidates and finding that the “ideological distribution of bureaucratic appointments broadly reflects the president’s preferences”).[/footnote] elite polarization rather than mass polarization is the more relevant phenomenon for our purposes. By most measures, the partisan sorting of elites accelerated in the mid- to late-1970s and has continued to pick up steam ever since.[footnote]See, e.g., Christopher Hare & Keith T. Poole, The Polarization of Contemporary American Politics, 46 Polity 411, 415 (2014) (observing—on the basis of voting patterns among members of the House and Senate—that “the parties began to diverge in the mid-1970s and this trend has continued unabated into the most recent Congress”); Marc J. Hetherington, Resurgent Mass Partisanship: The Role of Elite Polarization, 95 Am. Pol. Sci. Rev. 619, 622–23 (2001) (observing—also on the basis of congressional voting patterns—that polarization reached a “trough” in the mid-1970s but began a “steady rise” starting with the Carter years and a “second spike” in the last two years of George H.W. Bush’s presidency).[/footnote] On most accounts, the partisan realignment of the South—with white conservative Democrats moving to the Republican Party—was an important but nonexclusive factor contributing to partisan sort.[footnote]See Hare & Poole, supra note 73, at 417 (“[T]he southern realignment does not fully account for the increase in polarization. The Republican Party became much more conservative across all regions of the United States.”); Gary C. Jacobson, Partisan Polarization in American Politics: A Background Paper, 43 Presidential Stud. Q. 688, 691 (2013) (stating that the “main source” of polarization was “the partisan realignment of the South” but noting that “[c]onservative whites outside the South also moved toward the Republican Party, while liberals became overwhelmingly Democratic”).[/footnote] Whatever the causes, few scholars would dispute that ideology is a stronger predictor of partisan affiliation today than it was twenty-five to thirty-five years ago.[footnote]See, e.g., Joseph Bafumi & Robert Y. Shapiro, A New Partisan Voter, 71 J. Pol. 1, 3 (2009) (explaining that “changes have given way to an electorate that is more strongly driven by liberal/conservative ideological concerns” today).[/footnote] The following figure provides one window on this partisan sort. The figure uses DW-NOMINATE ideal point estimates—estimates of legislators’ ideological preferences based on their roll call voting records—to track partisan sort over the past fifty years.[footnote]These data were derived from Party Medians from DW-NOMINATE Congresses 1–113, Voteview (Mar. 23, 2015), http://legacy.voteview.com/pmediant.htm [http:// perma.cc/3Y52-E74N]. DW-NOMINATE arranges legislators on a -1 to 1 scale based on their roll call voting records, placing legislators with similar voting records close together on the scale. Id. The measure does not consider the content of the roll call votes. Id. Once the algorithm is run for a large set of votes, the optimal placement of legislators along with scale makes it obvious to any political observer that legislators placed near -1 are very liberal and those placed near 1 are very conservative. See Keith T. Poole & Howard Rosenthal, Congress: A Political-Economic History of Roll Call Voting 23–24 (2000).[/footnote] The solid red line charts the median Republican House member’s ideal point from 1963 (close to the nadir of polarization in the twentieth century) through 2013; the dotted light blue and dashed dark blue lines report those values for the median southern and non-southern Democratic representative, respectively. The figure illustrates two reasons why partisan sort has occurred: (1) the Republican Party became more conservative over a half century; and (2) southern (but not northern) Democrats became more liberal.[footnote]As the figure shows, the trend toward polarization is asymmetric, with greater movement away from the middle among Republicans than among Democrats. For further discussion, see Nolan McCarty, Keith T. Poole & Howard Rosenthal, Polarized America: The Dance of Ideology and Unequal Riches 11 (2006) (noting that “Republicans in the North and South have moved sharply to the right,” while “Northern Democrats . . . don’t look sharply different from the Democrats of old”).[/footnote] The latter phenomenon is attributable to the fact that a large number of white conservatives in the South have changed their party affiliation from Democratic to Republican, leaving a more liberal and largely African American Democratic Party in southern states.[footnote]See Nelson W. Polsby, How Congress Evolves 151 (2005).[/footnote] Figure 1: Party Polarization in the House of Representatives Figure 2 provides another perspective on partisan sort. It shows the gap in DW-NOMINATE scores for the median Democratic and Republican members of the House between 1963 and 2013. The takeaway from Figures 1 and 2 is clear: The parties have become more ideologically distinct over the past fifty years. The increase has been nearly monotonic, with polarization accelerating in the second half of this period. Figure 2: Ideological Distance Between the Parties in the House The phenomenon of partisan sort has clear implications for the ideological composition of multimember agencies subject to PBRs. In the late 1970s and early 1980s, when both parties remained relatively ideologically heterogeneous at the elite level, one might expect the effect of PBRs on ideological composition to be muted. A Democratic President making a cross-party appointment would not be forced to choose from among conservatives; likewise, a Republican President would likely be able to find qualified Democrats who were not liberals. But as partisan sort accelerated over the last two decades of the twentieth century and the beginning of the twenty-first century, one might expect the effect of PBRs on ideological composition to grow stronger. That is, insofar as supply-side constraints account for the effect of PBRs on the ideological composition of multimember agencies, one might expect that effect to increase as supply-side constraints become more binding. Accordingly, one way to assess the plausibility of the supply-side theory is to look at whether the effect of PBRs on the ideological composition of multimember agencies has increased over time. If it has, then that finding would place the supply-side theory on firmer footing. Further discussion of this point is deferred until Part II, which explains tests conducted to determine whether the effect of PBRs on ideological composition is temporally dependent. But first, section I.C.2 considers whether PBRs might affect ideological composition through an alternative causal mechanism. 2. The Demand-Side Theory . — Once a President has selected a nominee for an agency post, she must “sell” that nominee to the Senate. The confirmation hurdle imposes a potential demand-side constraint on the President’s ability to stock multimember agencies with ideologically sympathetic appointees. Senators from the opposition party may use what leverage they have to ensure that the President selects a “bona fide” member of the opposition party when a PBR forces a cross-party appointment.[footnote]See Neal Devins & David E. Lewis, Not-So Independent Agencies: Party Polarization and the Limits of Institutional Design, 88 B.U. L. Rev. 459, 488 (2008) (“[O]pposition Senators regularly use holds and other delaying strategies to pressure the President to appoint party loyalists to slots held by opposition-party members.”).[/footnote] To satisfy senators—the “consumers” in the market metaphor—the President may have to honor the spirit of PBRs when choosing cross-party commissioners. Insofar as demand-side constraints compel Democratic Presidents to choose conservative Republican commissioners and Republican Presidents to name liberal Democrats, one might expect these constraints to be most binding under conditions of divided government. Presumably, the Senate is more likely to stop a Democratic President from appointing a liberal Republican to satisfy a PBR when the Senate is Republican-controlled than when the Democrats have a majority (and vice versa when it is a Republican President seeking to place a conservative Democrat on a commission). To be sure, even a minority opposition party in the Senate has historically been able to stop a nominee from being confirmed through the filibuster, provided that it could muster forty-one votes. But with the abolition of the filibuster for agency nominees in November 2013,[footnote]See Paul Kane, Reid, Democrats Trigger ‘Nuclear’ Option; Eliminate Most Filibusters on Nominees, Wash. Post (Nov. 21, 2013), http://www.washingtonpost.com/ politics/senate-poised-to-limit-filibusters-in-party-line-vote-that-would-alter-centuries-of- precedent/2013/11/21/d065cfe8-52b6-11e3-9fe0-fd2ca728e67c_story.html (on file with the Columbia Law Review ).[/footnote] the minority party in the Senate no longer has this weapon in its arsenal. The demand-side theory, like the supply-side theory, leads us to expect that the effect of PBRs on the ideological composition of multimember agencies would vary across time. In particular, we might expect the effect to be strongest when the White House and Senate are controlled by different parties, such that the Senate can easily block a cross-party appointee who is ideologically sympathetic to the President. We might also expect the effect to be especially weak in the few periods when the President’s party has held a filibuster-proof Senate majority. Our dataset encompasses two such periods—both under President Obama. First, the Democrats held sixty Senate seats from July 2009 (after Democrat Al Franken was sworn in as senator from Minnesota[footnote]See Mark Zdechlik, Franken Brings 60th Vote to Senate Dems, Minn. Pub. Radio (July 1, 2009), http://www.mprnews.org/story/2009/07/01/franken60 [http:// perma.cc/K8FR-MWCU].[/footnote]) until January 2010 (when the Democrats lost a Senate seat from Massachusetts following Republican Scott Brown’s victory in a special election[footnote]See Michael Cooper, G.O.P. Senate Victory Stuns Democrats, N.Y. Times (Jan. 19, 2010), http://www.nytimes.com/2010/01/20/us/politics/20election.html (on file with the Columbia Law Review ).[/footnote]). And second, after the Senate eliminated the filibuster for confirmation votes on nominees to non-Supreme Court appointments in November 2013, the Democrats retained a narrow majority until January 2015 (55-45, including two independents who caucused with the Democrats). The next Part examines whether PBRs do indeed affect the ideological composition of multimember agencies—and, if so, why. Our approach allows us to test both the supply-side and demand-side theories, and to reach tentative conclusions as to why PBRs might have real-world consequences. II. An Empirical Analysis of Partisan Balance Requirements This Part presents an empirical analysis of the effect of partisan balance requirements on the ideological composition of multimember agencies. Section II.A describes our data and explains our research design. Section II.B presents initial results. Section II.C considers supply-side and demand-side explanations for the efficacy of PBRs. A. Data and Research Design To explore the relationship between the partisan and ideological outlooks of appointees to agencies with partisan balance requirements (PBR agencies), we first must define the population of PBR agencies. Then, we must identify both the partisan identification and ideological orientation for each appointee to these PBR agencies. This section details our process for collecting data on (1) PBR agencies, (2) appointee partisanship, and (3) appointee ideology. 1. Agencies with Partisan Balance Requirements . — PBR agencies share one essential attribute: a restriction, grounded in statute or a deeply entrenched norm, on the number of individuals from the same political party that may be appointed to the agency’s multimember governing board. We examined appointments to twenty-three agencies[footnote]These twenty-three PBR agencies are: the Commodity Futures Trading Commission (CFTC), Consumer Product Safety Commission (CPSC), Defense Nuclear Facilities Safety Board (DNFSB), Equal Employment Opportunity Commission (EEOC), Export-Import Bank (Ex-Im), Farm Credit Administration (FCA), Federal Communications Commission (FCC), Federal Deposit Insurance Corporation (FDIC), Federal Election Commission (FEC), Federal Energy Regulatory Commission (FERC), Federal Labor Relations Authority (FLRA), Federal Maritime Commission (FMC), Federal Trade Commission (FTC), Merit Systems Protection Board (MSPB), National Credit Union Administration (NCUA), National Labor Relations Board (NLRB), National Mediation Board (NMB), National Transportation Safety Board (NTSB), Nuclear Regulatory Commission (NRC), Postal Regulatory Commission (PRC), Securities and Exchange Commission (SEC), Surface Transportation Board (STB, formerly the Interstate Commerce Commission), and U.S. International Trade Commission (USITC).[/footnote] that meet this definition and that were in existence between 1979 and 2014.[footnote]See Datla & Revesz, supra note 1, at 784 & n.89; George A. Krause & Anne Joseph O’Connell, Replication Data for: Experiential Learning and Presidential Management of the U.S. Federal Bureaucracy: Logic and Evidence from Agency Leadership Appointments (2015) [hereinafter Krause & O’Connell, Replication Data], http://dataverse.harvard.edu/file.xhtml?fileId=2681304&version=RELEASED&version=.2 (on file with the Columbia Law Review ); David C. Nixon, The Independent Regulatory Commissioner Data Base, http://www2.hawaii.edu/~dnixon/IRC/index.htm [http://perma.cc/XK2Y-2E2B] (last updated Mar. 10, 2005).[/footnote] With the exception of the NLRB, these agencies’ PBRs are imposed by statute.[footnote]See, e.g., 42 U.S.C. § 2000e-4(a) (2012) (mandating that the EEOC “be composed of five members, not more than three of whom shall be members of the same political party”).[/footnote] (We classify the NLRB as a PBR agency based on a firmly entrenched norm favoring partisan balance on that body.[footnote]See infra notes 127–131 and accompanying text.[/footnote]) The structural features of these agencies vary considerably. Table 1 reports the presence or absence of design features in these agencies that, along with PBRs, are commonly considered to be indicia of agency independence.[footnote]See Rachel E. Barkow, Insulating Agencies: Avoiding Capture Through Institutional Design, 89 Tex. L. Rev. 15, 18 (2010) (outlining some of the design features that act as “more robust checks against agency capture under asymmetrical political conditions than the use of traditional factors alone”); Datla & Revesz, supra note 1, at 784 & n.89. An agency’s multimember structure is another common feature of independent agencies. See Barkow, supra, at 17; Datla & Revesz, supra note 1, at 792–97. Because all PBR agencies are by definition headed by multiple members, Table 1 omits this feature. Among these twenty-three agencies, the NMB is an outlier in two respects. First, its power to bind outside parties is limited. The NMB can compel railroads or airlines and their employees to arbitrate labor disputes if certain conditions are met; if one party refuses to arbitrate, the NMB can authorize a labor strike or a lockout by management. 45 U.S.C. § 155 (2012). Second, its structure includes only three indicia of independence: fixed terms, PBRs, and multimember boards. Nonetheless, we think the NMB deserves inclusion. Given the historical significance of labor disputes in the railroad industry and the importance of the transportation sector to the national economy, the NMB’s powers, though limited, are nontrivial. See generally Gabriel Kolko, Railroads and Regulations (2014). Further, fixed terms and PBRs—both of which the NMB has—are considered particularly important indicia of independence. See Paul R. Verkuil, The Purposes and Limits of Independent Agencies, 1988 Duke L.J. 257, 259. Nonetheless, excluding the NMB from our analysis does not materially change the results.[/footnote] For-cause removal protection prevents the President from removing an appointee at will, requiring that the President show cause to do so before the expiration of the appointee’s term.[footnote]See Barkow, supra note 87, at 27.[/footnote] For instance, FTC commissioners may be removed only “for inefficiency, neglect of duty, or malfeasance in office.”[footnote]42 U.S.C. § 5841(e); see also Humphrey’s Ex’r v. United States, 295 U.S. 602, 627–32 (1935). However, neither Congress nor the courts have defined these broad terms constituting good cause, see Bowsher v. Synar, 478 U.S. 714, 729 (1986), which raises the possibility that removal protections would not constrain a motivated President. See Lawrence Lessig & Cass R. Sunstein, The President and the Administration, 94 Colum. L. Rev. 1, 110–11 (1994) (discussing the vagueness of the statute’s reference to “good cause” and how the President could use this to retain “a large degree of removal and supervisory power”).[/footnote] Fixed terms in office further insulate appointees to PBR agencies from the White House and thus provide a second indicium of agency independence.[footnote]See Peter L. Strauss, An Introduction to Administrative Justice in the United States 15 (1989) (explaining how fixed terms for appointees to PBR agencies can separate them from “presidential influence”).[/footnote] The ability to pursue litigation independent of the Justice Department (litigation authority); bypass centralized OMB review and submit budgets directly to Congress (bypass authority”); and craft policy through formal adjudication, which is less susceptible to political interference than rulemaking (adjudication authority), all foster agency independence as well.[footnote]Datla & Revesz, supra note 1, at 799–804 (discussing litigation authority); id. at 804–08 (discussing bypass authority); id. at 808–12 (discussing adjudication authority).[/footnote] Finally, qualification requirements—for example, that members of the NTSB possess transportation-safety-related professional credentials and experience[footnote]49 U.S.C. § 1111(b) (2012).[/footnote]—constrain the President’s hand in appointments.[footnote]See William G. Howell & David E. Lewis, Agencies by Presidential Design, 64 J. Pol. 1095, 1098–99 (2002). We exclude U.S. citizenship requirements, which impose de minimis constraints on appointments, from our classification of qualification requirements.[/footnote] Table 1: Characteristics of Included PBR Agencies[footnote]See Henry B. Hogue, Cong. Research Serv., RL33886, Statutory Qualifications for Executive Branch Positions 26–30 tbl.A-2 (2015), http://fas.org/sgp/crs/misc/ RL33886.pdf [http://perma.cc/C572-XUD7] (identifying agencies with qualification requirements); Datla & Revesz, supra note 1, at 786 tbl.1 (identifying agencies with statutory removal protection); id. at 790 tbl.2 (identifying agencies with fixed terms); id. at 800 tbl.5 (identifying agencies with litigation authority); id. at 804 tbl.6 (identifying agencies with bypass authority); id. at 809 tbl.7 (identifying agencies with adjudication authority).[/footnote] Agency Removal Protection Fixed Terms Litigation Authority Bypass Authority Adjudication Authority Qualification Requirements CFTC ● ● ● ● ● CPSC ● ● ● ● ● ● DNFSB ● ● ● ● EEOC ● ● ● Ex-Im ● ● FCA ● ● ● ● FCC ● ● FDIC ● ● ● ● FEC ● ● ● ● ● FERC ● ● ● ● ● ● FLRA ● ● ● ● FMC ● ● ● ● FTC ● ● ● ● MSPB ● ● ● ● ● NCUA ● ● ● ● NLRB ● ● ● ● NMB ● NRC ● ● ● NTSB ● ● ● ● ● PRC ● ● ● ● ● SEC ● ● ● ● STB / ICC ● ● ● ● ● ● USITC ● ● ● ● As Table 1 shows, these agencies vary considerably in terms of their particular mix of most of the other indicia of independence—except for the fact that all PBR agencies also feature statutorily mandated fixed terms for their commissioners or board members. That feature is remarkably common; among all executive agencies that are typically classified as independent, only three relatively minor agencies do not require set terms for their leaders.[footnote]Datla & Revesz, supra note 1, at 790 tbl.2 (listing the Pension Benefit Guaranty Corporation, U.S. Trade and Development Agency, and U.S. Agency for International Development as the only currently existing independent agencies without specified tenure for appointees).[/footnote] 2. Appointee Party Identification . — Having identified the set of PBR agencies for this study, we turn to determining the partisan affiliations of appointees to these agencies. For eleven agencies (the Consumer Product Safety Commission (CPSC), EEOC, FCC, FEC, FERC, FTC, NLRB, NTSB, NRC, SEC, and Surface Transportation Board (STB)), we obtained the name, date of nomination, and partisan identification for each individual appointed between 1979 and 2008 from an executive-appointee biographical dataset compiled by Professors George Krause and Anne Joseph O’Connell.[footnote]See Krause & O’Connell, Replication Data, supra note 84. Krause and O’Connell’s dataset reports whether each appointee “has [the] same or strongly presumed partisan affiliation as [the] appointing president.” See Anne Joseph O’Connell & George A. Krause, Variable Codebook for Manuscript & Supporting Information Document (2015), http://dataverse.harvard.edu/file.xhtml?fileId=2681274&version=1.2 (on file with the Columbia Law Review ); see also George A. Krause & Anne Joseph O’Connell, Compliance, Competence, and Bureaucratic Leadership in U.S. Federal Government Agencies: A Bayesian Generalized Latent Trait Analysis 6–7 (Sept. 19, 2011) (unpublished manuscript), http://pdfs.semanticscholar.org/9f96/d968768c293aad47539d38d2a0a977aa85a6.pdf?-_ga=2.225073178.1651911087.15054242832085157714.1505424283 [http://perma.cc/CB5E-MQMR] (describing the database).[/footnote] We then extended this analysis through 2014 via Congressional Research Service (CRS) reports containing the same information for appointments made between 2000 and 2014.[footnote]See generally Michael Greene, Cong. Research Serv., R43893, Presidential Appointments to Full-Time Positions on Regulatory and Other Collegial Boards and Commissions, 112th Congress (2015); Henry Hogue et al., Cong. Research Serv., R43238, Presidential Appointments to Full-Time Positions on Regulatory and Other Collegial Boards and Commissions, 111th Congress (2013); Henry B. Hogue & Maureen Bearden, Cong. Research Serv., R41463, Presidential Appointments to Full-Time Positions on Regulatory and Other Collegial Boards and Commissions, 110th Congress (2010); Henry Hogue et al., Cong. Research Serv., RL34744, Presidential Appointments to Full-Time Positions on Regulatory and Other Collegial Boards and Commissions, 109th Congress (2009); Henry Hogue et al., Cong. Research Serv., RL32742, Presidential Appointments to Full-Time Positions on Regulatory and Other Collegial Boards and Commissions, 108th Congress (2005); Henry B. Hogue, Cong. Research Serv., RL30910, Presidential Appointments to Full-Time Positions on Regulatory and Other Collegial Boards and Commissions, 107th Congress (2003); Rogelio Garcia, Cong. Research Serv., RL 30476, Presidential Appointments to Full-Time Positions on Regulatory and Other Collegial Boards and Commissions, 106th Congress (2001). We excluded from our analysis the nine appointees whom the CRS Reports classified as “independent.”[/footnote] To assess the reliability of both data sources, we examined the extent to which the two sources include identical information for appointees during the period from 2002 to 2008 for which the two sources overlap. Inter-observer reliability between the two data sources is extremely high; the appointee names, nomination dates, and partisan affiliations reported in the two sources in this period were identical. For the other twelve agencies, we built an original dataset that includes the name, date of nomination, and party affiliation for each individual appointed as a board member or commissioner from 1979 through 1999. We drew information on name and date of nomination from congressional records and determined partisan identifications from a variety of other sources, including newspaper reports around the time of nomination and obituaries of now-deceased appointees. Once again, we extended this analysis through 2014 via CRS reports. 3. Appointee Ideology . — Finally, we employ the Database on Ideology, Money in Politics, and Elections (DIME) to determine the ideological preferences of each appointee.[footnote]Adam Bonica, Codebook for the Database on Ideology, Money in Politics, and Elections (DIME) (Version 2.0) (2016) [hereinafter Bonica, Codebook for DIME], http://dataverse.harvard.edu/file.xhtml?fileId=2865308&version=2.2 (on file with the Columbia Law Review ).[/footnote] Developed by Adam Bonica, DIME contains over 130 million political contributions made by 14.7 million individuals and 1.7 million organizations to over 80,000 political action committees and candidates in federal, state, and local elections between 1979 and 2014.[footnote]Id. at 3 (providing statistics for the most recent update to DIME). Contribution data are derived from the FEC and several good-government organizations. See Bonica, Ideological Marketplace, supra note 11, at 370 (explaining where the records comprising the databases were collected from).[/footnote] DIME leverages these data to generate ideology-based scores, referred to as Campaign Finance Scores (CFscores), for all donors and recipients during this period.[footnote]The use of spatial scores to measure the preferences of political actors is well established in political science. See Bonica, Ideological Marketplace, supra note 11, at 368. Their application to legal scholarship has become increasingly common in recent years. See, e.g., Lee Epstein & Eric A. Posner, Supreme Court Justices’ Loyalty to the President, 45 J. Legal Stud. 401, 408–11 (2016) (using statistical analysis of voting by Supreme Court Justices to demonstrate a “loyalty effect” that suggests Justices feel “personal loyalty to the president” that appointed her or him, “not loyalty to the groups (like parties) or ideas (like the Constitution)”); Nicholas O. Stephanopoulos, Spatial Diversity, 125 Harv. L. Rev. 1903, 1905–07 (2012) (using spatial scores analysis to identify issues in political gerrymandering). In particular, DIME’s CFscores have been employed or cited in a wide variety of social-scientific and legal scholarship since their debut in 2013. See, e.g., Adam Bonica, Adam S. Chilton & Maya Sen, The Political Ideologies of American Lawyers, 8 J. Legal Analysis 277, 279 (2016) (employing DIME data to explore the ideology of American lawyers); Abby K. Wood & Douglas M. Spencer, In the Shadows of Sunlight: The Effects of Transparency on State Political Campaigns, 15 Election L.J. 302, 307–08 (2016) (using DIME data to determine that campaign disclosure requirements had a negligible effect in chilling speech and deterring political participation).[/footnote] The intuition behind DIME is simple: Individuals’ decisions to contribute to political campaigns constitute revealed preferences concerning their political views.[footnote]Bonica, Ideological Marketplace, supra note 11, at 367.[/footnote] The destination and amount of an individual’s campaign contributions disclose information regarding that individual’s ideological “ideal point.” DIME’s algorithm uses this information on all of an individual’s donations between 1979 and 2014 to place donors along a scale based on their history of campaign contributions.[footnote]For a technical explanation of DIME’s estimation strategy, see id. at 369–70. Essentially, DIME places donors along a unidimensional scale so as to minimize the distance between donors with similar patterns of political giving. DIME then places recipients along a unidimensional scale so as to minimize the distance between recipients with similar donor pools. That some donors also are recipients provides a “bridge,” enabling DIME to place donors and recipients on the same scale. Likewise, that some donors contribute to both state and federal campaigns and that some recipients run for both state and federal office provide additional bridge observations. These bridges enable DIME to create a single ideological scale for a massive number of individuals across thirty-six years of politics. See Adam Bonica & Michael J. Woodruff, A Common-Space Measure of State Supreme Court Ideology, 31 J.L. Econ. & Org. 472, 476–77 (2015) (presenting a “method to construct ideological measures for state Supreme Court justices from campaign finance records”). Bonica has subjected DIME to several tests to evaluate the validity of the measure. Concerning the measure’s external validity, CFscores for members of Congress correlate closely with the scores assigned to these legislators by DW-NOMINATE, a well-established method of estimating legislators’ ideal points based on their roll-call voting records. See Bonica, Ideological Marketplace, supra note 11, at 370–71 (reporting that r = 0.92 for the bivariate correlation between these two estimates). Concerning internal validity, for individuals that were active donors or recipients throughout the period from 1979 to 2014, the static CFscores derived from their activity during the entire period are closely correlated with CFscores derived from these individuals’ activity during select years within this range. See id. at 373 (reporting that R 2 = 0.97 for the bivariate correlation between these two estimates).[/footnote] The scale ranges from -2 (assigned to the most liberal donors) to 2 (assigned to the most conservative donors).[footnote]See Bonica, Ideological Marketplace, supra note 11, at 369.[/footnote] The mean CFscore for donors is 0 and the standard deviation is 1.[footnote]Id.[/footnote] To obtain each appointee’s CFscore, we searched the DIME project’s dataset on appointees to federal agencies,[footnote]Adam Bonica, Executive Appointees to Federal Agencies, Database on Ideology, Money in Politics, and Elections (DIME) (2015) [hereinafter Bonica, Executive Appointees, DIME], http://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/ DVN/O5PX0B [http://perma.cc/DKY4-HLEJ]; see also Adam Bonica, Jowei Chen & Tim Johnson, Senate Gate-Keeping, Presidential Staffing of “Inferior Offices,” and the Ideological Composition of Appointments to the Public Bureaucracy, 10 Q.J. Pol. Sci. 5, 7 (2015) [hereinafter Bonica, Chen & Johnson, Senate Gate-Keeping] (using a “formal model that examines how Senate committees and the presidential staffing of ‘inferior offices’ affect the ideological distribution of appointed public bureaucrats”).[/footnote] as well as the project’s full database on all contributors to federal, state, and local elections between 1979 and 2014.[footnote]Bonica, Codebook for DIME, supra note 98.[/footnote] For individuals with common names or names with multiple entries in the full database, we cross-checked the individual’s profession, employer, and place of residence as reported in DIME with the appointee’s biographical information available via online sources—For example, biographical sketches on current employers’ websites, LinkedIn profiles, or obituaries—to determine whether the individual located in DIME and the appointee were the same person. This method yielded CFscores for 80% of appointees (578 of 722) to the twenty-three PBR agencies[footnote]See supra note 83 and accompanying text.[/footnote] we identified. This 80% match rate is comparable to the success rates of similar studies that employ DIME to determine the ideologies of members of other elite groups.[footnote]See Adam Bonica, Avenues of Influence: On the Political Expenditures of Corporations and Their Directors and Executives, 18 Bus. & Pol. 367, 375 (2016) (finding an 83% match rate for Fortune 500 CEOs and board members); Adam Bonica, Adam S. Chilton, Jacob Goldin, Kyle Rozema & Maya Sen, The Political Ideologies of Law Clerks, 19 Am. L. & Econ. Rev. 96, 105 (2017) (reporting a 66% match rate for former Supreme Court clerks); Bonica, Chen & Johnson, Senate Gate-Keeping, supra note 105, at 20 (reporting a 72% match rate for a subset of appointees under Presidents Clinton and George W. Bush).[/footnote] Readers may wonder whether campaign contributions truly capture donors’ political preferences. Donors may instead choose to give for a range of strategic reasons—for example, to encourage the election of politicians sympathetic to the donors’ business interests or to curry favor with existing officeholders in exchange for access, votes, or other services.[footnote]See Sanford C. Gordon, Catherine Hafer & Dimitri Landa, Consumption or Investment? On Motivations for Political Giving, 69 J. Pol. 1057, 1058 (2007) (describing donors’ varied potential motives).[/footnote] This critique may seem particularly powerful for our research design, given that all donors in our sample actually did receive something of value—an executive appointment—at some point during the thirty-six-year period in which they made at least one campaign contribution. Our response to this concern is twofold. First, we rely on a large body of political science research finding that individuals make political contributions sincerely—not strategically.[footnote]See, e.g., Bonica, Ideological Marketplace, supra note 11, at 370; Michael J. Ensley, Individual Campaign Contributions and Candidate Ideology, 138 Pub. Choice 221, 230 (2009); Nolan McCarty & Lawrence S. Rothenberg, Commitment and the Campaign Contribution Contract, 40 Am. J. Pol. Sci. 872, 875–81 (1996).[/footnote] Survey results indicate that donors consider the recipient’s ideology to be the most important factor when deciding whether to give, placing ideology far above more strategic considerations such as electability, incumbent status, and ability to influence the candidate.[footnote]See Michael Barber, Donation Motivations: Testing Theories of Access and Ideology, 69 Pol. Res. Q. 148, 154 & tbl.1 (2016).[/footnote] Accordingly, the dominant view among political scientists considers campaign contributions as a form of consumption among “political hobbyists” rather than an investment by strategic actors.[footnote]See Stephen Ansolabehere, John M. de Figueiredo & James M. Snyder, Jr., Why Is There So Little Money in U.S. Politics?, 17 J. Econ. Persp. 105, 105 (2003) (concluding that campaign contributions are best understood not as political investment but as a type of consumption good). See generally Eitan Hersh, Political Hobbyism: A Theory of Mass Politics (Mar. 17, 2017) (unpublished manuscript), http://www.eitanhersh.com/ uploads/7/9/7/5/7975685/hersh_theory_of_hobbyism_v2.0.pdf (on file with the Columbia Law Review ) (arguing that for many Americans, “political participation is not appropriately described as motivated by duty, but is more akin to a hobby”).[/footnote] Second, insofar as a divergence between donors’ ideological preferences and the ideal points reflected by their campaign contributions introduces measurement error into our analysis, we find it hard to explain why such a divergence would account for the change over time that we document in the next section. (Below we consider whether the easy observability of campaign contributions in the Internet age might affect our results, and we find strong evidence that it does not.[footnote]See infra notes 166–168 and accompanying text.[/footnote]) B. Results 1. Descriptive Statistics . — With these data on appointee ideology and partisan identification in hand, we turn to analyzing the effects of PBRs on the composition of agency leadership ranks. Table 2 provides an overview of the appointees included in our data: 578 appointees to twenty-three agencies over thirty-six years. Table 2: Overview Appointees by President Appointing President Co-Party Appointees Cross-Party Appointees Total Appts. Appts. per Year Total Appts. Appts. per Year All (1979–2014) 362 6.6 216 4.0 Democratic Presidents 155 5.9 96 3.8 Republican Presidents 207 7.2 120 4.3 Carter 11 5.5 9 4.5 Reagan 56 7.0 38 4.8 G.H.W. Bush 39 9.75 17 4.3 Clinton 75 9.4 48 6.0 G.W. Bush 112 14.0 65 8.2 Obama 69 11.5 39 6.5 n = 578. Includes appointees to twenty-three agencies from 1979 to 2014. Carter years include 1979 to 1980 only; Obama years include 2009 to 2014 only. Unsurprisingly, all Presidents during this period tended to appoint their fellow party members to PBR agencies; 63% of appointees (362 of 578) belonged to the same party as the appointing President.[footnote]Interestingly, the proportion of appointees that share the President’s party label almost perfectly mirrors the no-more-than-three-of-five requirement that is typical among independent agencies. See Datla & Revesz, supra note 1, at 776.[/footnote] This tilt is not the result of Presidents disregarding PBR statutes.[footnote]The imbalance was greatest during the presidency of George H.W. Bush, who was not known for his sharp partisan elbows. See Jon Meacham, Destiny and Power: The American Odyssey of George Herbert Walker Bush 390 (2015).[/footnote] Rather, most partisan balance statutes do not require perfect balance but permit a bare majority of members to be affiliated with the same political party—one possible explanation for the tilt.[footnote]See, e.g., 15 U.S.C. § 78d(a) (2012) (mandating that no more than three out of the five SEC commissioners be affiliated with the same party); id. § 2053(c) (mandating that no more than three out of the five CPSC commissioners be affiliated with the same party). But see, e.g., 52 U.S.C. § 30106(a)(1) (Supp. II 2015) (requiring exact partisan balance on the FEC).[/footnote] Another contributor to the imbalance may be that co-partisan appointees strategically retire before the end of their terms when a friendly President is in office (leading to more openings for co-party slots during any given presidency).[footnote]Cf. Lee Epstein & Jeffrey A. Segal, Advice and Consent: The Politics of Judicial Appointments 37 (2005) (summarizing studies showing that appellate judges strategically time their retirements to increase the likelihood that like-minded successors will replace them). Note that the number of appointees per year who appear in DIME is somewhat lower for the Carter and Reagan Administrations than for later presidencies. We attribute this trend to three factors. First, public disclosure of federal campaign contributions did not begin until 1972, and DIME’s coverage starts in 1979. See Federal Election Campaign Act of 1971, Pub. L. No. 92-225, § 304, 86 Stat. 3, 14–15 (1972) (codified as amended at 52 U.S.C. § 30104). Thus, Carter and Reagan appointees who made their last campaign contributions before the late 1970s will generally not appear in DIME. Second, the $200 reporting threshold has not changed since 1980, despite intervening inflation. See Federal Election Campaign Act Amendments of 1979, Pub. L. No. 96-187, sec. 104, § 10(b)(3), 93 Stat. 1339, 1352 (1980) (codified as amended at 52 U.S.C. § 30104). Individuals who made campaign contributions in more recent years thus are more likely to exceed the threshold and trigger a public disclosure. Finally, one agency was established partway through the period from 1979 to 2014, resulting in a small increase in the number of positions to be filled. See National Defense Authorization Act, Fiscal Year 1989, Pub. L. No. 100-456, § 1441(a), 102 Stat. 1918, 2076 (1988) (codified as amended at 42 U.S.C. § 2286 (2012)) (establishing the Defense Nuclear Facilities Safety Board). Excluding the Defense Nuclear Facilities Safety Board from the analysis does not materially alter any of our results.[/footnote] Table 3 contains descriptive statistics concerning the ideologies of appointees to PBR agencies, arranged by President and by whether the appointee and appointing President shared a partisan affiliation. For each President, Table 3 also reports differences in the mean ideology of co-party versus cross-party appointees.[footnote]The final column in Table 3 reports the p value for a series of two-sample Kolmogorov-Smirnov (KS) tests for differences in distributions. Low p values counsel in favor of rejecting the null hypothesis that the distribution of CFscores for co-party appointees and the distribution of CFscores for cross-party appointees were drawn from the same underlying distribution. All of the reported p values meet—and, for all but the Carter appointees, far exceed—the conventionally accepted p Table 3: Appointee Ideology, by President Appointing President Mean CFscore (Standard Deviation) Diff. in Means [footnote]Difference in means refers to the absolute value of the difference in mean CFscore for co-party appointees and mean CFscore for cross-party appointees.[/footnote] (95% Conf. Interval) KS Test p-value Co-Party Appointees Cross-Party Appointees Democratic Presidents -0.913 (0.521) 0.774 (0.535) 1.687 (1.549, 1.826) 0.000 Republican Presidents 0. 848 (0.368) -0.492 (0.808) 1.340 (1.182, 1.497) 0.000 Carter -0.494 (0.749) 0.301 (0.828) 0.795 (0.011, 1.601) 0.047 Reagan 0.797 (0.539) -0.005 (0.794) 0.802 (0.496, 1.108) 0.000 G.H.W. Bush 0.846 (0.422) -0.162 (0.900) 1.008 (0.497, 1.520) 0.000 Clinton -0.875 (0.408) 0.747 (0.494) 1.622 (1.447, 1.798) 0.000 G.W. Bush 0.873 (0.213) -0.830 (0.615) 1.703 (1.546, 1.861) 0.000 Obama -1.015 (0.563) 0.902 (0.458) 1.917 (1.719, 2.116) 0.000 n = 578. Includes appointees to twenty-three agencies from 1979 to 2014. Absolute values of differences in means between co-party and cross-party appointees reported. Differences in means calculated via Welch’s t-tests, appropriate for samples with skewed distributions drawn from independent populations. Carter years include 1979 to 1980 only; Obama years include 2009 to 2014 only. During the entire period from 1979 to 2014, the mean CFscore for all appointees was 0.078 (sd = 0.961). The key takeaway from Table 3 emerges from the column displaying the difference in means between co-party and cross-party appointees. During the Carter Administration, the point estimate for the ideological gap between co-party and cross-party appointees to PBR agencies is 0.795, which indicates a modest ideological difference (less than one standard deviation) between Carter’s co-party and cross-party appointees.[footnote]See Bonica, Ideological Marketplace, supra note 11, at 369 (stating that the mean CFscore for donors is zero and the standard deviation is one).[/footnote] This gap grows steadily in subsequent administrations, reaching 1.917 in the Obama years—almost two-and-one-half times the size of the gap under Carter. The widening occurs over Democratic as well as Republican administrations. Accordingly, at first glance the most notable aspect of this difference is its growth over time, rather than its association with any particular party or presidency. 2. Co-Party Versus Cross-Party Appointees . — What do the ideologies of appointees to PBR agencies look like? More specifically, how do the views of individuals appointed to Democratic seats by Democratic Presidents differ from those appointed to Democratic seats by Republican Presidents? And are the patterns similar for Republican appointees? To answer these questions, Figure 3 compares the ideological distributions of all four categories of nominees. Democratic officials appointed by Democratic Presidents appear in dark blue, whereas Democrats appointed by Republicans appear in light blue; likewise, Republicans appointed by Republicans appear in dark red, whereas Republicans appointed by Democrats appear in light red. Because, for this analysis, the focus is on recent appointment behavior, the figure is limited to the period from 2000 to 2014. Figure 3: Ideologies of Appointees to PBR Agencies, 2000–2014 CFscore

n = 347. Includes appointees made during the period from 2000 to 2014 to twenty-three PBR agencies. Figure 3 reveals a striking feature of recent appointments to PBR agencies: Democrats appointed by Democratic Presidents have views virtually identical to those of Democrats appointed by Republican Presidents, and the same holds true of Republican appointees. Presidents Clinton and Obama did not name liberal Republicans to cross-party positions on PBR agencies; neither did President George W. Bush appoint conservative Democrats to cross-party seats. Instead, recent Presidents abided by the spirit of PBRs in those agencies’ organic statutes. Presidents’ willingness to appoint their ideological opponents to cross-party seats appears to be a recent development. The following several figures compare the ideologies of co-party and cross-party appointees beginning in 1979. Figure 4(a) displays the ideological distance between the CFscore of the appointing President and the mean CFscore of that President’s appointees to PBR agencies. For a Democratic President, a negative value indicates that the mean appointee is more liberal than the appointing President, and a positive value indicates that the mean appointee is more conservative. For a Republican President, the reverse is true: Negative values indicate that the mean appointee is more conservative than the appointing President, and positive values indicate that the appointee is more liberal. Figure 4(b) reports similar information by year. The bars emanating from the point estimates in both figures denote one standard deviation in each direction from the relevant mean. Figure 4(a): Ideological Distance Between Appointees to PBR Agencies and Their Appointing Presidents President

n = 578. Point estimates, denoted as solid shapes, signify mean distance between appointees’ CFscore and the appointing President’s CFscore. The solid circles represent the mean CFscore for co-party appointees, whereas the solid triangles represent the mean CFscore for cross-party appointees. Positive values for appointees in a Democratic (Republican) administration signify that the mean appointee is more conservative (liberal) than the appointing President. Vertical bars extend one standard deviation above and below the mean. The appointing President’s CFscore is standardized to y = 0. Carter years include 1979 to 1980 only; Obama years include 2009 to 2014 only. Figure 4(b): Ideological Distance Between Appointees to PBR Agencies and Their Appointing Presidents, by Year Year

n = 578. Point estimates, denoted as solid shapes, signify mean distance between appointees’ CFscore and the appointing President’s CFscore. The solid circles represent the mean CFscore for co-party appointees, whereas the solid triangles represent the mean CFscore for cross-party appointees. Positive values for appointees in a Democratic (Republican) administration signify that the mean appointee is more conservative (liberal) than the appointing President. Vertical bars extend one standard deviation above and below the mean. The appointing President’s CFscore is standardized to y = 0. The point estimates for Figures 4(a) and 4(b) show that cross-party appointees are more conservative than co-party appointees across all Democratic administrations and more liberal than co-party appointees in all Republican administrations. This gap increases, in fits and spurts, throughout the period of study. During the Carter Administration, for instance, the CFscore for the mean appointee to a Democratic seat was 0.159 points to the left of President Carter’s CFscore, whereas the mean appointee to a Republican seat had a CFscore that was 0.636 points to the right of President Carter—a gap of 0.795 points. By the Obama Administration, the mean Democratic appointee was 0.637 points to the right of President Obama, while the mean Republican appointee was 2.471 points to the right of President Obama—a gap of 1.834 points. In other words, the ideological gap between co-party and cross-party appointees within the same party widened over time. The associated standard deviations in both figures provide further support for this finding. In Figure 4(a), the standard deviation bars for co-party and cross-party appointees overlap during the Carter, Reagan, and George H.W. Bush Administrations, indicating that a sizable number of co-party and cross-party appointees are ideologically similar. The bars then pull apart in the Clinton, George W. Bush, and Obama years, signifying significantly less ideological overlap between co-party and cross-party appointees during these presidencies. The standard deviation bars in Figure 4(b) tell a similar story. Figures 4(a) and 4(b) appear to demonstrate a time trend, with the ideological distance between cross-party appointees and their appointing President increasing over time. Figure 5, below, provides a better sense of this time trend. The figure plots the difference in ideological score between the appointing President and cross-party appointees over the thirty-six-year study period and includes a linear regression line. The positive slope suggests the presence of a trend component. Figure 5: Time Trend in Ideological Distance Between the Appointing President and Cross-Party Appointees The supply-side theory suggests that we should observe steady growth in the ideological gap between cross-party appointees and their appointing President over time, consistent with the pattern of partisan sort shown in Figure 2. By contrast, the supply-side theory gives us no reason to expect any similar changes over time in the gap between co-party appointees and their appointing President. None of the six Presidents in our study period (Carter through Obama) exhibited ideological preferences that were radically out of step with the mainstream of his own party. Accordingly, we would expect these Presidents to be able to find co-partisans whose views roughly matched their own. Visually, Figures 4 and 5 align with the supply-side theory’s predictions: Cross-party appointees have grown further apart from the appointing President, while co-party appointees have not. 3. Ideological Consistency . — Our analysis thus far treats appointees’ ideologies as time-invariant. The DIME master database assigns each donor a single CFscore based on that individual’s total contributions during the period from 1979 to 2014, and we adopt this measure as a proxy for ideology in the analysis above. Yet we are mindful that individuals may evolve in their thinking, or that they may donate strategically prior to their appointment so as to send a signal of ideology that differs from their true preferences. Fortunately, the available data allow us to observe whether donation patterns change post-appointment in ways that would suggest either ideological drift or strategic giving. Although ideological self-categorization is stable for most people throughout their lives,[footnote]Donald Green, Bradley Palmquist & Eric Schickler, Partisan Hearts & Minds: Political Parties and the Social Identities of Voters 28 (2002).[/footnote] it may be less consistent among members of the political class from which appointees are drawn. The notion that elite actors display ideological fluidity over their careers is familiar to students of judicial behavior. Observers of the Supreme Court have pointed to a so-called “Greenhouse effect”—a trend of Supreme Court Justices “drift[ing] away from the conservatism of their early votes” and toward the more liberal preferences of “cultural elites.”[footnote]See Lawrence Baum & Neal Devins, Why the Supreme Court Cares About Elites, Not the American People, 98 Geo. L.J. 1515, 1518 (2010). The “Greenhouse” in question is Linda Greenhouse, the longtime Supreme Court correspondent for the New York Times. The term can be traced back to a 1992 speech by Judge Laurence Silberman, a Reagan appointee to the U.S. Court of Appeals for the D.C. Circuit, who claimed that Greenhouse’s reporting pressured Justices to take more liberal stances in order to garner favorable coverage from the Times . See Martin Tolchin, Press Is Condemned by a Federal Judge for Court Coverage, N.Y. Times (June 15, 1992), http://www.nytimes.com/1992/ 06/15/us/press-is-condemned-by-a-federal-judge-for-court-coverage.html (on file with the Columbia Law Review ).[/footnote] While the Greenhouse effect generally refers to the evolution of appointees’ sincere preferences over time, a related possibility is that prospective appointees may misrepresent their preferences prior to confirmation and then reveal their true selves after they are sworn in.[footnote]On “stealth nominees” at the Supreme Court, see Michael Comiskey, The Supreme Court Appointment Process: Lessons from Filling the Rehnquist and O’Connor Vacancies, 41 PS: Pol. Sci. & Pol. 355, 357 (2008).[/footnote] For example, a Democratic President may “sell” a nominee to the Senate as a genuine Republican, but the individual may turn out to be—after confirmation—a liberal who registered as a Republican and donated to conservative candidates for the purpose of positioning himself for a potential cross-party appointment. To return to our market metaphor, a potential appointee who looks like a peach from the opposition party’s perspective may turn out to be a lemon—and consumers (here, senators) have incomplete information regarding which nominees fall into which of these categories until after the transaction. To test the ideological consistency of appointees over time, we examined the donation patterns of commissioners in the eight years before and after their initial appointments. We looked at commissioners’ cycle-specific scores derived from the biennial DIME contribution datasets, which include CFscores for donors in each two-year election cycle.[footnote]We thank Adam Bonica and Kyle Rozema for their assistance in this analysis.[/footnote] Naturally, these files are less comprehensive than the DIME master database for the period from 1979 to 2014; to be included in one of the election cycle-specific files, an individual must donate to a sufficient number of candidates or PACs to generate a CFscore for that cycle. Creating cycle-specific scores enables us to chart commissioners’ ideal point estimates over time and identify any changes in behavior around the time of their initial appointments. Here, we display results for the FCC, which in several respects is the archetypical agency with a PBR: The FCC has the modal number of commissioners (five), the modal term length (five years), and, like many of the twenty-three agencies in our study, was created during the New Deal era. Figure 6 displays our results for both Democratic and Republican appointees selected by both Democratic and Republican Presidents. To compare commissioners’ ideal point estimates before and after appointment, the x axis in both figures denotes the number of years since each commissioner was first appointed to the FCC (negative numbers thus represent pre-appointment years). Figure 6: Longitudinal Change in FCC Commissioners’ Ideal Point Estimates Years Since Initial Appointment

If liberals registered as Republicans and donated to conservative candidates so they could be sold by a Democratic President to Senate Republicans as genuine cross-party appointees, then we would expect to see the light red line (Democratic President, Republican appointee) trending downward in later years, as appointees who posed as conservatives pre-confirmation revealed their true liberal selves. Likewise, if conservatives registered as Democrats and donated to liberal candidates so they could be sold by a Republican President to Senate Democrats, then we would expect to see the light blue line (Republican President, Democratic appointee) trending upward in later years, as appointees who posed as liberals revealed their true conservative selves. And if appointees tended to drift in a single direction (a Greenhouse effect at multimember agencies), then we would expect the lines to be roughly parallel with similar slopes. We observe none of these phenomena in Figure 6. For the most part, commissioners display consistent ideologies over time. Concerning those commissioners for whom we have ideal point estimates both before and after their initial appointments, we see only slight, inconsistent movement toward more extreme positions post-appointment. 4. PBR Agencies Versus the Rest of the Executive Branch . — To provide broader context for this secular divergence in the ideological composition of PBR agencies, we compare ideological preferences of appointees to PBR agencies with the ideological preferences of other high-level executive branch policymakers.[footnote]High-level officials encompass heads and commissioners of other independent agencies, department secretaries, and second- and third-level leaders (typically deputy secretaries and assistant secretaries or undersecretaries) in executive departments. All positions require Senate confirmation.[/footnote] Placing PBR agency appointees and other officeholders on the same scale allows us to assess the extent to which the observed trend in PBR agencies reflects or departs from changes in the ideological composition of executive appointees writ large. For this analysis, we leverage a dataset created by Professors Adam Bonica, Jowei Chen, and Tim Johnson, which contains DIME-based CFscores for all available executive appointees between the Reagan and Obama Administrations.[footnote]Bonica, Executive Appointees, DIME, supra note 105; see also Bonica, Chen & Johnson, Senate Gate-Keeping, supra note 105, at 15. To allow for longitudinal comparisons across the entire period, we exclude agencies that operated during only part of this period, for example, the African Development Foundation, Corporation for National and Community Service, Federal Hospital Insurance Trust Fund, and U.S. Information Agency.[/footnote] The solid curves in Figures 7(a)–(e) depict the ideological distributions of PBR appointees across five presidential administrations; the dashed curves depict these distributions for other high-level, Senate-confirmed officials.[footnote]Figures 7(a)–(e) were generated using the CFscores for 2,318 appointees.[/footnote] Figures 7(a)–(e): Ideological Distributions of Appointees to PBR Agencies and Other Agencies Figure 7(A): Reagan CFscore

Figure 7(B): G.H.W Bush CFscore

Figure 7(C): Clinton CFscore

Figure 7(D): G.W. Bush CFscore

Figure 7(E): Obama CFscore

Figures 7(a) through 7(e) show a remarkable development. Figure 7(a) reports that during the Reagan Administration, the ideological distribution of appointees to PBR agencies—including both co-party and cross-party appointees—maps closely onto the ideological distribution of high-level executive appointees. Both distributions are unimodal, clustered around a conservative President’s own CFscore. Figure 7(b) provides a similar picture for the George H.W. Bush Administration. A change, however, is discernible in the Clinton years. Although most Clinton appointees to both PBR agencies and other high-level executive positions are clustered around the same left-of-center mode, a right-of-center local maximum has begun to develop—but only for appointees to PBR agencies. This bimodality becomes more pronounced in the George W. Bush Administration (naturally, with the locations of the global and local maxima flipping, as the White House switches from Democratic to Republican control). The trend continues in the Obama Administration; by the Obama years, the distribution of appointees to PBR agencies is almost completely bimodal, with a cluster of appointees located approximately at -1 and a second cluster—only slightly smaller—roughly centered around +1. By contrast, the distribution of other high-level appointees remains single-peaked, with its mode located near President Obama’s CFscore. To translate these trends from picture to prose: Our findings suggest that PBRs had very little effect on the ideological composition of agencies during the Reagan and George H.W. Bush Administrations. The ideological distribution of PBR agency appointees closely matches the distribution of appointees to positions not covered by PBRs. Starting with the Clinton Administration, however, a distinct PBR effect appears. President Clinton appointed a fair number of conservatives to PBR agencies but named very few conservatives to other posts. The PBR effect grows even more pronounced under President George W. Bush and President Obama. These results confirm our thesis that PBRs do indeed bite, but that they have come to bite only relatively recently. 5. Statutory Versus Informal PBRs . — Our analysis thus far has lumped the NLRB with other PBR agencies, even though the NLRB is not subject to a formal PBR. The House in 1947 passed legislation that would have added a formal partisan balance requirement to the statute governing NLRB membership, but the party balance provision was (for reasons not explained in the legislative history[footnote]H.R. Rep. No. 80-510, at 36–37 (1947) (Conf. Rep.) (noting the party balance provision was included in the House bill but not in the conference agreement, and offering no explanation for this omission).[/footnote]) omitted from the version of the bill that passed the Senate and became law.[footnote]See Labor Management Relations Act, 1947, Pub. L. No. 80-101, sec. 101, § 3(a), 61 Stat. 136, 139.[/footnote] Nonetheless, there has been a “tradition” since the Eisenhower years that Presidents have filled no more than three of the NLRB’s five seats with members of their own party.[footnote]See Matthew M. Bodah, Congress and the National Labor Relations Board: A Review of the Recent Past, 22 J. Lab. Res. 699, 700 (2001) (discussing the unwritten tradition of bipartisanship and restraint that characterized appointments to the NLRB); James J. Brudney, Isolated and Politicized: The NLRB’s Uncertain Future, 26 Comp. Lab. L. & Pol’y J. 221, 244 & n.109 (2005) (“Notwithstanding this silence, a tradition has developed of appointing both Democrats and Republicans to the Board, with the President’s party holding a three-to-two majority of the seats and also the chair.”). The last time that the NLRB had more than three members from the same party was August 1956, when four Republicans and one Democrat served on the Board; President Eisenhower then named a Democrat to replace a retiring Republican. See Members of the NLRB Since 1935, NLRB, http://www.nlrb.gov/who-we-are/board/members-nlrb-1935 [http://perma.cc/L7GB-PTN3] (last visited Oct. 2, 2017).[/footnote] The informal partisan balance requirement at the NLRB is an example of what Professor Adrian Vermeuele has called a “convention of agency independence”: an unwritten norm that has come to govern the behavior of political actors with respect to the agency.[footnote]See Adrian Vermeule, Conventions of Agency Independence, 113 Colum. L. Rev. 1163, 1166 (2013).[/footnote] The case of the NLRB allows us to examine whether a partisan balance requirement that emerges from convention has the same effect on ideological composition as a PBR set forth by statute. Table 4 compares the ideologies of appointees to the NLRB with the ideologies of appointees to the twenty-two agencies with statutory PBRs for which we have obtained data from 1979 to 2014. Under Democratic Presidents, cross-party appointees to the NLRB appear, on average, to be slightly less conservative than cross-party appointees to agencies with statutory PBRs, but this difference falls far short of statistical significance ( p = 0.945). Likewise, under Republican Presidents, cross-party appointees to the NLRB appear on average to be slightly less liberal than cross-party appointees to agencies with statutory PBRs, but again, this difference is not statistically significant ( p = 0.849). In short, we find little to suggest that the difference between the NLRB’s partisan balance convention and the statutory PBRs applicable to other agencies has any effect on ideological composition—the NLRB looks much like the statutory PBR agencies in terms of the ideological preferences of appointees. Table 4: Mean Ideology of Appointees to NLRB and Statutory PBR Agencies NLRB Statutory PBR Agencies Co-Party Appointees Cross-Party Appointees Co-Party Appointees Cross-Party Appointees Democratic Presidents -1.106 (0.309) 0.753 (0.774) -0.898 (0.531) 0.776 (0.520) Republican Presidents 0.681 (0.760) -0.341 (1.052) 0.863 (0.309) -0.500 (0.797) n = 578. Includes appointments made to the NLRB and twenty-two other agencies during the period from 1979 to 2014. Cells report mean CFscores and, in parentheses, standard deviations. Indeed, when we focus on twenty-first century appointments, we find evidence that the partisan balance convention at the NLRB has as much, if not more, bite than statutory PBRs elsewhere. Since 2000, Republicans appointed by Democratic Presidents to the NLRB have been, on average, even more conservative—and Democrats appointed by Republican Presidents to the NLRB even more liberal—than those Presidents’ cross-party appointees to agencies with statutory PBRs.[footnote]Further, when the analysis is limited from 2000 to 2014, the results also lend support to the conventional wisdom that the NLRB is an unusually politicized agency. See Brudney, supra note 131, at 223–24; Samuel Estreicher, ‘Depoliticizing’ the National Labor Relations Board: Administrative Steps, 64 Emory L.J. 1611, 1613 (2015) (explaining that, in reference to the NLRB, “the perception of a ‘politicized’ agency seems stronger than ever”); Julius G. Getman, The NLRB: What Went Wrong and Should We Try to Fix It?, 64 Emory L.J. 1495, 1496 (2015) (“The Board has become a controversial, often politicized, agency whose best efforts are denounced by politicians and often overruled by the Supreme Court.”). For each President–appointee combination, the mean NLRB appointee is more extreme than the corresponding cell in the table for the other agencies with PBRs.[/footnote] This conclusion, however, comes with an important caveat. Our dataset includes CFscores for only six cross-party appointees—three Democrats appointed by President George W. Bush and three Republicans appointed by Presidents Clinton or Obama—between 2000 and 2014. This relative lack of data on cross-party appointees to the NLRB may not be coincidental. The Board has suffered from persistent, politically motivated vacancies during the past several administrations.[footnote]See Mark Landler & Steven Greenhouse, Vacancies and Partisan Fighting Put Labor Relations Agency in Legal Limbo, N.Y. Times (July 15, 2013), http:// www.nytimes.com/2013/07/16/us/politics/vacancies-and-partisan-fighting-put-labor-relations-agency-in-legal-limbo.html?mcubz=0 (on file with the Columbia Law Review ) (describing the Senate’s refusal to confirm President Bush’s and then President Obama’s nominees).[/footnote] Whereas Congress permits appointees to other agencies to serve beyond the expiration of their terms until their successors are confirmed,[footnote]See, e.g., 7 U.S.C. § 2(a)(2)(A) (2012) (providing that a CFTC commissioner “shall hold office for a term of five years and until his successor is appointed . . . except that he shall not so continue to serve beyond the expiration of the next session of Congress subsequent to the expiration of said fixed term of office”); 15 U.S.C. § 78d(a) (2012) (providing similar restrictions on term lengths for the SEC); 42 U.S.C. § 7171(b)(1) (2012) (FERC); 47 U.S.C. § 154(c) (2012) (FCC).[/footnote] NLRB members are not authorized to serve in an acting capacity.[footnote]See 5 U.S.C. §§ 3349b, 3349c(1)(A) (2012) (specifying that appointees to multimember independent agencies cannot serve in a holdover or acting position unless otherwise statutorily authorized); 29 U.S.C. § 153(a) (2012) (specifying that there is no provision in the NLRB’s organic statute for continuation of service beyond the length of the fixed term).[/footnote] That NLRB seats remain vacant until filled may provide senators with an additional incentive to delay acting on nominations to the Board (beyond their usual incentives to do so for agencies when such delay leaves the existing appointees temporarily in place). Accordingly, it is possible that partisan gamesmanship is occurring with respect to cross-party appointees to the NLRB—it just occurs at an earlier stage in the appointments process and thus is not captured by the CFscores for those individuals that make it onto the Board. In sum, the partisan balance norm that exists by “convention” or “tradition” at the NLRB appears to be driving a gap between co-party and cross-party appointees that is broadly similar to the gap between co-party and cross-party appointees at agencies with statutory PBRs. The bite of PBRs does not appear to depend on whether these provisions are written into law. But whether we are dealing with a statutory PBR or a PBR that arises from convention, the puzzle remains as to why these requirements do anything at all to shape the ideological composition of multimember agencies. The next section turns to that question and to the implications of our results for various theories of PBRs’ efficacy. C. Market Forces and Appointee Selection This section reevaluates this Article’s two main causal explanations for the efficacy of PBRs—the supply-side theory and the demand-side theory—in light of our empirical analysis. It also considers (and largely rejects) alternative hypotheses that might account for the topline results and time trends observed above. 1. Supply-Side Constraints . — Our results are broadly consistent with the supply-side theory, which holds that Presidents are constrained in their selection by the ideological composition of the pool of potential cross-party appointees. As the political class from which the memberships of boards and commissions are typically drawn becomes more ideologically polarized along party lines, it becomes increasingly difficult for a President to identify competent individuals whose ideological preferences track the President’s own but whose party affiliation does not. Presidents are reactive in this narrative; they appoint bona fide cross-party members because these are the only competent potential appointees that “the market” supplies. The time trends in Figures 4(a) and 4(b) support this theory. The appointment of bona fide cross-party members gradually becomes more prevalent from the late 1970s through the mid-2010s, while simultaneously, the political classes gradually became more ideologically distinct along party lines. While it might seem surprising that a motivated Democratic President cannot find a liberal Republican for a cross-party seat or that a motivated Republican President cannot find a conservative Democrat, the supply-side theory becomes more plausible if one supposes that Presidents care about competence as well as ideological alignment. To be sure, a Democratic President might still be able to find someone in a nation of more than 300 million people who shares the President’s liberal views but registers as an independent or a Republican, and a Republican President might still be able to find someone who shares that President’s conservative views but registers as an independent or a Democrat. Yet while we observe past Presidents drawing cross-party appointees from the pools of liberal Republicans and conservative Democrats when these pools were much larger, we see relatively few cross-party appointments of these types in recent years. 2. Demand-Side Constraints . — The demand-side theory, by contrast, finds much weaker support here. Recall that this theory views senators as the “consumers” in the confirmation market to whom the President must “sell” her favored nominees. One could imagine that Democratic Presidents may want to appoint liberal Republicans to Republican-designated seats and that Republican Presidents may have the inverse desire, but that opposition-party senators use their advice-and-consent authority to block the appointments of moderates to party-specific seats. The demand-side theory therefore holds that the presence of bona fide cross-party commissioners on PBR agencies is attributable to pressure from cross-party senators. If this narrative is accurate, one would expect Presidents to appoint genuine cross-partisans to cross-party seats only when cross-party senators are sufficiently powerful to compel Presidents to do so. By contrast, when cross-party senators do not have enough votes to block a party-member-in-name-only nominee, Presidents will have free rein to appoint milquetoast partisans to cross-party seats. Essentially, one would expect the ideological distance between appointing Presidents and cross-party appointees to be positively correlated with cross-party senators’ power to deny confirmation to disfavored nominees. When do cross-party senators have the power to deny confirmation? One possibility is that the ability to block cloture—and thus deny a vote on a nominee—provides cross-party senators with a means of ensuring that the Presidents appoint bona fide cross-party members to PBR agencies. This explanation, however, cannot explain the time trend in Figures 4(a) and 4(b)—that PBRs have come to exert stronger effects over time. During virtually the entire study period, both parties held a sufficient number of seats to block a nominee under Senate filibuster rules.[footnote]For almost the entire period from 1979 to 2014, confirmation of executive branch nominees required the assent of sixty senators, which is the number needed to invoke cloture under Senate Rule XXII. See Standing Rules of the Senate, S. Doc. No. 113-8, Rule XXII, at 15–17 (2013) (providing that three-fifths of senators are required to invoke cloture, i.e., to vote to limit further consideration of a pending question, thereby permitting the question to be put to a majority vote). On November 21, 2013, the Senate reinterpreted Rule XXII to allow fifty-one votes to invoke cloture for nominees to positions other than the Supreme Court. Walter J. Oleszek, Cong. Research Serv., R44709, Changing the Senate Cloture Rule at the Start of a New Congress 8–9 (2016), http://fas.org/sgp/crs/misc/R44709.pdf [http://perma.cc/NER9-Y6WR]. Accordingly, a unified opposition party holding at least forty-one Senate seats prior to that date, or fifty-one seats thereafter, could block a nominee. See id. (describing the Rule XXII reinterpretation and the subsequent need for only fifty-one votes to confirm a nominee other than one to the Supreme Court). This new interpretation applies only if the vote occurs pursuant to a specific procedural posture. Id. The opposition party needs fifty-one votes, rather than fifty, to block a nominee after November 21, 2013, because, in the event of a fifty-fifty tie, the Vice President would vote for confirmation.[/footnote] Yet, as Figures 4(a) and 4(b) show, cross-party appointees did not become ideologically distinct, at conventionally accepted levels of statistical significance, until late in the period. Further, a recent change in Senate rules allows for a simple test of the Senate’s potential role in giving PBRs their bite—and the results suggest that role is limited. On November 21, 2013, the leadership of the Senate Democratic majority exercised the “nuclear option” and altered the chamber’s interpretation of Senate Rule XXII to reduce the number of votes needed to invoke cloture for nominees to positions other than Supreme Court justiceships from sixty to fifty-one (or fifty, with the Vice President as a tiebreaker).[footnote]See Oleszek, supra note 137, at 8–9.[/footnote] For only the second time during the study period, a unified cross-party (here, the Republicans) did not possess sufficient votes to prevent a Senate majority from invoking cloture on a confirmation vote for a nominee to a PBR agency. A Senate-focused theory of PBRs’ bite would predict that President Obama’s appointees to cross-party seats would tend to be more liberal after November 21, 2013. In reality, however, appointees to cross-party PBR seats in the twelve months after this date are 0.179 points more conservative than appointees to cross-party PBR seats in the prior twelve months—a difference that is in line with the long-term year-to-year increases in ideological distance between the President and cross-party appointees displayed in Figure 4(b). The other period during which a unified cross-party (again, the Republicans) did not possess sufficient votes to block a Senate majority from invoking cloture was July 7, 2009 through February 4, 2010.[footnote]See Monica Davey & Carl Hulse, Franken’s Win Bolsters Democratic Grip in Senate, N.Y. Times (June 30, 2009), http://www.nytimes.com/2009/07/01/us/politics/ 01minnesota.html?mcubz=0 (on file with the Columbia Law Review ) (stating that Franken’s seating would give the Democrats sixty seats in the Senate, which is sufficient to overcome a filibuster); Janet Hook, Republican Scott Brown of Massachusetts Sworn in to Senate, L.A. Times (Feb. 4, 2010), http://articles.latimes.com/2010/feb/04/nation/la-na-brown5-2010feb05 (on file with the Columbia Law Review ) (stating that Brown’s seating would reduce the Democrats’ numbers to fifty-nine, depriving them of their ability to avert a Republican filibuster).[/footnote] That period was bookmarked, on one end, by the seating of Senator Al Franken of Minnesota as the sixtieth Democratic senator following a long recount, and, on the other end, by the seating of Senator Scott Brown, a Republican from Massachusetts, after a special election that reduced the Democratic caucus to fifty-nine.[footnote]See Davey & Hulse, supra note 139; Hook, supra note 139.[/footnote] During that interval, the Obama Administration’s appointees to cross-party seats were 0.111 points more liberal than those in the period between President Obama’s inauguration and Senator Franken’s swearing in. This slightly more liberal tilt in cross-party appointees during a period in which Democrats controlled the White House and held a supermajority in the Senate hints at a demand-side effect. Trends following Senator Brown’s seating, however, cut against the demand-side theory. Following Brown’s seating, Senate Democrats saw their filibuster-proof supermajority end. The demand-side theory predicts that cross-party appointees would trend conservative based on Senate Republicans’ relatively greater power during this period. Yet in the twelve months following Brown’s seating, cross-party appointees were 0.093 points more liberal than they were during the interval in which the Democrats held a supermajority. This slight liberal trend continued through the date on which Senate Democratic leadership exercised the nuclear option. Table 5 compares the actual trends in cross-party appointee ideology around these demand-side pivotal moments with the effects that we might predict on the basis of the demand-side theory. (In all intervals in the table, cross-party appointees are Obama appointees to Republican seats.) In only one of the three intervals—roughly the odds one might expect from three coin flips—does reality line up with the predictions generated by the demand-side theory. While we cannot entirely rule out the possibility that the opposition party’s ability to block a nomination has a modest effect on the ideological composition of multimember agencies, the evidence from these Obama-era shocks casts significant doubt on the demand-side story. Table 5: Using Demand Shocks to Assess the Demand-Side Hypothesis Intervals Obama Inaugurated–Franken Seated Franken Seated–Brown Seated Brown Seated–Nuclear Option Exercised Nuclear Option Exercised–End of Democratic Majority Demand-side theory predicts cross-party appointees will be . . . N/A more liberal than previous period more conservative than previous period more liberal than previous period Cross-party appointees actually were . . . N/A more liberal than previous period more liberal than previous period more conservative than previous period But perhaps the demand-side mechanism is subtler than simply whether a completely unified cross-party could prevent a floor vote. After all, senators rarely utilize their Rule XXII prerogative for nominees to executive branch positions.[footnote]See Bonica, Chen & Johnson, Senate Gate-Keeping, supra note 105, at 10 (noting that “many scholars have viewed floor votes as an afterthought” in the context of executive branch appointments and that very few nominees fail floor votes).[/footnote] Of those executive branch nominations that are reported out of committee, only 4% die on the floor, a figure that includes failures to pass a cloture motion to end a filibuster, filibuster threats, and individual senators’ holds on nominations.[footnote]Id.[/footnote] By contrast, Senate committees fail to report out approximately 20% of nominations, which effectively ends these nominees’ chances of confirmation in most cases.[footnote]Id.[/footnote] The role that majority-party-dominated committees play in the process suggests that a Senate party’s influence over nominations is far broader—and perhaps harder to define—than simply the binary matter of whether the party (if completely unified) possesses forty-one votes to deny cloture. To test the hypothesis that the power of the cross-party in the Senate influences the President’s choice of moderates or ideologues for cross-party seats on PBR agencies, we regress the ideological distance between each cross-party appointee and the appointing President (“CFscore Difference”[footnote]Recall that positive values of CFscore in a Democratic (Republican) administration indicate that the mean appointee is more conservative (liberal) than the appointing President.[/footnote]) on the cross-party’s seat share in the Senate at the time of each appointment (Model 1). As an alternative specification, we also regress CFscore Difference on whether the cross-party held a Senate majority at the time of the appointment (Model 2); “Divided Government” is coded as 1 if the cross-party held at least fifty-one Senate seats.[footnote]To create both Cross-Party Seat Share and Divided Government, independent senators who caucus with a particular party are classified as members of that party.[/footnote] Finally, Models 3 and 4 add a time-trend variable into the mix. “Time” signifies the number of years since the inception of the series in 1979 that a given cross-party appointee was confirmed. Including both “Time” and either “Cross-Party Seat Share” or “Divided Government” in these final two models enables us to pit the supply-side and demand-side hypotheses directly against each other. Table 6: Senate Composition and Cross-Appointee Selection Model 1 Model 2 Model 3 Model 4 Cross-Party Seat Share -0.036 (0.027) N/A -0.027** (0.009) N/A Divided Government N/A (0.227) N/A -0.202* (0.100) Time (Years Since 1979) N/A N/A 0.043*** (0.006) 0.044*** (0.005) n = 216. Unit of analysis: cross-party appointees to twenty-three agencies during the period from 1979 to 2014. Carter years include 1979 and 1980 only; Obama years include 2009 to 2014 only. Dependent variable: CFscore Difference. Robust standard errors clustered at the two-year Congress-level are given in parentheses. Table includes fixed effects for each presidency (baseline category: Carter Administration). *** p p p Table 6 shows no support for the hypothesis that the cross-party’s power in the Senate encourages the President to choose cross-party appointees to PBR agencies whose views are further from the President’s own. The coefficient estimates for “Cross-Party Seat Share” and “Divided Government” are negative in all four models (and statistically significant in two of the four), indicating the President selects cross-party appointees who are closer to her ideal point when the cross-party is more powerful in the Senate. These negative estimates are precisely the opposite of the demand-side theory’s predication: that a stronger cross-party in the Senate should pull cross-party appointees in its direction and away from the President’s ideal point. By contrast, the coefficient estimates for “Time” are positive and statistically significant in both of the models in which they appear (Models 3 and 4). As the series progresses—and as partisan sort proceeds—cross-party appointees steadily become more extreme. When tested against the demand-side theory, the supply-side theory retains its explanatory power. None of this is to suggest that individual senators are irrelevant to the selection of cross-party appointees.[footnote]See, e.g., Floyd Norris, Independent Agencies, Sometimes in Name Only, N.Y. Times (Aug. 8, 2013), http://www.nytimes.com/2013/08/09/business/independent-agencies-sometimes-in-name-only.html (on file with the Columbia Law Review ) (noting that senior senators have been “instrumental” in the selection of SEC commissioners in recent years).[/footnote] In particular, the leader of the opposition party in the Senate (the Minority Leader when the President’s party controls the Senate, or the Majority Leader when it does not) often plays an outsized role in the process.[footnote]See id. (noting that, in the early 2000s, the expectation was that the opposition-party leaders would choose cross-party nominees).[/footnote] For example, Senate Republican leader Bob Dole sent recommendations for cross-party appointments to President Clinton throughout Clinton’s first term.[footnote]See Kamen, supra note 58; see also Dina Elboghdady, Owen Sees ICC Post as Capital Opportunity, Orange County Reg., Feb. 5, 1995 (on file with the Columbia Law Review ) (noting that Clinton’s cross-party appointee to the ICC had been recommended by Dole); Lynn Stevens Hume, Dole Urges Clinton to Name Unger, Banking Committee Lawyer, to SEC Post, Bond Buyer, Sept. 19, 1995 (on file with the Columbia Law Review ) (noting that Dole had recommended that Clinton name a Republican Senate aide to a cross-party seat on the SEC); John Maggs, GOP Lawmakers Oppose Bragg as ITC Chairman, J. Com. (May 25, 1994), http://www.joc.com/gop-lawmakers-oppose-bragg-itc-chairman_19940525.html (on file with the Columbia Law Review ) (noting that other Republican lawmakers were upset with Dole’s recommendation for a cross-party seat on the USITC).[/footnote] Dole’s successor as Senate Republican leader, Trent Lott, continued the practice[footnote]See, e.g., Senate Spat over Lott’s FCC Picks, Wired (Apr. 10, 1997), http://www.wired.com/1997/04/senate-spat-over-lotts-fcc-picks (on file with the Columbia Law Review ) (detailing Senator Lott’s recommendations for Republican FCC commissioners).[/footnote]—in one case clashing with fellow Republican Senator Orrin Hatch regarding the respective roles of the leader and Senate committee chairs in cross-party appointments.[footnote]See Michael Grunwald, Sentencing Panel’s Own Terms Are Up, Wash. Post (Aug. 28, 1998), http://www.washingtonpost.com/archive/politics/1998/08/28/sentencing-panels-own-terms-are-up/da055846-230d-448f-b33c11f596cb1f4d/?utm_term=.47875b15ee4f (on file with the Columbia Law Review ) (noting the conflict between Lott and Hatch). The Sentencing Commission, an advisory panel, is located in the judicial rather than executive branch and so is excluded from our empirical analysis. About the Commission, U.S. Sentencing Comm’n, http://www.ussc.gov/ [http://perma.cc/ZNZ9-Z47M] (last visited Oct. 2, 2017).[/footnote] Senate Democratic leaders Tom Daschle and Harry Reid frequently forwarded names for cross-party appointments to President George W. Bush.[footnote]See, e.g., Jenny Anderson, Democrats Push S.E.C. Official for Commission Seat, N.Y. Times (May 18, 2005), http://query.nytimes.com/gst/fullpage.html?res= 9E0CE2DA1639F93BA25756C0A9639C8B63 (on file with the Columbia Law Review ); Judith Burns, Daschle Recommends Hispanic-American Candidate for SEC, Dow Jones News Serv. (Jan. 7, 2002) (available in full on Lexis Advance and on file with the Columbia Law Review ); Michele Heller, In Brief: Daschle Recommends Goldschmid for SEC, Am. Banker (Jan. 30, 2002), http://americanbanker.com/news/in-brief-daschle-recommends-goldschmid-for-sec (on file with the Columbia Law Review ); Mark Wigfield, Sen. Daschle Recommends Democratic Aide for FCC Post, Dow Jones News Serv. (Nov. 15, 2001) (available in full on Lexis Advance and on file with the Columbia Law Review ).[/footnote] According to one account, Senate Republican leader Mitch McConnell made a particularly concerted effort to advance conservative ideologues for cross-party seats under President Obama, whereas earlier leaders often selected nominees on the basis of home-state connections or recommendations from other caucus members.[footnote]See Daniel Foster, Agent McConnell, Nat’l Rev. (June 3, 2013), http:// www.nationalreview.com/sites/default/files/uploaded/20130603.pdf (on file with the Columbia Law Review ).[/footnote] Most recently, a spokesman for Minority Leader Charles Schumer said in March 2017 that Senate Democrats “intend to assert our prerogative on nominees as always has been done,”[footnote]Kelcee Griffis, Trump Nixes Rosenworcel Nomination to Return to FCC, Law360 (Mar. 2, 2017), http://www.law360.com/articles/897483/trump-nixes-rosenworcel-nomination-to-return-to-fcc [http://perma.cc/8KCD-UE82] (internal quotation marks omitted) (quoting a Schumer spokesman).[/footnote] and Schumer has advanced several Democrats for cross-party appointments in the months since President Trump took office.[footnote]See, e.g., David Lieberman, President Trump Renominates Jessica Rosenworcel to FCC, Deadline (June 14, 2017), http://deadline.com/2017/06/president-trump-renominates-jessica-rosenworcel-fcc-1202113089 [http://perma.cc/NR34-XS4X] (noting that Schumer recommended a cross-party nominee for President Trump); Harper Neidig, Schumer Recommends Consumer Advocate for FTC, Hill (May 8, 2017), http:// thehill.com/regulation/finance/332535-schumer-recommends-consumer-advocate-for-ftc [http://perma.cc/X9NA-RAD8] (noting that Schumer recommended someone to fill the Democratic opening on the FTC).[/footnote] But while the leader of the opposition party in the Senate no doubt plays an important role in the selection process, the widening ideological gap between co-party and cross-party appointees should not necessarily be attributed to Senate leadership. First, the role of Senate leaders is purely advisory. As a formal matter, any provision that vested a Senate leader with the statutory power to choose an executive branch officer would violate the Appointments Clause of Article II.[footnote]See Buckley v. Valeo, 424 U.S. 1, 127–37 (1976) (striking down a provision in the Federal Election Campaign Act that purported to authorize the President pro tempore of the Senate and Speaker of the House to choose four members of the FEC).[/footnote] And as a practical matter, Presidents do not always follow the suggestions of Senate leaders. For example, then-Senate Minority Leader Dole complained that President Clinton often chose “Clinton Republicans” for cross-party seats and ignored Dole’s recommendations.[footnote]Kamen, supra note 58.[/footnote] Likewise, then-Senate Majority Leader Reid complained in 2008 that although he and President George W. Bush “were able to work cooperatively on [cross-party] nominations” through most of Bush’s second term, “there has been a notable shift in the President’s stance on these nominations from one of cooperation to intransigence,” with the Bush White House rejecting several of Reid’s suggestions during the final year of the Bush presidency.[footnote]Martin Kady II, Reid Letter Shows White House Rejected Nominee Compromise, Politico: Politico Now Blog (Feb. 28, 2008), http://www.politico.com/blogs/ politico-now/2008/02/reid-letter-shows-white-house-rejected-nominee-compromise-006621 [http://perma.cc/K89G-G8E4] (quoting a letter from Reid to White House Chief of Staff Joshua Bolton). Reid’s letter also made the surprising—and flatly false—claim that “I have the statutory authority to make recommendations to the President for Democratic positions on independent boards and commissions.” Id. (emphasis added).[/footnote] Second, when Senate leaders recommend nominees, they do so in the shadow of the White House’s range of acceptable options. Thus, even when the President follows the opposition party leader’s recommendation, it would be a mistake to attribute the appointment to the opposition party leader alone. Cross-party appointments often come at the end of a back-and-forth process whereby the leader proposes a nominee to a cross-party seat and the White House either accepts the nomination or offers a counterproposal.[footnote]Telephone Interview with Ronald Weich, Former Chief Counsel to Senator Reid (Aug. 29, 2017) (call notes on file with the Columbia Law Review ) ; see also, G. Calvin MacKenzie, Innocent Until Nominated: The Breakdown of the Presidential Appointments Process 33 (2011) (describing “[p]ainstaking negotiations between the White House and Senate stakeholders” regarding nominees to the FCC in 1997, as well as similar bargaining between President Clinton and Senate Republican leaders regarding a cross-party appointee to the FEC in 1999).[/footnote] These negotiations are sometimes bound up in bargaining between the White House and senators over issues unrelated or tangential to the appointment in question.[footnote]Email from Senator Jon Kyl, Former Senate Minority Whip, to Daniel Hemel, Assistant Professor of Law, Univ. of Chi. Law Sch. (Aug. 25, 2017) (on file with the Columbia Law Review ) (“[T]he minority is always looking for leverage, and nominations for anything work.”).[/footnote] Third, insofar as cross-party appointments are driven by the opposition party’s Senate leader, we would expect that leader’s influence to be at its peak when the opposition party holds more seats in the Senate. Yet as illustrated by Table 6, our data do not bear out that prediction. Remarkably, the extent to which Presidents choose genuine cross-partisans for multimember commission seats has no apparent relationship to the political power of the opposition party leader. Fourth and finally, while the median Senate Democrat and median Senate Republican have grown farther apart ideologically over the course of our study period, we see much less ideological movement among the parties’ respective Senate leaders. Figure 8, below, compares the CFscores for the Senate Democratic and Republican leaders, denoted with bold lines, with the median appointees to Democratic and Republican seats on agencies with PBRs; appointees to co-party seats are denoted with thin, solid lines, whereas appointees to cross-party seats are denoted with dotted lines. As the figure shows, changes over time in the median CFscores of cross-party appointees bear little discernable relationship to changes in the CFscores of Senate leaders. For example, Mitch McConnell, who led the Senate Republicans through the Obama years, has a CFscore that is 0.030 points lower (less conservative) than Howard Baker, the Senate Republican leader under Carter, whereas the median cross-party (Republican) appointee under Obama is 0.386 points more conservative than the median cross-party appointee under Carter. In other words, Republicans appointed by Democratic Presidents grew much more conservative even while the Senate Republican leader (as measured by CFscore) grew slightly less so. Likewise, Harry Reid, who led the Senate Democrats during George W. Bush’s second term, has a CFscore that is only 0.194 points lower than that of Robert Byrd, the Senate Democratic leader under Reagan, while the median CFscore for George W. Bush’s cross-party (Democratic) appointees is 1.190 points lower than the median CFscore for Reagan’s cross-party appointees. That is, Democrats appointed by Republican Presidents grew substantially more liberal even while the CFscore for the Senate Democratic leader barely budged. In sum, we find little evidence to suggest that the divergence between co-party and cross-party appointees over our study period can be explained by an increase in the power of opposition party senators or Senate leaders in the selection process. Figure 8: Senate Leaders and Appointees to Agencies with PBRs 3. Search Costs . — Along with increased partisan sort, another broad social phenomenon that may have increased PBRs’ bite occurred during this period: an information technology revolution that slashed the cost of obtaining data on nominees. Forty years ago, learning a potential appointee’s detailed political views might require using one’s Rolodex to contact members of the appointee’s social or professional circle and investing the time and political capital to convince these mutual connections to speak candidly (and, naturally, as a prerequisite, one would need to have the connections and status to pull this off). Today, one can procure similar information with a few keystrokes—as we did. As a result, a wide variety of interested parties—including Senate staffers, activists, and journalists—can inexpensively obtain information regarding the ideologies of potential appointees. In light of the reduction in the costs associated with obtaining this information, perhaps the President’s ability to present conservative Democrats and liberal Republicans as stealth nominees has diminished. While this search-costs hypothesis has some appeal, we are skeptical that it is driving our results for three reasons. First , this hypothesis understates the amount of information concerning potential appointees’ donations that was publicly accessible throughout this period. The FEC has a long-established reputation for facilitating public access to its data.[footnote]See R. Sam Garrett, Cong. Research Serv., R44318, The Federal Election Commission: Overview and Selected Issues for Congress 17 (2015), http:// fas.org/sgp/crs/misc/R44318.pdf [http://perma.cc/KLU2-MHMU] (“[T]he FEC generally is praised for its role in publicizing campaign finance data.”); Jeremy Gaunt, A Rocky Decade Later, the FEC Gets Little Respect, Campaign Prac. Rep., Feb. 11, 1985, at 1, 2 (“[M]ost . . . are willing to credit the agency with at least one major success. . . . [T]he FEC has built on earlier attempts to make campaign finance data open to public scrutiny and has made disclosure of campaign dollars an accepted and expected part of the electoral process.”).[/footnote] For virtually the entire period, the FEC considered reports on receipts and FEC-generated indices of these receipts to be subject to the Freedom of Information Act and thus endeavored to provide “the fullest possible disclosure” to the public.[footnote]See 11 C.F.R. §§ 5.2(a), 5.4(a)(1) (2017). For versions of the rule in effect between 1980 and 2010, see 65 Fed. Reg. 9201, 9207 (Feb. 24, 2000) (to be codified at 11 C.F.R. pt. 5); 52 Fed. Reg. 39,213 (Oct. 21, 1987); 52 Fed. Reg. 23,638 (June 24, 1987); 50 Fed. Reg. 50,778 (Dec. 12, 1985); 45 Fed Reg. 31,292 (May 13, 1980). But see 44 Fed. Reg. 33,368 (June 8, 1979) (stating that FOIA does not apply to reports of receipts and related indices).[/footnote] In 1980, Congress required the FEC to make reports submitted by campaigns “available for public inspection” and to “develop a filing, coding, and cross-indexing system” to facilitate public access.[footnote]Act of Jan. 8, 1980, Pub. L. No. 96-187, sec. 109, § 311(a)(3)–(4), 93 Stat. 1339, 1362.[/footnote] The FEC met this obligation that same year.[footnote]Telephone Interview with Senior Pub. Affairs Specialist, supra note 9.[/footnote] Since 1996, the FEC has provided to the public comprehensive electronic records of campaign contributions dating back to the late 1970s.[footnote]Press Release, FEC, supra note 10.[/footnote] In 1998, the FEC debuted an internet-based search function that allows users to search the FEC’s online database for specific donors by name.[footnote]Id.[/footnote] Access to campaign contributions data likely was more difficult prior to the current era of searchable online databases. Still, interested parties had the means to access this information throughout the study period. Even during the years in which identifying contributors required reviewing indices of campaign contributions, the stakes to Presidents, senators, and affected interest groups of placing a “disloyal” appointee on a board or commission likely were sufficiently high—and the resources available to these individuals and entities sufficiently deep—to outweigh these relatively modest search costs. Second , political donations are but one of several ways to convey an individual’s partisan affiliation. Presidents and senators could employ many other screens to divine potential appointees’ partisan affiliations: for instance, whether an individual consistently votes in party primaries, has a record of volunteering for political campaigns, or participates in a political party or party-connected group. Indeed, the secular decline in mass participation in civic activity over the past several generations suggests that, in some respects, it may have been easier to discern an individual’s partisan affiliation in 1979 than in 2014.[footnote]Further, to the extent that members of Congress are consulted in executive appointments, the fraying personal connections between legislators and their individual constituents present an additional impediment to evaluating a potential appointee’s partisan loyalty that is more significant in the later years of the study period. Compare Richard F. Fenno, Jr., Home Style: House Members in Their Districts 31–35 (1978) (looking at how House members “showed a good deal of personal attentiveness to their districts” to “enlarge their political support at home”), with Richard F. Fenno, The Challenge of Congressional Representation 10–12 (2013) (noting the continued lack of scholarly attention to how and if representatives connect with their constituents to gain their trust and support).[/footnote] Third , we construct what should be a relatively easy test for the search-costs hypothesis—and obtain null results. This test exploits the FEC’s introduction of a search function for its online database of campaign contributions on July 21, 1998, which constituted the single greatest reduction in search costs during this period. If the search-costs hypothesis is doing any significant work, we would expect to see it here. Yet we do not. Model 1 in Table 7, below, is similar to the regression models presented in Table 6, above, except that here, we add a dummy variable denoting whether the nomination occurred after the search function for the FEC’s online database became publicly accessible on July 21, 1998.[footnote]Observations in Table 7 are limited to the 145 cross-party appointees for whom we were able to obtain the exact date of nomination.[/footnote] A positive, statistically significant estimate for this coefficient would support the alternative hypothesis that greater access to information reduced the White House’s ability to get a “cross-partisan in name only” past opposition party senators. Model 1, however, shows a null result. Model 2 tests a slight variation on the alternative hypothesis. What if interested parties gradually began to adapt to the FEC website’s new search function, whether because they learned about the function over time through word of mouth or because old-guard staffers were secularly replaced with more internet-savvy political operatives? Instead of a discrete jump in effect size following the rollout of the search function, we would expect the time trend in the dependent variable to pick up speed after July 21, 1998. In other words, we would expect a positive coefficient on the interaction of the “date of nomination” and “nomination after July 21, 1998” variables, which tests for this phenomenon. As Model 2 shows, however, the coefficient on the interaction term is statistically insignificant and, in fact, weakly negative.[footnote]We also ran a series of placebo tests, substituting July 21 of the years 1993 to 1997 and 1999 to 2007 in place of the “Nomination after July 21, 1998” variable and interaction-term component. These dates have no special relevance; we included them simply to determine whether the coefficient estimates in models including these dates differ in any meaningful way from the estimates reported in Table 7. The results of these placebo tests are substantially similar to those reported in Table 7 (although the “Divided Government” variable lacks statistical significance in most of these placebo models, as expected). To determine whether a discontinuity in cross-appointee ideology exists around July 21, 1998, we constructed a regression-discontinuity model. However, a relative lack of observations around the July 21, 1998 cut-point impeded the analysis; using any reasonable set of parameter assumptions, the rdbwselect function in R recommended bandwidths of approximately two years—hardly a knife-edge. Using this bandwidth, the model reported a statistically significant, positive discontinuity at the cut-point. But we do not put much stock in this result; given the unacceptably large bandwidth size, the model likely is picking up the well-documented time trend and misreporting it as a “knife-edge” discontinuous jump. Finally, we constructed a regression-kink model, which tests for a discontinuous change in slope, rather than an upward shift in the regression curve, at around this date. The theory here is that interested parties may have learned about (and learned how to use) the FEC’s searchable database gradually, and therefore we might expect a change in the rate of growth in the ideological gap after this date, rather than an abrupt increase. This model reported null results, but, once again, serious bandwidth selection issues prevent firm conclusions.[/footnote] In other words, we find little evidence that the ideological distance between the President and cross-party appointees increased at a faster pace after the FEC’s website search rollout. Table 7: Effects of FEC Search Functionality on Cross-Appointee Selection Model 1 Model 2 Divided Government -0.304 (0.155) -0.332 (0.170) Date of Nomination 0.042** (0.015) 0.045* (0.017) Nomination after July 21, 1998? 0.036 (0.272) 40.636 (98.771) Date of Nomination * Nomination after July 21, 1998? (interaction term) N/A -0.020 (0.049) n = 145. Unit of analysis: cross-party appointees during the period from 1979 to 2014 for whom we know the exact date of nomination (a subset of the dataset used in previous models). Dependent variable: CFscore Difference. signifies p p p To sum up so far: Our results are consistent with a supply-side story in which partisan sort starting in the 1980s tightened the relationship between party affiliation and ideology, and so the preferences of cross-party appointees increasingly diverged from those of the appointing President. Our results are less consistent with a demand-side story in which the power of the opposition party in the Senate (or the influence of its leader) drives change over time in the ideological composition of multimember agencies. Finally, we find no support for the hypothesis that an Internet-induced reduction in search costs has prevented recent Presidents from filling cross-party seats with individuals whose campaign contributions reveal that these appointees themselves are out of step with their party of registration. 4. Identity Signaling. — Finally, the notion that appointment-seeking members donate solely to convey that they are loyal party members may bear on this analysis. Consider, for instance, a highly regarded product safety lawyer angling for a seat on the CPSC. Her qualifications are unimpeachable, but her involvement in partisan politics is either trivial or dated.[footnote]This example is loosely based on Marietta Robinson, a product safety lawyer with a long career representing both plaintiffs and defendants. Robinson ran as a Democrat for a Michigan Supreme Court seat in 2000, served in several nonpartisan positions in the 2000s, and was appointed by President Obama to a Democratic seat on the CPSC in 2013.[/footnote] She may therefore benefit from donating to her party’s candidates solely to convey that she is a member of that party. This rationale stands apart from the sincere-versus-strategic dichotomy discussed in section II.B.3. Although this behavior obviously is not a sincere expression of closely held beliefs, neither is it strategic in the sense that it is designed to misrepresent an individual’s ideological preferences. It simply broadcasts: “I am a member of this team.” If this motivation for political giving is widely held, then changes in appointees’ ideal point estimates over time might not capture genuine changes in their political views. Rather, if members of the political class donate mechanistically to candidates on their party’s ticket, and the ideological composition of the ticket changes over time as a result of partisan sort among candidates, then the observed longitudinal changes in appointees’ ideal point estimates would be epiphenomenal. Phrased another way, whereas a Washington lawyer could show loyalty to the Republican Party in 1970 by donating to Nelson Rockefeller, today she might do so by giving to Ted Cruz. Neither action is necessarily infused with much meaning; they both convey party loyalty—and, perhaps, nothing more.[footnote]Cf. Václav Havel, The Power of the Powerless, in The Power of the Powerless 23, 27 (John Keane ed., 1985). Amusingly, the Washington lawyer’s behavior is reminiscent of Havel’s classic description of shopkeepers in communist Czechoslovakia: "I think it can safely be assumed that the overwhelming majority of shopkeepers never think about the slogans they put in their windows, nor do they use them to express their real opinions. That poster was delivered to our greengrocer from the enterprise headquarters along with the onions and carrots. He put them all into the window simply because it has been done that way for years, because everyone does it, and because that is the way it has to be. If he were to refuse, there could be trouble." Id.[/footnote] We acknowledge that partisan sort among candidates may encourage donors interested solely in signaling their party loyalty to support more extreme candidates in 2014 than in 1979. But we caution that this theory’s explanatory power is limited. First, it cannot account for the fact that co-party and cross-party appointees differed markedly throughout the 1980s and 1990s. That Democrats selected by President Carter exhibit sharply different donation activity than Democrats selected by President Reagan casts doubt on the notion that appointment-seekers’ ideal point estimates merely reflect the ideal points of their party’s candidates. Second, and somewhat to our surprise, we found that appointees to PBR agencies routinely make contributions to candidates from both political parties. Even among appointees to the high-profile, highly politicized FCC, a third of appointees in our dataset (eleven out of thirty-three) donated to both Democratic and Republican candidates. If likely nominees are using campaign contributions solely to signal that they are on “Team Blue” or “Team Red,” the signals they are sending are remarkably noisy. Third, even with the trend toward polarized parties, donors who give exclusively to candidates from one party still have a wide variety of options from which to choose. In 1980, the mean Republican candidate on a general election ballot for state or federal office had a CFscore of 0.770, while the mean Democratic candidate’s score was -0.365. (Recall that a higher CFscore indicates greater conservatism.) By 2014, those means were 1.046 and -0.985, respectively. But the standard deviations around these means also increased between 1980 and 2014: from 0.368 to 0.573 for Republicans and from 0.529 to 0.668 for Democrats.[footnote]See Adam Bonica, All Recipients, 1979–2014, Database on Ideology, Money in Politics, and Elections (DIME) (2015), http://dataverse.harvard.edu/file.xhtml?fileId= 2865309&version=RELEASED&version=.2 [http://perma.cc/DKY4-HLEJ].[/footnote] In other words, party-loyal donors in 2014 could donate to a candidate who adopted the posture of a 1980-style Democrat (or 1980-style Republican) and who was still located within one standard deviation of the current party mean. That Democratic (Republican) appointees did not give to those more moderate candidates, and instead chose to give to more liberal (conservative) candidates, provides further support for the claim that appointees themselves have grown more extreme. III. Implications and Extensions We can now say with some confidence that PBRs are more than paper tigers—that they do indeed lead Presidents to choose cross-party appointees with divergent ideological preferences. Our results have clear implications for the leading accounts of PBRs in the political science and administrative law literatures. Our results also suggest avenues for future research on the effects of PBRs. This last Part considers those implications and potential extensions of our project. A. The Monitoring Account One perspective on PBRs in the political science and administrative law literature emphasizes the relationship between PBRs and the cost of monitoring agency actions. PBRs can reduce monitoring costs for Congress in three ways. First, minority party commissioners are likely to sound a “fire alarm” if the majority on the commission embarks on a potentially controversial course of action.[footnote]Barkow, supra note 87, at 41 (“[W]hen an agency is composed of members of different parties, it has a built-in monitoring system for interests on both sides . . . [and is] more likely to produce a dissent if the agency goes too far in one direction. That . . . serves as a ‘fire alarm’ that alerts Congress . . . .” (footnote omitted)). The origins of the “fire alarm” theory trace back to work by political scientists Mathew McCubbins and Thomas Schwartz. See Mathew D. McCubbins & Thomas Schwartz, Congressional Oversight Overlooked: Police Patrols Versus Fire Alarms, 28 Am. J. Pol. Sci. 165, 166 (1984).[/footnote] These “fire alarms” make it easier for lawmakers to keep an eye on agencies: Instead of reviewing every agency action individually, Congress can focus on instances in which minority party members dissent from a commission decision.[footnote]Alexander Bolton, Collegial Leadership Structures, Ideological Diversity, and Policymaking in the United States 7 (Aug. 15, 2015) (unpublished manuscript) (on file with the Columbia Law Review ) (arguing that ideological diversity and “minority votes can serve to reduce the oversight costs of principals, giving commission chairs incentives to moderate policies”).[/footnote] Second, beyond simply sounding an alarm, minority commissioners might further reduce the costs of congressional oversight by providing lawmakers with information about the consequences of agency actions as well as ways to overturn those actions.[footnote]Id. The presence of minority commissioners may also encourage majority commissioners to provide greater information than they otherwise would. Even the prospect of a dissent may serve an information-forcing function, compelling commissioners in the majority to provide more fulsome explanations for their decisions in an effort to dampen the potential dissenters’ fire alarm. Cf. Catherine M. Sharkey, State Farm “with Teeth”: Heightened Judicial Review in the Absence of Executive Oversight, 89 N.Y.U. L. Rev. 1589, 1605, 1657–60 (2014) (arguing that the prospect of inspector-general investigations, congressional hearings, or judicial review may compel agencies to provide better-reasoned explanations for their actions).[/footnote] Third, and relatedly, the same fire alarms that alert Congress to questionable commission decisions also alert the courts.[footnote]Bolton, supra note 173, at 13.[/footnote] This third mechanism can be considered as part of a multipronged strategy of congressional control: Congress creates agencies subject to PBRs, Congress also provides for judicial review of agency actions, and minority commissioners alert the courts to instances in which agency actions deviate from statutory directives. The monitoring account might lead us to expect that lawmakers would be most likely to impose PBRs when they are most worried about the executive branch straying from their own policy preferences. Consistent with this expectation, Professor David Lewis observes that the percentage of new agencies with appointment limitations such as PBRs is highest in periods of divided government.[footnote]See David E. Lewis, The Politics of Presidential Appointments: Political Control and Bureaucratic Performance 60–66 (2008) (“The number of appointees is also likely to be higher when the president and Congress have similar views about policy.”).[/footnote] Note, though, that while Lewis’s finding is consistent with the monitoring account, it does not confirm that account. Even if opposing party lawmakers seek to reduce monitoring costs by imposing PBRs on new multimember agencies, their efforts may prove fruitless. What we can say is that certain empirical findings make the monitoring account appear more plausible. For the monitoring account to be accurate, a necessary but not sufficient condition is that PBRs actually lead to ideologically diverse agencies with minority members who will alert Congress and the courts if agency leaders stray from legislative preferences or statutory directives. In that respect, our finding that the ideological preferences of cross-party appointees in recent years have indeed diverged from the preferences of the appointing President supports the monitoring account’s claims. Importantly, however, the fact that PBRs produce ideological diversity in multimember agencies does not prove that PBRs succeed in reducing monitoring costs for Congress and the courts. First, members whose ideological preferences diverge from the White House occupant’s may still exhibit loyalty toward the President who appointed them. Professors Lee Epstein and Eric Posner have documented a powerful “loyalty effect” among Supreme Court Justices, who are more likely to vote with the Solicitor General when the President who appointed them remains in office.[footnote]Epstein & Posner, supra note 100, at 402–03.[/footnote] This loyalty effect is distinct from the phenomenon of Supreme Court Justices tending to support the administration when the President is of the same party as the Justice’s appointer; the loyalty effect is person-specific rather than party-specific. A similar loyalty effect among members of PBR agencies might lead a conservative Republican to support the policies of a Democratic President who appointed her (and likewise for a liberal Democrat appointed by a Republican President). A loyalty effect among cross-party appointees might undermine the monitoring value of ideological diversity at multimember agencies because minority party members might be reluctant to interfere with the agenda of a President to whom they feel gratitude. Second, even in the absence of a loyalty effect, agency members whose ideological preferences diverge from the President’s might do little to reduce monitoring costs for Congress and the courts. Minority party members might be excluded from access to information about agency decisionmaking, or they might be reluctant to blow the whistle on their majority party colleagues, or their “fire alarms” might go unheeded by lawmakers and judges. Evaluating the strength of the monitoring account thus requires more than simply knowing how PBRs affect the ideological composition of multimember agencies. We would need to know whether the ideological composition of multimember agencies has observable effects on interactions between agencies and Congress, and between agencies and courts.[footnote]One of us has begun a project to study the effectiveness of congressional oversight of administrative agencies using empirical methods. See Brian D. Feinstein, Congress in the Administrative State, 95 Wash. U. L. Rev. (forthcoming 2018) (manuscript at 5), http://ssrn.com/abstract=2943074 (on file with the Columbia Law Review ). Studying the relationship between ideological composition and congressional oversight is a potentially fruitful path for future research.[/footnote] While the analysis above cannot answer all questions regarding the monitoring account’s accuracy, our topline results do provide guidance for researchers studying this subject. Most significantly, our findings suggest that the effect of PBRs on monitoring costs may be time-variant: How PBRs affected interactions between agencies and Congress—and between agencies and courts—in the 1980s is not necessarily predictive with regard to later years. Our results suggest that the effects of PBRs on the ideological composition of multimember agencies from the Clinton presidency onward are quite unlike the effects under Presidents Carter, Reagan, and George H.W. Bush. Accordingly, the follow-on consequences for monitoring costs may be different for later periods than for earlier ones. B. The Deliberation Account A second account of PBRs (which is distinct from, but not inconsistent with, the monitoring account above) draws from the social psychology literature on group polarization. Group polarization occurs “when an initial tendency of individual group members toward a given direction is enhanced following group discussion.”[footnote]Daniel J. Isenberg, Group Polarization: A Critical Review and Meta-Analysis, 50 J. Personality & Soc. Psychol. 1141, 1141 (1986).[/footnote] Numerous studies have documented this phenomenon.[footnote]Early experiments involved eighteen- and nineteen-year-old students in Paris who were asked about their attitudes toward France’s then-President, Charles de Gaulle, and the United States. Roger Brown, Social Psychology 223–24 (2d ed. 1986). Initial attitudes toward de Gaulle were “mildly favorable”; after group discussion, sentiments toward the President became even more favorable. Id. Initial attitudes toward the United States were “mildly negative”; these views tended to become more negative following discussion. Id. In both cases, group deliberation pushed group members toward the extreme. Id. Another study presented subjects with traffic felony scenarios and asked them to rate defendants’ guilt (from “definitely not guilty” to “definitely guilty”). David G. Myers & Martin F. Kaplan, Group-Induced Polarization in Simulated Juries, 2 Personality & Soc. Psychol. Bull. 63, 63–64 (1976).The subjects then discussed their impressions and rerated the defendants. The discussion process consistently led subjects to become more extreme in their views. Id. With respect to scenarios in which most subjects initially considered the defendant to be guilty, group discussion strengthened their confidence in that view. Id. When most subjects initially considered the defendant to be innocent, discussion likewise pushed subjects even further toward the innocence extreme. Id.[/footnote] Professor Cass Sunstein, who is largely responsible for bringing the group polarization literature to the attention of administrative law scholars, highlights two explanations for the observed patterns of polarization. One emphasizes “social comparison”: People want to be perceived favorably by their peers, and so when surrounded by others with liberal (conservative) views, they seek favor by adjusting their own views to be more liberal (conservative).[footnote]See Sunstein, supra note 13, at 88–89.[/footnote] A second explanation stresses “limited argument pools”: Individuals are influenced by the arguments they encounter, and an individual confronted with only liberal (or only conservative) arguments is likely to move further in that direction.[footnote]See id. at 89–90.[/footnote] Both of these explanations might suggest that while individuals surrounded by others with like-minded views will grow more extreme, individuals exposed to a diversity of viewpoints may become more moderate.[footnote]See id. at 89, 118–19.[/footnote] This diversity-leads-to-depolarization hypothesis draws some support in the social psychology literature, though this effect is stronger for group members confronting a problem for the first time than for group members dealing with familiar and much-debated questions.[footnote]See Brown, supra note 180, at 226.[/footnote] Sunstein argues that the social psychology literature on group polarization and depolarization produces potentially useful insights for agency design. He writes: An independent agency that is all Democratic, or all Republican, might polarize toward an extreme position, likely more extreme than that of the median Democrat or Republican, and possibly more extreme than that of any member standing alone. A requirement of bipartisan membership can operate as a check against movements of this kind.[footnote]See Sunstein, supra note 13, at 103 . In a similar vein, Professor Rachel Barkow writes: "[A] partisan balance requirement . . . can avoid extremely partisan decisions . . . . As a wealth of empirical research demonstrates, a group composed solely of ideologically like-minded people tends toward extreme decision making. Liberals and conservatives alike become more liberal and conservative, respectively, when they deliberate only with like-minded people. . . . [A] commission of five members all of the same party would be even more polarized than one in which a bare majority is of the same party." Barkow, supra note 87, at 40–41.[/footnote] The verbs “might” and “can” are important: Extrapolations from laboratory experiments to the real world of multimember federal agencies raise questions of external validity. Moreover, even in the absence of PBRs, agency commissioners may be exposed to a rich pool of arguments from sources other than their colleagues. Commissioners likely encounter arguments from lawyers who practice before the agency, lobbyists who seek to influence agency action, think tanks inside the Beltway, and—on especially high-profile issues—news media. Nonetheless, the literature on group polarization offers a potential explanation as to how PBRs might affect agency decisionmaking. Here, as elsewhere, ideological diversity may depolarize.[footnote]To be sure, partisan balance requirements theoretically could have the opposite effect. Considering the possibility of partisan balance requirements for judicial panels, Professors Cass Sunstein and Thomas Miles note that “[p]erhaps both Republican and Democratic appointees would conceive of themselves, to a somewhat greater degree, as political partisans, simply because the requirement of mixed composition would suggest as much.” Cass R. Sunstein & Thomas J. Miles, Depoliticizing Administrative Law, 58 Duke L.J. 2193, 2228 (2009).[/footnote] Like the monitoring account, the deliberation account is difficult to verify directly. What can be said is that for the deliberation account to be accurate, it must first be the case that PBRs have the immediate effect of producing ideological diversity. In this respect, our findings are supportive of the deliberation account.[footnote]See supra section II.B.[/footnote] Increasing party polarization appears to have the perhaps-unexpected effect of increasing the ideological heterogeneity of multimember agencies. No longer do Democratic Presidents fill seats on boards and commissions with like-minded liberals who differ only in their party registration. No longer do Republican Presidents stock multimember agencies with conservative Democrats and conservative Republicans. Instead, we observe PBRs leading to multimember agencies with a mix of liberal Democratic and conservative Republican members. That members are afforded opportunities to deliberate is a second necessary condition for the deliberation account. If decisions are preordained based on the views of the chair or those of an autonomous professional staff, with the multimember structure merely serving as a rubber stamp, then deliberation either will not occur or will be irrelevant to policy outcomes. Although the evidence is limited, it seems that this second condition is indeed met at most multimember agencies. Commissioners are generally able to hire their own dedicated staff, and thus have the capacity to critically evaluate proposals from the chair or others and to suggest informed alternatives.[footnote]See Marshall J. Breger & Gary J. Edles, Established by Practice: The Theory and Operation of Independent Federal Agencies, 52 Admin. L. Rev. 1111, 1168 (2000) (describing members’ ability to select and supervise their own staffs).[/footnote] Moreover, most multimember agencies are required to undertake official action only at or after a formal meeting with a quorum, which affords at least an opportunity to deliberate.[footnote]See id. app. at 1236–94 (listing meeting requirements).[/footnote] But even assuming that these preconditions are met—namely, that PBRs foster ideological diversity and that members have opportunities to deliberate—it does not necessarily follow that deliberation actually occurs, much less that minds are changed. The question of whether ideological diversity on multimember agencies has any effect on decisional outputs still remains unanswered. But while this Article does not seek to answer that question, it may be possible in future research to assess the deliberation account in further detail. For example, one might ask whether agencies that are more ideologically diverse are also less likely to be reversed by courts, or less likely to commit “infractions” of other kinds (that is, actions that elicit criticism from inspectors general, the Government Accountability Office, or major newspapers).[footnote]On the measurement of agency “infractions,” see Feinstein, supra note 178 (manuscript at 24–25).[/footnote] One might also look at whether ideological diversity on multimember agencies affects voting patterns of individual agency members. Analogously, Professors Thomas Miles and Cass Sunstein have found evidence suggesting that on three-member circuit court panels, the presence of an ideological minority has a moderating effect on the votes of judges in the majority.[footnote]Thomas J. Miles & Cass R. Sunstein, Do Judges Make Regulatory Policy? An Empirical Investigation of Chevron , 73 U. Chi. L. Rev. 823, 852 (2006) (“[B]oth Democratic and Republican appointees show far more political voting patterns when they are sitting on unified panels. When the panels are divided, the role of politics is greatly dampened.”).[/footnote] Professor Richard Revesz has documented a similar phenomenon in environmental law cases decided by the D.C. Circuit.[footnote]Richard L. Revesz, Environmental Regulation, Ideology, and the D.C. Circuit, 83 Va. L. Rev. 1717, 1719 (1997) (sharing the results of a study on judicial decisionmaking in the D.C. Circuit that found “a judge’s vote . . . is greatly affected by the identity of the other judges sitting on the panel”).[/footnote] While these authors use party affiliation as a proxy for ideology, our approach of using appointees’ campaign contributions can allow for a richer analysis of ideological diversity and depolarization in the agency context. Down the road, we hope to examine whether ideological diversity leads agency members to gravitate together or pull apart. C. The Obstruction Account A third and final perspective on partisan balance requirements posits that PBRs unconstitutionally interfere with presidential control over the executive branch. The Justice Department under President George H.W. Bush made a brief argument to this effect in a 1989 Office of Legal Counsel memorandum: According to the memo, PBRs violate Article II’s Appointments Clause[footnote]See U.S. Const. art. II, § 2, cl. 2 (“The President . . . shall nominate, and by and with the Advice and Consent of the Senate, shall appoint . . . all . . . Officers of the United States . . . but the Congress may by Law vest the Appointment of such inferior Officers, as they think proper.”).[/footnote] because “[t]he only congressional check that the Constitution places on the President’s power to appoint ‘principal officers’ is the advice and consent of the Senate.”[footnote]Common Legislative Encroachments on Executive Branch Auth., 13 Op. O.L.C. 248, 250 (1989).[/footnote] The Office of Legal Counsel listed PBRs as one of “ten types of legislative provisions commonly included in proposed legislation that weaken the Presidency” and that the executive branch should “consistently and forcefully resist[].”[footnote]Id. at 248, 250. In a separate vein, law professor Jamin Raskin has argued that PBRs violate the First Amendment and the principle of equal protection because they discriminate on the basis of party affiliation. See Jamin B. Raskin, “A Complicated and Indirect Encroachment”: Is the Federal Election Commission Unconstitutionally Composed?, 52 Admin. L. Rev. 609, 623–26 (2000).[/footnote] A more full-throated version of this argument emerges from recent work by Ronald Krotoszynski and coauthors. They rely on the Supreme Court’s 2010 decision in Free Enterprise Fund v. Public Company Accounting Oversight Board ,[footnote]561 U.S. 477 (2010).[/footnote] a case involving a five-member Board created to regulate auditing of securities issuers. While the Public Company Accounting Oversight Board (PCAOB) is not subject to a PBR, its members were protected by “dual for-cause removal limitations”: They could be removed only by the SEC and only for cause, and the SEC commissioners could be removed by the President only for cause.[footnote]Id. at 486–87. The SEC’s organic statute actually imposes no limitation on the removal of commissioners. However, the parties in Free Enterprise Fund both agreed that SEC commissioners are protected by an unwritten for-cause removal limitation. See id.[/footnote] The Court, in an opinion by Chief Justice Roberts, held that “the dual for-cause limitations on the removal of Board members contravene the Constitution’s separation of powers.”[footnote]Id. at 492.[/footnote] But the opinion arguably stands for the broader proposition that the President must have “meaningful oversight and control powers over independent federal agencies.”[footnote]Krotoszynski et al., supra note 15, at 948.[/footnote] Krotoszynski and coauthors argue that PBRs, when coupled with for-cause removal limitations for agency members, potentially violate the broader separation of powers principle embodied in the Free Enterprise Fund decision. They write: Statutory partisan balance requirements quite literally force Presidents to rely on political enemies to carry out their executive duties. . . . This is especially troubling considering the increased polarization present in American political parties. . . . Essentially, statutory partisan balance requirements foster a politically polarizing environment at the heads of independent agencies. . . . [T]hese statutory partisan balance requirements force Presidents to carry out their executive duties with contentious and highly polarized agency heads. . . . [and] preclude a President from appointing a sufficient number of agency commissioners in agreement with her political philosophy to overcome the debilitating effect of partisanship.[footnote]Id. at 991–92.[/footnote] Notably, Krotoszynski and coauthors do not claim that PBRs are always unconstitutional. They write: If Congress requires the President to appoint political opponents to an independent federal agency, but does not also entrench such persons with a fixed term of office or good cause protection against removal, it is difficult to see how a partisan balance requirement on these facts would significantly impede the President’s ability to oversee and direct the agency’s operations.[footnote]Id. at 1002.[/footnote] In their view, it is the combination of PBRs and for-cause removal protections, rather than either of these provisions individually, that renders a large swath of administrative agencies unconstitutional. They conclude: “Congress should have the choice of mandating partisan balance requirements or insulating principal officers who serve on agency heads from removal; it should not be permitted to impose both conditions concurrently . . . .”[footnote]Id. at 1005.[/footnote] There is, on reflection, something curious about the claim that for-cause removal protections are constitutionally allowable but that the combination of PBRs and removal protections raises problems. Consider the case of the SEC, a five-member body whose commissioners serve staggered five-year terms, with one term ending in June of every year.[footnote]15 U.S.C. § 78d (2012).[/footnote] Suppose there were no PBR and so a two-term President (say, President Obama) would have opportunities to fill every position on the Commission. Assuming that President Obama could get all of his nominees confirmed, he would be able to stock the Commission with like-minded members during his eight years in office (indeed, by year five). President Trump would then inherit an SEC with five liberal Democratic commissioners. Each June, President Trump would be able to appoint a conservative Republican to the expiring term, but it would not be until June 2019—more than halfway through his first term—that he could achieve a conservative Republican majority on the SEC. At least arguably, partisan balance requirements make Presidents less “rel[iant] on political enemies” insofar as they ensure that Presidents inherit commissions with only a bare majority of members hailing from the opposite party. In the example above, President Trump would be able to achieve a conservative Republican majority on the SEC the first time that a Democratic-appointed commissioner’s term expires. In an analysis of twelve multimember agencies subject to PBRs, Professors Neil Devins and David Lewis find that following a change in party control of the White House, it takes an average of thirteen to fourteen months for the new President to obtain a majority of co-partisans on a commission.[footnote]Devins & Lewis, supra note 79, at 470. Devins and Lewis analyze data from 1921 to 2005. Id.[/footnote] In other words, if historical patterns hold, it will take President Trump an average of thirteen to fourteen months from inauguration—until March or April of 2018—before a majority of commissioners are Republican. By contrast, it takes an average of twenty-six months for a new President to appoint an absolute majority of members—until March of 2019 before a majority of commissioners are Trump appointees.[footnote]Id.[/footnote] Thus, the extent to which PBRs make a President reliant on commissioners whose ideologies differ from the President’s own views depends critically on whether PBRs actually push Democratic Presidents to appoint conservative Republicans (and, vice versa, whether PBRs actually push Republican Presidents to appoint liberal Democrats). If PBRs impose only a weak constraint, such that a Democratic President can appoint liberals to both Democratic and Republican seats, then it will take longer for a new conservative Republican President to fill a commission with conservatives. If PBRs impose a stronger constraint on ideology, then a new conservative Republican President will be able to assemble conservative majorities on commissions much more quickly. From this perspective, our results should be reassuring to those who worry that multimember agencies subject to PBRs whose members enjoy for-cause removal protection will interfere with the President’s pursuit of her policy agenda. As PBRs become increasingly effective at generating ideological diversity on multimember agencies, they may in fact do less rather than more to obstruct the sitting President. Conclusion Our topline result can be summarized in a single sentence: Since the mid-1990s (but not in the decade and a half before), partisan balance requirements have shaped the ideological composition of multimember agencies by forcing Presidents to fill cross-party seats with appointees whose preferences diverge from their own. In other words, partisan balance requirements indeed bite, but they have come to do so only relatively recently. These findings are broadly consistent with supply-side developments: As ideology and party identification have become more closely correlated, Presidents have found it more difficult to identify competent cross-party appointees whose policy preferences they share. And as a consequence, the decades-long trend toward partisan sort in American politics has led to multimember boards and commissions composed of very few cross-party appointees sharing the ideology of the President who appointed them. Democratic Presidents may still be on the hunt for liberal Republicans to appoint to cross-party seats (and Republican Presidents for conservative Democrats), but with fewer of these creatures left in the wild, Presidents increasingly are compelled to select their ideological opponents for cross-party seats. Our results suggest a novel twist on the standard narrative of partisan polarization. The conventional wisdom holds that increased partisan polarization has strained the architecture of government. At agencies subject to PBRs, however, the effects of partisan sort may be salutary. Partisan sort may lead Presidents to select bona fide cross-party members, who in turn serve as in-house monitors and counterbalance tendencies that might drive groups to go to extremes. We cannot yet say definitively that PBRs reduce the costs borne by Congress and the courts in watching over multimember agencies. Nor can we confidently say that PBRs improve deliberative processes inside agencies. This study is a first step toward resolving those questions, but it is not a final answer. What we can conclude, based on the data analyzed here, is that PBRs appear to enhance the ideological diversity of multimember boards and commissions throughout the federal bureaucracy. At a time of heightened concerns regarding partisan polarization in Washington and across the country, that in itself is no mean feat.