The Conventional View

There is an enormous literature on the punitive turn, but for our purposes, it is sufficient to summarize what we take to be the conventional view. That view can be represented in three premises and a conclusion.

1. 1. Anxiety over changes to the racial order caused white Americans to fear crime.

2. 2. Rising white fear of crime led white (especially, white Republican) politicians to support punitive policies.

3. 3. White politicians’ support of punitive policies caused an increase in the rate of incarceration and policing.

• ⸫ White racial backlash explains mass incarceration.

Not everyone who argues something like the conventional view defends each of these claims in this specific order. The accounts that come closest are found in the work of Michelle Alexander (Reference Alexander2010; Reference Alexander2020), Katherine Beckett (Beckett Reference Beckett1997; Beckett and Sasson Reference Beckett and Sasson2003), and Vesla Weaver (Reference Weaver2007). Some authors argue that public panic was engineered by politicians rather than triggered by challenges to the racial order; that is, “frontlash,” rather than “backlash,” in Weaver’s terms. For our purposes, not much changes if one believes that politicians led and the public followed. But our presentation inverts that order, given that polls, interviews, and internal campaign documents show that the public was growing worried about crime before political elites like Goldwater or Nixon (or Johnson) began to emphasize it (Enns Reference Enns2016, 49–69).

More generally, we suspect that this story is what most scholars would invoke to defend the idea that American racism explains American mass incarceration (e.g., Cole Reference Cole1999; Hagan Reference Hagan2012; Loury Reference Loury2008; Mauer Reference Mauer2006; Simon Reference Simon2007). Even accounts that dispute that the protagonists were Republicans have the same architecture: a challenge to the racial order, followed by white backlash, and then a punitive turn that has little to do with crime (Hinton Reference Hinton2016; Murakawa Reference Murakawa2014). The thesis that mass incarceration is but the latest episode in an “antiblack punitive tradition” defines what Hinton and Cook (Reference Hinton and Cook2020) call a “burgeoning wave of literature” inspired by Alexander (Reference Alexander2010). This account also resonates with the quantitative literature on prisons and policing, in which “racial threat” theory has been a dominant paradigm (Feldmeyer and Cochran Reference Feldmeyer, Cochran, Unnever, Gabbidon and Chouhy2018).

The Challenge

In recent years this view has come under challenge from various quarters (e.g., Pfaff Reference Pfaff2017). This article is concerned with one kind of challenge. Recent work by Forman and Fortner suggests that black Americans supported policies that contributed to the growth in incarceration. Fortner (Reference Fortner2013; Reference Fortner2015) has shown that civic leaders in Harlem supported the Rockefeller drug laws, considered by many to be the model for the War on Drugs. Similarly, Forman (Reference Forman2012; Reference Forman2017) has documented that a majority-black legislature in Washington, DC, consistently passed tough-on-crime policies in the 1980s.Footnote ¹

To establish exactly how these findings challenge the conventional view, it helps to distinguish between a weaker and stronger form this challenge could take. We refer to these as “weak” and “strong” revisionism, respectively.

The weaker challenge is to the generalizability of the conventional view. Defenders of that view argue that the rise in punitiveness had very little to do with the facts of crime. If this view were to furnish a general theory of punitiveness, it would suggest that crime becomes politicized for reasons that have little to do with real trends in crime. But Forman and Fortner’s work suggests that this cannot explain what happened in black politics over this period. Thus, if they are right, it suggests that the general architecture of the conventional view does not travel.

We emphasize that this weaker challenge is not issued to a straw man. Several defenders of the conventional view argue that the rise in crime and its concentration in black communities were constructed rather than real social problems. At various points, Alexander (Reference Alexander2020), Murakawa (Reference Murakawa2014), Hinton (Reference Hinton2016), Muhammad (Reference Muhammad2011), and Eubank and Fresh (Reference Eubank and Fresh2022) express skepticism that crime was high, that it was rising, or that it was higher in black communities than in white ones. Alexander (Reference Alexander2020, 8–9) opens The New Jim Crow with this observation: “Crime rates in the United States have not been markedly higher than those of other Western countries… [and today] they have dipped below the international norm.” Murakawa (Reference Murakawa2014, 3) writes that, over the second half of the twentieth century, “The United States did not face a crime problem that was racialized; it faced a race problem that was criminalized.” Hinton (Reference Hinton2016, 7) similarly argues that the rise in crime was in part an artifact of new reporting practices and attributes the idea that black Americans were overrepresented in the ranks of offenders to a “statistical discourse about black criminality” (19) that replaced the biological racism of the prewar period. Muhammad (Reference Muhammad2011, 277) is also skeptical that black Americans were more likely to commit crime than white Americans, in light of the “invisible layers of racial ideology packed into the statistics.” More recently, Eubank and Fresh (Reference Eubank and Fresh2022, 8) argue that crime rates index the priorities of the state, rather than reflecting a real social problem to which the state responds: “Measures of crime are not objective indicators; rather, what gets logged as a crime… is at the discretion of the… carceral state.”

If the rise in the crime rate to high levels and its concentration in black communities were simply the result of where the state chose to cast its gaze, it could not have been crime—as a real social problem—that transformed black politics. This is the reason that those who make this argument today, after Forman’s and Fortner’s interventions, tend to attribute black punitiveness to elites pandering to the white mainstream (Alexander Reference Alexander2020; Jefferson Reference Jefferson2023; Taylor Reference Taylor2016).

Weak revisionism thus requires restricting the domain of the conventional view, calling its general usefulness into question. It suggests that crime is (or can be) a real social problem, and this fact will have important political consequences under some conditions. That said, it is not a challenge to the specific ambition of the conventional view, which is to explain mass incarceration. In other words, one can defend the premises enumerated earlier without having to deny the facts of crime. Beckett (Reference Beckett1997, 14), for instance, does not dispute that there was an increase in crime nor that crime is a real problem, “particularly for the poor and nonwhite.” Similarly, Weaver (Reference Weaver2007, 233) criticizes crime skeptics for relying on “questionable” evidence and throwing “the baby out with the bathwater.” For these authors, the important point is not that crime was stable, but that its rise cannot explain why crime became politicized in white America. The rise in crime might explain changes in black public opinion and black political behavior—although, until the work of Forman and Fortner we had little scholarly evidence of these changes—but if it was the changing racial order that detonated the changes in white politics, there would be no need to revise the earlier premises. The white public and white politicians were a supermajority of the relevant decision makers. Therefore, it is their changing attitudes and behavior that were decisive (Beckett and Francis Reference Beckett and Francis2020).

Thus, the stronger challenge is to the usefulness of the conventional view as an explanation for mass incarceration. To support strong revisionism, one must cast doubt on one or more of the earlier premises by showing that it was not white anxiety about black advancement that caused the transformation of white opinion and behavior, or by showing that this transformation was not the cause of the rise in the rate of incarceration. For their part, Forman and Fortner disagree about whether their evidence licenses weak or strong revisions to the conventional view. Forman (Reference Forman2017, 12) credits the conventional view with significant insights, but Fortner (Reference Fortner2015, 23) argues that “mass incarceration had less to do with white resistance to racial equality and more to do with the black silent majority’s confrontation with the ‘reign of criminal terror’ in their neighborhoods.”

Table 1 organizes the different positions in this debate. The conventional view proposes that racial backlash drove the punitive turn. If this were to furnish a general theory of punitiveness, it would suggest that backlash also explains the changes in black politics (via, say, the politics of respectability). If Forman and Fortner are right in arguing that it was crime and not the white backlash that transformed black politics, it supports weak revisionism. Finally, if any of the premises enumerated earlier are false—because it was crime and not backlash that transformed white politics—it licenses strong revisionism.

Table 1 Different Views of the Key Cause of the Punitive Turn

Our Contribution

Several issues in this debate remain unresolved. First, Forman and Fortner rely on qualitative data like media coverage, transcripts of legislative hearings, reports from government agencies, and statements made by officials and activists. These data are difficult to standardize for over-time, cross-place, cross-group, or within-group comparisons. We do not know how well their inferences would survive the use of conventional data and methods to measure public opinion or voting patterns. Scholars of public opinion have consistently found that black Americans are less punitive than white Americans, that white Americans with negative racial attitudes are more punitive than others, and that white elites use racial cues to trigger fear about crime (Bobo and Johnson Reference Bobo and Johnson2004; Green, Staerklé, and Sears Reference Green, Staerklé and Sears2006; Mendelberg Reference Mendelberg1997; Stephens-Dougan Reference Stephens-Dougan2020; Unnever and Cullen Reference Unnever and Cullen2007; Valentino, Hutchings, and White Reference Valentino, Hutchings and White2002). It is unclear how to reconcile the revisionist view of black politics with this work.

Second, both Forman and Fortner make causal inferences about the effect of black political representation on punitive outcomes, but this is difficult to do via narrative history. Their conclusions sit uneasily with the quantitative scholarship that shows a correlation between Republican control and punitive outcomes at the state level (Jacobs and Jackson Reference Jacobs and Jackson2010) and the long line of scholarship documenting “racial threat effects” in criminal justice (e.g., Campbell, Vogel, and Williams Reference Campbell, Vogel and Williams2015; Eubank and Fresh Reference Eubank and Fresh2022).

Third, Forman and Fortner’s evidence comes from case studies. There are good reasons to focus on Washington, DC, and Harlem: they are places where black people and their elected officials wielded substantial power. But the punitive turn was a national phenomenon that unfolded mostly in places in which black Americans had much less power. More than half lived in the South. It is fair to wonder whether Forman and Fortner’s inferences generalize.

Fourth, Forman and Fortner disagree about whether black punitiveness was driven by black elites or by what Fortner calls the “black silent majority.” As table 1 suggests, if black punitiveness were a ploy by black elites to curry favor with a punitive white mainstream, it could be considered a collateral consequence of racial backlash. One weakness of relying on what was published in newspapers or argued by community activists is that it makes it difficult to dismiss the charge that punitiveness was elite-led.

The main contribution of this article is to see what new evidence can teach us about these issues. The data we collected allow us to examine levels and trends in black public opinion between 1955 and 2014, assess trends in black politicians’ voting behavior in the House of Representatives between 1948 and 2017, and draw tentative causal inferences about the impact of black elected officials at the height of public concern about crime. In what follows we provide general support for weak revisionism: black Americans turned punitive and anxious about crime, and black congressmen supported key pieces of punitive legislation. Our evidence further suggests that these patterns are not explained by class. Less educated black Americans were consistently very anxious about crime and no less punitive or anxious than their more educated counterparts. Our findings thus support the inference that the rise in crime transformed black politics.

But this raises an important question: What should we make of the conventional explanation of mass incarceration in light of these findings? Should revisionism be weak or strong? Here, our conclusions remain more cautious, but we present some evidence which does not align with the conventional account of white politics, as well. We find that white Americans were consistently less anxious about crime than black Americans and that they turned punitive gradually over decades, and not suddenly in the late 1960s. We find that most of the increase in punitive voting by white politicians also happened in the most liberal parts of the United States. Finally, and more tentatively, we estimate that the effect of greater black representation (and thus, of less non-black representation) in the early 1990s was to increase the rate of incarceration and policing.

The final section, “Interpretation,” brings our findings together. We first consider an inconsistency in our results and then present general evidence in support of Forman’s observation that black Americans were also more likely to demand that governments fight the root causes of crime. Finally, in the conclusion, we argue that, even if strong revisionism is supported by the evidence, it would not be correct to conclude that mass incarceration was simply “democracy at work.”

Methods

The debate sparked by Forman and Fortner raises two types of questions about public opinion. The first (which bears on weak revisionism) concerns black America. What did black Americans believe about rising crime and punishment, and how did their views change over the period? The second (which bears on strong revisionism) concerns white America. Did the mounting fear of crime among white Americans reflect white anxiety about black empowerment, or was it more plausibly explained by a rise in crime?

Existing work on public opinion does tell us something about these questions, but we make three improvements. First, we gather new data. Our results draw on 176 nationally representative surveys of the US public administered between 1955 and 2014. These surveys together ask 39 questions to roughly 300,000 respondents (251,000 of whom were white and 34,000 black). Enns (Reference Enns2016) used data like these to identify an overall rise in the public’s punitiveness over this period. But we disaggregate the responses by race, which allows us to estimate levels of punitiveness among black and white respondents, as well as trends over time.Footnote ²

Second, in contrast to other work (Duxbury Reference Duxbury2021; Enns Reference Enns2014; Reference Enns2016), we group the 39 questions into three dimensions identified by Zimring and Johnson (Reference Zimring and Johnson2006): crime anxiety (questions about fear of crime and how much should be spent on fighting crime), punitiveness (questions about the death penalty, suspects’ rights, and the harshness of courts), and mistrust (questions about confidence in existing criminal justice institutions). This helps illuminate patterns that are obscured when all the questions are collapsed.

Third, because these questions have idiosyncratic effects on the probability that a respondent will answer affirmatively, and because they are asked inconsistently, simple averages over time are a poor estimate of actual trends. Thus, we fit a simple multilevel model to these data and use the resulting estimates to predict the probability of an affirmative response to a generic question in each dimension over time.

(1) $$ \Pr \left({y}_{i,d}=1\right)={\displaystyle \begin{array}{l}{logit}^{-1}\Big({\beta}^0+{\beta}^{race}{RACE}_i\\ {}+\hskip2px {\beta}^{sex}{SEX}_i+{\beta}^{ed}{ED}_i+{\beta}^{age}{AGE}_i\\ {}+\hskip2px {\alpha}_{q\left[i\right]}^{question}+{\alpha}_{s\left[i\right]}^{division}+{\alpha}_{t\left[i\right]}^{year}\\ {}+\hskip2px {\alpha}_{j\left[i\right],k\left[i\right]}^{race.\ sex}+{\alpha}_{j\left[i\right],m\left[i\right]}^{race.\ age}\\ {}+\hskip2px {\alpha}_{j\left[i\right],l\left[i\right]}^{race.\ ed}+{\alpha}_{j\left[i\right],s\left[i\right]}^{race.\ division}\\ {}+\hskip2px {\alpha}_{j\left[i\right],t\left[i\right]}^{race.\ year}\Big)\end{array}} $$

This model includes a set of demographic controls (e.g., $ {\beta}^{race}{RACE}_i $ ), accounts for question-level idiosyncrasies by fitting question-level random effects ( $ {\alpha}_{q\left[i\right]}^{question} $ ), and allows the effect of race to vary via race-interactions in the random effects.Footnote ³ We ran other models, but both less and more complex models yielded similar results, so we report the model here that fits best in each dimension.Footnote ⁴ We use the estimates from this model to predict the probability of a positive response to the average question in each dimension for each race X sex X age X education X census division cell in the population in a given year, where the “average” question is the one with the median random effect. Our estimate of the national, race-specific, or any subgroup average is the weighted average of these probabilities, where the weights are given by the share of the relevant population represented by each cell. We thus combine multilevel regression with poststratification (“MRP”), as does Duxbury (Reference Duxbury2021).Footnote ⁵

Results

The Black Public

Figure 1 shows the percentage of black and white respondents giving anxious, punitive, or mistrustful responses. Perhaps our most striking result is that consistently high proportions of black Americans answered that they were anxious about crime. Very large majorities of black respondents said that that not enough was being done to combat crime and that more should be spent on law enforcement, and slimmer but still significant majorities answered that they worried about crime or felt inadequately protected. Similarly, average black responses to many questions were notably punitive. When asked whether courts were harsh enough, whether they favored harsher sentencing laws, or whether they supported President Clinton’s 1994 crime bill, large majorities responded punitively. In several other questions we examined, about half of all black respondents giving clear answers answered punitively. Where they did not, their answers were mostly to questions that gave respondents alternatives (life without parole instead of the death penalty, solving problems of poverty and unemployment rather than using all available force to fight crime, attacking social and economic problems rather than improving law enforcement). We return to this later. Finally, Figure 1 also shows high levels of mistrust among black Americans. Large numbers consistently expressed a lack of confidence in criminal justice institutions.

Figure 1 Average Responses to Questions about Anxiety, Punitiveness, and Mistrust

This graph plots the share of white and black respondents giving anxious, punitive, or mistrustful responses to 39 questions asked between 1955 and 2014. The bars show bootstrapped 95% CIs.

Second, consider the differences between average responses in black and white America that are visible in figure 1 but are shown clearly in appendix figure A2. Black Americans were not just reporting high levels of anxiety about crime: they generally reported higher levels of anxiety about crime than did white respondents. Where the differences were slight or ran in the other direction, they were in response to questions about what ought to be done about crime; for example, “Please indicate whether you would like to see more or less government spending [on] the police and law enforcement.” Figure 1 does also show that black Americans were consistently less punitive than white Americans. Even if they were more concerned about crime, they were not typically as punitive as white Americans about what ought to be done about it. We say more about how to interpret this in what follows. Finally, we find that black Americans were more mistrustful of—that is, had less confidence in—police and the legal system.

Finally, consider trends in the average level of punitiveness, anxiety, and mistrust. Figure 2 plots the probability of an affirmative answer to a generic question in each dimension. We find that the average black American became more punitive, more anxious, and more mistrustful over this period. In each case, the probability of an affirmative response rose in the 1960s and 1970s, peaked in the 1980s and the 1990s, and declined after 2000. The fact that answers in these three domains trend together and peak in the 1980s and 1990s suggests that their lack of confidence reflects, at least in part, black Americans’ concerns about crime and not just concerns about criminal justice overreach (Kennedy Reference Kennedy1998; Leovy Reference Leovy2015; L. Miller Reference Miller2013).

Figure 2 Trends in Anxiety, Punitiveness, and Mistrust

This graph plots trends in anxiety, punitiveness and mistrust, obtained via the procedure described in the methods section. The lines are loess smooths through these estimates. Appendix Figure A3 shows raw levels.

To summarize, our data suggest that black Americans exhibited high levels of crime anxiety (that were, if anything, higher in black America than in white America), growing levels of punitiveness about crime (that seem to have reached absolutely high levels in the 1980s and 1990s), and growing and high levels of mistrust in criminal justice institutions. Admittedly, not every finding challenges the hypothesis that the rise in crime was political artifice and the rise of punishment nothing but racial backlash. It would not, on that view, be surprising to see high and rising levels of mistrust in black America, as well as a white–black gap in punitiveness and mistrust. But we seek an explanation that can make sense of all these findings together (Fairfield and Charman Reference Fairfield and Charman2022). In light of high levels of anxiety and the growing and high levels of punitiveness, we think the more likely explanation is that, over this period, black politics was being transformed by the rise in crime. This transformation did not leave black Americans without reservations about what should be done about crime, as we emphasize in what follows. But it seems that one can only account for the general contours of black opinion by accepting that concern for crime was not simply politically constructed.

The Black Elite?

Our results thus seem to support weak revisionism. But one “conventional” way of making sense of black crime anxiety and punitiveness is to attribute them to black elites. The Civil Rights Movement greatly expanded the political power of African Americans. Many more black Americans entered government, where they found themselves needing to pander to powerful whites. Both Alexander (Reference Alexander2010) and Forman (Reference Forman2017) criticize black leaders for being blind to the effects of the punitive turn on ordinary black Americans. But perhaps the clearest exponent of this view is Taylor (Reference Taylor2016, 98), who argues that “the call for law and order in Black communities… reflected the difference between being in power and being outside power.” Elsewhere, Taylor explains the Congressional Black Caucus’ (CBC) support for the War on Drugs by claiming that black elected officials were not aligned with their constituents (96) and attributes the CBC’s cosponsoring of “conservative law-and-order politics… [not to] political weakness but entrenchment in Beltway politics” (100). On this view, punitiveness in the black community is a form of respectability politics (Higginbotham Reference Higginbotham1994; Jefferson Reference Jefferson2023). Alexander (Reference Alexander2020, 282) herself invokes this idea to explain why civil rights organizations have been reluctant to take up the rights of victims of the carceral state.

Nationally representative polls are too small to include the political class at the center of Taylor’s argument (2016), but one way to test the hypothesis is to search for evidence of an elite versus non-elite gap in punitiveness. We thus compared average levels of anxiety, punitiveness, and mistrust felt by black high school dropouts versus those felt by black college graduates. Because levels of education are confounded by age and cohort, we limited the analysis to surveys between 1980 and 2000 and to respondents who were 45–64 at the time the survey was given. We also limited it to surveys that had at least 15 respondents in both groups, which is why some questions shown in figure 1 are not visible in figure 3.

Figure 3 Average Responses by Education in Black America

This graph plots responses given by black high school dropouts and black college graduates. We limit the analysis to years between 1980 and 2000, to respondents who were age 45–64, and to surveys for which we had at least 15 respondents in both groups. Bars represent bootstrapped 95% CIs.

Figure 3 shows minimal support for the respectability politics hypothesis. First, levels of anxiety among non-elite blacks were very high. Of those who gave clear answers, large majorities answered that they thought too little money was being spent on law enforcement (or, conversely, that more money should be spent on fighting crime).Footnote ⁶ Second, we find no clear-cut evidence of significant differences in levels of anxiety, punitiveness, and mistrust between the two groups. If, as Taylor (Reference Taylor2016) argues, black elected officials voting punitively were out of touch with ordinary black people, we would expect these differences to be much more pronounced. Yet figure 3 shows that differences are difficult to discern. Appendix figure A6 estimates this adjusted education gap based on equation 1. It further confirms that these differences are slight, if they exist at all. Across all questions and all years, and adjusting for other covariates, more educated black Americans report marginally higher levels of crime anxiety, marginally lower levels of punitiveness, and marginally higher levels of mistrust. Only the last of these differences is statistically significant (at $ \alpha =0.10 $ ).

The White Public

Thus, the first lesson is the weak revisionist one: racial backlash cannot account for black public opinion. Yet, as we have emphasized, this does not show that the conventional view cannot succeed on its own terms. After all, that view is not principally a thesis about black public opinion. Even if the shape of black public opinion better fits the facts of rising and falling crime, white opinion might still reflect growing anxieties about the changing racial order.

It is important to emphasize here that some of what we find does fit the conventional view. For instance, we find that whites are consistently more punitive than blacks. We also find that this gap was narrow and perhaps even not obvious in the late 1950s but grew significantly over the 1960s and 1970s (see appendix A5). American racism seems the only plausible explanation of why white Americans are today more likely to be punitive than black Americans, despite being less likely to be victims of crime. Racial backlash is also the most plausible explanation for the fact that this punitiveness gap grew in the 1960s and 1970s, as the racial order was being challenged.

Yet, our results also raise some serious questions. Consider trends in punitiveness. As figure 2 shows, white punitiveness rises during the 1960s and 1970s, as the conventional view would expect. But it continues to rise until the 1990s and then declines gradually in the late 1990s and 2000s. The conventional view identifies the Civil Rights Movement as the detonating cause of changing attitudes. But, as figure 4 shows, race-related protests and riots per capita peaked roughly 25 years before the peak in white punitiveness. Punitiveness was in fact at historically low levels when the Civil Rights Movement was at its peak, and it only reached its highest level long after civil rights mobilizations had faded in strength. Similar patterns are evident in white anxiety about crime and white mistrust of criminal justice institutions. White anxiety about crime peaks in the mid-1990s well after the Civil Rights Movement had declined. And white mistrust of criminal justice institutions grows in step with the rise in crime, rather than falling as the conventional view might lead us to expect.

Figure 4 Punitiveness, Protests, and Crime over Time

This graph shows trends in race-related protests and riots (per capita), white punitiveness, and violent crime and homicide between 1950 and 2014. The units here are standard deviations above and below the mean over this period.

Figure 4 also suggests that trends in punitiveness are closely correlated with trends in homicide and crime. This does not confirm the causal inference that crime caused these changes. However, the absence of any correlation between protests and white public opinion suggests that something other than the ebb and flow of racial conflict must be doing most of the explanatory work. Even though the absence of a correlation between X and Y cannot establish that X did not have a causal effect on Y, it suggests that most of the variance in Y must be explained by something other than the variance in X. The power of the conventional view is greatly diminished if something other than backlash explains the rise in white punitiveness.

How can we square the fact that whites are more punitive than blacks with the observation that backlash cannot explain over-time trends in white punitiveness? One possibility is that cross-sectional and longitudinal variance have different explanations (Pickett Reference Pickett2019). Facts that account for cross-group variation are not necessarily good explanations of national trends, any more than the fact that taller people earn more money than shorter people means that economic growth can be explained by an increase in average height. Note that if trends in white public opinion have little to do with racial backlash, it would support strong and not just weak revisionism. To furnish an explanation of mass incarceration, the conventional view cannot simply be an account of why white people have been consistently more punitive than black ones. It must explain why white people grew more punitive after the 1960s.

Methods

The debate raises two analogous questions about politicians: How did black politicians vote? And how did white politicians vote? To answer these, we analyze voting patterns in the House of Representatives. What share of black members of Congress, non-black Republicans, and non-black Democrats voted for key legislation over this period?Footnote ⁷ How did their voting patterns evolve over time?

To do this, we manually coded the 46 most prominent roll-call votes relating to federal crime policy between 1968 and 2015. In the 35 bills and amendments that increased mandatory minimum prison sentences or gave more power and resources to prosecutors and police, a “Yea” vote was coded as a punitive vote. For those bills and amendments that promised to reduce sentence length or severity, restrict police or prosecutorial power, or provide alternatives to incarceration, a “Nay” vote was coded as a punitive vote.

This hand-coded sample contains only a subset of the relevant votes in the House. We identified 952 other relevant votes after 1945, using the Congressional Research Service policy areas and issue codes that best identified our hand-coded subset (Lewis et al. Reference Lewis, Poole, Rosenthal, Boche, Rudkin and Sonnet2022). The principal challenge was to identify the direction of a punitive vote in the votes that we did not code ourselves. We solved this problem by training a model to predict the direction of a punitive vote in this sample using the information on the average ideology of “Yea” and “Nay” voters in our hand-coded dataset, measured by the average two-dimensional DW-NOMINATE scores. To pick the best model, we split 70% of the hand-coded dataset into a training set and 30% into a test set, experimenting with four classifiers. We picked the model that performed best across 10 iterations of this process. This was a straightforward logistic regression in which the cutoff for the outcome was a predicted probability of 0.50. This model correctly identified the direction of punitive voting in 84% of the bills using just the average DW-NOMINATE scores. This yielded a final sample of 998 votes that we use to summarize levels and trends in punitive voting.

Results

Black Politicians

Advocates of the conventional view argue that punitive legislation represented a revanchist reaction to the Civil Rights Movement. If this were all one knew about the punitive turn, one might expect that black politicians—whose arrival into office in significant numbers was one of the signal achievements of the Civil Rights Movement—would have opposed the passage of these racially motivated policies.

Accordingly, we do find that black members of Congress have been consistently less punitive than non-black Democrats (who have in turn been consistently less punitive than non-black Republicans). As figure 5 shows, between the late 1960s and 2000, there is a visible difference between the share of black and non-black Democrats voting for punitive legislation.

Figure 5 Trends in Punitive Voting in the Federal House of Representatives

This graph plots trends in punitive voting in the House. The lines are loess smooths fit to the average percentage of members of each group who voted punitively on bills in a given year. The sample includes 998 bills. Because very few bills predate 1960, we start the graph at 1960 (and the point in 1960 represents the average of all pre-1960 bills).

But two other results suggest something closer to weak revisionism. First, an absolute majority of black members of Congress voted in favor of each of the major federal crime bills of the punitive turn: the Omnibus Crime Control of 1968, the Comprehensive Crime Control Act of 1984, and the Violent Crime Control and Law Enforcement Act of 1994. A majority also consistently supported bills that increased mandatory minimum sentences for those at the center of public outrage, such as drug dealers in the 1980s and sex offenders in the 2000s. For instance, a majority of black members of Congress (65%) voted in favor of the Anti-Drug Abuse Act of 1986, which imposed the notorious 100–1 disparity in sentencing for crack versus powder cocaine.

Second, figure 5 also shows a discernible increase in black support for punitive legislation. Between the 1940s and 1960s, spanning 36 relevant votes in the House, an average of about 30% of black members cast punitive votes. In the 1970s, this increased to 42% (spanning 106 punitive votes), and in the 1980s, it was 43% (spanning 94 punitive votes). One can frame these trends in another way. The presidents most associated with racial backlash were Richard Nixon and Ronald Reagan. Under their presidencies, we might expect that black members would have been especially averse to supporting punitive legislation. Yet the share of black members voting for punitive legislation was higher under Nixon than Johnson (37% vs. 32%), and even higher under Reagan (45%). As figure 5 shows, by the end of this period—around the time that incarceration rates in the United States had reached their highest levels—there was only a very small gap between black and non-black Democrats in the House.

Again, one might think that these votes simply register the preferences of a comprador class (Taylor Reference Taylor2016). But our public opinion data suggest this is the more unlikely explanation. Black elites were in fact consistently more mistrustful of criminal justice institutions than black non-elites (and, although the difference was not statistically significant, less punitive). Further, as Forman (Reference Forman2017) argues and as we show later, black elected officials often voiced reservations about their support for punitive legislation. Later we use a case study of the 1994 Crime Bill to suggest that black members of Congress voted in this way because they faced such pressure from their constituents that they preferred to do something imperfect about crime than to do nothing at all.

White Politicians

We thus find general support for the idea that the rise in crime transformed the behavior of black politicians. But, as before, this on its own does not challenge the conventional explanation of mass incarceration. It is possible that black politicians were concerned about crime, but that the behavior of white politicians is still best explained by the politics of backlash.

Here our evidence is limited, but we sound one note of skepticism. If the conventional view were true, the rise in punitiveness should have been especially pronounced among the politicians at the leading edge of racial backlash. Figure 5 challenges that inference, because it suggests that the rise in punitive voting—like the rise in punitiveness among the public—was bipartisan and thus was a general feature of non-black voting in the 1960s and 1970s.

Figure 6 shows this in another way. Here we ignore partisan differences and split the non-black sample into two groups: the five states that Barry Goldwater won in 1964 (AZ, LA, MS, AL, GA, SC), and the five states in which Lyndon Johnson won by the largest margin (MI, MA, CA, PA, NY). The graph shows rising punitiveness not just in the Goldwater states but also in these liberal states. As the conventional view might anticipate, the rise in punitiveness is more pronounced in the first group, but punitive voting rises considerably among non-black representatives from the most liberal states as well. Between 1960 and 2018, punitive voting rises by about 105% in the Goldwater states and about 84% in the most liberal states. Put another way, around 80% of the rise in punitiveness in Goldwater states also occurs in these most liberal states.

Figure 6 Trends in Punitive Voting and the Southern Strategy

This graph plots trends among non-black politicians in the 998 votes described in figure 5. The solid line shows trends in punitive voting for non-black politicians from the five states that voted for Goldwater in 1964, and the dashed line shows trends in the five states that Johnson won by the largest margin.

Some have amended the conventional view to account for facts like this one. Both Hinton (Reference Hinton2016) and Murakawa (Reference Murakawa2014), for instance, argue that backlash was bipartisan (but see Usmani Reference Usmani2017). Yet these trends might also be explained by the facts noted earlier: crime was rising, and voters around the country were increasingly open to punitive suggestions about what should be done about it.

Methods

Finally, if the punitive turn were mainly explained by racial backlash, a decline in the political influence of the protagonists of this backlash (white Americans) should have dampened it. If it were not, and it was instead mainly explained by the rise in crime, a rise in the influence of those who were most affected by crime (black Americans) should have accelerated it. Thus, the conventional view and strong revisionism yield contrasting hypotheses about the effect of black enfranchisement. Which is correct?

We seek to answer this question by exploiting an instance of federal intervention into the state-level electoral process, following Ueda (Reference Ueda2005). (Here we do something similar to Eubank and Fresh [Reference Eubank and Fresh2022] but in a different setting.) The Voting Rights Act of 1965 empowered the federal government to intervene in state elections to ensure minority representation. In 1982 Congress amended the act to prohibit voting schemes that result in minority vote dilution. Subsequent Supreme Court decisions simplified the criteria for overturning discriminatory schemes. As a result, when it came time to redraw districts after the 1990 census, states were under pressure to maximize the number of districts in which minorities would form a majority of voters. By the elections of 1992, a total of 83 new majority-black electoral districts had been created, a 25% increase from 1990 (Grofman Reference Grofman2003, 18–19). The resulting influx of black politicians into state and federal legislatures has been characterized as “the single largest increase in black representatives in US history” (Kim Reference Kim2002, 65).

Two facts about these changes bear emphasizing. First, although black mobilization inspired federal intervention into state elections, the sharp increase in the number of elected officials immediately after 1990 was the result of redistricting following the census, the timing of which was exogenous to black protest (Grofman Reference Grofman2003, 16). Second, redistricting affected those states covered by the Section 5 provisions of the Voting Rights Act more dramatically than others. In unaffected states, the percentage of black legislators increased slightly but at a rate continuous with pre-1990 trends. By contrast, we find a sharp discontinuity in the affected states between 1990 and 1995, reflecting the post-1990s influx (see appendix figure A7).

To estimate the impact of this influx, we exploit the fact that redistricting affected only a subset of all states and for a confined period of time.Footnote ⁸ Specifically, we compare trends before and after redistricting in states that were subject to it to trends before and after in states that were not.

$$ {DV}_{st}=\left({RD}_s\times {PD}_t\right)\theta +{\delta}_s+{\mu}_t+{\sigma}_d\times {\mu}_t+{\varepsilon}_{st} $$

In this equation, $ {DV}_{st} $ denotes either of our dependent variables, the number of prisoners and police officers per 100,000 population, observed in state $ s $ and year $ t $ . $ {RD}_s $ is an indicator variable that takes the value 1 in redistricted states and 0 otherwise. $ {PD}_t $ is an indicator variable that takes the value 1 in the period after redistricting. $ \theta $ is thus our estimate of the effect of redistricting on our dependent variable. $ {\delta}_s $ denotes a set of state-fixed effects, and $ {\mu}_t $ denotes a set of year-fixed effects.

Of course, redistricting hit a nonrandom sample of US states. Thus, the principal threat to identification is the possibility that redistricted states were already different in ways that would have led to different outcomes across the two groups. To account for this, we include the state- fixed effects and a term that estimates flexible trends over time within each census division (i.e., the interaction of year-fixed effects and divisional dummies, denoted by $ {\sigma}_d\times {\mu}_t $ ). We lack the degrees of freedom to control for flexible trends at the state level, but in running robustness checks, we estimate a model with a linear state-level trend line, as well as a model that, following Eubank and Fresh (Reference Eubank and Fresh2022), is limited just to the states that Katznelson and Mulroy (Reference Katznelson and Mulroy2012) identify as having had Jim Crow laws (i.e., states that are most similar to those that experienced redistricting). We do not include other time-varying controls, such as the crime rate, because they would affect our estimate in uninformative ways. There is no way of distinguishing the bias of omitting any relevant controls (which we would want to avoid) from the bias of controlling relevant consequences of redistricting (which we would not). Appendix figure A8 reports results from other specifications, including one that includes these controls.

Results

Figure 7 shows that our estimate of the impact of federally mandated redistricting on both prisons and police is positive. We estimate that redistricting caused a 0.88 standard deviation increase in the incarceration rate (95% CI: 0.23–1.54) and a 0.54 standard deviation increase in the officer rate (95% CI: 0.05–1.02). These estimates represent increases of 122 prisoners and 20 police per 100,000 people, respectively.

Figure 7 Estimated Impact of Black Political Representation on Punitive Outcomes

This graph plots estimates of the effect of black political representation on the policing rate (number of police officers per 100,000) and the incarceration rate (number of prisoners under state or local jurisdiction in a state per 100,000), which we obtain by exploiting federally mandated redistricting enabled by the 1990 census.

Figure A8 illustrates results from a range of robustness checks. Our results are robust to most but not every specification. When adding state-specific linear time trends or limiting the sample to Jim Crow states, our estimate of the impact of redistricting on policing is still positive but statistically indistinguishable from zero at conventional levels of significance. (Our estimate of the impact on imprisonment remains large and statistically significant.) Notably, we find no evidence of a “conventional,” negative effect on these same outcomes.

What Is the Counterfactual?

In one way, our results are not readily reconciled. On the one hand, in the federal House of Representatives, black elected officials voted less punitively than non-black Democrats. Had all Democrats voted as they did, outcomes in the House would have been less punitive. This suggests that black elected officials had an attenuating impact. On the other hand, our analysis of redistricting suggests that more black representation would have yielded more punitive outcomes.

The regression results may be spurious. But if they are not, one possibility is that the inconsistency in these results is an artifact of American federalism. The first estimate is based on evidence from Congress, whereas the second is supported by evidence from the states. Yet, we should not expect the politics of crime and punishment to look the same at the federal and state levels. Federal-level politicians have a wider array of policy tools at their disposal. They can fight crime not just by appropriating funding for police and prisons but also by enacting social policy. This kind of advocacy is more difficult at lower levels, where governments are constrained by institutional precedent and the mobility of their tax base (Peterson Reference Peterson1981). It would thus have been easier for members of the Congressional Black Caucus to avoid the trap of “punishment-or-nothing” politics than it would have been for the black elected officials who entered state legislatures in the early 1990s.

Rather than trying to pin down the central tendency of black politics, it may just be worth observing that black politics has obviously spanned the distribution that these two results suggest. Public opinion data showed that black people were both more anxious about crime and less punitive about what governments should do about it. The second result implies that, in some places and at some times (e.g., at the height of worries about crime in the early 1990s), elevated levels of anxiety in the black population may have prevailed over their lesser punitiveness. Where black elected officials had a punitive impact, it was likely because higher levels of crime anxiety among their constituents made these elected officials particularly likely to bring crime to the attention of the legislatures in which they served. Had these black elected officials never found themselves in power, white legislatures may have paid less attention to crime and thus passed fewer punitive crime bills. Kennedy (Reference Kennedy1998) argued that something like this obtained in the era before black enfranchisement, where the common complaint was that black neighborhoods were consistently underprotected; see also L. Miller (Reference Miller2013) and Leovy (Reference Leovy2015). By contrast, the first result implies that, at other places and at other times, the lesser punitiveness of the black population prevailed over their higher levels of crime anxiety. It makes sense that this would be more likely to occur at higher levels of government, where black politicians had more ability to respond to anxiety about crime with an array of nonpunitive tools.

Penal and Social Policy

Our findings register the profound effect of crime on black public opinion and black politics. Punishment, we have argued, can sometimes be a response to a real problem of public concern. For Fortner (Reference Fortner2015), this observation suggests a relatively simple amendment: the center of gravity of black politics was further (perhaps, much further) to the right than has been understood. But, as Forman (Reference Forman2017) emphasizes, punishment was not the only action that the black public or even black politicians proposed to take in response to crime. He argues that black Americans were also distinctive in a second dimension: they were also more likely than the white public and politicians to demand that the government spend resources to combat the social and economic causes of crime. Here, we offer some general evidence for Forman’s view.

First, three of the questions we examined asked respondents what they thought of programs to tackle the root causes of crime and violence. Large majorities expressed support for this agenda. Of those giving clear-cut answers, around 69% of African American respondents answered that the best way to solve crime would be to attack the social and economic problems that lead to it (95% CI: 65%–72%), 56% answered that it is more important to get people in prison started on the right road than to punish them (95% CI: 50%–62%), and 88% said that the best way to combat urban unrest and rioting would be to solve poverty and unemployment (95% CI: 86%–90%).Footnote ¹⁰

Second, black elected officials seem to have had analogous preferences. L. Miller (Reference Miller2008) and Forman (Reference Forman2017) present evidence from analyses of local politics in Philadelphia and Washington, DC, respectively. Our evidence comes from the legislative history of Clinton’s 1994 crime bill. The Congressional Black Caucus (CBC) played an important role in shaping early drafts of the bill, having blocked passage of a draconian version passed by the Senate (and supported by Clinton) in the fall of 1993. But this was not because there was no concern about crime in the black community at the time. In fact, Jesse Jackson’s National Rainbow Coalition convened a three-day conference about crime in January 1994. There, leaders of the CBC expressed mixed support for punitive solutions while emphasizing that fighting crime would also require funds for crime prevention. The next year, CBC members of the House Judiciary Committee helped draft a new bill that was both less punitive than the Senate’s version from the previous fall and funded crime-prevention programs targeted at poor neighborhoods. They also fought to add a “racial justice” provision, which allowed death row inmates to challenge a sentence as racially discriminatory. As this bill made its way through the House, black elected officials consistently opposed amendments designed to make it more punitive and less preventive in orientation.

Eventually, however, the CBC was outmaneuvered. Its interventions did not survive legislative wrangling. The Clinton administration realized that it did not need the CBC to pass the bill and instead courted moderate Republicans. The resulting compromises involved a reduction of funding for crime-prevention measures that CBC members had fought to include. Critics of these measures focused on a small number of earmarks for “midnight basketball” programs in inner-city neighborhoods, and funds for crime prevention came to be increasingly identified by both sides of the debate as a “black” issue (Wheelock and Hartmann Reference Wheelock and Hartmann2007). Yet, notably, even as the CBC saw some of their most valued provisions stripped from the bill, a majority of members continued to support it. When the House voted on the final bill, the CBC voted 24–12 in favor.

This final vote lays bare the bind facing black advocates (see also Forman Reference Forman2017, 204). It also illustrates our earlier point that the level of government mattered. On the one hand, black politicians made the drafting and refining of the bill more than a simple, “punishment-or-nothing” issue. They introduced funds for crime prevention, which would have been much more difficult at lower levels of government. Yet two-thirds of the caucus members voted in favor of legislation about which they had many reservations. Even though the bill did not have a punishment-or-nothing history, it ended up a punishment-or-nothing vote. And when confronted with this kind of choice, a large majority of the CBC decided that the dividends of doing something about crime outweighed their reservations about a bill that exaggerated their punitiveness and flattened their preferences. Their bind mirrored the plight of their constituents. Even though large majorities of black Americans in this period expressed their support for fighting the root causes of crime, a September 1994 Gallup poll records that 63% (95% CI: 55%–71%) of black respondents were in favor of the bill, and only 18% (95% CI: 12%–25%) were opposed.