An Analysis of 287(g) Program Adoption and Support for Sanctuary Policies

Abstract

Much attention has been paid to sanctuary policies and local-level immigration enforcement recently. Still, there is a dearth of scholarship on why specific policies are pursued at the local and state levels. This study examined the theoretical conditions that produce anti-sanctuary policies at the state-level and 287(g) program membership. QCA analyses revealed that demographic and political conditions, alongside increases in violent crime, were the most salient causal conditions for explaining the adoption of anti-sanctuary policies. However, these same conditions were poorly explained adopting the 287(g) program. States that experienced significant increases in their Hispanic population alongside either increases in Republican voting during presidential campaigns or increases in violent crime were conjunctive and sufficient explanations of states adopting anti-sanctuary policies. This provides support for the ethnic and racial threat hypothesis as well as evidence as to the role of politics in formulating local immigration enforcement decisions.

Appendix

Conditions

For the creation of anti- and pro-sanctuary as conditions, Pro classifies bills and statutes that are in support of or directly provide for sanctuaries for immigrants residing in the United States unlawfully. This encapsulates everything from outright statements of legal sanctuary for undocumented immigrants to the prohibition of local law enforcement from working with or serving as federal immigration enforcement authorities. This does not cover, however, officers enforcing a duly authorized warrant by a judge. Anti is the opposite, whereby a specific bill or statute explicitly forbids the legal provision of sanctuary for unlawful immigrants or interference with the administration of immigration law by federal authorities. Based on these codings, several measures were generated. Each of these categorizations is determined by looking at the specific language used within a given statute or executive order. For example, Colorado is coded as “Pro-Sanctuary” in that the statute (§ 24–76.6–102(2)) says the following: “A law enforcement officer shall not arrest or detain an individual on the basis of a civil immigration detainer request.” Conversely, a state like “Florida” is coded as “Anti-Sanctuary” given that is statutes (§ 908.101-§ 908.104) state: “A state entity, law enforcement agency, or local governmental entity may not adopt or have in effect a sanctuary policy” as well as “A law enforcement agency shall use best efforts to support the enforcement of federal immigration law. This subsection applies to an official, representative, agent, or employee of the entity or agency only when he or she is acting within the scope of his or her official duties or within the scope of his or her employment.” Almost all of these bills and statutes are at the state level (see Table 14).

Fuzzy-Coding

More information on conditions included within the models is presented within the paper; however, each of these conditions required a decision about how to fuzzy-code them.

Political Conditions

President denotes to what extent a state shifted toward voting for the Republican Party during this period and by how much (in %). As such, 2% is used as the barometer for a significant shift in voting at the state-level during this period, as the national average was a little over 1%, yet some states had drastic (20% <) shifts in either directions during this period, so 2% ensures the condition will not be too sensitive to this reality. State political shifts also need to be controlled for. Data from the National Conference of State Legislature’s (NCSL, n.d.) State Partisan Composition database provided for a state condition—State Politics—to be coded as (1) Democrat to Republican, (2) Split to Republican, (3) Split to Democrat, (4) Republican to Democrat, and (5) Same. These are fuzzy coded as “0” (No change), “0.5” (S → R OR S → D), and “1” (D → R), respectively.³

Demographic Conditions

In terms of demographic conditions, two are used: (1) the percent change in the foreign-born population of a given state and (2) the percent change of the Hispanic population of a given state. The first represents what share of a state’s population was born outside of the United States, whereas the second merely serves as a proxy for the presence of the fastest growing demographic change in America. These conditions denote the share of the foreign-born and Hispanic populations in a given state in 2019, relative to 2010 (Foreign and Hispanic). For example, if the foreign-born population in state A is 10% in 2010, but is 12% in 2019, there was a 20% increase during this period. These measures allow for the approximation of demographic changes at the state-level while allowing for static and dynamic measurements to be included. The cut-off threshold for inclusion in the Foreign born condition is 5.75% (0.5) given that this is the national-level increase for this period (see ACS, n.d.). The same is true of the Hispanic population, but the metric is 23.8% (e.g., for Hispanic).

Other Conditions

One economic measure is also included. This measure—GDP—constitutes the percent change from 2010 to 2019 in real GDP at the state level (BEA, n.d.). This is a continuous fuzzy-coded condition whereby 1.78% is the threshold value (average amount of change at the national-level during this period, according to the World Bank), with 0% and 5% being the markers for full exclusion and inclusion, respectively.⁴ One index measure of violent crime was also included. The rate of crime (per 100,000 people) is derived for each state for 2010 and 2019 from the FBI’s Uniform Crime Reports data (UCR, n.d.). As such, changes in violent crime from 2010 to 2019 are represented as Violent Crime, with a threshold value for membership in the condition set at − 9.3%, given that this was the average national decline during this period.

Analytic Strategy

This study uses qualitative comparative analysis (QCA). This technique allows for the examination of different conditions’ influence on various outcomes. This technique is not inferential, but instead allows for causal explanations for different phenomena by way of examining the role that differing conditions and combinations of conditions have on specific outcomes. QCA is a robust technique for small-n studies and enables researchers to assess the causal relationship—not effect size—of a given condition on producing a specific outcome using Boolean algebra and minimization. In this study, the central outcome of interest is Sanctuary, in that the presence of the outcome (1) constitutes an anti-sanctuary position at the state level. This also means that “0” constitutes the absence of the outcome, and “0.5” denotes the threshold for membership in neither category (e.g., states with no sanctuary laws at all).

Several conditions are fuzzy-set conditions, in that membership in a given case ranges based on a theoretically relevant scale of inclusion. Instead of a simple dichotomization of a measure, fuzzy-set calibration enables researchers to include more nuance within a given analysis. For example, President represents a fuzzy-coded condition, whereby “0” equals there was more than a 2% decrease in support for the GOP in presidential elections during this period (2012–2020), “0.33” represents that there was a “0 to 2%” decrease in support for the GOP; “0.67” denotes a “0 to 2%” increase in support during this period; and “1” classifies whether a state had a more than 2% increase in support for the GOP during this period. In this manner, “0.33” represents partial inclusion, but cases are still more out than in. However, cases with a score of “0.67” are more in than out in terms of membership within that set. Most measures are coded in this manner. For a robust description of how each measure was fuzzy-coded (e.g., calibrated), please see Tables 13 and 14.

Put simply, the more conditions are introduced, the more difficult it becomes to find consistency across the different conditional combinations (“recipes”) in producing the outcome of interest. However, too few conditions runs the risk of producing weak or meaningless recipes that ignore causally relevant factors. This is also why the foregoing conditions (see above) are derived from the discussion of the literature.

The final steps require the generation of necessity analyses and truth tables. The former denotes which conditions are necessary (but not sufficient) for the presence of the outcome. If conditions are necessary, then they are a superset of the outcome and should not be used in a truth table. Truth tables represent rows of cases whereby specific conditions are present or absent in the production of that outcome. In essence, a truth table depicts the conditional combinations that produce the outcome in question. From there, Boolean minimization procedures are used to denote which expressions can be reduced to produce less complex combined expressions. Consider the following from Ragin and colleagues (2017, p. 32):

If two Boolean expressions differ in only one causal condition yet produce the same outcome, then the causal condition that distinguishes the two expressions can be considered irrelevant and can be removed to create a simpler, combined expression. Essentially this minimization rule allows the investigator to take two Boolean expressions that differ in only one term and produce a combined expression.

Each of these steps are repeated not just for the presence of the outcome (Sanctuary) but also for the absence of the outcome, as the latter can aid in the understanding of the causal mechanisms that produce the outcome in question (~ Sanctuary) (see Schneider & Wagemann, 2010 for an explanation).

Only conditions representing change over time are used. We do not believe this array biases the proceeding analyses for several reasons. First, the 287(g) program only included 29 participants in 2009 (Capps et al., 2011), 32 in January of 2016 (Kandel, 2016), and 95 by the end of 2019 (ICE, 2020). However, there are almost 142 today. This signifies the massive explosion of the program under the Trump administration. Second, the same phenomenon occurs with state and local sanctuary laws too. Thus, the massive policy changes surveyed here occur almost entirely after the demographic, economic, and behavioral changes examined. Nevertheless, some of the change variables—namely, the political ones—arguably better articulate how the magnitude of change—not just that change occurred—caused some of the outcomes we examine.

Table 13 Fuzzy-coding decision

Table 14 State sanctuary policies