Decarceration and…? Local Policy Responses to State-Level Criminal Justice Reform

Public Safety Realignment

In 2011, the state of California was ordered by the U.S. Supreme Court to reduce its prison population quickly and substantially because of the intense health risks posed to incarcerated people by severe overcrowding and inhumane conditions (Bird and Grattet 2014; Valenzuela 2012). This order resulted in Public Safety Realignment, a package of legislation that sought to stymie the inflow of people to state prisons by diverting those convicted of nonviolent, nonsexual, nonserious (or non–non–non) offenses to county jails and expanding probationary/split sentencing for these crimes (Bird, Nguyen, and Grattet 2022; Latessa, Johnson, and Koetzle 2020; Lofstrom and Raphael 2016). Functionally, Realignment resulted in a large decrease in the number of people incarcerated in California's prisons and a smaller (and more variant) decrease in the overall number of incarcerated people at the state level (27,000 prisoners, ∼17,000 incarcerated people total from 2011 to 2012; Bird, Nguyen, and Grattet 2022; Lofstrom and Raphael 2016, 201–2; Mauer and Ghandnoosh n.d., 7). Table 1 summarizes the specific legislation included in the package. ²

OPEN IN VIEWER

Table 1.

California's Public Safety Realignment Package.

Specific Law	Description of Law's Intent
AB 109	Provides a dedicated and permanent revenue stream to the counties—funding became constitutionally guaranteed by California voters under the passage of Proposition 30 in 2012. Allows nonviolent, nonserious, and nonsexual offenders to serve their sentence in county jails instead of state prisons (counties can contract back with the State to house local offenders, except for parole revocations). No inmates are transferred from state prisons to county jails. No state prison inmates are released early. All felons sent to state prisons prior to the implementation of Realignment will continue to serve their entire sentence in state prisons. All parole revocations are served in county jails instead of state prisons and can only be up to 180 days.
AB 111	Gives counties additional flexibility to access funding to increase local jail capacity for the purpose of implementing Realignment.
AB 94	2011 Realignment Legislation Addressing Public Safety; lowers the county's required contribution from 25 percent to 10 percent and additionally requires California Department of Corrections and Rehabilitation and the Corrections Standard Authority to give funding preference to those counties that relinquish local jail construction conditional awards and agree to continue to assist the state in siting reentry facilities.
AB 118	Establishes the Local Revenue Fund 2011 (Fund) for counties to receive the revenues and appropriate funding for 2011 Public Safety Realignment.
SB 89	Dedicates a portion ($12) of the Vehicle License Fee (VLF) to the Fund (freed up VLF previously dedicated to DMV administration and VLF that was previously dedicated to cities for general purpose use).
SB 87	Provided counties with a one-time appropriation of $25 million to cover costs associated with hiring, retention, training, data improvements, contracting costs, and capacity planning pursuant to each county's AB 109 implementation plan.
AB 117	Amendments to AB 109.

I describe the criminal landscape in California to illustrate both the salience of criminal justice as a policy priority in that state and the reasons Public Safety Realignment came about. California is much like the rest of the United States in its crime and incarceration landscape; the crime “surge” of the 1960s–1980s was followed by an incarceration surge in the late 1980s and 1990s (Harris and Cremin 2024; Lofstrom and Martin 2025). Data show that the constellation of new and more punitive sentencing laws resulted in a large drop in the number of people in California incarcerated in jail (from about 425 per 100,000 people to about 200 per 100,000 people in the early 1990s), while the number of people incarcerated in prisons remained high (Harris and Cremin 2024). As in the rest of the country, the overall crime rate in California has decreased substantially since its peak in the 1990s. ³ The public health risks that overcrowding in prisons creates are not unique to California either. In the last twenty-five years, 26 states have been subjected to court orders mandating improvements in the healthcare and living conditions inside prisons (Nam-Sonenstein 2015). Although California is the only state to have been ordered by the court to reduce its prison populations, many states and the federal government have passed several policies where decarceration is the primary goal (Kimble and Grawert 2019; Mauer and Ghandnoosh n.d.; Schrantz and DeBor 2018). The implications of this study are, therefore, wide-reaching.

Broader Institutional Background

To make the full scope of the Realignment changes clearer, I offer a brief description of the relevant institutions and intergovernmental relationships here. First, it is important to specify the differences between prisons and jails in the United States. Prisons are long-term facilities holding only people who have been convicted of criminal offenses. Long-term, in this case, refers to the length of time a person spends incarcerated: Most people in a prison are serving sentences longer than a year. Each prison is run by the Federal Bureau of Prisons, a state's department of corrections, or a private company (usually contracted to the state). Jails are short-term facilities where people are held both before and after conviction. People are typically incarcerated in a jail for a year or less, and jails are typically locally operated (by a sheriff, police chief, or other administrator), although some are privately owned and some are operated by two or more independent jurisdictions (Bureau of Justice Statistics ⁴ ). The state of California has 31 state-run prisons, 12 federal prisons, and over 60 jails. For more information on the political history of California's prisons, I refer the reader to Ruth Wilson Gilmore's Golden Gulag (2007).

Second, understanding the constellation of local government types in the United States is essential for this study. The U.S. Census designates two broad categories of local governments: general purpose (those governing counties, municipalities, and townships) and special purpose governments (school districts, special districts, and regions). This study's inquiry is focused on municipal (city) and county governments, so I will describe these further here. Counties are typically established by the state for “performing state-mandated duties” (National League of Cities ⁵ ). The scope of county governments’ responsibilities and power has grown as the U.S. population has become more urbanized; this led almost all counties (98 percent) to establish their own, independent charters over the last several decades, but the scope of these is still constrained by state governments. City governments are hugely variant in their structure and size, but all have similar powers and functions. Power-sharing between cities and counties differs hugely across the United States and is largely determined by population size and delineated in formal power-sharing agreements (Lubell, Henry, and McCoy 2010). ⁶

Finally, I offer a brief description of the mechanics of the U.S. criminal–legal system, beginning with arrest and ending with incarceration. ⁷ It is critical to acknowledge that there are many additional steps and complexities I do not attend to here for the sake of clarity, that moving through the criminal–legal system is a reflection of the actions of many people, especially legal officials, and that the following description is focused on people who are eventually convicted of a crime (National Academies of Sciences, Engineering, and Medicine 2022). When an adult has been arrested, they are typically incarcerated in a jail, where they are held until charges are filed or they are released. If charges are filed, the person can either be diverted to a support program or will move through a prosecution process in a court (Johnson and Ali-Smith 2022). If the person is charged with a felony (a more serious crime) and is convicted, they are typically sentenced to probation (supervision outside a prison) or to incarceration in a prison. If the person is charged with a misdemeanor (less serious crime), they are typically sentenced to probation or to incarceration in jail. Once incarcerated, a person has almost no control over where they are held or whether they are transferred to another facility during their sentence.

Utility in the Current Study

Public Safety Realignment substantially altered the carceral landscape in California. Of most relevance to this study is the fact that the state, using its institutional power over the counties, shifted responsibility over people convicted of all but the most serious crimes to county administrators and jails (Bird and Grattet 2014; Lofstrom and Raphael 2016). The policy package also included legislation that reclassified certain crimes to become newly eligible for split-sentencing, or a combination of a shorter period of incarceration with a subsequent period of probation (Mauer and Ghandnoosh n.d.). While Realignment's mandate was from the state to the counties, the reduced number of incarcerated people (whether diverted from supervision entirely or through increased noncarceral sentencing) was a highly salient outcome to city officials, as it necessarily translated to an increased number of people with criminal histories on streets. This is evident in the council meeting transcripts I analyze in the section “LocalView Analysis,” where local officials express concerns about undue “burden on local police departments,” describe transitioning to “alternatives to custody,” and argue that they can’t “arrest our way out of our crime statistics.”

This paper's research question is: How do city policymakers alter their spending behavior after state-initiated criminal justice reforms? Realignment has already provided a fruitful opportunity for exploring the impact of criminal justice reforms on outcomes such as crime rates (Lofstrom and Raphael 2016), recidivism (Bird and Grattet 2014; Bird, Nguyen, and Grattet 2022; Lofstrom, Raphael, and Grattet 2014), and judicial behavior (Ball and Weisberg 2014; Weisberg and Quan 2014). However, a critical question that has gone largely unaddressed in existing research is how city policymakers responded to this state-mandated increase in control over the Realignment population and the resultant drop in the number of incarcerated people in general. Answering this question is essential because a key tenet of Realignment was that “managing lower-level offenders in locally run, community-based programs using evidence-based practices” is more effective than incarceration in improving individual-level outcomes (Bird and Grattet 2014; Latessa, Johnson, and Koetzle 2020; Petersilia et al. 2014, 2). Did the design, resources, and implementation structures of Realignment create feedback effects that reflected this tenet? This paper explores this by empirically testing whether and how city spending changed in response to these elements.

Uniform Crime Reports

I expect spending behavior at the city level to vary based on how high their average arrests per capita were in the five years leading up to Realignment. This is the treatment variable for my difference-in-differences analysis. I use the UCRs for this variable because they include the number of arrests by precinct and crime type, allowing me to identify the number of arrests by city. Figure A.1 in the Supplemental Appendix shows the distribution of average arrests per capita across my sample. While most of the cities in my dataset have an average arrests per capita value between 0 and 0.2, there are some outliers. Thus, I use a binary “high-arrests” variable as the primary treatment in my difference-in-differences regressions to ensure these cities do not skew the results, and include an alternative specification using continuous arrests in Supplemental Appendix Table A.8.

I include all types of arrests in my measure because I expect cities’ overall crime rates to be most salient to local officials. However, I also include an analysis in Supplemental Appendix A using only qualifying (nonviolent, nonsexual, nonserious) crimes per capita as the independent variable to test whether the results persist under this specification (Supplemental Table A.9). The most liberal version includes all nonviolent, nonsexual, nonserious crimes. The medium definition excludes the sale and possession of opiates and synthetic drugs. The conservative definition includes only nonviolent property crimes; drug crimes related to marijuana and other nonopiate, nonsynthetic drugs; crimes related to liquor, drunkenness/disorderly conduct, and vagrancy; and crimes listed as “other.” The drug-only definition includes only nonopiate, nonsynthetic drug crimes, to try to capture only low-level drug offenses. While these classifications are based solely on the most serious crime type associated with the arrest and do not capture the actual severity of the crime itself (i.e. type of felony), arrests are still commonly used as the measurement for city crime rates, given that people are not always incarcerated in the place where they were arrested.

Again, the base assumption is that policymakers will be aware of how many people under their jurisdiction are arrested, and that those with proportionately higher shares of arrestees (those that are more “treated”) will respond differently to reduced incarceration than those with lower shares (those that are less “treated”).

Spending

The outcomes of interest for the difference-in-differences analysis are local spending on social policy and criminal justice. I use spending data from 2000 to 2016 from the Census’ Government Finance Database. My primary analysis is focused on total social policy spending per capita, but given the evidence that local policymakers exercise more control over certain policy areas, I explore heterogeneity across four social policy categories (health and welfare, economy and employment, housing and land use, and education) in Supplemental Appendix Table A.4. Supplemental Appendix Table A.1 contains detailed descriptions of the specific expenditure line-items included in the social policy expenditure variables, separated into four subcategories defined in table Supplemental Appendix Table A.4. For example, health and welfare includes the distribution of cash benefits to individuals and vendors for public assistance programs like Medicaid (health insurance for people with limited income) and Supplemental Security Income (a means-tested cash assistance program for people with disabilities and senior citizens), additional public health expenditures, and expenditures for hospitals. I chose to include these specific line items in the broader health and welfare category because they incorporate expenditures that are both directly redistributive in nature and expenditures that generally target public welfare/well-being. Including both types of expenditures across my four major categories was important because this paper's primary inquiry is whether the design of Realignment's policy provisions impacted local government spending behavior, specifically by making nonenforcement responses seem more pragmatic.

Since most expenditures for redistributive programs and policies come from the federal or state level and are often allocated directly to service providers, rather than to individuals themselves, it was important to incorporate expenditure categories in my overall measure that corresponded to areas over which local governments have more control. This also motivates my inclusion of the law enforcement expenditures dependent variable—87 percent of law enforcement expenditures come from local governments (Urban Institute 2021a). My law enforcement expenditures variable includes all expenditures for corrections, judicial expenditures, and expenditures for police protection, although I also include a specification in which each of these subcategories and all nonjudicial expenditures are dependent variables in separate regressions (Supplemental Table A.5). Including criminal justice spending as an outcome also allows for a more thorough exploration of the ways Realignment's design shaped cities’ responses to reduced incarceration; spending may increase in neither, either, or both policy areas.

To create a proportionate measure of expenditures, I calculate cities’ total expenditures per capita, social policy expenditures per capita, and criminal justice expenditures per capita. All cities reporting zero (indicating no expenditures) or with missing values for total expenditures were dropped. Any cities that reported total expenditures and reported zeros or missing for the relevant expenditure categories remained coded as zero or missing respectively. Table 2 provides summary statistics for the aggregated spending variables. It shows that California cities generally spend less per capita on average than other cities on social policy and criminal justice in both the pre- and post-Realignment periods. This may be due in part to California's expansion of Medicaid coverage under the Affordable Care Act in 2010 and affirms the need for the exploration of variation within California cities and the inclusion of the expenditure to revenue ratio alternative dependent variables in my empirical analysis (Golberstein et al. 2015).

OPEN IN VIEWER

Table 2.

Summary Statistics for Expenditures Per Capita in California and NonCalifornia Cities.

		Pre-Realignment		Post-Realignment
	Spending Category	Mean	St. Dev.	Mean	St. Dev.
California cities	Total	1.6725	1.2114	1.9894	1.5841
	Criminal justice	0.1774	0.0702	0.1834	0.0658
	Social policy	0.2201	0.0902	0.1803	0.0993
Other cities	Total	2.0003	1.4940	2.3602	1.7111
	Criminal justice	0.1411	0.0723	0.1477	0.0739
	Social policy	0.1561	0.1550	0.1415	0.1488

I use the share of total expenditures allocated to social policy and criminal justice as alternative dependent variables in Supplemental Appendix Table A.6, to ensure that any effects I observe for social policy and criminal justice expenditures are specifically related to Realignment, rather than just an overall increase or decrease in total spending. Finally, because the expenditure data cannot be disaggregated into its source funds, I use ratios of social policy and criminal justice expenditures to total revenues as alternative dependent variables in Supplemental Appendix Table A.7 to ensure my results are not driven by increases in revenues during this period.

Political Ideology

As discussed in the third section, I expect that diffusion of Realignment's priorities is more likely in cities that are more ideologically aligned with the state. At the time of Realignment's passage, the Democratic Party held over 60 percent majorities in California's state Senate and Assembly, and the governor changed from Republican to Democrat in 2011 (Arnold Schwarzenegger to Edmund Brown Jr.). Therefore, I characterize ideological alignment with the state as being more liberal. I include data on political ideology in my analysis, expecting more conservative cities to spend less on social policy to begin with and policymakers in more conservative cities to be less likely to respond to Realignment by increasing their social policy spending (H2). To test this, I use data from Tausanovitch and Warshaw's (2022) measure of city-level ideology, in which scores greater than zero indicate more conservative areas, and scores less than zero indicate more liberal areas (Tausanovitch and Warshaw 2013, 333). Specifically, I use the ideology scores from their 2004–2011 survey to capture cities’ political environment in the period around the passage of Realignment. The final sample contains over 30,000 observations for almost 3,000 cities across 1,100 counties in the United States, with 55 of California's 58 counties represented in the data.

LocalView

The basic premise of this paper is that local government officials were aware of Realignment and the potential impact a reduction in incarceration would have on their cities and took some action in response to this anticipated impact. I use data from the LocalView Public Meetings database, which contains transcripts from hundreds of public meetings from across the United States, to provide a descriptive analysis of whether and how local governments were responding to Realignment. I combined the 2010–2014 LocalView datasets and identified all city council meetings where any one of 10 key words/phrases ¹¹ appeared in the meeting transcript. From this, I created two datasets: First, a dataset containing the number of meetings by place by year, which includes the total number of meetings, the number and share of meetings in which at least two key words/phrases were mentioned, and the number of meetings where each of the 10 key words/phrases was mentioned. Second, I created a dataset containing the details of each meeting where at least two key words were mentioned, isolating the parts of the meetings in which the key words were discussed to conduct a more thorough content analysis. I did this because, for example, of the cities where the string “109” was mentioned in conjunction with another key word, only California cities in the post-Realignment period were actually discussing AB 109.

In total, there were 57 California cities or counties represented in the data, which had a total of 464 meetings where any of the 10 key words/phrases appeared in the transcript. On average, at least two key words were mentioned in 26 percent of the meetings in California per year. All the cities and counties in the LocalView data were represented in my primary dataset, although two cities were excluded from my regression analyses due to population size (under 10,000 people). These areas were also a mix of high-and low-arrest places as defined in my primary dataset. Again, while these 57 cities are a small portion of the over 400 cities in California, the transcripts from these meetings and comparative shares in topics indicate that Realignment was a salient policy change to local officials. The full analysis is presented in the section “LocalView Analysis.”

LocalView Analysis

At the most basic level, Table 3 shows the share of meetings during which the word “realignment” was mentioned, split by places in and outside of California, the pre- and post-Realignment periods, and high- and low-arrest places. As expected, there is an uptick in California places and differences across both high- and low-arrest places—no meetings had the word “realignment” prior to the passage of Realignment (2010 and 2011 meetings), while 45 meetings (1.4 percent of all meetings) mentioned it in some capacity from 2012 to 2014. In total, 0.1 percent (39) of these meetings took place in cities with a high share of arrests in the five years prior to Realignment. Interestingly, cities with low shares of arrestees mentioned Realignment in a higher share of meetings than cities with high shares of arrestees. As for cities outside California, there was also an increase in the number of meetings where the word Realignment appeared in the transcript after the passage of Realignment in both high- and low-arrest places. While most of these mentions had to do with structural or budgetary realignments, some cities were aware of what was happening in California. For example, a council member in Sioux Falls, SD (a low-arrest city) brought up the policy as a point of interest in a 2013 municipal council meeting:

California has enacted [these] numbers simply by setting people free…It has…sent a lot of nonserious offenders to county jails instead of state prisons in a policy called Realignment…They were operating at over 200% capacity of their state penitentiaries, now they’re down to a mere 135%…it also says that crime has gone up in the places where they’re being released.

OPEN IN VIEWER

Table 3.

Frequency of Meetings in Which Realignment Was Mentioned.

Place	Time	Arrest Share	Number of Meetings
Places in California	Pre-Realignment	High	0
		Low	0
	Post-Realignment	High	0.001
		Low	0.009
Places in other states	Pre-Realignment	High	0.003
		Low	0.004
	Post-Realignment	High	0.004
		Low	0.005

Of more importance to this paper is whether there is evidence of policy feedback. That is, do local officials actually cite the design, resource effects, or interpretive effects of Realignment as impacting their policy choices when responding to Realignment? The meeting transcripts in the LocalView data do contain some descriptive evidence in support of feedback effects. For example, in a 2012 municipal council meeting, the mayor of West Covina, CA, said:

…our community will be changed forever by this is state law…we’re releasing all the prisoners, that's the new realignment, and no one realizes the impacts when they say crime's going up…You arrest them, you put them in jail, they [state officials] release them…That's why over the next five years you’re going to see crime skyrocket with Realignment and what the state government is doing to reduce their budget by releasing all the prisoners. They have no one to watch over the ones they’re releasing. They’re supposed to be managed but they don’t have anyone hired to actually supervise them, and they’re putting it all on the burden on local police departments. That's going to crush our Police Department….

This official is clearly pointing to the design of Realignment's component policies as putting undue pressure on localities, with the lack of resources as a key problem in cities’ ability to respond effectively to the reduction in incarceration. However, while the concerns about a reduction in incarceration were echoed in many of the other meeting transcripts from cities in California, not all officials were strictly focused on the implications for law enforcement. Supporting the empirical results of this paper, some officials saw the design of Realignment as specifically creating a need to move away from a primarily enforcement-based response to the policy. For example, in a meeting of the Stockton Planning/Zoning Commission in March of 2012, one official said:

…the reality is if we think we can just arrest our way out of our crime statistics we’re fooling ourselves because four years ago…it was like a revolving door at the jail because we were…overcapacity and now with Realignment that's going to happen even faster, so we can’t take the position that we’re going to keep on bringing as many officers in as we can…[and] arrest [criminals] into submission…that isn’t going to work…we have to have these other community partners there to figure out what programs are working and how do we keep them from entering into this system….

In addition to evidence that the design of the policy and its resource effects directly influenced local officials’ responses to Realignment, the local view transcripts also evidence interpretive effects—that the type of criminal targeted by Realignment impacted officials’ responses to the policy. An official in Susanville, CA, said in a meeting in April of 2014:

…there are some bad people [who do] bad things and they need to be in prison. There are other people are basically good folks and make mistakes and you need to deal with in a different way. Putting them in jail is not always effective; it's expensive and it doesn’t work…punishment doesn’t work, so what we are trying to do with our Realignment is to come up with some programs…[for example] we’re setting out to pay a portion of our probationers to come in and work on high school diplomas and such…More than half of our probationers don’t even have a high school [diploma].

Several other cities mentioned community partnerships they had developed since the passage of Realignment to improve the monitoring and transition of people who were convicted of crimes that qualified for noncarceral sentencing. These partnerships may have been a target for the increased social policy expenditures observed in the main results of this paper. Additionally, Realignment was occasionally brought up in conjunction with meetings around California ordinances relating to housing, especially restrictions on new builds relating to affordability and the number of households that would be served (i.e. multifamily vs. single-family units). This is interesting considering the finding in Supplemental Appendix A that much of the change in social policy spending observed after Realignment was driven by increases in spending on housing and land use policies. In part, the transcripts from LocalView indicate that this may have been as much a function of cities focusing on “where to put” newly nonincarcerated people, indicating the salience of a transition away from incarceration brought on by Realignment. Overall, these data provide support for both mechanisms proposed in the third section; the design of Realignment's component policies, specifically its resource and interpretive effects, led policymakers in California, especially those in high-arrest cities, to interpret Realignment as requiring a response more focused on prevention and reentry support than on enforcement and social control. The qualities of the target population also appear to have played a role in this interpretation.

Difference-in-Differences Results

Figure 1 contains the results for my primary specification, in which cities’ expenditures per capita is the dependent variable and the primary independent variable of interest is the interaction between being in the post period (after 2011), being in California, and a high average arrests per capita from 2006 to 2010. The top (red) coefficient is the result for all expenditures per capita, the middle (blue) is the result for criminal justice expenditures per capita, and the bottom (gray) is the result for social policy expenditures per capita. The results indicate that cities in California with a high share of arrests per capita spent significantly more on social policy and overall than those with lower-arrest shares. Interestingly, these increases in spending were not accompanied by a significant change in spending on criminal justice, corroborating the descriptive evidence in the previous section that policymakers in California cities may have felt that increasing spending on law enforcement would not be a pragmatic response to Realignment.

OPEN IN VIEWER

Figure 1.

The impact of realignment on per capita spending at the city level.

Delving more deeply into this result, Figure 2 contains a plot of the three coefficients in Figure 1, as well as coefficients for three additional spending categories: operational expenditures for infrastructure, capital outlays for infrastructure, and all other spending. Descriptions of the specific expenditures included in these categories are listed in Supplemental Appendix Table A.2. For example, operational expenditures for infrastructure include spending for regular highways, local fire protection, general public buildings, natural resources, sewerage, solid waste management, sea and inland port facilities, water supply, electric power, and gas supply. As the plot shows, the increase in California cities’ total expenditures after Realignment does appear to be driven primarily by an increase in spending on social policies—the coefficients for criminal justice expenditures, infrastructure, and any remaining expenditures are all small and insignificant on their own.

OPEN IN VIEWER

Figure 2.

The impact of realignment on per capita spending at the city level—expanded.

One may be concerned that these results are due to an increase (or decrease) in funds, especially intergovernmental revenue (IGR), which incorporates grant funding for municipalities from federal, state, and other local governments. To ensure the results in Figure 1 are due specifically to the interaction between Realignment and treatment, I rerun the specification shown in equation (1), but with local, state, and federal IGRs as separate dependent variables. The results, listed in Supplemental Appendix Table A.3, show that Realignment, in conjunction with the high-arrest treatment, did not have a significant effect on IGRs, increasing the certainty that the observed increase in social policy expenditures in Figure 1 was due to a change in local policymakers’ prioritization after Realignment based on its anticipated impact.

In alignment with hypothesis 2, political ideology does appear to mitigate the relationship between Realignment and city spending. The results in Table 3 include both the results for the main outcome of interest ( β 5 ), and the coefficient for the additional interaction between being in California, in the post-Realignment, having a high share of arrestees per capita, and political ideology ( β 6 in equation (2)). The results in row two of Table 3 are similar in magnitude and significance to those in Figure 1—places in California with a high share of arrestees per capita spend significantly more overall and on social policy per capita than other places. The first row, which contains the coefficients of interest in this model, shows that when accounting for the level of conservatism in a city, the sign for spending on all three expenditure variables is negative and not statistically different from zero. To better parse this result, I also include the linear combination and its associated p value—the results show the levels of spending (total, criminal justice, and social policy per capita) are comparatively lower for more conservative cities, affirming that the main result on social policy is driven by a significant increase in spending in nonconservative places (Table 4).

OPEN IN VIEWER

Table 4.

The Impact of Realignment on Per Capita Spending With Ideological Interaction.

	Total Expenditures	Criminal Justice	Social Policy
Post × CA × High Arrests × Conservative	−0.183	−0.008	−0.071
	(0.116)	(0.013)	(0.061)
Post × CA × High Arrests	0.186**	0.003	0.088*
	(0.092)	(0.011)	(0.049)
N	30,981	30,981	30,981
R 2	0.901	0.831	0.899
Lin. comb.	0.002	−0.005	0.017
p value	.976	.484	.641

Note. Standard errors are in parentheses. Each cell in the table is from a distinct regression, with standard errors clustered at the city level.

*p < .1, **p < .05, ***p < .01.

The event studies in Figure 3a–c further illuminate my results. While the years pre- and post-Realignment are grouped in increments of two in these figures, I also include one-year event studies in Supplemental Appendix Figure A.2a – A.2c. There is no evidence of a relationship between arrests and spending in California cities prior to Realignment, satisfying the parallel trends assumption for difference-in-differences. Interestingly, while the results in Figure 3a–c align with those in Figure 1, there is heterogeneity in the significance of the results across years, suggesting the effects of Realignment on cities’ expenditures may have been short-lived. For example, the results in Figure 1 showed that Realignment led to an increase in social policy expenditures when interacted with a city's share of arrests. The coefficients shown in Figure 3a are in alignment with these results in sign, but lose their significance when disaggregated by year. However, the results in Figure 3b show significant decreases in criminal justice expenditures zero–one and two–three years after Realignment, while the results in Figure 1 are not statistically different from zero. Taken together, these results provide additional support for the hypothesis that cities in California actively responded to Realignment by decreasing their spending on law enforcement in the short term and increasing their spending for social policies on average (H1).

OPEN IN VIEWER

Figure 3.

Two-year event studies: the impact of realignment on per capita spending at the city level. (a) Total expenditures. (b) Criminal justice expenditures. (c) Social policy expenditures.

As discussed in the third section, I substitute alternative measures for spending and arrests in Supplemental Appendix A to corroborate my results. Supplemental Tables A.6 and A.7 contain the results for my alternative specifications of expenditures: as shares of total expenditures and as a ratio of expenditures to total revenue, respectively. These results provide further evidence that cities in California generally did not respond to Realignment through increased law enforcement; the significant increase in social policy expenditures persists in the specification using expenditure shares as the dependent variable, and there is a significant decrease in criminal justice expenditures in the ratio specification. Supplemental Table A.4 contains the results for each social policy category and suggests that the effects observed in Figure 1 may be driven by increases in spending toward specifically housing and land use policies. Supplemental Table A.5 contains the results for each criminal justice expenditure category and suggests that while there appears to have been a significant reduction in judicial expenditures for high-arrest cities in California, the null results on corrections and policing expenditures drove the overall null result observed in Figure 1. While drawing strong conclusions about these subcategorical results is difficult given the small size of expenditures per capita in certain categories, these results do align with existing research on local policy-making that suggests policymakers focus their energy on areas over which they have most control, including housing and land use policies, and affirm the main result that expenditures did not change significantly in the more enforcement-related categories. Again, taking spending as an indicator of priority, the significant increase in housing and land use policies we observe in Supplemental Table A.4 may indicate this area was prioritized by local policymakers.

Supplemental Tables A.8 and A.9 contain the regression results for my alternative specifications of arrests. The results in Supplemental Table A.8, where I use a continuous measure of average arrests per capita rather than the binary high-arrest variable in my primary specification, have similar results to Figure 1, although the coefficient on social policy expenditures loses its significance. Given the distribution of the arrests variable in Supplemental Figure A.1, this suggests that Realignment did impact cities’ spending behavior, but only in cities that were more “treated,” a finding in alignment with H1. Finally, I empirically test whether the specific targeting of people who committed non–non–non offenses mattered for city policymakers’ responses. Supplemental Table A.9 shows the regression results for four increasingly conservative alternative definitions of qualifying arrests. The only significant coefficient is on the conservatively defined qualifying arrests variable, which corroborates the descriptive evidence in the third section that there were feedback effects due to the construction of the target population.

I also repeat my primary analysis using several subsamples that capture more similar cities on multiple dimensions to ensure that my results are not driven by the wide heterogeneity across U.S. cities and states. First, I limit my sample to only cities in Western states (as defined by the U.S. Census). The results in Supplemental Table A.10 are similar to those in Figure 1 but are not statistically significant. However, in Supplemental Table A.11, I incorporate propensity score weighting into my model based on race, political ideology, and income. The results from these regressions are similar in magnitude and significance to those in Figure 1, indicating that those results are not driven by heterogeneity in my sample of cities. This is further corroborated in Supplemental Table A.12, where I remove cities in states with the strongest preemption laws to create a sample of cities with more similar legal environments. The results in this table are also similar in magnitude and significance to those in Figure 1—spending on social policy remains significantly higher post-Realignment for California cities with a high number of arrests per capita in relation to other places.