Abstract
Sanctuary policies are generally explained as the outcome of a conflict between desires for openness and restrictionist impulses. In this view, economic factors are exogenous forces that interact with ideological commitments. In contrast, we contend that some of these economic factors follow from ideological commitments and are therefore not exogenous forces. The economic factors follow from ideology because governments that choose to enact sanctuary policies also favor higher levels of land-use and labor-market regulation. We show that: (1) declines in the relative size of the non-citizen population pre-date the sanctuary policies and are confined to counties that ultimately adopt the sanctuary policies; and (2) reduced access to housing and labor markets predicts both sanctuary policy adoption and negative changes in the relative size of the non-citizen population. Thus, the outcomes from policy choices of left-of-center governments on land-use and labor-market regulation directly contravene the apparent purpose of the sanctuary policy.
Keywords
Introduction
For more than 40 years, the economic and social effects of undocumented immigrants in the U.S. have been the focus of public debate. If anything, the debate has intensified in recent years. One important source of this increased friction has been President Donald Trump. In a January 25, 2017 executive order, President Trump declared that it was the policy of the executive branch to construct a more extensive wall on the border between the U.S. and Mexico, expedite reviews of apprehended individuals for eligibility to remain in the U.S., and facilitate cooperation between federal, state, and local law enforcement in enforcing federal immigration policies (Trump 2017).
Trump’s order was one in a series of federal efforts, dating back to the 1980s, aimed at coordinating state and local criminal law enforcement and federal immigration law enforcement (Lasch et al. 2018). The legal framework for this coordination is established in the Illegal Immigration Reform and Responsibility Act of 1996. The Act instituted the so-called 287(g) program which deputizes local law enforcement to aid the federal government in enforcing federal immigration law. Activity under the program accelerated in the wake of the September 11 attacks and a 2002 U.S. Department of Justice legal opinion that argued local law enforcement had “inherent authority” to enforce federal immigration laws (Lasch et al. 2018).
One key avenue through which local authorities enforce federal immigration laws is immigration detainers. These detainers, issued by U.S. Immigration and Customs Enforcement (ICE), instruct local law enforcement officials to keep detainees in custody for up to 48 business hours beyond the time they would have otherwise been released (Immigration and Customs Enforcement 2017a). Such additional detention allows ICE (following notification from local authorities) to assume custody of a “removable alien.” However, in 2014, a federal trial court decided that local authorities were not required to comply with ICE detainer requests. 1 As a consequence, the number of local governments that declined federal 287(g) detainer requests surged. In 2013, no U.S. counties enacted policies which restricted cooperation with ICE (i.e., declined detainer requests). In 2014, 144 U.S. counties enacted policies which restricted cooperation with ICE (see Table 1). For the purposes of the following analyses, we consider any local policy to decline these detainer requests a “sanctuary policy.” 2
Sanctuary Status Adoption by Year.
Note. All U.S. counties that adopted a sanctuary policy appear in the table above. The analysis reported below drops counties with a population under 100,000, as well as counties that adopted a sanctuary policy prior to 2011 or after 2015. We identify adoptions using Section III: Table of Jurisdictions that have Enacted Policies Which Restrict Cooperation with ICE in the February 17th ICE Declined Detainers Report. ICE compiles this data using “public announcements, news report statements, and publicly disclosed policies.” This report can be found at: https://www.ice.gov/doclib/ddor/ddor2017_02-11to02-17.pdf
Public officials justify these policies on both pragmatic as well as moral grounds (Lasch et al. 2018). Pragmatic justifications for sanctuary policies focus on concerns that using local police to enforce federal immigration laws will diminish trust and cooperation between police and citizens, especially among citizens who are recent immigrants (Lasch et al. 2018). Moral justifications, by contrast, focus on discrimination. Inquiries into the causes of sanctuary policies generally frame the analysis in moral terms—as a conflict between cosmopolitan desires for openness and restrictionist impulses that seek to preserve cultural homogeneity (e.g., Collingwood and Gonzalez O’Brien 2019; Ramakrishnan and Wong 2010; Walker and Leitner 2011). In this view, economic factors are exogenous forces that may interact with latent ideological commitments to produce sanctuary policies.
By contrast, we contend that some of these economic factors follow from ideological commitments and are therefore not exogenous forces. The economic factors follow from ideology because governments that choose to enact sanctuary policies also favor higher levels of land-use and labor-market regulation. While this regulation is ostensibly aimed at reducing income disparities between rich and poor, decreasing congestion, limiting environmental degradation, or protecting worker rights, it also reduces access to housing and labor markets.
This reduced access has a disproportionate effect on U.S. residents with fewer family connections and lower earning potential. Because undocumented immigrants, and the broader group of non-citizens, generally have fewer U.S.-based family connections and a lower earning potential than citizens, they are more sensitive to local regulations that reduce access to housing and labor markets. The reduced access to housing and labor markets then reduces the undocumented immigrant and non-citizen population share.
Thus, the preferences typical of jurisdictions with left-of-center governments produce the pairing of sanctuary policies and declining relative non-citizen population shares. Stated alternatively, the outcomes from policy choices of left-of-center governments on land-use and labor-market regulation directly contravene the apparent purpose of the sanctuary policy (i.e., offering refuge to non-citizens, including undocumented immigrants). Because both the regulations that are reducing the non-citizen population share and the sanctuary policies are the result of conscious decisions, officials face a tradeoff. The character of the tradeoff differs however based on what we assume officials understand regarding the effects of land-use and labor-market regulation.
We support this theory by showing that (1) declines in the relative size of the non-citizen population pre-date the sanctuary policies and are confined to counties that ultimately adopt the sanctuary policies; and (2) reduced access to housing and labor markets predicts both sanctuary policy adoption and negative changes in the relative size of the non-citizen population. We use county-level data from the largest counties in the U.S. (i.e., population of 100,000 or greater) and measure access to housing and labor markets using median rent, unemployment rate, and homeownership rate. High regulation will generally raise rents and unemployment, while lowering the homeownership rate.
We then show that the wage for unskilled workers (i.e., the dropout wage), rents, unemployment, and homeownership predict both sanctuary policy adoption and changes in the percentage of non-citizens. Moreover, these results persist even after adding controls for ideological and demographic factors employed in the literature to explain sanctuary policy adoption (e.g., Hedrick 2011; Ramakrishnan and Wong 2010; Walker and Leitner 2011). Because some regulatory policy is enforced at the state level, we also check the results using state-level data.
Ideally, we would also show that these effects hold not only for non-citizens but also for undocumented immigrants (a subset of non-citizens). However, county-level data on undocumented immigrant populations is virtually non-existent. Nevertheless, we contend that non-citizens serve as an excellent proxy for undocumented immigrants. We argue below that: (1) non-citizens have similar income and education characteristics to undocumented immigrants and as a consequence will respond in similar ways to rates of unemployment, homeownership, unskilled wage, and rents; (2) undocumented immigrants are a substantial percentage of non-citizens (49%); and (3) the county-level correlation of the non-citizen and undocumented populations is high (r = 0.73). In addition, we note that if the goal of sanctuary policies is motivated by a desire for “openness” (vs. “restrictionism”) or to foster cultural diversity, the change in the percentage of non-citizens is a relevant assessment of the effect of sanctuary policies.
Literature Review
We argue that: (1) sanctuary policy adoptions typically occur in counties that also favor higher levels of land-use and labor-market regulation; (2) higher levels of land-use and labor-market regulation reduce access to housing and labor markets; and (3) the literature on the determinants of sanctuary policies generally fails to recognize the implications of items (1) and (2). As a consequence, the literature on the determinants of sanctuary policies treats a series of economic variables as exogenous to ideology when they are, in fact, endogenous. Following this logic, we first review the literature on the effects of state and local regulations on housing and labor markets. Then, using the insights of the housing and labor market regulation literature, we offer a critique of the literature that examines the determinants of sanctuary policies. Finally, we consider the literature on attitudes regarding immigration and the literature on the determinants of sanctuary policies and show that reconciling the disparate results across these literatures is possible if we recognize that key economic variables in the analysis of the determinants of sanctuary policies follow from ideology.
State and Local Regulatory Policies and Their Effects on Housing and Labor Markets
A long series of analyses of both land-use regulation and labor-market regulation support the view that these regulations: (1) are generally associated with left-of-center political beliefs; and (2) increase housing prices and unemployment (or reduce employment) and thereby reduce access to housing and labor markets. These negative effects of regulation on housing prices and unemployment do not imply, however, that these regulations should be abolished as they may deliver other types of social benefits. For instance, land-use regulation may deliver environmental benefits and labor-market regulation may protect consumers.
As most land-use regulation occurs at the local level, most studies of these regulations analyze either variations across cities or metro areas or variations across municipalities within a metro area (Gyourko and Molloy 2015). For instance, recent analyses of the effect of land-use regulation on housing prices and new construction have examined eastern Massachusetts (Glaeser and Ward 2009; Zabel and Dalton 2011), suburban Maryland (Wrenn and Irwin 2015), California cities (Jackson 2016; Quigley and Raphael 2005), Florida cities (Ihlanfeldt 2007), and all US metro areas (Saks 2008).
In restricting land use, governments employ an almost dizzying array of regulations including minimum lot sizes, minimum quality standards, density restrictions, building height restrictions, urban-service boundaries, and restrictions on rezoning and upzoning. To make sense of the regulatory landscape, Glickfeld and Levine (1992) sort California land-use regulations into 14 classes and two catch-all categories. Similarly, Ihlanfeldt (2007) sorts Florida land-use regulations into 13 categories. Facing this difficult array of regulations, researchers have attempted to capture “regulation” using a series of tactics.
Some researchers focus on a single type of land-use regulation (Glaeser and Ward 2009; Zabel and Dalton 2011) while others attempt a measure of the combined effect of regulations (Ihlanfeldt 2007). One method to assess the combined effect of regulation is to simply sum the total number of regulations in place in a particular municipality (Ihlanfeldt 2007; Jackson 2016; Quigley and Raphael 2005). Another method measures the length of regulation-induced construction delays (Wrenn and Irwin 2015).
The origin of these land-use regulations has also attracted considerable attention. Saiz (2010) shows that topography explains variations in regulation across cities while Glaeser and Ward (2009) report evidence suggesting that historical density explains variations in regulation. Kahn (2011) shows that liberal ideology (measured as the sum of the shares of voters registered with Democratic, Peace and Freedom, and Green Parties) also explains variations in regulation, even after accounting for topography and density. Kahn analyzes data from 317 California cities from 2000 to 2008 and, after controlling for city characteristics, metro-area fixed effects, metro- area-specific time trends, and endogeneity, finds that that a 10 percentage-point increase in the city’s liberal share is associated with a 25 percent decrease in building permits.
In general, analyses of the effect of land-use regulations find they reduce new construction and raise house prices by decreasing the elasticity of housing supply (Gyourko and Molloy 2015) or shifting the housing supply curve to the left (Zabel and Dalton 2011). 3 Zabel and Dalton (2011) explain that these changes in supply may occur because of: (1) direct restrictions on housing supply (e.g., building permit quotas); (2) direct increases in construction costs (e.g., minimum quality standards); or (3) indirect increases in construction costs (e.g., permitting delays).
Because endogeneity may contaminate results (e.g., higher house prices could induce more regulation), the more recent literature employs designs to mitigate such concerns. Measuring the effect of regulation on prices rather than quantities poses additional challenges. If close housing substitutes are available in nearby municipalities, land-use restrictions that reduce housing supply in one municipality may fail to raise house prices as consumers simply move to nearby municipalities they perceive as close substitutes.
Considering the effect of land-use regulations on house prices, Ihlanfeldt (2007) uses cross-sectional data from Florida cities to show that higher levels of land-use regulation increase house prices. After controlling for endogeneity, the number of substitute locations, property characteristics, neighborhood, and jurisdictional variables, he finds that an additional land-use regulation raises house prices by 7.7 percent. Glaeser and Ward (2009) and Zabel and Dalton (2011) examine the effect of changes in minimum lot size regulations using data from eastern Massachusetts. Using cross-sectional data, Glaeser and Ward (2009) find that an increase of one acre in the minimum lot size raises house prices by about 8.5 percent. However, they find that this effect disappears after they control for household demographics and density. This suggests that there is no price effect from regulation when the control location is a good substitute. Building on this analysis, Zabel and Dalton (2011) use panel data and control for fixed effects and substitute locations. They find that an increase from the 10th to the 90th percentile minimum lot size raises house prices by about 13 percent.
Saks (2008) uses an alternative strategy to identify the effect of land-use regulation on house prices. She employs panel data from U.S. metro areas and constructs a regulation index for these metro areas to estimate housing supply elasticity changes in the presence of land-use regulation. She shows that higher values of the regulation index are associated with smaller supply responses to a housing demand shock.
Like land-use regulation, labor-market regulation takes many forms. Unlike land-use regulation, most labor-market regulation in the U.S. is enacted and enforced at the state and federal level. Important areas where labor-market regulation varies across states include: the minimum wage, occupational licensing, and collective bargaining. Nevertheless, many municipalities regulate labor markets. As of January 2020, 51 cities and counties, including Washington, DC had local minimum wage laws. 4 Kleiner and Krueger (2013) report that 29 percent of the U.S. workforce was licensed in 2008, but only 23 percent of the workforce held a state-issued license. In addition, a cursory search suggests that at least some municipalities regulate entry into occupations. For instance, New York City issues licenses for series of occupations. 5
Analyses of the determinants of state minimum wage increases show a strong relationship between ideology and the minimum wage (Ford et al. 2012; Waltman and Pittman 2002; Whitaker et al. 2012). We see similar evidence for the determinants of unionization. Ford et al. (2012) and Whitaker et al. (2012) use state-level panel data to test the relative importance of political ideology, cost of living explanations, and other institutional factors as explanations of changes in minimum wages. Measuring political ideology using votes by congressional representatives in the U.S. House and Senate for each state delegation, Ford et al. (2012) find that political ideology predicts both the probability of a minimum wage increase and the magnitude of the increase. By contrast, competing explanations have small and inconsistent effects. Whitaker et al. (2012) produces similar results using a measure of ideology constructed from citizen survey responses. While no recent studies examine the link between unionization and ideology, nine of the ten most heavily unionized states voted for Clinton in the 2016 presidential election (Alaska was the only exception) and the correlation between Clinton vote share and unionization rate is 0.51 (p = .0002). 6
The literature on the effect of the minimum wage on unemployment and labor markets more generally is enormous. Moreover, the existence and magnitude of the effects of the minimum wage on unemployment remain a subject of controversy among economists. In a review of the recent literature, Neumark (2017) concludes that the disparate results follow from differences in identification strategies. Analyses that employ geographically proximate designs (e.g., Dube et al. 2010) find no effect (or small effects) from minimum wages on employment while analyses that employ other designs (e.g., synthetic controls) find that the minimum wage reduces employment (e.g., Neumark et al. 2014).
The relation between unionization and unemployment is less controversial. Vedder and Galloway (2002) report that the state median unemployment rate over the period from 1964 to 1999 is associated with higher levels of unionization; a 10 percentage-point increase in the unionization rate implies a 0.7 percentage-point increase in the median unemployment rate. Farber (2005) shows that that unionization rates are in turn associated with a series of state-level regulations that govern union security and collective bargaining rights.
Finally, a wide range of evidence drawn from analyses of individual occupations shows that licensing restrictions reduce employment. Cathles et al. (2010) reports that states with a more stringent licensing law for funeral directors (i.e., ready-to-embalm laws) have 17 percent fewer funeral directors (measured on a per-capita basis). Wanchek (2010) and Timmons and Thornton (2010) report similar results for dental hygienists and barbers. Pagliero (2010) shows that increases in bar exam difficulty consistently predict an increase in median entry-level salaries for lawyers. However, he finds only weak evidence that bar exam difficulty reduces the number of lawyers. Thornton and Timmons (2013) show a similar pattern of results for state licensing of massage therapists.
Determinants of Policies Regarding Undocumented Immigrants in the U.S.
Analyses of the causes of sanctuary policies do not consider that higher levels of land-use and labor-market regulation reduce access to housing and labor markets and instead generally focus more narrowly on ideology. In this view, ideology determines openness which in turn causes sanctuary policies. Economic factors may diminish or exacerbate openness by creating friction or conflict. This friction then causes anti-immigrant sentiment (or ideology) and makes sanctuary policies less likely. That is, economic factors are an exogenous force that interacts with latent ideological commitments. For instance, Walker and Leitner (2011) maintain that education and economic security reduce anti-immigration sentiments. These sentiments may in part result from wage competition between natives and immigrants (Ramakrishnan and Wong 2010). Because undocumented immigrants are primarily unskilled, this wage competition most strongly affects unskilled workers. Collingwood and Gonzalez O’Brien (2019) adopt a similar framework in an analysis that does not consider sanctuary policies per se but rather the number of sanctuary-related bills introduced into state legislatures.
Rather than consider the economic factors to be exogenous, we maintain that the key economic factors in this analysis are determined by ideology. Governments that choose to enact sanctuary policies also favor higher levels of land-use and labor-market regulation. While this regulation is ostensibly aimed at reducing income disparities between rich and poor, decreasing congestion, limiting environmental degradation, or protecting worker rights, it also reduces access to housing and labor markets and raises unemployment. Thus, the higher levels of unemployment observed in sanctuary locations are not causing changes in sentiment or ideology, but rather the unemployment is a consequence of ideology.
Consistent with the view that key economic factors in this analysis are determined by ideology, Walker and Leitner (2011) find higher unemployment and lower homeownership rates increase the probability of a sanctuary policy. Similarly, Collingwood and Gonzalez O’Brien (2019) find that higher unemployment reduces the number of bills opposing sanctuary. Because: (1) liberal ideology favors land-use and labor-market regulation as well as sanctuary policies; and (2) such regulation increases house prices and unemployment, we observe a positive association between unemployment and sanctuary policies and a negative association between homeownership and sanctuary policies. If economic factors had an exogenous effect on ideology (as the literature contends), we would expect that higher unemployment and lower homeownership would diminish openness by creating friction or conflict and reduce the probability of a sanctuary policy (or raise the number of bills opposing sanctuary)—the exact opposite of what Walker and Leitner (2011) and Collingwood and Gonzalez O’Brien (2019) observe.
Of course, land-use and labor-market regulations do not alter all economic factors. Hedrick (2011) finds that increases in per-capita income and employment in the service economy increase the likelihood that U.S. cities adopt a sanctuary policy but increases in poverty rates reduce the likelihood that U.S. cities adopt a sanctuary policy. Ramakrishnan and Wong (2010) find agricultural employment share, overcrowded households (percentage of total), and poverty rates had no effect.
While the literature treats economic factors as exogenous forces that act on latent ideological commitments, other (non-economic) factors intend to capture ideological commitments directly. For instance, Ramakrishnan and Wong (2010) argue the proportion of Republicans in a region is “a proxy for political ideology and issue preferences at the local level” while Latino share of the population is a measure of the “potential electoral strength of Latinos” to enact more liberal immigration policy. Walker and Leitner (2011), by contrast, contend that concentrations of immigrants and the pace of change in immigrant population increase the salience of local immigration policy.
Evidence that Latino population share or growth in the Latino population share affects sanctuary policies is mixed. In an analysis of the determinants of sanctuary policies in U.S. cities over the period 1990 to 2008, Hedrick (2011) finds no evidence that either the percentage of the population that is foreign-born, the level of recent immigration, or the percentage of the population that is Latino influences the adoption of sanctuary policies. Similarly, Ramakrishnan and Wong (2010) find Latino share of citizens and growth in the Latino population had no effect.
By contrast, Walker and Leitner (2011) and Collingwood and Gonzalez O’Brien (2019) find significant effects from growth of the Latino population. Comparing only municipalities with either a sanctuary policy or a policy intended to sanction or deter undocumented immigrants, Walker and Leitner (2011) find that the rate of growth of foreign-born residents rather than the level of foreign-born residents in the population is associated with a lower probability of a sanctuary policy. In their analysis of bill introductions, Collingwood and Gonzalez O’Brien (2019) find that rising Latino populations reduce the number of pro-sanctuary bills.
For other ideological and cultural factors, Walker and Leitner (2011) find that suburban locations, Republican vote share, and lower levels of educational attainment reduce the probability of a sanctuary policy. Collingwood and Gonzalez O’Brien (2019) find higher Republican vote share reduces the number of pro-sanctuary bill introductions. Finally, Ramakrishnan and Wong (2010) find that higher municipal population, lower Republican majorities, higher protest activity, and location outside the Southern U.S. are associated with a higher probability of a sanctuary policy. Interestingly, agricultural employment share, overcrowded households (percentage of total), and the number of anti-immigrant organizations had no effect.
Reconciling the Determinants of Sanctuary Policies and Immigration Attitudes
The literature on the determinants of sanctuary policies fails to recognize that key economic factors follow from ideology. This failure produces an odd disjuncture between the results from analyses of sanctuary policy determinants and results from analyses of attitudes toward sanctuary policies and immigration.
Across analyses of sanctuary policies and analyses of attitudes toward undocumented immigration, we see support for the view that ideological considerations are important drivers of outcomes. For instance, Casellas and Wallace (2020), Collingwood and Gonzalez O’Brien (2019), Berg (2009), and Chandler and Tsai (2001) show political conservatism/Republican party identification has a negative effect on attitudes regarding sanctuary policies/undocumented immigrants. This result aligns with analyses of the determinants of sanctuary policies which show that Republican vote share is negatively associated with sanctuary policies (Ramakrishnan and Wong 2010; Walker and Leitner 2011).
By contrast, we see lack of alignment on employment status between the analyses of sanctuary policies and analyses of attitudes toward undocumented immigration. Walker and Leitner (2011) find higher unemployment rates increase the probability of a sanctuary policy, but the literature on attitudes regarding undocumented immigration finds no effect for employment status (Berg 2009; Casellas and Wallace 2020).
One way to reconcile the apparent contradiction on unemployment is to consider economic indicators like unemployment as reflections of ideological differences or norms. Restated, ideological differences are reflected in both voting patterns (Republican or Democrat) and differences in labor-market and land-use regulation. The regulation is ostensibly aimed at reducing income disparities between rich and poor, congestion, environmental degradation, or protecting worker rights. However, it also reduces access to housing and labor markets. As the regulation reduces access, unemployment at the county level rises. In this view, an individual’s employment status is irrelevant, and surveys of individuals will show no link between an individual’s employment status and their immigration attitudes. Instead, a higher unemployment rate in a particular county reflects higher levels of labor-market regulation. This view that preferences for regulation cause changes in economic variables at the aggregate level is also consistent with Walker and Leitner’s finding that lower levels of homeownership are associated with pro-immigration policies—higher levels of land-use regulation raise housing prices and reduce homeownership rates.
Thus, our explanation of the determinants of sanctuary policies explicitly recognizes that variations in economic factors across counties may reflect regulatory policy choices fueled by ideological differences (i.e., endogenous economic factors). That is, in addition to economic factors that intend to capture the wage competition noted by Ramakrishnan and Wong (2010) (i.e., exogenous economic factors), the economic factors in our model also intend to capture the consequences of regulatory decisions that are often difficult to measure directly and are based on confidence in market mechanisms to solve social problems.
These regulatory decisions then induce changes in net benefits of locating in a particular county by changing labor-market outcomes and living costs for non-citizens and undocumented immigrants. We capture the net benefits of particular counties and the location decisions that follow from the differences in net benefits using wages, rents, housing access, and unemployment rates.
Therefore, we follow Ramakrishnan and Wong (2010), Hedrick (2011), and Walker and Leitner (2011) and use similar or identical ideological, demographic, and economic indicators to discover the determinants of sanctuary policies but add economic factors that capture the consequences of land-use and labor-market regulations. Finally, we note that larger county governments are more likely to have the capacity to challenge the federal government on immigration issues and public perceptions of the value of more immigration are likely influenced by crowding or congestion in the county. Consequently, we add measures to capture these effects.
Data and Methods
We examine the determinants of sanctuary policies at the county level. The choice of counties as the unit of analysis follows from our definition of sanctuary. Our measure of sanctuary status is a local policy to decline detainers issued by ICE to keep detainees in custody for up to 48 business hours beyond the time they would otherwise have been released. Because the U.S. court and prison system is primarily organized at the state and county level and not at the municipal level, most detentions are handled by county officials. 7 Nevertheless, the literature suggests that the majority of labor-market regulation (but not land-use regulation) is enforced at the state level. We prefer the county-level analysis because it captures the heterogeneity within states on the key independent variables. For instance, the standard deviations of housing rents and the unemployment rates within California (σ CA rents = 0.24 and σ CA unemployment = 0.039) are higher than the standard deviations of housing rents and unemployment rates across all counties in the data set (σ rents = 0.20 and σ unemployment = 0.025). Nevertheless, if we repeat the county-level analyses reported below using state-level data, we find the same basic results. 8
In addition to the question of the appropriate geographic unit, we must also consider the appropriate scope of “sanctuary.” The literature defines sanctuary policies in a series of ways. Wong (2017) defines sanctuary policies as those that decline 287(g) detainer requests. Similarly, Casellas and Wallace (2020) solicit responses to the question of whether “Local police should report undocumented immigrants to federal immigration authorities if they are arrested” (i.e., deny 287(g) detainer requests) to measure attitudes toward sanctuary policies. Gonzalez O’Brien et al. (2019) and Hedrick (2011), by contrast, define sanctuary policies somewhat more broadly as including those that: (1) restrict the ability of local law enforcement to ask suspects about their immigration status; or (2) forbid cooperation with federal immigration authorities (e.g., ICE).
Ramakrishnan and Wong (2010) and Walker and Leitner (2011) use even broader definitions of pro-immigrant policies that include variously: (1) items 1 and 2 from Hedrick (2011); (2) resolutions and mandates that express opposition to immigration raids; (3) instructions to decline 287(g) detainer requests; (4) policies granting local rights to undocumented immigrants (e.g., voting rights and ID cards); and (5) resolutions in support of a path to legalization. Because broad definitions increase the risk that county characteristics may cause some “sanctuary” policies but not others, we follow Wong (2017) and Casellas and Wallace (2020) and define sanctuary policies as policies to decline 287(g) detainer requests.
To determine sanctuary status, we compiled data from Immigration and Customs Declined Detainer Reports (Immigration and Customs Enforcement 2017b). ICE compiles this data using “public announcements, news report statements, and publicly disclosed policies.” The reports show that 171 counties adopted a sanctuary policy between 2008 and 2017. From Table 1, we see that 144 of 171 (84.2%) of the counties that adopted a sanctuary policy did so in 2014. For our analysis, we drop counties with a population of under 100,000 because sanctuary policy adoption is concentrated among larger counties and we lack some wage data for these smaller counties.
We also drop counties that enacted a sanctuary policy before 2011 and after 2015. The two counties that adopted a sanctuary policy in 2008 were dropped from the dataset to ensure the independent variables pre-date the policy adoption variable. In addition, we drop the nine counties that adopted sanctuary policies in 2016 or 2017 to reduce the time lag between the independent variables and sanctuary policy adoption. Nevertheless, our results do not vary based on whether we include these nine counties in the analysis. If we repeat the analysis reported in Tables 6 and 7 with these nine observations, we obtain the same basic results.
Of the counties that adopted a sanctuary policy within the period of analysis, 68 had a county population of under 100,000. Dropping these counties leaves 92 treated counties. The 92 counties enacted their sanctuary policies in a variety of ways, including county sheriff decisions and policies, county resolutions and ordinances, and county jail policies. Of the 567 counties included in the dataset (92 treated and 475 untreated), 16.23 percent adopted a sanctuary policy. Using this data, we coded a dummy variable that took a value of 1 for counties that adopted a sanctuary policy between 2011 and 2015, and 0 otherwise. The geographic distribution of these counties is reported in Table 2.
Geographic Distribution of Sanctuary Counties in the Data.
Note. States are allocated to regions based on Census Bureau classifications at: https://www2.census.gov/geo/pdfs/maps-data/maps/reference/us_regdiv.pdf
Significant at .1. **Significant at .05. ***Significant at .01.
To estimate propensity to adopt a sanctuary policy, we regress (using probit) the dummy variable indicating the presence of a sanctuary policy on ideological, demographic, and economic indicators. These economic indicators may reflect wage competition (i.e., exogenous economic factors) or regulatory policy choices fueled by ideological differences (i.e., endogenous economic factors). Following Ramakrishnan and Wong (2010), we use percentage of the 2008 vote to Barack Obama as “a proxy for political ideology and issue preferences at the local level.” We include Latino share of the population as a measure of the “potential electoral strength of Latinos” to enact more liberal immigration policy and to capture the salience of local immigration policy (Walker and Leitner 2011).
We test a series of industry employment share variables to capture wage competition (Ramakrishnan and Wong 2010). 9 Because unskilled workers may see undocumented workers as a threat, high concentrations of unskilled workers in certain industries may be able to organize to prevent sanctuary policies. In a similar way, minority-owned businesses often occupy market niches (Glazer 1998) and as a result increases in immigrants may represent a competitive threat.
The remaining economic variables included in the analysis reflect regulatory policy choices fueled by ideological differences (i.e., endogenous economic factors). In particular, ideological differences cause differences in labor-market and land-use/housing regulation. This regulation is generally aimed at reducing income disparities between rich and poor, reducing congestion externalities and environmental degradation, and protecting worker rights. However, the regulation imposes additional burdens on employers and developers and deters hiring and construction. We include median rent, unemployment rate, and homeownership rate to capture the effects of this regulation.
Finally, we include population, median family income, the poverty rate, and population density as predictors of sanctuary policies. We include population to capture government capacity. Larger governments and governments in jurisdictions with higher median family income will be more likely to have resources to oppose federal government policy preferences on immigration and offer support services. In a similar way, government capacity is lower when poverty rates are higher. We include population density to capture crowding effects. In areas with high levels of congestion, the desire for more immigration may be weaker.
Because high crime rates may stoke fears regarding undocumented immigrants, we also test whether crime rates (property crime rate, violent crime rate, murder rate) increase the probability of a sanctuary policy. We also tested a series of racial demographic measures (e.g., proportion White, change in Latino population), educational attainment (bachelor’s degree share, associate’s degree share), per-capita welfare expenditures, foreign-born population share, population growth, change in the unemployment rate, median age, percentage of the population over 65 years of age, proportion female, and a series of interaction variables. Insignificant variables were dropped from the analysis because of constraints related to sample size.
Data for these variables were gathered from the American Community Survey (ACS; Census Bureau 2018), the Decennial Census in both 2000 and 2010, the Survey of Business Owners 2007, the USA Today election results in 2008, and the Quarterly Census of Employment and Wages (Bureau of Labor Statistics 2018).
To support the claim that counties with high rents, high unemployment rates, low rates of homeownership, and low unskilled wage rates will be especially unattractive to non-citizens (including undocumented immigrants), we consider changes in the relative size of the non-citizen population. We use ACS data on non-citizen population. The ACS produces annual county-level estimates of the number of non-citizens. This non-citizen count includes undocumented immigrants as well as legal immigrants. Ideally, we would also examine the effect of the rates of unemployment, homeownership, unskilled wage, and rents on changes in the relative size of the undocumented immigrant population. Unfortunately, county-level data on undocumented immigrant populations is virtually non-existent.
In support of using only the relative size of the non-citizen population, we point out that: (1) non-citizens have similar income and education characteristics to undocumented immigrants; (2) undocumented immigrants are a substantial percentage of non-citizens (49%); and (3) the county-level correlation of the non-citizen and undocumented populations is high.
Passel and Cohn (2009) report that for 2008, 47 percent of undocumented immigrants had less than a high-school degree. Current Population Survey data for 2008 shows that about 39.6 percent of non-citizens had less than a high school degree. By contrast, the percentage for U.S. natives was 10.2 percent. Similarly, the median household income in 2008 for undocumented immigrants was $36,000 (Passel and Cohn 2009) while the median incomes for non-citizens and natives were $40,406 and $65,319, respectively (U.S. Census Bureau 2009). Because of the similarity in educational attainment and household income for non-citizens and undocumented immigrants, we expect that they will react in similar ways to unemployment, homeownership, unskilled wage, and rental rates.
In addition to similar educational attainment and household income, undocumented immigrants constitute nearly half (49%) of the non-citizen population (U.S. Department of Homeland Security 2017). 10 ACS data for 2010 shows that the population of non-citizens was 22.0 million while Hoefer et al. (2011) reports a population of 10.8 million undocumented immigrants in 2010 (49% of non-citizens). The undocumented immigrant and non-citizen populations also show similar spatial concentrations. The Migration Policy Institute (2020) has produced estimates of the undocumented immigrant population for a small number of U.S. counties in 2016. Matching their counties with counties in our data yields 168 matches. The correlation between the population proportion that is non-citizens and the population proportion that is undocumented for 2016 is 0.73 (p < .00001).
Finally, we contend that even without these strong links between non-citizens and undocumented immigrants, associations between sanctuary policies and changes in the relative size of non-citizen populations would be noteworthy as sanctuary policies are often justified based on moral appeals to “openness” and fostering cultural diversity.
Figure 1 shows the trend over time in the average county percentage of non-citizens from 2007 to 2017 for both counties that ultimately enact sanctuary policies and those that do not. The percentage of non-citizens in the average sanctuary county is generally higher than the average non-sanctuary county. These data also show that for both sanctuary and non-sanctuary counties, the percentage of non-citizens rises from 2009 to 2010. This may be the result of recalibration from the 2010 decennial census. More importantly, the percentage of the population that is noncitizens falls over time (with the exception of the 2010 bump) in the sanctuary counties and this decrease accelerates after about 2014 when most of the sanctuary counties enact their policies (see Table 1). By contrast, the trend in the non-citizen population percentage in the non-sanctuary counties is relatively stable over time.
Noncitizen proportions for sanctuary and non-sanctuary counties 2007 to 2017.
Results
To test for statistical significance in the observed trends in the percentage of non-citizens for sanctuary and non-sanctuary counties that we observe in Figure 1, we run fixed-effects regressions of the non-citizen population percentage on a yearly time trend for both non-sanctuary (n = 475) and sanctuary (n = 92) counties over the period from 2007 to 2017. Table 3 defines each of the variables employed here and in subsequent analyses, while Table 4 reports the results of the fixed-effects regressions of the non-citizen population percentage on a yearly time trend. In each regression, we cluster the robust standard errors at the county level. From column 1 of Table 4, we see that the percentage of non-citizens in non-sanctuary counties increases each year by 0.007 percentage points (p = .1). By contrast, column 2 of Table 4 shows that the percentage of non-citizens in sanctuary counties decreases each year by 0.063 percentage points (p < .0001).
Variable Names and Definitions.
Fixed-Effects Regressions on Percentage of Non-citizens for Both Sanctuary and Non-sanctuary Counties 2007 to 2017.
Note. Dependent variable: Percentage of Non-citizens in the county population for county i and year t. Robust standard errors clustered at the county level in parentheses.
Significant at .1. **Significant at .05. ***Significant at .01.
Column 3 of Table 4 adds a term that interacts year with a dummy for all years in which a sanctuary policy is in place (most are enacted in 2014) to the column 2 specification. From these estimates, we see that the percentage of non-citizens in sanctuary counties decreases each year by 0.05 percentage points (p < .0001) up until the sanctuary policy is enacted. After the policy is enacted, the percentage of non-citizens in these sanctuary counties falls by an additional 0.11 percentage points annually (p = .018).
Because the trend in the percentage of non-citizens in sanctuary counties falls over time and accelerates after the sanctuary policies take effect, we may infer that: (1) the sanctuary policies do not induce non-citizens to remain in sanctuary locations; (2) other factors are reducing the proportion of non-citizens in sanctuary locations. This suggests that the next step is to examine the determinants of the sanctuary policies with an eye toward factors that may cause outmigration of non-citizens and undocumented immigrants. The factors that predict outmigration of non-citizens should also predict sanctuary policies.
Table 5 reports the means and standard deviations for a series of county-level variables intended to capture the desirability of a location and hence the likelihood of in-migration or out-migration. We also include variables intended to capture government capacity as well as political ideology that may make voters more likely to support sanctuary policies. Column 1 reports the means and standard deviations of the variables for all the counties in the dataset, while columns 2 and 3 report the means and standard deviations of sanctuary counties (i.e., counties that adopted a sanctuary policy) and non-sanctuary counties (i.e., counties that did not adopt), respectively. The fourth column reports the difference in the means between sanctuary and non-sanctuary counties. Below the mean difference, in parenthesis, is the reported t-statistic for the two-sample t-test of the difference in means between sanctuary and non-sanctuary counties.
Means and Standard Deviations for Adopters and Non-adopters.
Note. For columns 1 through 3 standard deviations appear below the mean value in parentheses. The final column reports the difference in means for adopters and non-adopters with the t-statistic for a two-sample t-test.
Significant at .1. **Significant at .05. ***Significant at .01.
Among the variables intended to capture differences in net benefits across counties and therefore drive location decisions, we see large differences that suggest housing and labor-market conditions are more favorable for new entrants in non-sanctuary counties. In sanctuary counties, median rents and the unemployment rate are significantly higher, while the homeownership rate is significantly lower than in non-sanctuary counties. Only wages for dropouts show no statistically significant differences.
Median rents are $152 dollars higher (sanctuary: $817; non-sanctuary: $665, p < .0001)—a 22.8 percent difference—and the unemployment rate is more than a full percentage point higher (sanctuary: 10.4%; non-sanctuary: 9.2%, p = .001) in sanctuary counties. In addition, homeownership rates are about 4.5 percentage points lower in sanctuary locations (sanctuary: 65.0%; non-sanctuary: 69.6%, p < .0001). By contrast, wages for workers with less than a high-school education (dropout wages) were nearly identical in 2010 across sanctuary and non-sanctuary counties (sanctuary: $20,110; non-sanctuary: $20,010).
Of the variables intended to capture ideology or political commitments directly, we see that sanctuary locations have significantly larger Democratic vote share and Latino population proportions. The relative size of the Latino population was nearly twice as large (sanctuary: 19.8%; non-sanctuary: 10.7%, p < .0001), and the vote share to Obama in the 2008 election was 9 percentage-points higher (sanctuary: 57.6%; sanctuary: 48.3%) in sanctuary counties. The remainder of the variables in the table serve as controls in the subsequent analysis to capture interest-group bargaining, relative electoral strength among interest groups, and government capacity. Discussion of these variables appears in Appendix A.
To determine estimated effect sizes from each of these Table 5 variables, we run probit regressions with sanctuary status as the dependent variable. Table 6 reports the results. The parameter estimates report the change in conditional probability of adopting a sanctuary policy when the value of one of the independent variables is changed, holding all other independent variables constant. Following the results displayed in Figure 1, column 1 of Table 6 simply tests the effect of non-citizen proportions in 2008 and changes in non-citizen proportions (2008–2010) on the propensity to adopt a sanctuary policy.
Probit Analysis of County Sanctuary Policy Adoption.
Note. Dependent variable: Adoption of a sanctuary policy, 1 = adopt, 0 = otherwise. Robust standard errors in parentheses.
Significant at .1. **Significant at .05. ***Significant at .01.
Column 2 reports a specification that includes variables that drive location decisions of non-citizens. Column 3 adds controls for the proportion of non-citizens and the homeownership rate to the column 2 specification. Column 4 drops the proportion of non-citizens from the column 3 specification and adds instead controls for ideology (vote share to Obama, proportion Latino) and other factors that aim to capture interest-group bargaining and government capacity (Leisure and Hospitality Employment, Population, and Population Density). 11 Finally, column 5 adds variables for interest-group bargaining and government capacity that show high correlations with variables included in column 4.
The Table 6 column 1 estimates show that, as Figure 1 suggests, sanctuary status is positively related to the proportion of non-citizens in 2008 but negatively related to the change in the proportion of non-citizens 2008–2010. A one percentage-point increase in the proportion of non-citizens in 2008 is associated with a 1.7 percent increase in the probability of adopting a sanctuary policy. A 0.1 percent decrease in the proportion of non-citizens from 2008 to 2010 is associated with a 1.2 percent increase in the probability of adopting a sanctuary policy. We may view the remainder of Table 5 as an exploration of the likely causes of changes in the proportion of non-citizens and therefore sanctuary status adoption.
The Table 6 column 2 estimates examine three factors that may drive changes in non-citizen proportions and sanctuary policy adoption: median rents, unemployment rate, and the wage for unskilled workers (i.e., high-school dropouts). The results show that a $100 increase in median rents is associated with a 5.6 percent increase in the probability of sanctuary policy adoption (p < .0001). Inclusion of controls across columns 3 through 5 has little impact on the magnitude of this estimate and it remains statistically significant (p < .001) across all specifications. 12
Like median rents, the unemployment rate remains statistically significant across specifications (p < .01). From column 2, we see that a one-percentage point increase in the unemployment rate is associated with a 2.5 percent increase in the probability of sanctuary policy adoption. However, inclusion of controls across columns 3 through 5 causes a modest decline in the magnitude of the estimate. From column 5, we see that a one-percentage point increase in the unemployment rate is associated with a 1.6 percent increase in the probability of sanctuary policy adoption (p = .012).
The wage rate for dropouts has a negative and statistically significant association with sanctuary status adoption (column 2), however, the estimate loses statistical significance once we control for the non-citizen proportion and the homeownership rate (column 3). The homeownership rate has a negative and statistically significant effect on the probability of sanctuary policy adoption. The column 3 specification suggests that a one percentage-point decrease in the homeownership rate increases the probability of sanctuary policy adoption by about 0.4 percent (p = .044). Inclusion of additional controls, reported in column 5, increases the magnitude of the estimate to about 0.8 percent (p = .001).
The proportion of the population that identifies as Latino is initially insignificant (see column 4; p = .29). However, adding a control for the proportion of firms with minority owners produces a statistically significant result for Latino population share. From column 5, we see that a one percentage-point increase in the Latino population share is associated with about a 0.5 percent increase in the probability of a sanctuary policy (p = .001). Not surprisingly, the two variables are highly correlated (r = 0.57, p < .0001). Thus, the result suggests that non-Latino minority-controlled firms oppose policies that are believed to increase the percentage of Latinos. The percentage of Latinos, once we control for these non-Latino minority-controlled firms, increases the probability of a sanctuary policy.
Taken together, the results suggest that higher rents and unemployment as well as lower homeownership and population density are associated with sanctuary policies. This is consistent with the view that ideology causes economic outcomes. That is, governments that choose to enact sanctuary policies also favor higher levels of land-use and labor-market regulation and this regulation reduces access to housing and labor markets.
Like Ramakrishnan and Wong (2010), we contend a positive effect from the 2008 vote share to Barack Obama on the probability of a sanctuary policy suggests “political ideology and issue preferences at the local level.” The Obama vote share, one measure of ideology, is positive and significant. From column 4, we see that a one percentage-point increase in the Obama vote share is associated with about a 0.3 percent increase in the probability of a sanctuary policy (p = .014). Adding controls for poverty, income, and the proportion of firms that are minority owned increases the estimate to 0.45 percent. The remaining variables serve as controls in the subsequent analysis to capture interest-group bargaining, relative electoral strength among interest groups, and government capacity. Discussion of the estimates of these control variables appears in Appendix A.
Because the literature suggests that most labor-market regulation is enforced at the state level, we repeat the analysis reported in Table 6 using state-level (rather than county-level) variables. This analysis produces the same basic results.
To support the theory that counties with high rents, high unemployment rates, low rates of homeownership, and low unskilled wage rates will be especially unattractive to non-citizens (including undocumented immigrants), we run fixed-effects regressions on the annual change in the percentage of non-citizens using data from 2007 to 2017 and cluster the standard errors at the county level. The dependent variable in these regressions, change in the percentage of non-citizens, is the percentage of the population that is non-citizens for county i in year t less percentage of the population that is non-citizens for county i in year t−1. The mean change in the percentage of non-citizens over the period is −0.00631 with a standard deviation of 0.268. We use the same independent variables as in Table 6. 13 We report these estimates in Table 7. To ensure against endogeneity, we use a two-year lag on each of the independent variables.
Fixed-Effects Regression Results on Annual Change in the Percentage of Non-Citizens in the County Population 2007 to 2017.
Note. Dependent variable: Annual Change in the Percentage of Non-citizens in the County Population = Percentage of the population that is non-citizens for county i in year t less percentage of the population that is non-citizens for county i in year t − 1. All independent variables are defined as above and include a two-year lag. Robust standard errors clustered at the county level in parentheses.
Significant at .1. **Significant at .05. ***Significant at .01.
If counties with high rents, high unemployment rates, low rates of homeownership, and low unskilled wage rates are especially unattractive to non-citizens, we should see reductions in the percentage of non-citizens as a result of these factors. If ideological commitments that favor regulation of land use and labor markets are associated with decreases in the proportion of non-citizens and sanctuary policies, then higher rents and unemployment rates as well as lower homeownership, and unskilled wage rates should predict an increase in the probability of sanctuary policy adoption and a decrease in the percentage of non-citizens. Restated, we expect the parameter estimates for median rent, unemployment rate, dropout wage, and homeownership to change sign between Tables 6 and 7.
As in Table 6, we begin with the factors that motivate location decisions of non-citizens and then add controls. Column 1 of Table 7 employs the same independent variables as column 2 of Table 6. Column 2 of Table 7 adds a control for the proportion of non-citizens and the homeownership rate to the column 1 specification (like column 3 of Table 6). Column 3 of Table 7 drops the control for non-citizens and adds a control for the proportion of the population that is Latino as well as the controls employed in column 5 of Table 6. (As above, the proportion of non-citizens and the Latino proportion are highly correlated.)
From column 1 of Table 7, we see that a $100 increase in the median rent implies a 0.07 percentage-point reduction in the percentage of non-citizens (p = .012). This estimate becomes statistically insignificant with the addition of a control for the proportion of non-citizens (see column 2). Adding instead a control for the Latino population proportion produces an estimate similar to the column 1 estimate. From column 1, we also see that a one percentage-point increase in the unemployment rate implies a 0.01 percentage-point reduction in the percentage of non-citizens (p < .0001). This estimate changes little across the three specifications reported in Table 6.
In contrast to median rent, the wage for dropouts and the homeownership rate are statistically significant in the column 2 specification, but not the column 3 specification. From column 2 of Table 7, we see that a one percentage-point increase in the homeownership rate implies a 0.018 percentage-point increase in the percentage of non-citizens (p < .0001) while a $1000 increase in the dropout wage implies a 0.007 percentage point increase in the percentage of non-citizens (p = .011). In addition, we see that controls for the proportion of Latinos and the proportion of non-citizens have a significant and negative effect on changes on the percentage of non-citizens.
From column 2, we see that a one percentage-point increase in the proportion of non-citizens (the level) implies a 0.16 percentage-point decrease in the percentage of non-citizens (the change) (p < .0001). From column 3, we see that a one percentage-point increase in the proportion of Latinos (the level) implies a 0.03 percentage-point decrease in the percentage of non-citizens (the change) (p = .001). Thus, even after we control for labor market and housing access as well as wages, the level of non-citizens (and Latinos) has a negative effect on changes in the percentage of non-citizens.
To once again check whether these results hold at the state level, we repeat the analysis reported in Table 7 using state-level (rather than county-level) variables. This analysis produces the same basic results as the county-level analysis.
Conclusion
The sanctuary policy controversy is generally framed as a conflict between cosmopolitan desires for openness and restrictionist impulses that seek to preserve cultural homogeneity (e.g., Ramakrishan and Wong 2010; Walker and Leitner, 2011). For instance, Walker and Leitner (2011) “hypothesize that local communities that value and respect cultural and racial diversity in their jurisdiction and in the national community at large are more likely to reject anti-immigration ordinances and/or favor pro-immigration measures—whereas local communities that value cultural homogeneity are more likely to support anti-immigration ordinances” (p. 158). In contrast to this hypothesis, we present evidence that suggests additional factors related to ideological commitments are at work and these ideological commitments are captured by economic indicators. That is, economic factors reflect regulatory policy choices (or ideology).
These regulatory policy choices generally affect labor and land-use/housing markets. The stated goals of the regulations are typically to reduce income disparities between rich and poor, decrease congestion, limit environmental degradation, or protect worker rights (or simply increase house prices). However, the regulations often impose additional burdens on employers and developers and deter hiring and construction. These added burdens reduce access to labor and housing markets and alter location decisions. Because non-citizens and undocumented immigrants generally have fewer U.S.-based family connections, and lower earning potential than citizens, non-citizens are more sensitive to local regulations that raise living costs and reduce access to housing and labor markets. In essence, economic factors that predict adoption of a sanctuary policy (e.g., median rents and unemployment) also make the county less attractive to non-citizens. To the extent that county voters generally have ideological commitments (or preferences) that favor both: (1) the cultural and racial diversity that non-citizens provide; and (2) relatively high levels of labor and housing market regulation, officials are inevitably forced into a tradeoff.
We support this theory by showing that: (1) declines in the relative size of the non-citizen population pre-date the sanctuary policies and are confined to counties that ultimately adopt the sanctuary policies; and (2) reduced access to housing and labor markets predicts both sanctuary policy adoption and negative changes in the relative size of the non-citizen population. We measure access to housing and labor markets using median rent, unemployment rate, and the homeownership rate. That is, sanctuary policies are associated with higher unemployment, lower homeownership rates, and higher median rents. In addition, positive changes in the relative size of the non-citizen population are associated with lower unemployment, higher homeownership rates and lower median rents. In each case, these effects also generally persist even after controlling for government capacity and other measures of ideology.
The character of the tradeoff between: (1) the cultural and racial diversity that non-citizens provide; and (2) relatively high levels of labor and housing market regulation differs based on what we assume regarding the motives of officials and their understanding of the effects of sanctuary policies and regulation. If we assume that officials do not generally appreciate the effects of their land-use and labor-market regulations on housing and labor-market access and that the officials aim to increase the welfare of non-citizens and undocumented immigrants, better information on the effects should better align policies and goals.
However, it is unlikely that officials do not understand that high levels of land-use and labor-market regulation reduce access to housing and labor markets and decrease the non-citizen population share. Zoning regulations are a significant portion of land-use regulations and it is well known that a significant body of case law assesses the allegation that zoning is a mechanism to exclude the poor (Stern 2020). Some states (e.g., New Jersey) even have a dedicated regulatory process to reduce exclusionary zoning. Nevertheless, it is possible that sanctuary policies are not a reaction to declines in the relative size of the non-citizen population; the regulation and sanctuary policies may simply follow from the same underlying left-of-center political preferences.
These political preferences, in turn, may reflect a desire to assemble a winning electoral coalition. Over time, officials learn that sanctuary policies increase their probability of electoral success based on feedback from elections and interactions with voters. Still, it remains unclear the extent to which the sanctuary policies are a reaction to an explicit recognition by officials that land-use and labor-market regulation is causing a decrease in the non-citizen population share. In this view, officials may seek to allay concerns of some voters that current land-use and labor-market regulations exclude non-citizens, not by reducing the regulations, but rather by enacting sanctuary policies that suggest concern for non-citizens (or undocumented immigrants).
Judging by the metric of changes in non-citizen population share, sanctuary policies have little impact on the welfare of non-citizens; non-citizen population share falls significantly faster in sanctuary counties following adoption of sanctuary policies (compared to the years prior to the policy adoption). In addition, our evidence suggests that the welfare of non-citizens in sanctuary locations is lower than in competing locations and, therefore, reductions in land-use and labor-market regulation should improve welfare for non-citizens; non-citizen welfare is low enough in sanctuary locations to foster migration decisions that decrease the non-citizen population share.
When officials decline to deregulate land use and labor markets and instead adopt sanctuary policies, they may be unaware that sanctuary policies have little value to non-citizens and undocumented immigrants. In this case, recognition that the policies fail to stem the reductions in the non-citizen population share may lead to a reconsideration of land-use and labor-market regulations. However, it is also possible that the combination of high-levels of land-use and labor-market regulation reflects fully informed choices by officials; officials do not wish to alter the level of land-use and labor-market regulation and adopt sanctuary policies regardless of their value to non-citizens or undocumented immigrants. Under such conditions, sanctuary policies have more value to liberals (or beneficiaries of the land-use and labor-market regulations) than the immigrant populations they are ostensibly intended to protect.
Indeed, we may rely on classic work by Schattschneider (1960) to clarify this mechanism. Schattschneider contends that the process of competition between leaderships is the foundation for democratic accountability. The competition causes conflict over public policy questions. However, the scope of conflict is an open question. The politically relevant questions are not foreordained, and the political competitors will offer competing narratives of the appropriate scope of the conflict with an eye toward enhancing their electoral prospects. Schattschneider explains that while there are several ways that competing leaderships (or political parties) may alter the scope of the conflict, one key method to expand the scope is to nationalize the issue. Nationalization allows the leadership to move freely among levels of government (local, state, and national) to find the level that offers the greatest leverage against the competing leadership.
Applied to the case of sanctuary policies, these policies are a means to protect the rents that accrue to insiders from land-use and labor-market regulation. A straightforward defense of current land-use and labor-market regulations is unlikely to build a winning electoral coalition. To protect these rents and build a winning electoral coalition, insiders (i.e., Democratic officials) introduce additional issues into the debate (i.e., sanctuary policies) and thereby expand the scope of the debate. These sanctuary policies nationalize the conflict by expanding the scope of a local conflict over access to housing and labor markets and enhance the electoral prospects of the leadership that favors the status quo on land-use and labor-market regulation.
Future research should focus on clarifying the political forces that cause the association between: (1) sanctuary policies; and (2) high levels of land-use and labor-market regulation. Evidence on the mental constructs that officials employ to assess the effects of land-use and labor-market regulation and sanctuary policies is one piece of this puzzle. In addition, other policies (e.g., increases in the minimum wage or proposals for health care coverage for undocumented immigrants) may serve to broaden the scope of conflict and protect the rents that follow from land-use and labor-market regulation. Consequently, analyses of policies similar to sanctuary policies would allow an additional test of the theory. These policies, ostensibly aimed at supporting groups harmed by the land-use and labor-market regulation, could be evaluated in comparisons across counties and over time for their correspondence with the level of land-use and labor-market regulation.
Footnotes
Appendix A
For completeness, this appendix discusses outcomes on control variables in Tables 5 and 6. These variables test for the effects interest-group bargaining, relative electoral strength of interest groups, and government capacity on sanctuary policy adoptions. The results show that each of these factors influences sanctuary policy adoptions. From Table 5, we see that the proportion of firms that are minority owned are higher in sanctuary locations (sanctuary: 15.2%; non-sanctuary: 12.0%, p = .003) while the proportion of total employment that is in leisure and hospitality is lower (sanctuary: 10.0%; non-sanctuary: 10.9%, p = .01). For the government-capacity variables, we see higher population (sanctuary: 743.79; non-sanctuary: 348.59, p < .0001) and median household income (sanctuary: 72.37; non-sanctuary: 66.96, p = .003) in sanctuary counties but no statistically significant differences in population density and poverty rates.
The probit analysis reported in Table 6 also shows that variables designed to capture possible interest-group bargaining, relative electoral strength among interest groups, and government capacity explain sanctuary policy adoptions. In the column 5 specification, we see that a one percentage-point increase in the proportion of firms that are minority owned is associated with a 0.79 percent decrease in the probability of a sanctuary policy (p < .0001). Like minority firms, workers in the leisure and hospitality industry may oppose sanctuary policies because of the threat of competition from undocumented workers. From column 5, we see that a one percentage-point increase in leisure and hospitality employment share is associated with a 1.7 percent decrease in the probability of a sanctuary policy (p < .0001).
The poverty, family income, and population variables intend to capture government capacity. Governments in areas with higher median family income, lower poverty, and higher population are much more able to deliver the services sought by recent immigrants (or non-citizens). As a consequence, they are more likely to enact sanctuary policies. A one hundred-thousand person increase in population is associated with roughly a 0.04 percent increase in the probability of a sanctuary policy (p = .1).
Unfortunately, interpretation of the influence of poverty and income on sanctuary policies is complicated by high correlations between poverty and income (r = −0.67, p < .0001), income and median rent (r = 0.81, p < .0001) and poverty and median rent (r = 0.50, p < .0001). While inclusion of income and poverty in the analysis has little effect on the median rent estimate, the correlations reverse the sign on the income estimate and inflate the magnitude of the poverty estimate. Nevertheless, we may conclude that the estimated effect of median rent on the probability of a sanctuary policy is robust to inclusion of controls for poverty and income.
Finally, we note that high densities may make cities less attractive to migrants and suggest overcrowding to current residents. The column 4 and 5 estimates suggest that an increase of 100 people per square mile in population density decreases the probability of a sanctuary policy by 0.23 percent to 0.31 percent (p = .003).
Acknowledgements
The authors would like to thank Richard Baker, Trevor O’Grady, Paul Vandegrift, and Daniel Bowen for helpful comments.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
