What a difference a decade makes. In 2006, 45% of Americans were worried a great deal about crime. 1 Alyssa Davis, In U.S., Concern About Crime Climbs to 15-Year High, Gallup (Apr. 6, 2016), []. By 2016, the number had jumped to 53%, the highest level since 9/11, which was the last time a majority of Americans had expressed that view. 2 See id. (showing 62% of Americans reported “personally worry[ing] about crime and violence” after 9/11). This increase in the level of fear buoyed Donald Trump to the presidency on his promise to restore law and order. 3 See, e.g., Jose A. DelReal & Sean Sullivan, Trump Ratchets Up Nationalist and Law-and-Order Rhetoric on Campaign Trail, Wash. Post (Sept. 22, 2016), []; see also Standing Up for Our Law Enforcement Community, White House, [] (last visited Mar. 15, 2017) (“The Trump Administration will be a law and order administration.”). But while fear of crime rose in 2016, crime itself had fallen sharply over that decade. Since 2006, the murder rate had dropped 15.5%, 4 See FBI, Table 1: Crime in the United States by Volume and Rate per 100,000 Inhabitants, 1996–2015, FBI: Unif. Crime Reporting, [] [hereinafter FBI, 2015 UCR Data] (last visited Feb. 8, 2017) (showing the murder rate dropped from 5.8 per 100,000 people in 2006 to 4.9 in 2015). violent crime had fallen 22.3%, 5 See id. (showing the violent crime rate dropped from 479.3 per 100,000 people in 2006 to 372.6 in 2015). and property crime had fallen 25.7% 6 See id. (showing the property crime rate dropped from 3,346.6 per 100,000 people in 2006 to 2,487.0 in 2015). —according to the latest FBI data of 2015. If the United States in 2015 experienced the same murder rate as it had in 2006 (5.8 per 100,000), almost 3,000 more murders would have taken place in 2015 than the 15,696 that did occur. 7 See id. How, then, did 2015 come to symbolize a nightmare of exploding violent crime that could aid the fortunes of the law-and-order candidate? Welcome to the uncertain and unpredictable world of crime in the United States.

Through 2014, the broad statistical picture of crime during the term of President Barack Obama was largely congenial. Despite some hiccups, every major crime category trended down in the first six years of his Administration. 8 See supra notes 4–7 and accompanying text (detailing declines in murders, violent crime, and property crime from 2006 to 2015). Talk of evidence-based criminal justice reform was in the air, resort to the ineffectual death penalty was waning (with a number of states entirely jettisoning the costly and controversial punishment), 9 Editorial, The Continuing Collapse of the Death Penalty, N.Y. Times (Dec. 26, 2016), (on file with the Columbia Law Review). and sensible proposals to reduce mass incarceration without creating risks to the public were debated. 10 See James Austin et al., Brennan Ctr. for Justice, How Many Americans Are Unnecessarily Incarcerated? 7 (2016),
files/publications/Unnecessarily_Incarcerated.pdf [] (“Of the 1.46 million state and federal prisoners, an estimated 39 percent (approximately 576,000 people) are incarcerated with little public safety rationale.”).

But then in October 2015, despite the fact that the 2014 murder rate was the lowest the United States had seen since 1957, FBI Director James Comey delivered a controversial speech at the University of Chicago Law School in which he highlighted the jump in murders that had begun earlier that year. 11 James B. Comey, Dir., FBI, Remarks Delivered at the University of Chicago Law School: Law Enforcement and the Communities We Serve: Bending the Lines Toward Safety and Justice (Oct. 23, 2015), [] [hereinafter Comey, Remarks at the University of Chicago]. Comey issued a dire prediction about an explosion of urban, black crime enabled by the withdrawal or hesitance of police under pressure from the Black Lives Matter movement, which he did not reference by name. 12 See id. Comey embraced the so-called “Ferguson Effect” hypothesis, which conservative columnist Heather Mac Donald had advanced a few months earlier, writing about “The New Nationwide Crime Wave” that was generated “as officers scale[d] back on proactive policing under the onslaught of anti-cop rhetoric.” 13 Heather Mac Donald, The New Nationwide Crime Wave, Wall St. J. (May 29, 2015, 6:27 pm), (on file with the Columbia Law Review). What a decade earlier might have seemed a dream year in crime was now painted as a crime nightmare.

Comey’s remarks were infuriating to both supporters of the movement and the police but were quickly embraced by Trump, who made the need to reestablish law and order a centerpiece of his political campaign. 14 See supra note 3 and accompanying text. Trump then proceeded to deliver a string of brazen deceptions about crime that could easily be disproved. Picking up on the race-and-crime theme, Trump tweeted a graphic entitled “USA Crime Statistics—2015,” which showed a black man with a gun and a set of statistics stating that 81% of murdered whites were killed by blacks. 15 Jon Greenberg, Trump’s Pants on Fire Tweet that Blacks Killed 81% of White Homicide Victims, PolitiFact (Nov. 23, 2015), []. The number is ludicrous since murder is overwhelmingly intraracial, 16 FBI, Expanded Homicide Data Table 2: Murder Victims by Age, Sex, Race, and Ethnicity, 2015, FBI: Unif. Crime Reporting,
victims_by_age_sex_and_race_2015.xls [] (last visited Feb. 12, 2017).
but such claims encourage belief in the false narrative that the overwhelming threat to whites comes from black criminals. Trump’s graphic cited the “Crime Statistics Bureau,” a nonexistent organization. 17 When confronted about this by Bill O’Reilly, Trump replied: “I retweeted somebody that was supposedly an expert. Am I gonna check every statistic?” Ezra Klein, Donald Trump Is Too Gullible to Be President, Vox (Mar. 15, 2016, 10:18 am), [].

Further, in the days leading up to the election, Trump repeatedly made incorrect claims about the murder rate in the United States. For example, at a campaign rally in Cedar Rapids, Iowa, on October 28, 2016, Trump announced: “You won’t hear this from the media: We have the highest murder rate in this country in 45 years. You don’t hear that from these people. They don’t want to talk about it. The highest murder rate in the United States in 45 years.” 18 Michelle Ye Hee Lee, Trump’s False Claim that the Murder Rate Is the ‘Highest It’s Been in 45 Years,’ Wash. Post (Nov. 3, 2016),
news/fact-checker/wp/2016/11/03/trumps-false-claim-that-the-murder-rate-is-the-highest-its-been-in-45-years/?utm_term=.5fcc59c7f179 [].
Day after day on the campaign trail, Trump repeated the identical false assertion that the murder rate is the highest it has been in forty-five years. 19 Id. For further fact-checking of Trump’s claims on crime, see Lori Robertson, Dueling Claims on Crime Trend, Fact Check: The Wire (July 13, 2016), []; see also Philip Bump, Donald Trump’s Understanding of the Historical Nature of Crime Is Upside Down, Wash. Post (Aug. 29, 2016), [].

Amazingly, the press was not willing or able to effectively refute his wildly inaccurate claims. In fact, while the jump in the murder rate was the highest in many years, the murder rate in 2015 was around 50% lower than at its peak in 1991. 20 See FBI, Crime in the United States by Volume and Rate per 100,000 Inhabitants, 1991–2010, FBI: Unif. Crime Reporting, [] (last visited Feb. 1, 2017) (showing a 9.8 per 100,000 murder rate in 1991); FBI, 2015 UCR Data, supra note 4 (showing a 4.9 per 100,000 murder rate in 2015). Prior to Obama’s presidency, the last time the U.S. murder rate was as low as it was in 2015 was in 1963—fifty-two years prior. 21 James Alan Fox & Marianne W. Zawitz, Homicide Trends in the United States, Bureau of Justice Statistics (Jan. 25, 2010),
pdf/htius.pdf [] (showing the homicide rate in 1963 was 4.6 per 100,000).
Under both Uniform Crime Reporting (UCR) and the National Crime Victimization Survey (NCVS) crime measures, the country was better off in 2015 than before Obama’s presidency: From 2008 to 2015, the murder rate decreased from 5.4 murders to 4.9 per 100,000, 22 FBI, 2015 UCR Data, supra note 4. UCR violent crime went from 458.6 to 372.6 per 100,000, 23 Id. and NCVS violent crime went from 25.3 to 18.6 per 1,000 persons age twelve or older. 24 Jennifer L. Truman & Rachel E. Morgan, Bureau of Justice Statistics, Criminal Victimization, 2015, at 18 app. tbl.1 (2016),
pdf/cv15.pdf [] [hereinafter Truman & Morgan, NCVS Data].
The property crime drop during the first seven years of the Obama Administration was even larger. 25 In the UCR, property crime dropped from 3,214.6 per 100,000 in 2008 to 2,487.0 per 100,000 in 2015. See FBI, 2015 UCR Data, supra note 4. In the NCVS, there was a drop from 142.6 to 110.7 victimizations per 1,000 households over the same period. Truman & Morgan, NCVS Data, supra note 24, at 20 app. tbl.5.

At the same time that the right-wing media and politicians were making provocative comments on race and crime and emphasizing the large jump in murders in 2015, 26 See supra notes 11–19 and accompanying text. the Brennan Center was downplaying the size and significance of this development. 27 See, e.g., Ames Grawert & James Cullen, What the Data Tell Us About Crime and the ‘Ferguson Effect’, Brennan Ctr. for Just.: Commentary (Mar. 7, 2016), http://www.
E2%80%99 []. The Brennan Center statements were in turn harshly criticized by police researcher Peter Moskos. Peter Moskos, Who You Gonna Believe?, Cop in the Hood Blog (Apr. 21, 2016),
2016/04/who-you-gonna-believe.html [] [hereinafter Moskos, Who You Gonna Believe].
Meanwhile, the New York Times blasted Comey for spreading “the false notion that the country is entering a crime wave that is some how [sic] related to the public backlash against police brutality.” 28 Editorial, Don’t Blame the Video, N.Y. Times (May 16, 2016), (on file with the Columbia Law Review).

Darrel Stephens, Executive Director of the Major Cities Chiefs Association and former police chief in Charlotte for nine years, spoke for many in law enforcement in stating, “I personally don’t really believe that police officers across America have pulled back from doing their job.” 29 Max Ehrenfreund & Denise Lu, More People Were Killed Last Year than in 2014, and No One’s Sure Why, Wash. Post (Jan. 27, 2016),
graphics/national/2015-homicides/ [].
According to the New York Times, the National Fraternal Order of Police criticized Comey for suggesting police “were afraid of doing their jobs.” 30 Don’t Blame the Video, supra note 28. Law Enforcement Leaders to Reduce Crime and Incarceration rejected Comey’s suggestion that the “viral video effect” was responsible for rising crime rates, calling this proposition “unfounded, and frankly, damaging to the efforts of law enforcement.” 31 Id. In May 2016, the White House Press Secretary agreed, explaining:

[T]here still is no evidence to substantiate the claim that the increase in violent crime is related to an unwillingness of police officers to do their job. . . . [T]he President’s point is that as we consider policy approaches to addressing those communities where we’ve seen an uptick in violent crime, we need to be making policy decisions that are based on facts and evidence, and not anecdotes. 32 Press Release, Press Sec’y Josh Earnest, White House Press Briefing (May 12, 2016), [].

A comprehensive analysis of all federal crime data for the year reveals that the full picture of crime in 2015 was more complex than Comey’s remarks suggested. 33 See infra Part II (undertaking a more detailed analysis of 2015 crime data). In contrast to the terrible news on murders, 34 See infra section II.A.1. the news about property crime in 2015 continued to be entirely positive, 35 See infra section II.A.3. and the news about violent crime overall was mixed and complicated 36 See infra section II.A.2. but again far more positive than Comey’s and Trump’s apocalyptic claims would have suggested. Moreover, they offered no empirical support for linking a nationwide explosion of violent crime to blacks. This Essay examines these complexities. It seeks to explicate what most Americans should have known about crime going into this presidential election, how challenging it is to know what is happening with crime right now, and whether we are witnessing a reversal of a long-term downward trend in crime.

Part I discusses the major trends in crime over the last half century and notes that, because of the enormous post-1993 crime drop, we are currently in a relatively good position in terms of crime—as long as the spike in murder does not persist. Part II takes a closer look at crime in 2015, and Part III distills the stylized fact that while the news for murder was quite bad and the news for property crime was quite good, there are conflicting signals about what really happened to violent crime in 2015. Nonetheless, we can say with confidence that Comey was incorrect in saying there was an overall violent crime explosion in 2015 (although one might speak in those terms for the jump in murders, but not with the extreme racial overtones that Comey and Trump advanced).

Having established the basic facts of crime in 2015, we then turn in Part IV to an examination of the major causal influences on crime in that year. Declining incarceration and police-employment rates, as well as growing markets for illegal drugs, were large stimulants to crime, particularly for murder, while a series of unique factors have led to lower property crime. Part V points out the imprecision in the Ferguson Effect hypothesis and notes that de-policing likely has little causal significance for the nationwide jump in homicides in 2015, even if some version of it can apply in selected cities at particular times. The Essay concludes, noting that Comey and Trump probably contributed to the increase in the public’s apprehension of crime, which likely aided Trump’s law-and-order candidacy.

I. The Major Trends in U.S. Crime over the Last Half Century

In his initial October 2015 speech and his continuing subsequent commentary, Comey—referencing the crack-epidemic-fueled jump in crime in the late 1980s—announced: “I fear we are facing another wave of violent crime and homicide, and our communities are once again in trouble . . . . Police chiefs say the increase [in homicide victims] is almost entirely among young men of color, at crime scenes in bad neighborhoods where multiple guns are being recovered.” 37 Comey, Remarks at the University of Chicago, supra note 11 (emphasis added).

To assess these claims, this Essay begins by considering the long-run trends in violent crime and murder to compare the state of crime today to the pre-1993 situation and then looks more closely at violent crime in 2015 (and beyond to the extent the limited data allows). It then addresses causal stories, including this Ferguson Effect hypothesis, but only after establishing the basic facts about murder and property crime and discussing the conflicting data on violent crime.

A. U.S. Crime Data—The UCR and NCVS

The FBI’s UCR compiles crime data from around 18,000 agencies into an annual, official report for the eight major index crimes. 38 See FBI, UCR Offense Definitions, FBI Unif. Crime Reporting Statistics, [] [hereinafter FBI, UCR Offense Definitions] (last updated Jan. 26, 2017). In 2015, 16,643 of the 18,439 UCR-eligible agencies submitted data. See FBI, CIUS Summary, FBI: Unif. Crime Reporting, [] (last visited Feb. 1, 2017). While the UCR data on homicides is quite reliable, the data on other crimes have some significant limitations. First, because the program is not mandatory, some agencies fail to submit any information, forcing the FBI to estimate the missing data. 39 FBI, About CIUS: Methodology, FBI: Unif. Crime Reporting,
crime-in-the-u.s/2010/crime-in-the-u.s.-2010/methodology [] (last visited Feb. 5, 2017) (cautioning users of UCR data that reporting problems require “the FBI to estimate crime counts for certain contributors”).
Second, among agencies that do submit, some of the reported information may suffer from errors or even intentional police department deception. 40 For an extended discussion of flaws in the UCR data, see generally Michael Maltz, Bureau of Justice Statistics, NCJ 176365, Bridging Gaps in Police Crime Data (1999), []. In the year after a UCR release, agencies can continue to submit updated, corrected data. These corrections are reflected in the national total when the following year’s UCR is released (that is, the 2015 national tables of the UCR contain adjustments for 2014). The adjustment to the 2014 murder numbers that the FBI issued when it published 2015 numbers lowered the 2014 murder rate from 4.47 per 100,000 people to 4.44, which had the effect of elevating the increase in the murder rate for 2015 from 9.6% to 10.0%—as the 2015 murder rate was 4.88 per 100,000 people. See FBI, Table 1: Crime in the United States by Volume and Rate per 100,000 Inhabitants, 1995–2014, FBI: Unif. Crime Reporting, [] (last visited Feb. 8, 2017) (showing 14,249 murders occurred in a population of 318,857,056 in 2014); FBI, 2015 UCR Data, supra note 4 (showing 14,164 murders occurred in a population of 318,907,401 in 2014 and 15,696 murders occurred in a population of 321,418,820 in 2015). Finally, the police themselves do not hear of all the crimes. For a variety of reasons, victims often choose not to report crimes, and this is especially true in areas with low trust of police. 41 Lynn Langton et al., Bureau of Justice Statistics, NCJ 238536, Victimizations Not Reported to the Police, 2006–2010, at 1 (2012),
pdf/vnrp0610.pdf [].

Thankfully, a second governmental data source, the NCVS, tries to address some of these shortcomings. The NCVS is a national survey on crime administered by the U.S. Census Bureau on behalf of the Department of Justice. 42 See Data Collection: National Crime Victimization Survey (NCVS), Bureau of Justice Statistics, [] [hereinafter BJS, NCVS Methodology] (last visited Feb. 5, 2017). It consists of responses by approximately 96,000 households composing a nationally representative sample. 43 Id. Participants answer questions about recent victimizations, as well as whether they reported the crime to the police. 44 Id. By collecting data from citizens themselves, the NCVS sidesteps the limitations of UCR counts based only on crimes known to police. Because of the relatively small sample, however, it can provide reliable annual crime estimates only at the national level.

While the goal of the NCVS and UCR data is to provide an accurate picture of violent and property crime in America, one needs to be mindful that they cover distinct subsets of crime. The most notable difference is for murder, which is the best-measured crime in the UCR but which does not show up in a victim survey for obvious reasons. 45 In addition, the UCR includes crimes against children under eleven, while the NCVS does not. See id. (noting data is collected only from persons twelve years old and above). The NCVS, meanwhile, includes simple assault and sexual assault, 46 See Truman & Morgan, NCVS Data, supra note 24, at 7 tbl.5. while the UCR index is limited to the more serious crimes of aggravated assault and rape. 47 See FBI, UCR Offense Definitions, supra note 38. Importantly, since the NCVS captures a great deal of the unreported crime that is not reflected in the UCR, 48 See FBI, The Nation’s Two Crime Measures, FBI Unif. Crime Reporting Statistics, [] [hereinafter FBI, Two Crime Measures] (last updated Jan. 26, 2017) (“The BJS established the NCVS to provide previously unavailable information about crime (including crime not reported to police), victims, and offenders.”). the NCVS counts of violent and property crime tend to be far higher than the UCR measures.

Finally, since the NCVS is based on a sample, 49 BJS, NCVS Methodology, supra note 42. it will have a predictable amount of sampling error, which leads the NCVS to calculate standard errors for its estimates. This means that a listed NCVS change in crime of, say, a drop of 3% with a standard error of 2 should be thought of as evidence that the actual change in crime that year is roughly between a drop in crime of 7% and an increase in crime of 1% (at the 95% confidence level). The bottom line, then, is that unless one is looking at murder, the reliability of any single-year movement in a UCR or NCVS crime category needs to be carefully evaluated.

B. Murder, Violent Crime, and Property Crime over Time

Figure 1: UCR Murder Rate

Figure 2: UCR Violent Crime Rate

Figure 3: NCVS Violent Crime Rate

The country’s remarkable post-1993 crime drop is reflected in both the UCR murder and violent crime measures, shown in Figures 1 and 2, as well as in the NCVS violent crime measures shown in Figure 3. According to the UCR, from 1993 to 2015, the U.S. murder rate fell 49%, 50 According to the UCR, the murder rate was 9.5 per 100,000 in 1993 and 4.9 per 100,000 in 2015. FBI, State-by-State and National Crime Estimates by Year, FBI Unif. Crime Reporting Statistics, (select “United States-Total,” “Violent crime rates,” and “1993” to “2014”) (on file with the Columbia Law Review) [hereinafter FBI, UCR Violent Crime Rate Search] (last visited Feb. 1, 2017); FBI, 2015 UCR Data, supra note 4. violent crime fell 49%, 51 The UCR violent crime rate was 747.1 per 100,000 in 1993 and 372.6 per 100,000 in 2015. FBI, UCR Violent Crime Rate Search, supra note 50; FBI, 2015 UCR Data, supra note 4. and property crime fell 45%. 52 The UCR property crime rate was 4,740.0 per 100,000 in 1993 and 2,487.0 per 100,000 in 2015. FBI, State-by-State and National Crime Estimates by Year, Unif. Crime Reporting Statistics, (select “United States-Total,” “Property crime rates,” and “1993” to “2014”) (on file with the Columbia Law Review) [hereinafter FBI, UCR Property Crime Rate Search] (last visited Feb. 1, 2017); FBI, 2015 UCR Data, supra note 4. According to the NCVS, meanwhile, the drops in violent and property crime (from 1993 to 2015) have been even more dramatic at 77% and 69%, respectively. 53 Truman & Morgan, NCVS Data, supra note 24, at 18 app. tbl.1 (showing a drop in violent crime from 79.8 to 18.6 per 1,000 persons from 1993 to 2015); id. at 20 app. tbl.5 (showing a drop in property crime from 351.8 to 110.7 per 1,000 households from 1993 to 2015). As I wrote in 1998: “We would achieve an enormous public policy victory if we could engineer a return to the low crime rates of the 1950s and early 1960s.” 54 John J. Donohue, Understanding the Time Path of Crime, 88 J. Crim. L. & Criminology 1423, 1428 (1998) [hereinafter Donohue, Understanding the Time Path]. Happily, that victory was achieved during the Obama Administration: In 2014, the murder rate, at 4.4 per 100,000, had dropped to the lowest level seen in the United States since 1957! 55 See Fox & Zawitz, supra note 21 (showing the homicide rate in 1957 was 4.0 per 100,000). Even with the jump in the murder rate that came the following year, the last time the U.S. murder rate was as low as it was in 2015 prior to Obama’s presidency was in 1963—fifty-two years prior. 56 See supra note 21 and accompanying text. This was astonishingly good news for the country. Well over 100,000 lives were saved by the huge post-1993 drop in the murder rate, and millions avoided being victimized by violent and property crime.

C. The Incomplete Success of the Great American Crime Drop

In one important respect, though, the success of the great American crime drop has not been as complete as the crime data would suggest. Whether we have returned to the murder rates of 1963 or 1957, the fact is that we have not yet achieved what I hoped for in 1998—the level of underlying criminality of the late 1950s or early 1960s. The ostensible contradiction is explained by the fact that while the observed murder rate has fallen to the level of a half century ago, the implicit murder rate is substantially higher today. 57 Of course, the same argument applies with regard to all crimes. Since murder is the only crime that has been accurately measured in the United States for over fifty years, the text focuses on this crime, without loss of generality. The United States responded to the massive crime increases of the late 1960s and 1970s by dramatically increasing the incarceration rate to the highest level the world has ever seen 58 See E. Ann Carson & Daniela Golinelli, Bureau of Justice Statistics, Prisoners in 2012: Trends in Admissions and Releases, 1991–2012, at 1 fig.1 (2014), http://www.bjs.
gov/content/pub/pdf/p12tar9112.pdf [].
and by substantially increasing public and private police forces and other anticrime measures. 59 See infra note 158 and accompanying text. With these suppressive forces at work, we have returned to an observed murder rate of an earlier time, but without these forces in play today, the murder rate would be substantially higher. 60 See infra notes 160–161 and accompanying text. Thus, the implicit murder rate of today—the murder rate that would exist if the criminal justice policy of, say, 1963 were currently in place—is substantially higher than the actual murder rate of today (or of 1963).

This discussion highlights that at any one point in time, the observed crime rate reflects the operation of many suppressing and stimulating influences on crime. Evanescent influences, whether benign or malign, may temporarily move the crime rate above or below any underlying time path, but they will not disrupt the longer-term trend. On the other hand, more sustained influences can alter the course of crime more powerfully, for good or ill.

D. American Misperceptions About Changes in Crime

Figure 4: Percentage of Americans Reporting Crime and NCVS Violent Crime Rate


Rivaling the astonishing crime drop is the level of ignorance Americans have about this good news. Figure 4 documents, for each year from 1993 to 2015, the percentage of Americans reporting that crime has gone up that year (in a Gallup poll 61 Justin McCarthy, More Americans Say Crime Is Rising in U.S., Gallup (Oct. 22, 2015), (on file with the Columbia Law Review). ), alongside the actual NCVS violent crime rate. 62 Truman & Morgan, NCVS Data, supra note 24, at 1 fig.1. Despite the long-term decline in violent crime, since 9/11 60%­ to 70% of Americans have consistently answered that crime has risen over the last year. 63 See McCarthy, supra note 61. Even with the broad crime drop experienced during the Obama Administration, 63% of Americans thought crime was higher in 2014 64 Id. —which, in reality, was the year properly extolled by Comey for its low crime rate. In the next year—which witnessed the biggest jump in murders in years—the percentage of Americans with the same opinion of rising crime increased but only to 70%. 65 Id.

It is not surprising that Americans are woefully ignorant about the level and trends of crime in the United States. Unscrupulous politicians and their supporters, the gun industry (using fear of crime to help sell guns and elect pro-gun legislators), and parts of the media constantly seek to scare the public with alarmist crime stories. 66 See supra notes 13–19 and accompanying text; see also Jarrett Murphy, How the Gun Industry Got Rich Stoking Fear About Obama, Nation (Aug. 22, 2012), [] (noting the gun industry’s incentive and practice of portraying “a picture of criminals constantly menacing our families and a government hellbent on taking our guns”). This degree of voter ignorance emboldens demagogues and increases their power to peddle untruths.

In addition to the distorting factors mentioned previously, the public perception may have been influenced by the increase in one particular type of highly publicized, shocking crime: mass shootings. The long-term downward trend in crime has been accompanied by a concurrent upward trend in mass shootings, from an average of 2.7 events per year in the 1980s to an average of 4.5 events per year from 2010 to 2013. 67 William J. Krouse & Daniel J. Richardson, Cong. Research Serv., R44126, Mass Murder with Firearms: Incidents and Victims, 1999–2013, at 14–15 (2015), []; Mark Follman, Yes, Mass Shootings Are Occurring More Often, Mother Jones (Oct. 21, 2014, 5:05 am), []. While the total number of deaths in these mass shootings has been too small relative to the overall homicide rate to overcome the downward trend in murders, mass shootings are particularly high-visibility events, endlessly reported on by the cable news networks, quite shocking to the public, and unsettling to the sense of public safety. 68 Notably, the average number of mass murders per person is twice as high among states with no background checks as it is for those with background checks (on average across states between 2009 and July 2015, 8.4 versus 4.0 mass shootings per 100 million person-years, a metric representing the cumulative number of years residents had lived in a state in a given year). See State Background Check Requirements and Mass Shootings, Everytown (Nov. 12, 2015), [] (last updated Aug. 31, 2016) (“[C]ontrolling for population, in states that require background checks for all handgun sales, there were 52 percent fewer mass shootings between January 2009 and July 2015.”). The striking disparity warrants further investigation to determine whether universal background checks would reduce the frequency of mass shootings. Horrific mass shootings—such as those perpetrated by white supremacists at a church in Charleston 69 Alan Blinder & Kevin Sack, Dylann Roof Found Guilty in Charleston Church Massacre, N.Y. Times (Dec. 15, 2016), (on file with the Columbia Law Review) (noting Roof killed nine members of a black church in Charleston in June 2015). and at Umpqua Community College in Oregon, 70 Julie Turkewitz, Oregon Gunman Smiled, Then Fired, Student Says, N.Y. Times (Oct. 9, 2015), (on file with the Columbia Law Review) (noting eight students and a professor were killed in an attack in Oregon in October 2015). and by ISIS sympathizers at Inland Regional Center in San Bernardino 71 Christine Hauser, San Bernardino Shooting: The Investigation So Far, N.Y. Times (Dec. 4, 2015), (on file with the Columbia Law Review) (noting fourteen were killed in December 2015). and at Pulse in Orlando 72 Gregor Aisch et al., What Happened Inside the Orlando Nightclub, N.Y. Times (June 12, 2016), (on file with the Columbia Law Review) (noting a gunman killed forty-nine in a June 2016 attack). —although small in number compared to the total number of homicides, generated widespread apprehension and rich material for the Trump campaign.

II. A Closer Look at Crime in 2015

Before we begin speculating about the cause of the striking and unwelcome 9.95% increase in the 2015 murder rate, 73 See FBI, 2015 UCR Data, supra note 4 (showing the murder rate was 4.9 per 100,000 in 2015 and 4.4 per 100,000 in 2014). it is necessary to analyze crime far more deeply than simply noting that a one- or even two-year jump in murder has occurred. First, it is important to remember that past claims that a crime jump portends a significant and sustained reversal in the benign trend in crime have often proven to be wrong. For example, in 2006 a wave of pessimistic news articles appeared that decried the end of the great crime decline. USA Today published a piece entitled “Cities See Crime Surge as Threat to Their Revival,” 74 Haya El Nasser, Cities See Crime Surge as Threat to Their Revival, USA Today (Jan. 24, 2007), []. while the Associated Press announced that “[v]iolent crime in the U.S. is on the rise, posting its biggest one-year increase since 1991.” 75 Associated Press, Violent Crime Rises in Some Big Cities, L.A. Times (June 13, 2006), [
Even a police advocacy group, the Police Executive Research Forum, contributed a report ominously titled “A Gathering Storm—Violent Crime in America.” 76 Police Exec. Research Forum, A Gathering Storm—Violent Crime in America (2006),
20storm%20-%20violent%20crime%20in%20america%202006.pdf [
In retrospect, we can see that those concerns of resurgent violent crime were unfounded. As Figures 1 to 3 illustrate, the 2006 crime increase gave way to a continued decline in both murder and violent crime more generally.

Second, it is important to look at all crime data to fully understand both changes in crime and possible causal explanations for those changes. The three short red (darker) portions of Figure 1 show the three years since 1977 with increases in the rate of murder that were comparable to that in 2015. 77 The highlighted red portions of Figure 1 (with murder growth in parentheses) are 1978 to 1979 (9.72%), 1985 to 1986 (8.62%), and 1989 to 1990 (9.01%). The most recent unusually high jumps in murders that occurred prior to 2015 were in 1990, 1986, and 1979. 78 See supra Figure 1. But an important fact that distinguishes these years from 2015 is that in all of the earlier high-murder-spike years, the country also experienced double-digit increases in UCR overall violent crime that were each greater in magnitude than the homicide changes. 79 The UCR violent crime growth rate was 11.3% from 1978 to 1979, 12.1% from 1985 to 1986, and 10.2% from 1989 to 1990. FBI, State-by-State and National Crime Estimates by Year, FBI Unif. Crime Reporting Statistics,
Search/Crime/State/StatebyState.cfm (select “United States-Total,” “Violent crime rates,” and “1970” to “2014”) (on file with the Columbia Law Review) (last visited Feb. 1, 2017).
In 2015, the UCR jump in violent crime was nowhere near that level, at only 3.0%. 80 See FBI, 2015 UCR Data, supra note 4.

A. UCR and NCVS Data on Murder, Violent Crime, and Property Crime in 2015

1. Murder in 2015.

Figure 5: Actual Versus Predicted Murder Rate from UCR Data 81 Figure 5 documents the path of the murder rate in the United States since 1976.

The eight years of the Clinton Administration stand out as perhaps the greatest period of sustained crime decline in the nation’s history, rivaled only by the drop ushered in by the ending of Prohibition in 1933. 82 Donohue, Understanding the Time Path, supra note 54, at 1447 n.40. The murder rate in the United States fell from 9.5 per 100,000 in 1993 to 5.5 in 2000, Bill Clinton’s final year in office—a whopping drop of over 42%. 83 FBI, UCR Violent Crime Rate Search, supra note 50. During the George W. Bush years, the period of a sharply falling murder rate ended. 84 Id. By his last year in office in 2008, the murder rate was virtually unchanged at 5.4 per 100,000. 85 Id. The downward trend in the murder rate was renewed under Obama, but the murder rate jumped up sharply in 2015. 86 FBI, 2015 UCR Data, supra note 4. Note that after 1993, the murder rate dropped year after year at a brisk 9% per year before slowing to a trend rate drop of 1.6% per year after the decline in the Clinton years. 87 FBI, UCR Violent Crime Rate Search, supra note 50. The fact that the post-2000 murder rate data fit so tightly on a downward linear trend—as reflected in the R-squared value of .95 for the entire time series—tells us that, as of 2015, even the dramatic one-year jump in the murder rate would not be enough to justify the conclusion that crime was suddenly on some sustained new upward trajectory. 88 Of course, many different trend lines could be drawn through the time series shown in Figure 5, but the depicted three-segment line is the figure that maximizes the R-squared value conditional on having only two break points. The resulting R-squared value is 0.95, which is a remarkably high value for such a sparse model. Only time will tell if a long-time reversal in the murder rate has begun.

Figure 1 shows that the murder rate in the United States in 2015 was 4.88 89 See FBI, 2015 UCR Data, supra note 4 (showing 15,696 murders occurred in a population of 321,418,820 in 2015). (Figure 5 shows a value of 1.6, which is the natural log of the murder rate), and the first time it had dropped below 5 per 100,000 in decades was in 2010, when it fell to 4.76. 90 See id. (showing 14,722 murders occurred in a population of 309,330,219 in 2010). While the increase in homicides in 2015 was alarming, the murder rate in 2015 stood out only in comparison to the exceptionally low murder rates of 4.52 and 4.44 of 2013 and 2014, respectively. 91 See id. (showing 14,319 murders occurred in a population of 316,497,531 in 2013 and 14,164 murders occurred in a population of 318,907,401 in 2014).

Nonetheless, the jump in the 2015 murder rate was unusually, indeed shockingly, large, representing roughly 1,500 more murders than we would have had if the low (given the unseemly high American standards) 2014 rate could have been preserved. That would be a highly unfortunate development even if it were quickly reversed. The big picture, however, is that the United States is nowhere near the level of killing of the early 1990s, when the murder rate was roughly twice the level of 2015.

Since murder is the rarest and hence most volatile of major crimes, a one-year jump can quickly return to trend, as it often has over the last forty years. But the still-incomplete evidence 92 Federal crime data for 2016 will not be available until late October 2017. from selected cities suggests that 2016 will again have an increase in the murder rate. Partway through 2016, crime analyst Jeff Asher aggregated murder data for seventy-nine cities with populations over 250,000. 93 Jeff Asher, A Handful of Cities Are Driving 2016’s Rise in Murders, FiveThirtyEight (Oct. 6, 2016, 7:00 am), []. For these cities, the estimated increase in murder count—not rate—is around 10.5% (compared to 14.7% for the same group in 2015). 94 Id. Chicago and Orlando (primarily due to the Pulse nightclub shooting that killed forty-nine) had a noticeable influence on this increase: If one excludes those two cities, the murder increase for the remaining seventy-seven cities is only 6.3%. 95 Id. In December 2016, the Brennan Center, looking at a smaller set of twenty-six of the thirty largest cities in America, projected that the 2016 murder rate would be 14% higher than in 2015 for those cities (which included the bloated murder jump in Chicago, diluted by fewer other locales). 96 Matthew Friedman et al., Brennan Ctr. for Justice, Crime in 2016: Updated Analysis 1 (2016),
2016_Updated_Analysis.pdf [].

Since the murder hike will be larger in the major urban areas than in the rest of the country, the combined information from these two datasets suggests that the murder rate will rise again in 2016, although probably less than in 2015. But even a 5% increase following last year’s 9.95% increase in the murder rate would be a source of substantial concern given the size of the added body count, particularly if this were to signal a structural break in crime rather than a temporary deviation around the previous benign trend. 97 Executive Director of the Major Cities Chiefs Association Darrel Stephens said, “I don’t know if we’ve got the beginning of a longer term trend in the change in violent crime . . . . I think it’s too soon for anybody to really know that for sure.” Gabriel Dance & Tom Meagher, Crime in Context, Marshall Project (Aug. 18, 2016, 2:20 pm), [].

2. Violent Crime According to the UCR and NCVS. — The UCR and NCVS show conflicting movements for 2015. The UCR shows violent crime rate increasing by 3%, 98 See FBI, 2015 UCR Data, supra note 4. while the NCVS shows the rate of violent crime declining by 7.5%. 99 See Truman & Morgan, NCVS Data, supra note 24, at 2 tbl.1 (showing the violent crime rate declined from 20.1 in 2014 to 18.6 in 2015). Recalling that the NCVS captures simple assault while the UCR does not (counting only aggravated assaults), 100 FBI, Two Crime Measures, supra note 48. one might speculate that the conflicting movements in the NCVS and UCR violent crime rates for 2015 could be explained by the considerably more numerous simple assaults dropping even as the more serious aggravated assaults rose. Fortunately, the NCVS also provides breakdowns of its violent crime category, and it shows that not only did both aggravated and simple assaults drop in 2015 but the percentage drop was substantially greater for aggravated assaults. 101 See Truman & Morgan, NCVS Data, supra note 24, at 2 tbl.1 (showing per 1,000 persons age twelve and older, the simple assault rate dropped from 12.4 in 2014 to 11.8 in 2015, and the aggravated assault rate dropped from 4.1 to 3.0 in the same period). Indeed, the NCVS drop in the rate of aggravated assaults in 2015 was roughly 27% (while the drop in the simple assault rate was about 4.8%). 102 Id.

We see the same basic story if we limit our focus to the following NCVS subcategories: (1) the serious violent crime rate fell almost 12%, (2) serious violent crime involving weapons fell almost 27%, and (3) serious violent crime involving injury fell almost 8%. 103 Id. On the other hand, owing to the size of the standard errors on these NCVS figures, 104 See supra text accompanying note 49. none of the NCVS violent crime drops I have referenced here are statistically significant at the .05 level. 105 Three NCVS violent crime drops were significant at the lower .10 level: rape or sexual assault, aggravated assault, and serious violent crime involving weapons. See Truman & Morgan, NCVS Data, supra note 24, at 2 tbl.1. Consequently, while the NCVS sample on its face suggests that all of these violent crime rates fell in 2015, the NCVS states that there was no “statistically significant” drop in the rate of overall violent crime in 2015. 106 Id. at 1. The true level of violent crime would be captured in a 95% confidence interval that would suggest anything from a modest rise in violent crime (but certainly nothing comparable to the jump in the murder rate) to a decline that is substantially greater in absolute value than the murder rate increase. 107 See supra note 49 and accompanying text.

Interestingly, the NCVS provides data not only on total victimizations but also on the proportion of the population that suffers any violent crime during the year (the latter number counting not how many times someone was victimized in the year but just whether that person was victimized during the year). 108 While the NCVS recorded 5,006,620 violent crimes in 2015, only 2,650,670 individuals were victimized, underscoring that many Americans experience multiple victimizations within the same year (in separate incidents). See Truman & Morgan, NCVS Data, supra note 24, at 2 tbl.1, 12 tbl.9. To provide context, the total number of violent crime victims in 2015 was roughly the population of the city of Chicago—estimated at 2,720,546 persons in 2015. See QuickFacts: Chicago City, Illinois, U.S. Census Bureau, [http://perma.
cc/3R7H-D68W] [hereinafter QuickFacts: Chicago] (last visited Feb. 1, 2017).
The NCVS tells us that in 2015, 0.98% of all persons age twelve or older (roughly 2.7 million persons) experienced at least one violent victimization, which was a statistically significant drop from the level of 1.11% in 2014. 109 Truman & Morgan, NCVS Data, supra note 24, at 12 tbl.9. In 2015, it was thus the first time that NCVS data showed less than 1% of Americans experienced a violent crime victimization in the past year. Disaggregating this violent crime prevalence category, the drop in crime prevalence from 2014 to 2015 was also statistically significant at the .05 level for overall assault and at the .10 level for aggravated assault, domestic violence, and stranger violence. 110 Id.

If we limit our focus to subcategories of what the NCVS counts as serious violent crime (rape or sexual assault, robbery, and aggravated assault), the drop in crime prevalence from 2014 to 2015 was statistically significant at the .05 level for serious stranger violence and serious violent crime involving weapons. 111 Id. The drop in prevalence of property crime was also statistically significant at the .05 level (from 7.99% to 7.60% of the population). 112 Id.

3. Property Crime According to the UCR and NCVS. — While the UCR and NCVS violent crime numbers gave a conflicting picture of what happened in 2015, all of the national evidence on property crime pointed in one direction—down. The UCR reported that the property crime rate fell 3.4% in 2015, 113 FBI, Property Crime, FBI: Unif. Crime Reporting, [] (last visited Feb. 1, 2017). and the NCVS reported the property crime rate dropped 6.3%, which was statistically significant at the .05 level. 114 Truman & Morgan, NCVS Data, supra note 24, at 5 tbl.3 (showing a drop in property crime from 118.1 per 1,000 households in 2014 to 110.7 per 1,000 households in 2015).

B. State Analysis of UCR Data in 2015

Figure 6: UCR Crime Data by State

Figure 6 examines the pattern of changes in the three UCR crime categories in 2015 for all fifty states and the District of Columbia (recall there is no state data released for the NCVS because the sample size is too small for state-by-state analysis). 115 Initial data were obtained from FBI, Table 4: Crime in the United States by Region, Geographic Division, and State, 2014–2015, FBI: Unif. Crime Reporting, [http://] [hereinafter FBI, Crime by Region] (last visited Feb. 1, 2017).
The figure reveals that many more states experienced murder rate increases (thirty-seven) than decreases (ten) in 2015 (with four states showing no change), and in twenty-nine states and the District of Columbia the jump in the murder rate was at least 5%. On its face, this widespread pattern of murder rate increases calls into question the claim that the crime explosion was purely the black, urban phenomenon espoused by Comey at his 2015 University of Chicago speech and by others who have insisted on the presence of a nationwide Ferguson Effect. 116 See supra text accompanying notes 11–13. Interestingly, only twenty states had a jump in the violent crime rate of that magnitude in 2015 (thirty-seven states had some increase, with five showing no change), and only three had such a jump in property crime (nine had some increase). 117 FBI, Crime by Region, supra note 115. These changes are reflected in Figure 6.

According to UCR data, the five states with the highest increases in the violent crime rate in 2015 were Hawaii (24.0% increase), South Dakota (16.7%), Vermont (15.1%), Alaska (14.9%), and Wyoming (13.7%) 118 Id. —all states with strikingly low black populations. 119 Hawaii’s black population was 2.6%. QuickFacts: Hawaii, U.S. Census Bureau, [] (last visited Feb. 1, 2017). South Dakota’s black population was 1.8%. QuickFacts: South Dakota, U.S. Census Bureau, [] (last visited Feb. 1, 2017). Vermont’s black population was 1.3%. QuickFacts: Vermont, U.S. Census Bureau,
table/PST045216/50 [] (last visited Feb. 1, 2017). Alaska’s black population was 3.9%. QuickFacts: Alaska, U.S. Census Bureau, http://www.census.
gov/quickfacts/table/PST045216/02 [] (last visited Feb. 1, 2017). Wyoming’s black population was 1.4%. QuickFacts: Wyoming, U.S. Census Bureau, [] (last visited Feb. 1, 2017).
If one were to determine the worst deterioration in crime performance in 2015 using a metric of large increases in murder and violent crime as well as increasing property crime, the two worst states would be South Dakota, with rates of murder up 38.3%, violent crime up 16.7%, and property crime up 3.4%, and Alaska, with rates of murder up 43.6%, violent crime up 14.9%, and property crime up 2.1%. 120 FBI, Crime by Region, supra note 115. Note that while Comey was pointing toward an explosion in black inner-city crime as the reason for the 2015 crime jump, 121 See supra text accompanying notes 11. there was clearly much more going on with crime in 2015 than Comey addressed or can be explained by his police-retrenchment story. Comey’s black, urban crime rise story undoubtedly plays out in a few cities but does not ring true for the nation as a whole. 122 For example, the District of Columbia—where black residents compose 48.3% of the population—had a 51.4% increase in its murder rate but only a 2.1% jump in the violent crime rate and a 9.6% drop in the property crime rate. See FBI, Crime by Region, supra note 115; QuickFacts: District of Columbia, U.S. Census Bureau, http://www. [] (last visited Feb. 5, 2017). Many other states with low black populations also had substantially above-average jumps in the murder rate in 2015: Iowa (19.4% increase in murder rate; 3.5% black population), Nebraska (16.2%; 5.0% black population), New Mexico (15.9%; 2.6% black population), Oregon (16.2%; 2.1% black population), and Washington (16.8%; 4.1% black population). FBI, Crime by Region, supra note 115; see also QuickFacts: Iowa, U.S. Census Bureau, [http://perma.
cc/JJZ3-TBE2] (last visited Feb. 5, 2017); QuickFacts: Nebraska, U.S. Census Bureau, [] (last visited Feb. 5, 2017); QuickFacts: New Mexico, U.S. Census Bureau, [] (last visited Feb. 5, 2017); QuickFacts: Oregon, U.S. Census Bureau, [] (last visited Feb. 5, 2017); QuickFacts: Washington, U.S. Census Bureau, [] (last visited Feb. 5, 2017).

It seems highly implausible that the anomalous jumps in murder in these states and other low-black-population states can simply be explained by unleashed black, urban crime because of police retrenchment in the wake of protests over the killing of unarmed blacks. In fact, there is zero relationship between the percentage of blacks overall or young black males and the change in the murder rate across states in 2015. 123 This is true whether or not one weights by state population.

III. The Stylized Facts of Crime: 2015 to 2016

This Essay’s exploration of the full set of federal crime data from the UCR and NCVS for 2015 and the limited murder data from selected cities in 2016 underscores the fact that all of our knowledge about crime is probabilistic and that crime data can be uncertain and conflicting. This Part examines some of the uncertainties resulting from these data.

A. Do Murder Rates and Violent Crime Rates Always Move in Sync?

Police researcher Peter Moskos, though recognizing that the UCR increase in violent crime was far more modest than the large spike in the murder rate, dismissed its significance, noting that “violent crime and homicide are always correlated.” 124 Moskos, Who You Gonna Believe, supra note 27 (emphasis omitted) (critiquing the Brennan Center’s analysis of UCR crime statistics). He explained the “apparent dis­parity” as a reflection of “less reporting of crime, particularly in cities where crime is going up. Bodies get counted. Bruises less so.” 125 Id. (emphasis omitted). Furthermore, he noted, proactive policing can increase the reporting of crimes, since “some crimes are only recorded because police took the initiative and made an arrest.” 126 Id. (emphasis omitted).

The hypothesis that big jumps in murder might lead to under­counting of violent crime did not hold true in the three prior years of large murder rate hikes, 127 See supra Figure 1 (highlighting increases greater than 8.5% in the murder rate in 1979, 1986, and 1990). in which each measured violent crime increase exceeded that year’s murder rate increase. 128 See supra Figure 2 (tracking the increase in violent crime since 1975 and showing large jumps in the aforementioned years). Moreover, while Moskos’s claims seem reasonable, one need not rely only on police-collected UCR violent crime data. NCVS data are collected to deal precisely with this issue, and they do not support the view that the 2015 homicide increase is a better measure of the violent crime movement than the far-lower UCR figure. Indeed, the percentage drop in the NCVS violent crime data for 2015 is 7.5%. 129 Truman & Morgan, NCVS Data, supra note 24, at 2 tbl.1 (showing the rate declined from 20.1 in 2014 to 18.6 in 2015). If one can rely on this NCVS point estimate, the United States in 2015 may well have experienced increasing murders and falling overall violent crime. The United States certainly experienced less property crime (as both the UCR and NCVS con­firmed). 130 See supra section II.A.3. Even though the near-double-digit increase in the murder rate is of paramount concern, to ignore the major divergence in crime that emerged in 2015 between murder and all other crime is to miss the big picture. Something more complex than an overall expansion of crime is underway. There is not even a hint in either the UCR or the NCVS data that the long-term national decline in property crime is abating.

B. The Link Between Changes in Murder and Other Crimes

Figure 7: UCR Data on Murder, Violent Crime, and Property Crime and NCVS Data on Violent Crime and Property Crime

Figure 7 provides a heat map to visually track the changes over the period 1994 to 2015 in the five major crime measures we have discussed: UCR murder, violent crime, and property crime rates and NCVS violent and property crime rates. 131 Initial data for the UCR rates were obtained from FBI, 2015 UCR Data, supra note 4; FBI, UCR Property Crime Rate Search, supra note 51; FBI, UCR Violent Crime Rate Search, supra note 50. Initial data for the NCVS data were obtained from Truman & Morgan, NCVS Data, supra note 24, at 18 app. tbl.1, 20 app. tbl.5. The data are discussed supra in notes 50–53 and shown in Figures 1–3. Combined, these data sources again confirm that 2015 is a rather anomalous year from a historical perspective: 2015 was the only year since 1993 (the start of the modern NCVS) in which (1) an increase in the national UCR violent crime rate has coincided with a decrease in the NCVS violent crime rate, and (2) the jump in the murder rate even exceeded 5% (and it reached almost twice that level). Moreover, in fifteen of the twenty-two years shown in Figure 7, NCVS violent crime and UCR and NCVS property crime all fell—as they did in 2015—yet in none of those other fifteen years did murder and UCR violent crime rise. Figure 7 also illustrates the strong, continuing downward trend in UCR property crime: Since 1993, there has been only one year—2001—that UCR property crime did not fall (and only four such years for NCVS property crime).

Figure 8: UCR Data on Murder, Violent Crime, and Property Crime and NCVS Data on Violent Crime and Property Crime—Percent Changes in Crime Rates in 2015

Figure 8 summarizes what we know about the percentage change in 2015 in the rates of murder, violent crime, and property crime, the last two of which we have estimates for from both the UCR and the NCVS. 132 FBI, 2015 UCR Data, supra note 4; Truman & Morgan, NCVS Data, supra note 24, at 10 tbl.8. The murder rate is measured only in the UCR, not the NCVS. See supra note 45 and accompanying text. This figure highlights that any assessment of crime in 2015 that focuses only on the jump in murder misses a great deal of the full picture of crime in that year.

So we have a puzzle. The UCR tells us that murder rates rose sharply in 2015—the biggest increase in twenty-four years—yet the violent crime rate rose only modestly (up 3.0%), and the property crime rate fell (down 3.4%). 133 See FBI, 2015 UCR Data, supra note 4. Moreover, the NCVS tells us that violent crime and property crime both fell, by 7.5% and 6.3%, respectively, 134 Truman & Morgan, NCVS Data, supra note 24, at 10 tbl.8 (showing 20.1 and 18.6 rates for violent crime in 2014 and 2015, respectively, and 118.1 and 110.7 rates for property crime in 2014 and 2015, respectively). although at least with respect to the NCVS violent crime numbers some qualifications are needed, as I will discuss in section III.D. Since anomalous patterns tend to return to normal, it will be interesting to see whether the not-yet-released 2016 UCR and NCVS violent crime figures converge, whether the 2016 murder and violent crime rates converge, or whether either or both of these divergent patterns continue.

C. Maintaining a Healthy Skepticism

The confusing and uncertain nature of these data must impart a degree of humility to our pronouncements. If we do not even know what happened—for example, did violent crime rise by just a fraction of the rise in murders, or did it decline similarly to property crime?—we cannot be confident about any causal story. Moskos seems to have endorsed the view that as long as you know the murder data, you know all you need to know—or perhaps can know—about trends in crime. 135 See supra text accompanying note 124. Murder data are certainly privileged in the sense that they are the best-measured crime data we have, and homicide is the most serious crime category. But if all crime moved in approximate lockstep with murder, there would be no need to maintain the rest of the UCR crime data or have any NCVS.

There are certainly reasons to be skeptical of the nonhomicide UCR data. The most common components of UCR violent and property crime—aggravated assault and larceny—have been manipulated in certain cities. 136 For example, the Los Angeles Police Department (LAPD) misclassified approximately 14,000 serious assaults as minor offenses from 2005 through 2012, which artificially lowered violent crime rates. Ben Poston et al., LAPD Underreported Serious Assaults, Skewing Crime Stats for 8 Years, L.A. Times (Oct. 15, 2015, 4:47 am), []. Moreover, changes in reporting rates, as well as the aggressiveness of policing, may alter the count of crimes known to police (which is all that the UCR reports) even if crime itself does not change. Similarly, one must always ask whether a survey such as that provided by the NCVS has gone astray in a single year or suffered from some alteration in the response rate in a way that could skew the results. 137 If the decline in NCVS violent crime estimates for 2015 were caused by some differential nonresponse, it would mean that fewer individuals subject to violent crime were interviewed in 2015 than in 2014. With murder and UCR violent crime both showing increases in 2015, it seems more likely that any error in the 2015 NCVS violent crime data would have led to an understatement in crime. Without this conflicting pattern, I might even have had concerns about NCVS overstatement of violent crime in 2015, given election-year exaggerations, as reflected in the growing concerns the public felt about violent crime when respondents were interviewed. This overstatement could result if the high salience of crime increased respondents’ recall about victimizations, regardless of whether the victimizations had occurred in the relevant NCVS time frame. The NCVS asks about victimizations over the last six months, and respondents tend to report more such victimizations in the preceding three months, suggesting that they underreport the more distant (presumably less well-remembered) crimes.

While the NCVS has limitations—for example, it undersamples the homeless, and its respondents do not report fully accurately, especially underreporting domestic violence and forgetting some events 138 Patricia Tjaden & Nancy Thoennes, U.S. Dep’t of Justice, Full Report of the Prevalence, Incidence, and Consequences of Violence Against Women 31 (2000), [] (“Because they reflect only violence perpetrated by intimates that victims label as criminal and report to interviewers, estimates of intimate partner violence generated from the NCVS are thought to underestimate the true amount of intimate partner violence.”); Truman & Morgan, NCVS Data, supra note 24, at 14 (discussing the sample’s lack of homeless individuals). —Professor Chris Jencks notes that the NCVS, to a greater extent than the UCR, “is conducted in the same way every year, so most of [its] biases are likely to be constant. So while the [NCVS] almost certainly under­estimates the level of violence in America, it should be a quite reliable guide to trends in violence.” 139 Christopher Jencks, Is Violent Crime Increasing?, Am. Prospect (1991), []. Consequently, it would be unwise to simply reject NCVS evidence of a continuing violent crime decline (or at most a minor increase) in estimating the true level of violent crime in 2015. Whatever pull toward a 10% increase is exerted by the Moskos “violent-crime-moves-with-homicide” equation 140 See supra text accompanying notes 124–126. is offset by the counter­vailing NCVS data. If the NCVS violent crime estimates are accurate, this would directly contradict the Comey claim of exploding violent crime. While I give considerable, albeit guarded, weight to the NCVS violent crime evidence, one’s confidence would of course increase if additional years of data were to show a similar pattern.

IV. Considering the Impact of Major Factors on Crime

The 1990s witnessed a confluence of many factors that combined to have a large dampening effect on crime. The result was the sharpest and most abrupt crime shift in American history, which began in 1993. 141 See supra Figures 1, 2, 3 & 4. An array of diverse phenomena—increasing rates of incarceration and police, 142 See infra sections IV.A.1–.2. improved quality of policing, 143 For a discussion of how CompStat improved the quality of policing through better resource allocation, see infra note 161. shifts in illegal drug use and marketing, 144 See infra section IV.A.3 and arguably even abortion legalization 145 See infra section IV.B.1. and lead elimination in the 1970s 146 See Kevin Drum, Lead: America’s Real Criminal Element, Mother Jones (Feb. 11, 2016, 5:58 pm), [] (providing the lead-elimination theory of the crime drop). —finally overwhelmed the forces that were stimulating crime. As I have previously written:

With continuing crime reduction benefits coming in the 1990s from demographics, the economy, and incarceration along with a growing police presence, the USA was poised for a major improvement when it got two additional benign shocks—the abatement of the crime stimulant from the crack trade, and the sudden jolt of the crime-suppressing drop in unwanted births engineered via 1970s abortion legalization. . . . The lesson of the great American crime drop is that when six major factors press in one direction, crime moves. 147 John J. Donohue, Book Review, 10 Punishment & Soc’y 333, 336 (2008) (reviewing Franklin E. Zimring, The Great American Crime Decline (2007)).

To fully understand the crime pattern of 2015 and reflect on where crime may be headed, one needs to consider this larger array of factors. By 2015, many of the most potent factors that had contributed to the post-1993 crime decline were no longer contributing to crime reductions or had even reversed and were now putting upward pressure on crime. Sections IV.A and IV.B discuss the factors currently operating to stimulate crime, but I begin with four factors that could also plausibly explain the basic 2015 crime pattern of an increasing murder rate and divergence from other crimes. Section IV.C then discusses current factors tending to suppress crime. To the extent that the major influences that led to the post-1993 drop in crime have been reversed or have lapsed, there will be upward pressure on crime that will need to be confronted, either through restoration of past inhibiting factors or through initiation of new crime-suppressing approaches.

A. Factors that Differentially Promote Homicide and Hence Crime Divergence

1. Reduced Incarceration.

Figure 9: U.S. Incarceration Rates

A major contributing factor in the declining crime rates of the 1990s was the sustained, enormous increase in the country’s rate of incarceration. 148 See Carson & Golinelli, supra note 58, at 1 fig.1. As Figure 9 shows, however, the rate of incarceration peaked and has now begun to fall. 149 Initial data were obtained from Corrections Statistics Analysis Tool (CSAT)—Prisoners, Bureau of Justice Statistics, (select “Quick Tables” and “Prisoners under the jurisdiction of state or federal correctional authorities, December 31, 1978–2015”) (on file with the Columbia Law Review). This data series was created by taking a population-weighted average of the incarceration rates of all fifty states. Donohue, Incarceration Dataset (2017) (unpublished dataset) (on file with the Columbia Law Review). The 8.6% drop in the incarceration rate since a peak of 465 prisoners per 100,000 people in 2007 is now providing a stimulus to crime as more convicts—who typically have high recidivism rates 150 A long-term study by the U.S. Sentencing Commission tracked a cohort of federal prisoners released in 2005; over the next eight years, 49.3% were rearrested for violating supervision conditions or committing a new crime. Kim Steven Hunt & Robert Dumville, U.S. Sentencing Comm’n, Recidivism Among Federal Offenders: A Comprehensive Overview 5 (2016), []. —are returning to the streets than criminals are being taken off. 151 Troublingly, states struggling with heavy budgetary burdens imposed by mass incarceration may face legal constraints that limit their ability to release older criminals who have been sent away with controversial programs such as three-strikes laws. It could well reduce the safety of the public if these constraints discourage the release of older criminals and promote the release of more criminally active younger criminals. One could easily get an upward bump of 1.5% in the homicide rate from prison-population decreases of this magnitude.

Is there any evidence that would suggest that the declining level of incarceration might also contribute to the divergence in crime that we have found in 2015, with murder rising and other crime experiencing a far lower jump or even a decline? Interestingly, an earlier paper of mine with Professor Steven Levitt estimated that the likely crime-stimulating effect of a decrease in incarceration would be greatest for murder and an order of magnitude less for violent crime. 152 Donohue and Levitt’s Table 4 shows an elasticity of almost 25% for the impact of incarceration on murder but an elasticity only one-tenth as large for overall violent crime. John J. Donohue III & Steven D. Levitt, The Impact of Legalized Abortion on Crime, 116 Q.J. Econ. 379, 404 tbl.4 (2001). The impact of incarceration on property crime is also smaller than for murder and not statistically significant. Id. Thus, if the regression estimates from that paper were true today, the shrinking level of incarceration would elevate the murder rate in 2015 by 1.5% but would cause only a negligible upward push of one-tenth that level in violent crime. In other words, the decreasing incarceration rate would be differentially stimulating increases in murder and thereby increasing crime divergence (although still providing lesser stimulus to other violent and property crimes). 153 Of course, as with every aspect of crime, there is a debate on the magnitude of the impact of incarceration on crime. While I have used a consensus estimate for this elasticity, others believe the effect is far smaller or even negligible. A recent Brennan Center study argues that incarceration had a negligible impact on crime rates from the early 1990s through 2013. See Oliver Roeder et al., Brennan Ctr. for Justice, What Caused the Crime Decline? 4, 7 tbl.2 (2015), []. Table 1 of this Essay shows the type of evidence that the Brennan Center report relies on to reach this conclusion, but it is unclear whether the statistically insignificant Table 1 estimates for the impact of incarceration on crime are reliable. The other possibility is that the Table shows that given the endogeneity between incarceration rates and crime, one needs a stronger empirical tool than a panel data model to capture the true impact.

2. Declining Police Rates.

Figure 10: Police Officer Employment in the United States, Shown per 100,000 People

Important papers by Professor Levitt, 154 Steven D. Levitt, Using Electoral Cycles in Police Hiring to Estimate the Effect of Police on Crime, 87 Am. Econ. Rev. 270, 279–84 (1997). Professors Jonathan Klick and Alexander Tabarrok, 155 Jonathan Klick & Alexander Tabarrok, Using Terror Alert Levels to Estimate the Effect of Police on Crime, 48 J.L. & Economies 267, 275­–77 (2005). and most recently Professors Aaron Chalfin and Justin McCrary 156 Aaron Chalfin & Justin McCrary, The Effect of Police on Crime: New Evidence from U.S. Cities, 1960–2010, at 37–39 (Nat’l Bureau of Econ. Research, Working Paper No. 18815, 2013), (on file with the Columbia Law Review). have concluded that higher police employment has a dampening effect on crime, and, most strikingly, on murder. Chalfin and McCrary, for example, find elasticities of -0.67 for murder but only -0.34 for violent crimes and -0.17 for property crimes. 157 Id. at 30. As Figure 10 shows graphically, police employment of sworn officers has declined over the past few years. 158 Initial data were obtained from FBI, Police Employee Data, FBI: Unif. Crime Reporting, [] [hereinafter FBI, UCR Police Employment Data] (last visited Mar. 25, 2017). This data series was created by taking a population-weighted average of the police employment rates of all fifty states. John J. Donohue, Police Dataset (2017) (unpublished dataset) (on file with the Columbia Law Review). Police officer rates have fallen by 7.1% from a peak of 254 officers per 100,000 people in 2008. 159 FBI, UCR Police Employment Data, supra note 158. The decline in overall police employment has been somewhat greater. Since a peak of 368 police staff per 100,000 people in 2007, there has been a fall of 7.9% (as of 2014). Id. Using Chalfin and McCrary’s numbers, this drop would be associated with a 4.8% increase in murder but only a 2.4% increase in violent crime. 160 Using Donohue and Levitt’s numbers, one would expect the decline in the police force to lead to a 2.1% increase in the murder rate and to have far smaller effects on violent and property crime (perhaps an order of magnitude less)—again suggesting differential effects consistent with the patterns we have identified for 2015. See Donohue & Levitt, supra note 152, at 404 tbl.4 (finding an elasticity of -0.300 for murder and just -0.028 for violent crime).
Table 1 of this Essay also presents estimates for the impact of police on crime that also indicate the effect of police is greatest in stopping murder. The Table 1 estimates for the effect of police on violent and property crime are statistically insignificant, but they may be subject to the same infirmity I addressed in note 153 concerning the Table 1 elasticity estimates for incarceration—they make no correction for endogeneity.
Again, we see evidence that a major influence on crime would be differentially stimulating homicide and generating crime divergence.

Combining just the two factors of incarceration and police, one would expect the murder rate to be 6.3% and the violent crime rate to be 2.6% higher in 2015 (than they would have been without the recent declines in incarceration and police). These two factors alone explain two-thirds of the 2015 jump in the murder rate and almost 85% of the UCR violent crime rate increase. 161 See FBI, 2015 UCR Data, supra note 4, at tbl.1A. Society is not only interested in having an appropriate number of police employees and sworn officers but also wants to make sure they are able to work effectively to reduce or restrain crime. The Brennan Center report considering reasons for the long-term crime decline uses a regression analysis to argue that the introduction of CompStat led to a significant reduction of crime in adopting cities between 1994 and 2012. Roeder et al., supra note 153, at 10 tbl.4. The authors contend that CompStat has enhanced police effectiveness by providing timely information and allowing police to deploy officers more effectively and rapidly and to follow up in problematic areas. Id. at 66–67.
By 2015, more than 80% of the police departments in the nation’s fifty largest cities had adopted CompStat, although the rate of new adoptions has slowed considerably in recent years. Id. at 68 fig.28. If more cities continue to adopt this crime-fighting strategy or the police are able to increase their crime-fighting effectiveness as they work with this tool, then CompStat could continue to provide additional pressure reducing crime. Conversely, if no further expansion in the use of CompStat occurs and whatever benefits it confers have stabilized, this technology would be providing no additional stimulus to crime reduction (since the prior benefits would be baked into the current crime numbers).

3. Illegal Drugs—Stimulating Crime and Perhaps Divergence.

Figure 11: Heroin-Related Overdose Deaths in the United States

Figure 12: Overdose Deaths in 2015 in Pennsylvania, Ohio, New Hampshire, Kentucky, Connecticut, and New York City

The last sustained national increase in crime was triggered by the growth of the illegal crack-cocaine trade in the mid-1980s. 162 See Roland G. Fryer, Jr. et al., Measuring Crack Cocaine and Its Impact, 51 Econ. Inquiry 1651, 1651 (2013) (analyzing the effects of crack-cocaine prevalence in the 1980s on crime rates in the African American population). A new nationwide increase in drug addiction and abuse is underway in the United States, leading to an estimated 46,940 drug overdose deaths in 2014, 163 See Richard Rosenfeld, Nat’l Inst. of Justice, Documenting and Explaining the 2015 Homicide Rise: Research Directions 12 fig.4 (2016),
nij/249895.pdf [].
a figure that dwarfs homicide deaths. 164 See FBI, 2015 UCR Data, supra note 4 (showing 14,164 murders in 2014). In the past few years, opioid demand, in particular, has risen sharply. Figure 11 depicts the rise in heroin-related overdose deaths in the United States since 2011. 165 Nat’l Inst. on Drug Abuse, NIH, National Overdose Deaths from Select Prescription and Illicit Drugs (2017) (unpublished data set) (on file with the Columbia Law Review). The data set used to create Figure 11 was originally obtained from the Centers for Disease Control and Prevention’s online public database, CDC WONDER, [] (last visited Apr. 12, 2017). Some cities, like Chicago, have been hit particularly hard, with drug use expanding to suburban areas. In October 2015, seventy-four people with heroin overdoses flooded into Chicago hospitals within a seventy-two-hour window. 166 Tara Dodrill, Chicago Overdoses: Heroin Overdoses Send 74 to the Hospital in 72 Hours, Inquisitr (Oct. 5, 2015), []; see also CDC WONDER, supra note 165. According to the Centers for Disease Control and Prevention (CDC), per capita deaths from drug overdoses have more than doubled since 2000. 167 Rose A. Rudd et al., Ctrs. for Disease Control & Prevention, Increases in Drug and Opioid Overdose Deaths—United States, 2000–2014, 64 Morbidity & Mortality Wkly. Rep. 1378, 1378 (2016). From 2014 to 2015, the total drug overdose death rate increased by 11%, and the heroin overdose death rate increased by 21%. 168 See Rose A. Rudd et al., Ctrs. for Disease Control & Prevention, Increases in Drug and Opioid Overdose Deaths—United States, 2010–2015, 65 Morbidity & Mortality Wkly. Rep. 1445, 1445–46, 1450 tbl.2 (2016) (noting 52,404 people died from drug-related overdoses in 2015, up from 47,055 in 2014 and showing a 20.6% increase for the same period for the rate of heroin-related overdoses). Figure 12 shows the across-the-board increase in overdose deaths in 2015 for a group of states and cities. 169 See Drug Overdose Death Data, Ctrs. for Disease Control & Prevention, [] (last updated Dec. 16, 2016) (listing the number of deaths and percent changes in deaths by state); see also Denise Paone et al., N.Y.C. Dep’t of Health & Mental Hygiene, Epi Data Brief No. 74, Unintentional Drug Poisoning (Overdose) Deaths Involving Heroin and/or Fentanyl in New York City, 2000–2015, at 1 (2016)
doh/downloads/pdf/epi/databrief74.pdf [] (providing data for New York City).
To the extent that the increase in drug overdoses is a reflection of an expansion in illegal drug markets, one would expect crime to be stimulated, as it was with the emergence of crack in the 1980s.

Increases in illegal drug use, particularly in newly emerging or growing illegal markets, can promote deadly gang turf wars, as well as robberies and burglaries as drug users try to support their habits. 170 See Fryer et al., supra note 162, at 1655–56 (providing a synopsis of crack cocaine’s history). These changes would stimulate crime overall, but there is also a mechanism that would differentially stimulate murders but not other measured crimes. This results because the expansion of the illegal drug market provides opportunities for a booming business in robbing drug dealers or fighting over drug turf. 171 Speaking in early 2015, NYPD Chief of Detectives Robert Boyce said each of the seven drug-related murders so far that year involved the targeting of drug dealers for robbery: “These are not turf battles. These are ripoffs of marijuana dealers.” Rocco Parascandola et al., Murders up 20% in 2015 in Year-to-Year Comparison, NYPD Says, N.Y. Daily News (Mar. 3, 2015, 12:04 am), [http://perma.
For obvious reasons, drug dealers do not report such crimes and are unlikely to speak about them to NCVS interviewers. Since these crimes are likely to involve armed participants on both sides, deaths are inevitable, and only the presence of a body will ensure that the event is recorded in some federal crime statistic. This factor would perfectly mimic the observed crime pattern of increased murders and crime divergence (although in this scenario, it is only reported, not actual, violent crime that diverges, as Moskos argued 172 See Moskos, Who You Gonna Believe, supra note 27 (positing the increase in the homicide rate relative to the violent crime rate is because of underreporting crime). ).

4. Increasing Lethality of Assaultive Attacks. — One possible explanation for an increase in murder but not in overall violent crime is that physical violence has increased in lethality. This could happen if attacks that in the past would have led to only a nonfatal injury (e.g., an aggravated assault) have now become more likely to lead to death (hence the jump in murders). Conceivably, the increasing availability of lethal weapons and bullets 173 See Bureau of Alcohol, Tobacco, Firearms & Explosives, Firearms Commerce in the United States: Annual Statistical Update 2016, at 1 (2016),
resource-center/docs/2016-firearms-commerce-united-states/download [http://perma.
could be yet another factor pushing in the direction of more murders, while not stimulating overall violent crime—hence promoting crime divergence.

Of course, there are possible factors pushing in the opposite direction. Improving medical treatment of gun injuries would reduce rather than increase the lethality of violent crime, and any such recent trauma-treatment improvements would push in the opposite direction to our observed spike in murders (by increasing measured violent crime as murders declined). I suspect both of these conflicting factors are operating, but their relatively small and competing effects could well offset each other, so I refrain from considering them further.

B. Factors Stimulating Crime or Whose Previous Dampening Effect Has Lapsed

1. Declining Abortion.

Figure 13: Showing Peak Effective Rates for Murder, Violent Crime, and Property Crime

In our 2001 paper, Levitt and I argue that the reduction in unwanted children that resulted from the legalization of abortion had a significant dampening effect on crime as successive cohorts reached their high-crime years beginning in late adolescence. 174 Donohue & Levitt, supra note 152, at 380–82. The abortion–crime hypothesis has been controversial, and not all have accepted this causal relationship. See Christopher L. Foote & Christopher F. Goetz, The Impact of Legalized Abortion on Crime: Comment, 123 Q.J. Econ. 407, 407–09 (2008) (arguing updated metrics indicate the absence of an abortion–crime causal relationship). Nonetheless, nearly two decades of additional U.S. data, coupled with confirmatory studies from abroad, have strengthened the link between abortion and crime, so there may be some rethinking of this issue by past critics. See Abel François et al., Abortion and Crime: Cross-Country Evidence from Europe, 40 Int’l Rev. L. & Econ. 24, 30–35 (2014) (showing an annual decline of 12% to 40% for homicide and 23% to 43% for theft ten years after legalization’s effect begins).

Using the Donohue-Levitt index designed to capture the potential impact of abortion on crime (based on the summation of each cohort’s abortion rate times the proportion of arrests for that age), Figure 13 shows that the peak effective rates for murder (428) and violent crime (427) were reached in 2003, and property crime (426) in 2002. 175 Initial data were obtained from Data Center, Guttmacher Inst., (select “Data,” “Select All,” and “Abortion rate” and “Number of abortions”) (on file with the Columbia Law Review) (last visited Feb. 22, 2017) (providing abortion statistics), and FBI, Uniform Crime Reporting Program Data: Arrests by Age, Sex, and Race, Summarized Yearly, 2014, Inter-Univ. Consortium for Political & Soc. Research,
sortBy=7&archive=ICPSR&keyword%5B0%5D=Uniform+Crime+Reports&paging.startRow=1 [] (last visited Feb. 22, 2017) (providing dataset for arrest statistics). The effective abortion rate (EAR) is constructed as a summation of arrest-weighted lagged abortion rates. See the equation from Donohue & Levitt, supra note 152, at 394, for each state:
Effective_Abortioni = ∑aAbortion*t-a (Arrestsa/Arreststotal)〗
where t indexes year, and a indexes age. The weighting variable on the right captures the proportion of arrests for a given crime category—murder, violent crime, or property crime—that are of arrestees of age a. (There are thus three EARs, one for each crime category.) These arrest fractions are constant over all years and states and are calculated using the universe of arrests from 1980 to 1985. The Abortion variable represents the number of abortions per 1,000 live births. This variable construction allows for more crime-prone cohorts to contribute more to the EAR by weighting the abortion rates in their years of birth more heavily.
By 2015, these values had declined by roughly 15%: The effective abortion rates were down for murder (361), violent crime (358), and property crime (353).

The first three columns of Table 1 use this effective abortion rate to establish how powerfully increases in abortion are linked to decreases in the three crime categories (using a panel data analysis with state and year fixed effects for data from 1979 to 2014). Indeed, no other factor con­tained in the table even comes close to the enormous statistical significance of the abortion effect. Combining the information from Figure 13 and Table 1, the recent 15% reduction in the effective abortion rate would, mechanically applied, be associated with an increase in crime of almost two-thirds the coefficient estimates in Table 1. 176 To see this, consider the almost 18% increase in murder that is projected from a drop of 100 in the effective abortion rate. Since Figure 13 suggests the drop in the relevant abortion rate was roughly 65, one would expect this abortion decline to increase murder by about two-thirds of the 18% coefficient estimate, or about 12%.

While this mechanical application would suggest that the abortion rate drops would lead to crime increases across all three categories of perhaps 10% to 12%, that number may be exaggerated as an estimated impact on crime in 2015 for two reasons. First, one would expect the effect of the multiyear decline in abortion to accumulate over time rather than to be fully felt in 2015 since the abortion effect is not generated by changes in perceptions (which can move episodically) but rather by a change in the number of unwanted children reaching their high-crime years (which will change continuously). 177 See Donohue & Levitt, supra note 152, at 386. Second, reduced crime can be attributed to a reduction in the number of unwanted children more generally, not just to an increase in abortions from the 1970s to the 1990s: Reducing unwanted pregnancies, like increasing abortions, reduces the number of unwanted children. Recent declines in abortion were not just the product of state-imposed limitations resulting in increased births of unwanted children but rather were caused by reductions in unwanted pregnancies (in part resulting from earlier reductions in unwanted births). 178 See John J. Donohue III et al., The Impact of Legalized Abortion on Teen Childbearing, 11 Am. L. & Econ. Rev. 24, 26 (2009) (describing how legalized abortion led to fewer children likely to themselves have unwanted pregnancies).

If all of the abortion drop were caused by this benign trend, then the reduced level of abortion would not be currently stimulating more crime. Some of the abortion drop, though, may have been caused by state efforts to restrict access to abortions through various legal and policy changes. 179 States employ an array of measures to try to reduce the frequency of abortion. See Sean Sullivan, Where All 50 States Stand on Abortion, in Two Charts, Wash. Post: The Fix (Mar. 7, 2013),
where-all-50-states-stand-on-abortion-in-two-charts/ [].
Such restrictions would be crime-stimulative. 180 If we knew the relative proportion of the observed drop in abortion that falls into this crime-stimulative category, we could adjust our mechanical estimated effect of the decline in abortion on crime. Without this percentage, it is hard to provide a solid estimate of the likely impact of the recent abortion decline on crime. At the very least, it is plausible that the crime-reducing benefit of diminishing the number of unwanted children by expanded access to abortion is no longer providing an added stimulus to crime declines and might actually have reversed, leading to crime increases by 2015.

Of course, those who do not accept the described abortion–crime link would likely not want to include the effective abortion rate variable in the panel data regressions of Table 1. Accordingly, I eliminated that variable from columns 4 to 6 in the Table and otherwise reran the same crime regressions. This has the interesting effect of altering the estimated effects of laws allowing citizens to carry concealed handguns (right-to-carry (RTC) laws), suggesting that the continuing movement to allow more concealed carry is promoting violent and property crime but not homicide. In this event, the growth of RTC laws—such as the adoption of such a law in Illinois in 2013 181 See Ray Long et al., General Assembly Overrides Governor’s Veto of Concealed Carry Bill, Chi. Trib. (July 9, 2013, 10:10 pm),
local/breaking/chi-illinois-concealed-carry-story.html (on file with the Columbia Law Review).
—would be stimulating crime in a manner that promotes crime divergence, but the opposite type of crime divergence that we observed in 2015, when increased homicides were dominant.

2. Out-of-Wedlock Births.

Figure 14: Percentage of Out-of-Wedlock Births by Race/Hispanic Origin

Over time there has been tremendous growth in the number of children born out of wedlock. Specifically, as shown in Figure 14, the percentage of live births that occur to single women has grown from about 5% in the early 1960s, to about 25% in 1988 (a level that, in part, led Assistant Secretary of Labor Daniel Moynihan in the early 1960s to see the black family as in peril 182 See Office of Policy Planning & Research, U.S. Dep’t of Labor, The Negro Family: The Case for National Action 9 (1965),
Moynihan’s%20The%20Negro%20Family.pdf [].
), and to about 33% in 1999. 183 Child Trends, Births to Unmarried Women: Indicators on Children and Youth 3 fig.1 (2015),
Unmarried_Women.pdf [].
As of 2009, the figure was 41.0% for all U.S. births and 72.8% for blacks. 184 Id. This pattern of increasing out-of-wedlock births increases the likelihood that the children will experience a number of factors that have been associated with higher risks of subsequent criminal involvement: growing up in poverty and unstable homes without the benefits of the presence of two parents. 185 William S. Comanor & Llad Phillips, The Impact of Income and Family Structure on Delinquency, 2 J. Applied Econ. 209, 213 (2002).
A remarkable recent study finds that criminality is substantially elevated by certain types of suboptimal family dynamics. See Sanni N. Breining et al., Birth Order and Delinquency: Evidence from Denmark and Florida 11–13 (Nat’l Bureau of Econ. Research, Working Paper No. 23038, 2017), (on file with the Columbia Law Review) (finding second children have substantially higher rates of delinquency and criminality). Whatever disadvantages second children have relative to their older siblings, they are small relative to some of the disadvantages burdening poor inner-city children in unstable families.

Growth in the out-of-wedlock birth rate has been flat since approximately 2009; 186 See Child Trends, supra note 183, at 3 fig.1. however, children born out of wedlock prior to 2009 continue to enter their late-adolescent high-crime years. Therefore, the rate should still rise until children growing up in these circumstances prior to 2009 have aged out of the high-crime-age population. Thus, while I do not have a regression-based estimate for this effect, we have evidence that the past increases in out-of-wedlock births will put upward pressure on crime for perhaps the next ten years.

C. Factors Tending to Suppress Crime

The factors discussed above have all been stimulating crime in recent years. As we saw earlier, however, property crime continues to fall. 187 See supra section II.A.3. This means some potent factors are at work putting downward pressure on crime—or at least on property crime—and these factors continue to dominate any stimulating influence on property crime.

1. Increasing Use of Technology that Raises the Cost of Crime. — A number of technological innovations are operating to restrain crime. Professor Jennifer Doleac has recently written of the success of the expanding use of DNA databases to reduce not only murder and violent crime but property crime as well. 188 Jennifer L. Doleac, The Effects of DNA Databases on Crime, 9 Am. Econ. J.: Applied Econ. 165, 165–201 (2017). The continuing use and growth of these databases is one of the new factors restraining crime across the board.

The widespread use and availability of cell phones and surveillance videos have also been correlated to a decrease in criminal activity. 189 See Jonathan Klick et al., Mobile Phones and Crime Deterrence: An Underappreciated Link 2–3 (Univ. Pa. Law Sch. Inst. for Law & Econ., Research Paper No. 12-33, 2012), (on file with the Columbia Law Review). Before cell phones were popular, burglars might have cut phone lines before going into houses. Obviously, they believed that their criminal designs could be compromised if an occupant were to hear them and phone the police. Cell phones have ended that option for burglars. They facilitate calls for help, and their cameras allow license plates and other evidence to be gathered more easily, thereby elevating the burdens on criminals. Of course, new technology can also be employed by criminals to aid criminal enterprises, but the dampening effect on crime from the broadening use of technology has been the more dominant influence thus far.

2. The Economy. — While the relationship between the economy and crime is often unclear, to the extent that an improving economy has any dampening effect, the strengthening economy should have recently been operating to reduce crime and should continue to do so in the near future.

The year 2015 was unusually good in every economic dimension that one might think would be relevant to crime. Not only did median household income rise by a robust 5.2% from 2014 to 2015 (the first such increase since 2007), but the large job growth in 2015 enabled the number of individuals in poverty to decline by 3.5 million. 190 Bernadette D. Proctor et al., Income and Poverty in the United States: 2015, at 1 (U.S. Census Bureau, U.S. Dep’t of Commerce, Current Population Reports, Ser. No. P60-256(RV) (2016),
2016/demo/p60-256.pdf [].
While poverty declined in every group, blacks and Hispanics experienced the largest improvement. 191 Patricia Cohen, Millions in U.S. Climb Out of Poverty, at Long Last, N.Y. Times (Sept. 25, 2016), (on file with the Columbia Law Review). One would expect these developments to restrain crime, and perhaps they did for property crime and violent crime. Obviously, whatever restraining influence this substantial eco­nomic improvement was having on murders was overwhelmed in 2015 by the other factors that were stimulating murders. 192 I note in passing that the nature of media reporting and perhaps voter attention in a presidential election year was such that many more people were aware of the claims about the enormous increase in murders in 2015 than the enormous improvements in job growth and poverty reduction in the same year.

3. Factors Differentially Suppressing Property Crime. — The two previous factors will tend to suppress all types of crime, so their influence was outweighed in 2015 by factors that stimulated homicides. Three additional factors have been differentially suppressing property crime. The first two make traditional robbery and theft crimes less attractive: (1) the lesser carrying of cash with the ascendancy of credit cards, and (2) the lower cost of goods, which makes theft less attractive. Previously, criminals might have hoped to find cash in a burglarized home, but the move toward credit cards has reduced the likelihood that significant amounts of cash will be found (except in the homes of drug dealers, as I noted above). In the past, television sets and other appliances were also attractive to burglars if they could be resold quickly for a reasonable price. But as the price of these appliances falls and their size increases, their attractiveness to potential burglars falls. Today, the most alluring targets for nonprofessional burglars are probably jewelry and guns. 193 It has long been recognized—based on both the comments of burglars and social science research—that guns are a primary and highly prized target of burglars. Philip J. Cook & Jens Ludwig, Guns and Burglary, in Evaluating Gun Policy: Effects on Crime and Violence 74, 78 (Jens Ludwig & Philip J. Cook eds., 2003). The very reasons that make guns attractive—they are small, concealable, easily transported from the house (without being damaged), and valuable—are increasingly making television sets and other major appliances less appealing (they are large, hard to conceal and transfer, and ever less expensive).

The third factor is in part a response to the first two factors, and while it does not reduce crime, it changes the nature of property crime: Criminals are shifting away from stealing goods to stealing via the Internet. 194 M. Tcherni et al., The Dark Figure of Online Property Crime: Is Cyberspace Hiding a Crime Wave?, 33 Just. Q. 890, 893–98 (2016). Note that in 2014, an estimated 47% of Americans had their information stolen by hackers. Jose Pagliery, Half of American Adults Hacked This Year, CNN (May 28, 2014, 9:25 am),
05/28/technology/security/hack-data-breach/ []; see also Erika Harrell, Bureau of Justice Statistics, Victims of Identity Theft, 2014, at 1 (2015), [] (finding 7% of Americans age sixteen or older were victims of identity theft in 2014).
These losses are probably far less likely to be counted in federal crime statistics, in part because so many victims remain unaware of their losses. This Internet effect suggests that some of the reported property crime decline is not real, simply reflecting a change in the technology of theft. Nonetheless, it is probably less socially costly to have a $100 credit card theft than having a burglar steal $100 from a home. The latter will engender more fear and also carries the risk of escalation into violence.

The move to a cashless society—currently underway in some other countries 195 See Jon Henley, Sweden Leads the Race to Become Cashless Society, Guardian (June 4, 2016, 11:00 am), []. —certainly merits close attention as a possible further step that can be taken to reduce crime.

D. The Net Effect on Crime in 2015

Considering all the factors putting upward pressure on crime in 2015, it is not surprising that homicide experienced a considerable spike that year. This conclusion follows from a simple exercise: Use existing elasticity estimates and the percentage declines in police and prisoners to estimate the likely effect on murder and violent crime, and add in a plausible estimate for the impact of growing illegal drug markets. These three factors alone can essentially explain the roughly 10% higher murder rate observed in 2015 and the far-smaller jump in UCR violent crime. This suggests that murder and violent crime in 2015 would have roughly been at the 2014 level had these three factors not been operating. The end of the crime suppression generated by the increased abortion rate and the increase in out-of-wedlock births that has only recently stalled are other factors that are either still stimulating crime or whose previous suppressing effect has now been eliminated or reduced.

Of course, all sorts of complexities are embedded in these numbers: Which elasticity estimates are the best for estimating the impact of the declines in police and incarceration? How should one account for changes that occurred over a number of years in assessing the impact on murder in 2015? If the growing drug problem, decreasing amount of police and prisoners, and increasing number of out-of-wedlock births have pushed crime up and thereby cumulatively explain today’s murder and violent crime rate, that still leaves unanswered why the murder jump did not start earlier. In other words, if over the last five years those four factors operated in combination to push up the murder rate by, say, 2% per year, why did it jump by 10% in 2015? Shouldn’t the murder rate increase from 2014 to 2015 only be one-fifth that high? Why, then, wasn’t its rise in 2015 only 2% instead of 10%? Put differently, can continuous changes in causal factors ultimately lead to discontinuous impacts on crime?

From the sharp reversal in crime in 1993, it appears that this is exactly what happens. Conceivably, the initial decline in police in the years prior to 2015 had relatively little effect, but, at some point, a larger cumulative effect was felt. As the police force shrinks, at first one loses the presumably least-valuable police services, so the upward pressure on crime is small. When the losses continue and 5% or 10% of the police force is gone, the department has started losing some muscle and bone. If this tipping point or threshold effect story is correct, then by the time the police force had fallen by 7.1% and the incarceration rate had fallen by 8.6% in 2015, 196 See supra note 159 and accompanying text (discussing police enforcement employment statistics); see also supra notes 148–153 and accompanying text (discussing incarceration rates). that threshold had been met. 197 Since the deterrent effect decreases when criminals perceive a diminished risk of punishment, it is not surprising that factors such as declining police and incarceration can suddenly lead to abrupt jumps in crime as it may take time for potential criminals to perceive the slowly diminishing risk. When they do, the reduced deterrent effect of the continuously changing factor is felt. At that point, the criminal population finally perceived a lower police presence or the police force was stretched more thinly, thereby stimulating a rise in criminal behavior, which is less easily restrained. Similarly, lower incarceration levels meant more criminals for the increasingly short­handed police to grapple with, and the growing illegal drug markets imposed similar burdens.

The above discussion suggests that major crime shifts result from the accumulation of several factors pushing in one direction that eventually overwhelm forces pushing in the opposite direction. There may be a catalytic event that stimulates the shift or alters the perception of likely sanctions, but the power of a crime shift derives from the accumulation of forces, perhaps over a number of years. This is true about the dramatic post-1993 crime drop and will be true if 2015 heralds a sustained murder rate increase. Accordingly, 2015’s dramatic and sudden rise in homicides may have resulted from countervailing forces finally being overwhelmed by factors that had been putting upward pressure on crime over the course of many years.

Since the more fundamental factors that I have mentioned may largely explain the increase in murder, there is little scope for the Ferguson Effect thesis to have played a role in the nationwide increase (unless other factors depressing homicide are so great that one needs yet another element to again close the gap between the observed homicide rate and the rate that can be explained).

Note that the elasticity estimates for police and incarceration discussed in section IV.A both show that murder was most influenced by these factors, that violent crime came next, and that property crime was last. Hence it should not be surprising that this was the basic pattern seen for 2015 with a very large jump in murder but a large decline in property crime. The property crime decline must reflect the continuing power of previously mentioned forces putting particularly strong downward pressure on the theft of goods.

V. Further Thoughts on Policing

A. The De-Policing Hypothesis

1. The Ferguson Effect. — Given the enormous importance of the police in reducing crime, it is not implausible to speculate that some of the developments in the wake of various highly publicized shootings, protests, and police killings might be responsible for the dramatic murder hike of 2015 and the continuing rise in 2016. The causal story that Comey seemed to embrace was that “a chill wind blowing through American law enforcement over the last year” made “officers reluctant to get out of their cars and do the work that controls violent crime.” 198 Comey, Remarks at the University of Chicago, supra note 11.

Comey’s statement was widely taken to mean that the FBI Director had concluded that the protest against perceived police misconduct spearheaded by the Black Lives Matter movement was to blame for the added 1,500 deaths in the United States last year. 199 See supra text accompanying notes 13–15. Since Comey added that these deaths were not at the hands of the police and “almost entirely among young men of color,” 200 See Comey, Remarks at the University of Chicago, supra note 11. the subtext of Comey’s remarks was that even if Black Lives Matter had some legitimate complaints, the move­ment’s conduct precipitated many more deaths of young black men than would have been saved even if all the wrongful police shootings were eliminated. The “blame-the-victim” overtone of Comey’s musings surely explained some of the angry response it elicited.

But the implicit criticism of the police was also striking. A chill wind blows and suddenly the police aren’t doing their job, and as a result 1,500 people die? Such unsupported allegations could inflame both fears of black crime and white resentment, particularly during a presidential campaign, and especially coming a few months after the Dylann Roof killings that were motivated by the mistaken belief that blacks were slaughtering whites 201 Blinder & Sack, supra note 69 (noting Roof claimed his attack was motivated by “an epidemic of black-on-white crime”). —the same mistaken belief that Trump’s erroneous tweet was reinforcing. 202 See supra notes 17–19 and accompanying text.

Where is the evidence that the 1,500 additional deaths were “almost entirely among young men of color”?  The nature of the murder increases across states that I discussed above certainly raises questions about this claim, and indeed, the claim is refuted by homicide figures from the CDC indicating that nonblack homicides grew by almost nine percent in 2015. 203 See supra notes 118–120 and accompanying text. But even if all the added deaths were of black men, that would not prove the de-policing story. An alternative or comple­mentary hypothesis may be that when “the longstanding griev­ances and discontent with policing in African-American com­munities [were] acti­vated by controversial incidents of police use of force, chronic discontent erupt[ed] into violence.” 204 Rosenfeld, supra note 163, at 2. The riots in Baltimore in April 2015 following the death of Freddie Gray in police custody and in Ferguson, Missouri, four months later on the one-year anniversary of the shooting of Michael Brown (with more cities following in 2016 after other police shootings) seem to conform to that descript­ion. We know that this rage was spilling over into homicidal conduct in at least some cases because this was the explicit reason offered for some of the horrible police shootings of the last two years.

The de-policing hypothesis (perhaps supplemented by the “building-frustration-at-police-misconduct-finally-explodes” theory) is more plausible as an explanation for a short-term increase in crime in a specific city after an identifiable triggering factor 205 Following the launching of a Department of Justice investigation into the Seattle police department in 2011, claims of de-policing emerged, with reports of police failing to investigate crimes. Eric Scigliano, The D Word: Seattle Confronts the “Depolicing” Dilemma, Crosscut (Dec. 1, 2014), []. Similarly, a drop in arrests in Oakland in 2012 was attributed, by some, to de-policing following federal oversight. See Demian Bulwa, Arrest Totals Down Sharply in Oakland, SFGate (Dec. 1, 2012, 10:44 pm), []. than as an explan­ation for a widespread jump in murders across thirty-seven states. Some FiveThirtyEight articles on the topic show striking time-series graphs of arrests falling as crime rises. Such studies from a small number of cities that have experienced a traumatic event provide some plausibility to the existence of a de-policing effect in these locales. 206 See, e.g., Rob Arthur & Jeff Asher, Gun Violence Spiked—and Arrests Declined—in Chicago Right After the Laquan McDonald Video Release, FiveThirtyEight (Apr. 11, 2016, 3:53 pm), []; Reuben Fischer-Baum & Andrew Flowers, Charts: Baltimore Crime, Before and After Freddie Gray’s Funeral, FiveThirtyEight (June 16, 2016, 6:00 am), []. Causal attribution can be tricky even in these cases, however. During the riots following the death of Freddie Gray, one could imagine that crime could rise and arrests could fall if the police were scrambling to maintain order, especially given Baltimore’s shrinking police force. 207 Scott Malone, Baltimore Saw Steep Fall in Police Numbers as Murder Rate Soared, Reuters (July 7, 2016, 1:05 am), []. This may not be the product of the type of intentional de-policing that those who endorse the Ferguson Effect envision but simply the consequence of a police response to a riot in which police may be reassigned from other tasks to confront a mob. Of course, a Ferguson Effect could follow, but then the analyst would need to differentiate the portion of the post–Freddie Gray crime increase caused by the riot from that caused by the intentional police retrenchment.

Professor David Pyrooz and others conducted an empirical test for any kind of Ferguson Effect after the August 2014 shooting of Michael Brown. 208 David Pyrooz et al., Was There a Ferguson Effect on Crime Rates in Large U.S. Cities?, 46 J. Crim. Just. 1 (2016). The authors concluded that except for robbery there was no evidence to support the claim. 209 Id. at 4 (concluding “there is no systematic evidence of a Ferguson Effect on aggregate crime rates throughout the large U.S. cities represented in th[e] study”). The paper looked at monthly city crime data from eighty-one large cities for the period one year prior to the August 2014 shooting and one year after and did not find a statistically significant break in the trend in crime for murder, violent crime, or property but did so for robbery. 210 Id. at 5 (finding “the overall null Ferguson Effect was robust for both total and disaggregated crimes scales” but noting “robbery rates significantly increased in the study cities post-Ferguson”). Of course, whether the full effect of de-policing occurred within one year of the Brown killing remains unclear, so a different statistical model or just a longer post-treatment evaluation period might produce different results.

Finally, Professors Stephen Rushin and Griffin Edwards have an intriguing new paper that analyzes “all police departments that have been subject to federally mandated reform under 42 U.S.C. § 14141.” 211 Stephen Rushin & Griffin Edwards, De-Policing, 102 Cornell L. Rev. (forthcoming 2017) (manuscript at 7), (on file with the Columbia Law Review). Their major finding was that this “external regulation” was “associated with a statistically significant increase in the frequency of several crime categories—particularly property crimes . . . concentrated in the years immediately after the initiation of external regulation and diminished into statistical insignificance over time.” 212 Id. (manuscript at 40–41). The paper finds that the property crime rate elevation is substantial—a 25.4% increase. 213 Id. (manuscript at 46–48 & fig.7). This may suggest that as police departments are undergoing policy changes in response to federal regulation, their crime-reduction capacities are temporarily disrupted, particularly for property crimes. But of course since both the UCR and NCVS data showed property crime was declining in 2015, 214 See supra section II.A.3. this effect seems not to have been operating very powerfully that year. 215 Note that the difference-in-difference methodology employed in this Essay could not capture any national de-policing effect, because that would be operating on both the treated and untreated cities.

None of these studies provide evidence of a contagion effect that could explain the rising crime in cities nationwide. In effect, supporters of the Ferguson Effect hypothesis seem to conflate geographically and temporally local effects with broad national trends. Even reports showing that the cities experiencing the largest murder rate jumps have relatively high black populations is not particularly corroborating. 216 See Rosenfeld, supra note 163, at 10 (explaining the top ten cities contributing to the 2015 big-city homicide rise have nearly double the black populations of other cities). These cities may also be experiencing disproportionate drops in incarceration, police, and abortion, while having larger growth in illegal drug markets and out-of-wedlock births—all factors that can and should have been explored before announcing a highly controversial theory.

2. “De-Policing” in New York City and Chicago.

Figure 15: Homicides in New York City and Chicago

(2016 projected as of mid-November)


Figure 16: New York City Stop and Frisk Counts by Week

A very telling graph by Moskos, shown in Figure 15, 217 Peter Moskos, The Best of Times, The Worst of Times, Cop in the Hood (Nov. 18, 2016), [] [hereinafter Moskos, The Best of Times]. illustrates simultaneously how a major disruption in a police department led to a chillingly large jump in crime in Chicago in 2016, while effective policing substantially curtailed murders in New York City from 2011 through 2013.

a. New York City. — The dramatic murder decline in New York City after 2011 was accompanied by a drastic reduction in police stops, as shown in Figure 16. 218 Initial data were obtained from NYPD, The Stop, Question and Frisk Data, NYPD: Stop, Question and Frisk Report Database,
analysis_and_planning/stop_question_and_frisk_report.shtml [] (last visited Apr. 21, 2017). This series was created by aggregating the data into a weekly level. John J. Donohue, Stop and Frisk Dataset (2017) (unpublished dataset) (on file with the Columbia Law Review).
The best evidence from an important study by Professors John MacDonald, Jeffrey Fagan, and Amanda Geller shows that “crime reduction can be achieved with more focused investigative stops.” 219 John MacDonald et al., The Effects of Local Police Surges on Crime and Arrests in New York City, PLOS One, June 2016, at 1,
article/file?id=10.1371/journal.pone.0157223&type=printable [].
The good news on crime in New York City has persisted after 2013 under the liberal Mayor Bill de Blasio, despite many loud and insistent predictions that crime would explode if stop and frisk were to end. 220 Former Mayor Michael Bloomberg articulated a common view that New York City would not be able to fight crime effectively without the use of a stop-and-frisk policy. Michael R. Bloomberg, ‘Stop and Frisk’ Keeps New York Safe, Wash. Post (Aug. 18, 2013), []; see also Mac Donald, supra note 13. Indeed, the New York City experience underscores that without a careful definition of the term, claims about de-policing can be both misleading and dangerous. The New York City murder drop of 2011 to 2013 was achieved contemporaneously with a massive reduction in stops and misdemeanor drug arrests. 221 See Moskos, The Best of Times, supra note 217. The New York Civil Liberties Union publishes data on stops in New York City. See NYCLU, Stop-and-Frisk Data, supra note 218. For New York City drug arrests, see N.Y. Div. of Criminal Justice Servs., Adult Arrests: 2006–2015, [] (last updated Jan. 31, 2016). That may well be a type of de-policing we need more of.

b. Chicago. — Looking at the counts of murders for New York City and Chicago, the first noteworthy observation is that Chicago—with only one-third the population of New York City 222 As of July 2015, the U.S. Census Bureau estimated that New York City had a population of 8,550,405 and that Chicago had a population of 2,720,546. See QuickFacts: Chicago, supra note 108; QuickFacts: New York City, U.S. Census Bureau, [] (last visited Feb. 6, 2017). —has had more murders, and accordingly a higher murder rate, than New York City since 2012. 223 In Chicago, the number of murders was 473 in 2015, 415 in 2014, 419 in 2013, and 503 in 2012. Ryan Marx, Chicago Homicide Data Since 1957, Chi. Trib. (Mar. 2, 2016, 2:03 pm), (on file with the Columbia Law Review). By comparison, in New York, the number of murders was 352 in 2015, 333 in 2014, 335 in 2013, and 419 in 2012. Major Crime in New York City, 2009–2015, Newsday, [] (last updated Aug. 31, 2016). The disastrous murder spike that befell Chicago in 2016 is tragic and, much like in Baltimore, 224 See infra text accompanying notes 237–242. the product of some acute pathologies; however, more precision is needed to tie a change in the murder rate to a vague concept like de-policing.

Former Chicago Police Superintendent Garry McCarthy—whom Mayor Rahm Emmanuel fired amid controversy over the police shooting of unarmed teen Laquan McDonald in October 2014 225 Bill Ruthhart & David Heinzmann, Emanuel Dismisses Top Cop Garry McCarthy amid Pressure for Change, Chi. Trib. (Dec. 2, 2015, 6:35 am), http://www.chicagotribune.
com/news/ct-chicago-police-superintendent-garry-mccarthy-20151201-story.html (on file with the Columbia Law Review).
—said criticism of policing methods by local officials there had left cops “hamstrung.” 226 Shawn Cohen et al., Why Crime in Chicago Soars While New York Is Safer than Ever, N.Y. Post (Nov. 14, 2016, 10:49 pm), []. Such claims need to be carefully examined because police often raise them whenever police misconduct is identified and criticized. 227 For example, an LAPD recruit contended that after the release of the 1992 Rodney King video, massive de-policing occurred: “We make no arrests; we do not even try to . . . . Officers are beyond paranoia when it comes to using force. Our police force has been effectively emasculated.” William Dunn, Boot: An L.A.P.D. Officer’s Rookie Year 197 (1996). One must be highly skeptical of such claims in light of the actual experience of sharply dropping crime in Los Angeles after 1992. Specifically, the violent crime rate dropped in Los Angeles from 1992 to 1996 by 27%, and the murder rate declined by 32.8%. Initial data were obtained from FBI sources. Donohue, LA Police Dataset (2015) (unpublished dataset) (on file with the Columbia Law Review). Crime declines of this magnitude are unlikely to occur if there really is emasculation of the police force. Any effect of “de-policing” in Los Angeles at that time was likely modest, short-lived, and outweighed by countervailing downward pressures on crime.

The protean and imprecise nature of the concept of de-policing is shown by the claim of Chicago Police Department spokesman Anthony Guglielmi, who blames the decline in proactive policing in Chicago on a new form police must complete after certain interactions with the public. 228 See Arthur & Asher, supra note 206. The city’s August 2015 settlement with the American Civil Liberties Union over the department’s stop-and-frisk program mandated the use of a new form, called an investigatory stop report (ISR). 229 See Landmark Agreement Reached on Investigatory Stops in Chicago, ACLU (Aug. 7, 2015, 7:00 am), []. ISRs are longer than the contact cards they replace and, after some incidents, take substantially longer to complete. 230 Arthur & Asher, supra note 206. “The rules of the game changed on Jan. 1,” Guglielmi said. 231 Id.

But a review of arrest rates in Chicago concluded that these rates fell right after the release of the McDonald video (again, a riot-induced effect?) and actually rose after the ISR requirement went into effect. 232 See id. (noting “[t]he onset of the decline in arrests significantly predates the ISR, and arrests have actually increased since it was introduced, though they are occurring less frequently than they did in 2015”). Negative consequences could result from requiring officers to fill out forms, but it matters what the requirement seeks to effectuate. Obviously, if the police spend 10% of their time filling out forms, that would potentially cut the time they could be doing other police work by 10%. One always needs to consider the opportunity cost of any particular allocation of time. But if the forms are needed to effectuate lawful arrests or to prevent racially biased policing, then the costs of the forms have to be weighed against the benefits they provide. Moreover, the costs of filling out the forms could fall over time if a learning effect operates. It is a different matter, though, if the police simply dislike filling out the forms because they do not like to be monitored. The pure “opposition-to-monitoring” effect is supported by the cases in which police show their extreme unhappiness about being video-recorded, even if from a distance and unobtrusively. 233 For example, after the widely publicized shooting of Alton Sterling in July 2016, Baton Rouge police confiscated the nearby Triple S Food Mart’s security system, detained the store’s owner, and confiscated the phone he had used to record the shooting. Mohit Priyadarshi, Store Owner Who Filmed Alton Sterling’s Death Is Now Suing the Baton Rouge Police Department, Inquisitr (July 17, 2016),
store-owner-who-filmed-alton-sterlings-death-is-now-suing-the-baton-rouge-police-department/ []. In 2011, Georgia police slammed a girl’s head into a car after she videotaped the arrest of her cousin. Andrea Prichett & Annie Paradise, The Criminalization of Copwatching: Berkeley Copwatch Report on State Violence, Police Repression and Attacks on Direct Monitoring 16 (2011), []. In 2010, Maryland police arrested a man after he posted a video of a traffic stop on YouTube. N. Stewart Hanley, Note, A Dangerous Trend: Arresting Citizens for Recording Law Enforcement, 34 Am. J. Trial Advoc. 645, 645–68 (2011) (detailing Maryland police’s decision to arrest a man for violation of wiretapping laws after the video was posted to YouTube).
This may be a human reaction, but it is not a legitimate basis for objection because it is essential that all government officials, especially those given great powers and accorded great discretion, are monitored effectively.

3. Police Brutality and Compensatory Payments. — A sense of the severity of the problem in Chicago is conveyed by the fact that since 2004, the city has spent almost $642 million dealing with police-misconduct legal claims, with $106 million of that total from the last two years alone. 234 Andrew Schroedter, BGA Public Eye: City Pays a Price for Police Misconduct—$642 Million Since 2004, Chi. Sun-Times (Jan. 30, 2016, 6:45 pm), []. These settlements diverted government resources away from needed services that could improve high-crime communities. Chicago took out almost $100 million in bonds in 2014 to pay for police-misconduct cases. 235 Carrie Sloan & Johnaé Strong, Chicago Has Spent Half a Billion Dollars on Police Brutality Cases—It’s Impoverishing the Victims’ Communities, Nation (Mar. 11, 2016), [ JLR5-P4NS]. Indeed, the police have tried to cover up the nature of misconduct with certain settlements. For example, the Chicago police tried to suppress a video of the fatal police shooting of McDonald with a $5 million settlement and destroyed video footage of a nearby store’s security camera to hide the evidence of what happened. 236 The settlement contained a clause specifically prohibiting the release of the video; ultimately, however, a court ruled that the video had to be released. Curtis Black, How Chicago Tried to Cover Up a Police Execution, Chi. Rep. (Nov. 24, 2015), []. In late 2016, a special prosecutor was appointed to investigate the role of seven Chicago police officers in attempting to cover up the shooting. One of the attorneys involved, Michael Robbins, told the Chicago Tribune that “[t]he real problem here is the culture . . . that creates this code of silence and this almost unspoken instinctual conduct on the part of officers to shield one another from allegations of misconduct.” See Annie Sweeney et al., Top Cop Seeks to Fire 7 Officers for Lying About Laquan McDonald Shooting, Chi. Trib. (Aug. 18, 2016), (on file with the Columbia Law Review).

Chicago is not alone. From 2011 through 2014, Baltimore had 100 settlements or judgments based on police brutality or civil rights abuses. 237 Mark Puente, Undue Force, Balt. Sun (Sept. 28, 2014), http://data. [].
In many of these instances, police acted violently against bystanders after apparently minimal provocation. For example, in one convenience store, police stopped Jerriel Lyles without any discernible reason, then hit him in the face when he objected. 238 Id. Lyles’s case ended in a verdict against the officers and an award of $200,000. 239 Id. Starr Brown was trying to explain what she had witnessed to officers at the scene of a crime, when the situation escalated. 240 Id. Even after she declared she was pregnant, an officer threw her to the pavement. “I was tossed like a rag doll. He had his knee on my back and neck,” she said. 241 Id. Brown settled her case for $125,000. 242 Id. The high-profile death of Freddie Gray, who sustained spinal injuries in the back of a police van, led to a $6.4 million settlement. 243 Sheryl Gay Stolberg, Baltimore Announces $6.4 Million Settlement in the Death of Freddie Gray, N.Y. Times (Sept. 8, 2015),
09/09/us/freddie-gray-baltimore-police-death.html (on file with the Columbia Law Review).

According to a tally by the Wall Street Journal, payouts from police-misconduct cases in the ten cities with the largest police departments totaled $248.7 million in 2014, compared to $168.3 million in 2010. 244 Zusha Elinson & Dan Frosch, Cost of Police-Misconduct Cases Soars in Big U.S. Cities, Wall St. J. (July 15, 2015), (on file with the Columbia Law Review). The family of Eric Garner, a man who was killed by a police chokehold (and famously declared, “I can’t breathe”), negotiated a $5.9 million settlement from the city of New York. 245 J. David Goodman, Eric Garner Case Is Settled by New York City for $5.9 Million, N.Y. Times (July 13, 2015), (on file with the Columbia Law Review). A 2016 case in Michigan, meanwhile, led to a strikingly large award of $37 million when a man arrested for drunk driving was severely beaten by police at the jailhouse. 246 Tresa Baldas, $37M Awarded in Genesee Co. Police Beating Caught on Camera, Detroit Free Press (Nov. 6, 2016, 11:56 pm),
93287248/ []. The jury ruled against six officers for use of excessive force—damages included $10,420,000 for the use of unlawful force, $7,210,000 for future harm, and punitive damages for each officer, ranging from $2 million­ to $5 million. Amanda Emery, Attorney for Jail Guards Plans to Appeal $36.6M Verdict for Excessive Force, MLive (Nov. 3, 2016, 6:30 pm),
index.ssf/2016/11/attorney_for_jail_guards_plans.html [].

B. Policing in the Years Ahead

The relationship between the police and the community will never be without friction. Police are needed because not everyone obeys the law, and police must be the ones to arrest lawbreakers or impede their misconduct, neither of which will endear police to the people they apprehend or those whose preferences they thwart. Moreover, given the nature of their tasks in a country brimming with weapons, the police will always be making some deadly errors, even with the best of intentions, since split-second judgments at times must be made under situations of considerable stress. The police are also not given nearly enough credit in certain circles for their absolutely critical role in crime control.

But that said, the police also have to do better. The admittedly few officers who cause the most problems need to be ferreted out or quickly retrained, rather than protected by a tribal loyalty. No one should impede the police in the exercise of their legitimate authority, but the public should also be protected in its right to appropriately monitor the police without being threatened or physically accosted. Everyone will benefit if the relationship between the police and the community can be improved, but that will require steps from both sides that neither has been willing to take in the past.

An excellent example was set in October 2016 when the International Association of Chiefs of Police President Terrence M. Cunningham delivered a speech on the law enforcement profession and historical injustices, recognizing the importance of both “acknow­ledg[ing] and apologiz[ing]” for the role of law enforcement in the past while also creating an “atmosphere of mutual respect” necessary to work together to “ensure fairness, dignity, security, and justice.” 247 Terrence M. Cunningham, President, Int’l Ass’n of Chiefs of Police, Remarks at the 2016 IACP Annual Conference and Exposition: The Law Enforcement Profession and Historical Injustices (Oct. 17, 2016), [].
The speech received a standing ovation. Certainly, Cunningham was setting the right tone, and hopefully more police chiefs will be able to suffuse these insights throughout their departments. Americans’ confidence in the police was at a record high of 64% in 2004 but subsequently slipped to a low of 52%, reflecting some of the policing troubles of 2015 before rebounding modestly in 2016. 248 Frank Newport, U.S. Confidence in Police Recovers from Last Year’s Low, Gallup (June 4, 2016), (on file with the Columbia Law Review). At present, 25% of Americans have “a great deal” of confidence in the police, 31% “quite a lot,” 29% “some,” 13% “very little,” and 1% “none.” 249 Id. The combined 14% with very little or no confidence in the police is down from 18% in 2015, which was the worst assessment in Gallup’s history of rating the police. 250 Id.

Improved national data collection on policing would enhance account­­ability, and the Department of Justice has taken some promising steps in this direction by, for example, announcing the creation of a use-of-force database. 251 Press Release, U.S. Dep’t of Justice, Justice Department Outlines Plan to Enable Nationwide Collection of Use of Force Data (Oct. 13, 2016), []. The Obama Administration took steps to improve the quality and quantity of police data. In December 2014, Obama signed an executive order creating the President’s Task Force on 21st Century Policing, a committee tasked with identifying best practices and making recommendations. Exec. Order No. 13,684, 79 Fed. Reg. 76,865 (Dec. 18, 2014). In the Task Force’s first report, it recommended that law enforcement “collect, maintain, and report data . . . on all officer-involved shootings, whether fatal or nonfatal, as well as any in-custody death.” President’s Task Force on 21st Century Policing, Final Report of the President’s Task Force on 21st Century Policing 21–22 (2015), []. But relying on accountability systems that administer punishment after police misconduct has occurred may be less effective than taking proactive measures to prevent police misconduct in the first place. The University of Chicago Crime Lab has joined with the Chicago Police Department and the city of Chicago to reduce adverse police–citizen encounters and improve police–community relations by streng­thening support services for officers who display signs of elevated risk of future misconduct, using data-driven operational systems to identify and prioritize officers for available support. 252 Ted Gregory, U. of C. Researchers Use Data to Predict Police Misconduct, Chi. Trib. (Aug. 18, 2016, 6:45 am), (on file with the Columbia Law Review).

Steps can also be taken to reduce tensions with the police by finding ways to further improve police safety. Police are always aware that they are in danger from guns. And even though assaults (fatal and nonfatal) on the police have been trending down, they are still at a distressingly high level (almost a tenth of the police force is assaulted every year). 253 Under the Obama Administration, “the average number of police intentionally killed each year [fell] to its lowest level yet—an average of 62 deaths annually through 2015.” Christopher Ingraham, Police Are Safer Under Obama than They Have Been in Decades, Wash. Post (July 9, 2016), []. This number is down from about 101 annual police deaths during the Reagan years. Id. The number of assaults per 100 police officers is down from 12.7 in 2000 to 9.0 in 2014. Id. The primary cause of death to police officers from intentional assaults is from guns. 254 See Causes of Law Enforcement Deaths, Nat’l Law Enf’t Officers Mem’l Fund, (on file with the Columbia Law Review) (last updated July 18, 2016) (showing a total of 521 law enforcement deaths were attributed to gunshot wounds). Efforts to improve officer safety through body armor and effective training can lessen the threat they feel and thereby reduce the number of “quick trigger” events.

One important study that merits further investigation found that states with high rates of civilian gun ownership are more dangerous for the police. 255 David I. Swedler et al., Firearm Prevalence and Homicides of Law Enforcement Officers in the United States, 105 Am. J. Pub. Health 2042, 2045 (2015). The study examined data on the number of homicidal deaths of police in two groups of states with roughly equal number of police officers—the eight states with the lowest levels of gun ownership and the twenty-three states with the highest rates of gun ownership. 256 Id. at 2046 tbl.3. The study found that, over the period from 1996 to 2010, the rate of police deaths from homicide in the high-gun-prevalence states was three times as high as the rate in the low-gun-prevalence states. 257 Id. at 2042. In exploring whether gun prevalence is the causal factor in the higher police death rate, one would have to control for the fact that high crime rates could lead to both more guns and more police deaths. It might be worth studying whether gun control measures—universal background checks, reducing civilian firepower, and banning armor-piercing bullets—could enhance police safety.


This Essay shows that after one makes the simple observation that the murder rate rose sharply and property crime continued to fall in 2015, nothing else about crime is known with complete certainty. It is challenging, over one year after the fact, to not know exactly what happened with violent crime in 2015, even with two different data sources designed to measure this important social metric. Given the major discrepancies in UCR and NCVS violent crime data, it is unclear whether we should be applauding the news on violent crime or concerned about a somewhat worsening performance. It is harder still to figure out why one might see big jumps in murder, large drops in property crime, and an apparent substantial reduction in the prevalence of violent crime (even if more multiple victimizations). If there really were 300,000 fewer violent crime victims and 500,000 fewer property crime victims as well as 1,500 more homicides in 2015, that would be an unusual crime picture—and one that is strikingly divergent from the dire crime picture painted by the supporters of the Ferguson Effect theory.

Exaggerating the extent of the bad news on crime, emphasizing its racial dimension, and espousing a de-policing causal explanation without providing any empirical support is dangerous given the fearmongering about the race–crime connection that conforms to a frequent Republican presidential election playbook. Further, Comey’s statements on crime provided fodder to Trump, who jettisoned even the smallest pretense of concern with the actual facts about crime, with predictable results.

For the first time since 9/11, a majority of Americans (53%) in 2016 said they were personally worried “a great deal” about crime and violence. 258 Davis, supra note 1. But nothing about the present level of crime would have warranted more fear in 2016 than was seen any year after 9/11 during the Bush Administration, when crime was clearly worse. 259 See FBI, UCR Violent Crime Rate Search, supra note 50. The importance of that perceptual shift going into the presidential election is hard to exaggerate. This heightened apprehension provided enough of a political boost to Trump to at least raise the question whether this development alone, in a razor-thin presidential election in three states, was enough to influence the outcome. 260 The number of those worried a great deal about crime was only 43% in 2015 and 39% the prior year, which shows that the perceptual shift was enormous over the course of 2016. See Davis, supra note 1.

Looking forward, what can we expect in the years ahead concerning crime? For now, at least, the downward trend in property crime has yet to be disrupted. The path of violent crime and murder will in part depend on continuing developments with the major factors that I have mentioned. Future incarceration and police-force rates, as well as the prevalence of illegal drugs, will critically influence crime patterns, particularly because the effects of abortion and out-of-wedlock births may have stabilized.

If one wanted to immediately forestall any further bad news concerning the murder rate, a reasonable suggestion might be to return to Bill Clinton’s 1992 plan of putting an extra 100,000 police on the streets. This would undo the crime increases that have followed from the declining police employment and even facilitate further decreases in incarceration. At the margin, spending more on police and less on prisons seems to be a wise trade-off. 261 John J. Donohue, Fighting Crime: An Economist’s View, 7 Milken Inst. Rev. 46, 48–51 (2005); see also John J. Donohue III & Peter Siegelman, Allocating Resources Among Prisons and Social Programs in the Battle Against Crime, 27 J. Legal Stud. 1, 43 (1998). Perhaps if the economy continues to improve there will be resources for greater police hiring, but without federal support, almost certainly lacking in Congress at this point, police hiring will be difficult for financially strapped cities. Regardless of whether there are increases in the size of police forces, the steps I have alluded to about healing the police–community rifts are essential for promoting a sense of justice and will likely generate crime-reduction benefits as well. We are poised to see a historic adjustment in police–community relations, and some disruption in crime may occur until the process of adjustment has been completed. Hopefully, New York City has paved the way for understanding how effective use of police can reduce crime without resorting to unconstitutional stops that are not based on reasonable suspicion. Other jurisdictions should learn from that experience. At the end of the day, there is no reason that Chicago should have three times the murder rate of New York City.

Table 1