Abstract
The Great Recession of late 2007 through 2009 had profound negative economic impacts on the U.S. states, with 49 states experiencing revenue decreases in their 2009 budgets representing more than $67.2 billion USD. Also during this period, states enacted a record number of laws related to immigrants residing in their states. We make use of data from the National Conference of State Legislatures (NCSL) to examine punitive immigration policy enactment from 2005 to 2012 and conduct a state comparative study using cross-sectional time-series analysis to examine the potential ways in which the economic recession and changing demographics in the states have impacted punitive state immigration policy making. We hypothesize that although anti-immigrant anxieties are driven in part by economic insecurity, they are also impacted by the presence of a large or growing proportion of racialized immigrants. We find that increases in state Hispanic populations and state economic stressors associated with the recession have both led to a greater number of enacted punitive state immigration policies. In addition, we find that changes in the non-Hispanic white populations in the states are also impacting the expression of anti-immigrant attitudes in state policy during this period.
Introduction
The Great Recession provides a unique opportunity to examine the impact of economic factors on immigration policy making in the states, and more importantly, to examine the way in which economic and demographic factors may together create the conditions that lead to this specific type of public policy in the U.S. states. The nationwide recession beginning in December 2007 and officially ending in 2009 saw an increase in the national unemployment rate to a peak of 10.0% in October 2009 (Bureau of Labor Statistics, 2012). This recession was the deepest seen in the United States since the Great Depression, and its impact on the states was particularly hard. In total, 49 of the 50 states experienced general fund tax revenue decreases from fiscal year (FY) 2008–2009, representing more than $67.2 billion. State budgets contracted again in FY 2010 with 34 states experiencing general fund tax revenue losses, representing another $22 billion (National Association of State Budget Officers [NASBO] 2012). During this same time period, states experienced continued profound demographic change, with the continued increase in the Hispanic and Asian populations that began in the 1970s. Together, the Hispanic and Asian population grew more than 43% from 2000 to 2010, making up 23.2% of the total U.S. population in 2010 (Humes et al. 2011). Also during these years, state legislatures continued a trend begun before the recession by enacting legislation related to immigrants. In 2012, states enacted a total of 156 such laws (National Conference of State Legislatures [NCSL] 2013).
In terms of immigration policy response to the recession, scholars of international immigration have observed that the recession prompted receiving nations around the world to adopt policies to limit immigrant inflow, encourage immigrants to leave, and protect their native-born workforces (Fix et al. 2009). Others have noted changes in U.S. federal policy during the recession at the bureaucratic level, resulting in increased removal of unauthorized immigrants (Freeman and Tendler 2011). Although at least six previous published studies have empirically examined predictors of state-level immigration policy making across the 50 states in the 2000s (Boushey and Leudtke 2011; Chavez and Provine 2009; Marquez and Schraufnagel 2013; Monogan 2013; Nicholson-Crotty and Nicholson-Crotty 2011; Zingher 2014), none has examined the impact of the recession. Thus, to date, there has not yet been a scholarly empirical examination of the potential impact of the recession on punitive immigration policy making in the U.S. states. This is the gap this study is intended to address.
We contribute to the empirical literature on state immigration policy making by testing the specific impacts of the economic recession on the states, by using a time-series analysis and explicitly accounting for time, and by testing the impact of the presence of racialized populations in the states. We contribute to the field theoretically as well by combining disparate threads from the study of immigration policy to generate a theory of anti-immigrant anxieties in public policy to examine the ways in which the economic downturn and the changing demographics of the nation interact to impact state immigration policy making.
This article first examines the contemporary context of state immigration policy making. Next, we review the relevant previous empirical research and theoretical considerations. After describing the data and methods used in the study, we present the analyses and marginal effects of key predictors. We find that the presence of larger Hispanic populations along with the economic stressors in states brought on by the Great Recession increase the likelihood of states enacting punitive state immigration policies. In addition, we find that changes in state non-Hispanic white populations are also impacting state immigration policy making. In the concluding section, we discuss the implications of our findings for future research.
Literature Review
Although federal immigration reform has failed to materialize despite symbolic efforts to animate this discussion from the past two presidential administrations, there has been major action in this policy area among the nation’s states; some positive, though a great deal of it negative. For example, in 2011, Alabama enacted what many believe to be the most severe anti-immigrant state law on the books. Among other things, Alabama’s HB (House Bill) 56 requires police officers in the state to ascertain the immigration status of people stopped, detained, or arrested, and prohibits undocumented immigrants from receiving any public benefits at the state or local level, bars undocumented immigrant students from attending public institutions of higher education, and requires public elementary and secondary school officials to ascertain whether students are undocumented. 1 The Alabama law is more extreme than the hotly debated Arizona SB (Senate Bill) 1070, which has generated international attention. The laws enacted in these two states are symbolic of a larger trend across the country, as state immigration policy activity increased from just 300 proposed bills and 39 enacted laws in 2005 to more than 900 proposed bills and 156 enacted laws in 2012 (NCSL 2013). Policy making on immigration and immigrants in the U.S. states is not new, although the form it takes has changed based on the federal context. The state immigration policy making of the 2000s may simply be an extension of previous contemporary waves of state immigration policy making starting in the 1980s and 1990s (Newton and Adams 2009).
Federal legislation of the 1990s set the stage for the current wave of state immigration laws. PRWORA (Personal Responsibility and Work Opportunity Reconciliation Act) of 1996 provided states flexibility to exclude legal immigrants with less than five years in the country from Medicaid coverage (Agrawal 2008). Illegal Immigration Reform and Immigrant Responsibility Act (IIRIRA) of 1996 provided the option for state and local police to participate in enforcement of federal civil immigration law. States and localities began to take up this opportunity in the wake of 9/11 (Varsanyi 2008).
As a result of this policy boom, scholars have devoted a great deal of attention to the factors contributing to the passage of state immigration policy. The passage of state immigration policy making is attributed to state ideology (Boushey and Leudtke 2011; Chavez and Provine 2009; Monogan 2013; Zingher 2014), demographics (Boushey and Leudtke 2011; Marquez and Schraufnagel 2013; Zingher 2014), and state wealth (Boushey and Leudtke 2011; Marquez and Schraufnagel 2013). In addition, Nicholson-Crotty and Nicholson-Crotty (2011) find support for the role of special interests, state legislative professionalism (see also Boushey and Leudtke 2011), unionization levels (see also Marquez and Schraufnagel 2013), and the presence of undocumented populations (Nicholson-Crotty and Nicholson-Crotty 2011) in influencing state immigration policy making. Although these studies address immigration policy making writ large, we are primarily interested in the factors underlying the passage of anti-immigrant legislation; that is, immigration legislation that is either restrictionist or punitive in nature.
To this point, we locate several cross-sectional studies using data from the 1990s that find evidence that both economic factors and racial/ethnic factors predict public preferences for restrictionist immigration policy. In their examination using the 1992 and 1994 American National Elections Studies (ANES) data, Citrin et al. (1997) find that economic threat stimulated cultural anxiety in producing restrictionist immigration policy preferences. Similarly, Burns and Gimpel (2000) find that both negative stereotypes against minorities and perceptions of a negative economy were predictive of restrictionist immigration policy preferences among white respondents to the ANES 1992 and 1996. Hopkins (2010), using public opinion polls from 1992 to 2009, finds that restrictionist policy preferences are stimulated by local increases in the foreign-born population and national salience of the immigration issues. In other words, localities become “politicized” when foreign-born populations increase.
In addition, research using data from our time period of study suggests potential causal mechanisms linking economic concerns and changing demographics with attitudes toward immigrants, immigration policy preferences, and ultimately state policy making. Using a 50-state survey from 2005, Wallace and Figueroa (2012) find that recent state-level increases in the foreign-born population, along with low state economic growth and low minimum wage are associated with increased levels of perceived immigrant job threat at the state level. Using surveys from 2007 to 2008, Perez (2010) finds that negative attitudes toward Latino immigrants are associated with restrictive policy preferences toward both legal and undocumented immigrants. This is an important finding, as survey research has indicated that although the public’s attitudes toward immigration policy are somewhat divided, a large segment of the public believes that immigrants represent a burden to the country (Kohut et al. 2011).
Using a survey experiment conducted in 2010, Hartman, Newman, and Bell (2014) find that racial antipathy toward Hispanics plays a role in anti-immigrant sentiment and in support for anti-immigrant policies. Furthermore, they find support for the idea that anti-immigrant sentiment and anti-immigrant policy preferences act in part as “coded” expressions of anti-Hispanic prejudice. Although the forgoing analyses are instructive of the underlying factors associated with anti-immigrant public opinion, we are interested in the next step of the process: the transformation of anti-immigrant opinions into policy making and thus law.
Theoretical Framework: The Combined Influence of Economic and Population Demographics on Immigration Policy Enactment
Protectionist sentiment and resulting restrictionist/punitive state and local policies during times of economic hardship are not new. History provides examples of how state and local political institutions have effectively translated exclusionary public sentiment into public policy. The early twentieth century saw widespread state legislation barring legal noncitizens from entering professions licensed by the states (Fields 1933). In 1914, an Arizona initiative, ultimately overturned by the Supreme Court, barred private employers from employing noncitizens as more than 20% of their workforce (Haralambie 1977). During the Great Depression, at least seven states passed antimigrant legislation that prevented indigent migrant workers (including U.S. citizens) from entering their states to find work (Langer 2011). The Great Depression also led to repatriation of almost half a million Mexicans and Mexican Americans, ostensibly to open up employment opportunities for American citizens and reduce relief rolls (Hoffman 1974). The time period under examination in our study, 2005 to 2012, provides an opportunity to examine the specific role of economic factors on immigration policy in that they include both years of great economic expansion and the nation’s greatest economic contraction in contemporary history. Surely if economic threat is predictive in any way of restrictionist immigration policy, we expect to find evidence of that during this unique period of American history.
To buttress our historical anecdotes, we begin from the assumption that states have an active interest in immigration reform. At first glance, it may seem that states lack interest in policies related to immigration, given our federalist structure. However, this assertion is backed up by economic studies which generally find that, although the economic benefits of immigration accrue to the federal government, the costs are borne disproportionately by state and local governments (Espenshade and Belanger 1998; Fix and Passel 1994). Governors attest to such impact in that they have been active in demanding assistance from the federal government to defray costs associated with immigrants/immigration and flexibility to exclude immigrants from shared programs such as Medicaid (Skerry 1995; Suro 1994; Vernez 1992).
Flexibility in excluding immigrants at the state level strongly manifested itself in the 1980s and 1990s producing the official English language policy movement (Preuhs 2005; Schildkraut 2001; Tatalovich 1995). States continued exclusionary policy making targeted at racial/ethnic minorities and impacting immigrants in the form of anti-affirmative action initiatives and exclusionary state welfare policies in the 1990s (Hero and Preuhs 2006; 2007). Exclusionary state policies more explicitly targeting immigrants can be traced to California’s Proposition 187 passed in 1994, “arguably the progenitor of all contemporary grassroots local and state anti-immigration legislation” (Varsanyi 2008, 888).
We approach our analysis from the standpoint that the nationwide recession, beginning in December 2007 and officially ending in 2009, led to substantial financial stress across the American states, similar to other historic periods where we have seen anti-minority and anti-immigrant sentiment rise. All but one of the 50 states experienced general fund tax revenue decreases from FY 2008 to 2009, representing more than $67.2 billion in overall losses. State budgets contracted again in FY 2010 with 34 states experiencing general fund tax revenue losses, representing another $22 billion in losses (NASBO 2012). In short, the recent recession hit state’s budgets hard, leading bureaucrats and elected officials searching for cost saving measures, including restriction of resources to citizens.
In this context, our theory regarding restrictionist/punitive immigration policy making during the Great Recession is straightforward. We argue that during periods of economic contraction such as the Great Recession, 2 anti-immigrant anxieties increase among racial majorities. When anti-immigrant anxieties increase, concern over the distribution of public resources to nonnative populations surges to higher than normal levels. This is consistent with Citrin et al. (1997) who argue that beliefs about impact of immigration (job impact, tax impact, and cultural impact) are the cognitive links that mediate between economic threat and immigration policy preferences. In addition, we see some evidence of these attitudes among the general U.S. public during the period under examination. Pew Research Center reports that compared with 2007 and 2009, public anti-immigrant sentiment declined somewhat when measured in 2012. For example, in 2007, 75% of respondents agreed that the United States should restrict immigration more, compared with 69% in 2012; in 2009, 51% of respondents agreed that immigrants threaten traditional American values, compared with 46% in 2012 (Kohut et al. 2012). Thus, we see that among the general American public, the anxieties underlying anti-immigrant sentiment are present during the period of our study.
Policy makers view heightened anxieties as an opportunity. Due to electoral incentives, policy makers, most notably elected officials, jump at the chance to scapegoat immigrants as a primary reason for the economic contractions. Heightened anti-immigrant anxieties are used to sell immigration policy initiative and/or reforms and are used as the firebrand issue by which politicians increase their electoral margins. Sanchez (1997, 1013) points to opportunistic political elites, who take “full advantage of America’s long-standing fears of immigrants and foreigners when such a strategy can bring success at the polls.” In particular, he references California’s Governor Pete Wilson and President Herbert Hoover as examples. Likewise, in their examination of white respondents in nine statewide California surveys from 2006 to 2010, Newman and Johnson (2012) find that an increase in local Hispanic population and negative perceptions of the economy are associated with an increase in concern over immigration, and further that these concerns lead to lower approval ratings of state government and preferences for changes in state policy toward immigration. The authors speculate that . . . the passage of restrictive immigration policies and the incorporation of such positions on immigration into political platforms may reflect an attempt by office holders to counteract the deleterious effects of unchecked immigration and ethnic change, or at least the public’s perception of this, on their level of popularity and approval. (Newman and Johnson 2012, 431)
Finally, recent research has found that poor economic conditions (such as high unemployment) lead presidents to utilize more negative frames regarding immigration, including placing blame on immigration policies and immigrants for the rise in unemployment (Arthur and Woods 2013).
For clarity, we adopt a definition of anti-immigrant anxieties close to that of Sanchez’s (1997) concept of racialized nativism. In our conceptualization, nativism is an expression of, among other things, fear concerning the diversion of public resources to support undeserving immigrants. 3 We include under the heading of anti-immigrant anxieties concepts such as concerns by majority populations regarding the use of public resources like welfare or school systems by immigrant populations, and economic concerns revolving around job loss, and racial concerns revolving around “other-ness” and the violation of key American values (Dustmann and Preston 2007; Golash-Boza 2009; Sniderman, Brody, and Tetlock 1991).
Our theory of anti-immigrant anxieties in public policy operates at the state policy level. We also assume, like many scholars, that U.S. immigrant populations are racialized, indeed that immigration policy is a racialized issue (Ayers et al. 2009; Golash-Boza 2009, Golash-Boza and Hondagneu-Sotelo 2013; Loveman and Hofstetter 1984). We build on the work of public opinion scholars who find that increases in racialized populations 4 and perceived economic threat elicit public opinion effects, including increased concern among whites about immigration as a public policy issue, increased support for restrictionist immigration laws, and decreased approval of state government (Burns and Gimpel 2000; Citrin et al. 1997; Filindra and Pearson-Merkowitz 2013; Hartman, Newman, and Bell 2014; Hopkins 2010; Newman and Johnson 2012; Perez 2010; Wallace and Figueroa 2012). We propose that in the states, these public preferences lead opportunistic political parties and political elites to propose state legislation, and further that in receptive partisan/ideological environments of state legislatures, such policies are passed into law. Although we argue that heightened anti-immigrant anxieties are driven in part by economic insecurity, it is not expressed in the absence of a large or growing proportion of racialized immigrants. We hypothesize that although economic insecurity may be a trigger that leads to policy expressions of anti-immigrant anxieties, the presence of a racialized immigrant population is a necessary precondition. Conversely, economic insecurity in the presence of a relatively smaller racialized immigrant population will not necessarily lead to policy expressions of racialized nativism. We argue here that it is the impact of immigration-driven demographic change, driven by immigrants from Latin America and Asia, resulting in racial anxieties that contribute to anti-immigrant anxieties that in turn are expressed in state policy.
Hypotheses
This study is placed in the tradition of comparative state policy scholarship, examining the impacts of the Great Recession on state enactment of punitive immigration policy. Based on previous findings, we expect that political variables matter for enactment of punitive state immigration policy. In addition, we propose that the relationship between economic factors and racial factors leading to punitive immigration policy is more complex than previously acknowledged. We posit that indicators of economic stress in states and the presence/increase of racialized immigrant populations together lead to punitive immigration policy. This study tests a theory of anti-immigrant anxieties in state policy by examining the extent to which economic stressors in states brought on by the Great Recession and the presence/increase of racialized immigrant populations in the states has led to punitive immigration policy.
Based on these expectations, this study proposes the two following hypotheses:
To test these hypotheses, this study consists of a series of three multivariate analyses testing the impact of ideology, partisanship, state economic distress, Hispanic population, Asian population, and non-Hispanic white population indicators. 5 In addition, we test interactive effects between significant demographic factors, and between economic and demographic factors. Finally, we illustrate the estimated magnitude of the significant effects by presenting marginal effects of the significant predictor variables of interest.
Data and Method
In this study, we examine the predictors of state enactment of punitive 6 immigration policy between 2005 and 2012. Our unit of analysis is state-years, and the dependent variable is the number of punitive policies adopted during any given state-year. With 50 states and eight years in our constructed data set, we have a total of 400 state-years available for analysis. Given that the dependent variable in our analysis is a count variable of policies and we employ panel data, we make use of a cross-sectional negative binomial regression with random effects model using xtnbreg in Stata/IC 13.1 7 specifying panel data with xtset—identifying state as the panel variable and year as the time variable. 8 We specify the default standard error, conventionally derived variance estimator for generalized least-squares regression. Stata confirms that the panels are strongly balanced.
We build on and extend the previous empirical research on predictors of state immigration policy in four ways, thus contributing new knowledge to this substantively important policy area. First, we are explicitly modeling the impact of the Great Recession on the U.S. states as a predictor variable. Only three of the six previously published state comparative studies of state immigration policy making include the years of the recession (Marquez and Schraufnagel 2013; Monogan 2013; Zingher 2014), and none of these explicitly examines the impact of the recession. Second, we utilize a time-series analytic model that explicitly models time, as recommended by state policy scholars to address shortcomings in the larger body of comparative state policy empirical research generally (Blomquist 2007). Modeling time allows a more dynamic account of the factors impacting the policy outcome of interest—particularly important given the dramatic and relatively short-term fiscal effects of the Great Recession on state economies, variation that would be lost in a time-static modeling approach. Most previous studies on the predictors of state immigration policies have employed cross-sectional time-static analytic modeling, collapsing both the predictor variables and the outcome policies of interest over time (Boushey and Leudtke 2011; Chavez & Provine 2009; Marquez and Schraufnagel 2013; Monogan 2013).
Third, we make use of alternative specifications of demographic changes to clarify which groups and which changes are empirically contributing to immigration policy enactment in the U.S. states. The six previously published studies of predictors of state-level immigration policy making across the 50 states in the 2000s noted above have variously tested measures of the Latino population (Chavez and Provine 2009; Marquez and Schraufnagel 2013; Zingher 2014), the foreign-born population (Boushey and Leudtke 2011; Chavez and Provine 2009; Monogan 2013; Zingher 2014), and the undocumented population (Nicholson-Crotty and Nicholson-Crotty 2011). Given the high correlation among these overlapping but distinct demographic measures in our data set, 9 we find it appropriate to choose among these measures based on best model fit.
Finally, we include measures for white population as demographers have observed that rapid immigrant-driven demographic change is occurring at a time of slow growing and in some places decreasing white populations (Fry 2014). White population change viewed in light of the immigrant threat literature reviewed above leads us to expect that the majority white population may experience even greater anti-immigrant anxiety in places where the white population is decreasing, thus resulting in more anti-immigrant policies. We interact Hispanic population and white population measures to determine whether there is a conditional relationship between the two. Because Hispanics are increasing in population in almost every state-year in our analysis, 10 an interaction term will test if the effects of an increasing Hispanic population on state policy making are conditional upon changes in the white population.
Dependent Variable
We model the dependent variable as the count of total punitive immigration policies adopted by each state in each year from 2005 to 2012. We chose these years for their coverage of years of the Great Recession; 2005 to 2012 provides for a balanced data set with three years prior to the national recession (2005–2007), the almost two years during the national recession (2008–2009), and three years following the recession (2010–2012). 11 Data for the dependent variable were obtained from the NCSL Immigrant Policy Project database. We coded each piece of legislation in the NCSL data set as to direction—neutral, beneficial to immigrants, or punitive to immigrants—and use those policies coded as punitive to immigrants for this study. The dependent variable is limited to enacted policies, so excludes bills introduced and not passed and policy action taken independently by the executive or judicial branch. Although those other forms of policy may be important, we focus in this study on enacted legislative policy. Each of the 50 states in our analyses passed at least one punitive policy during the period included in this study.
As our interest is in discrete policy enacted rather than laws, we split omnibus legislation into multiple policies based on unique subtopic/direction combinations. By “omnibus legislation,” we mean large, multipart laws aimed at dealing with different aspects of the lives of immigrants, prime examples include Arizona’s SB 1070 passed in 2010 and Alabama’s HB 56 passed in 2011. We chose to count such omnibus law as multiple policies and to operationalize our dependent variable as policies rather than laws for two reasons. First, the separate policies contained in omnibus legislation concern different subtopics under immigration such as housing, employment, law enforcement, education, identification, and so on. In some cases, these larger pieces of legislation contain subtopic clauses of different direction (punitive, beneficial, neutral). Thus, from a practical coding perspective, counting them as separate policies is the only way to accurately capture direction. Second, operationalizing omnibus legislation in this way provides a weighting of sorts on the dimension of intensity that is otherwise absent from our coding scheme. Some concern exists among immigration scholars working with NCSL data about intensity of legislation—for example, is a single law passed in Arkansas in 2011 concerning health benefits (HB1428) equivalent to the single Alabama omnibus law also passed in 2011 containing many separate clauses? 12 Both are single laws, yet the greater intensity of intention to more severely restrict the lives of immigrants in Alabama seems clear. Counting single omnibus legislation such as HB 56 according to its separate “policies” allows us to address this concern in coding. See Appendix A for more details on our coding scheme.
Figure 1 illustrates the trend in our dependent variable, enacted restrictive/punitive state immigration policies from 2005 to 2012 in terms of total punitive policies enacted and mean policies enacted. We see that the peak for punitive immigration policy during this time period was 2011 when we count 131 such policies. In examining the data set, this peak appears driven by large omnibus legislation with many punitive provisions enacted in Utah, Indiana, Virginia, and Alabama in 2011.

Total and mean restrictive/punitive policies by year.
Predictor Variables
We chose predictor variables by considering those used in the political science comparative state policy literature broadly, as well as the theoretical and empirical literature on determinants of state immigration policy when available. We include predictor variables to represent state politics/ideology, economic conditions, and state demographic makeup. Summary descriptive statistics for all predictor variables are included in Appendix B, and sources and coding are detailed in Appendix C. The economic predictor and all racial/ethnic population variables are included in the models as both absolute measures and as a measure of change.
Economic indicators are of particular importance to our examination given our desire to test the potential impact of the Great Recession on our outcome of interest. Although the recession was a national phenomenon, states experienced the recession differently with some actually experiencing economic expansion while most experienced some level of economic contraction (Cushing 2011). We make use of a composite measure of state economic health produced by the Federal Reserve Bank of Philadelphia. The State Economic Coincident Index combines four state-level indicators to summarize economic conditions in a single statistic. Figure 2 illustrates the annual change in the state economic coincident index by year for the 400 state-years included in our analysis. We see that while the vast majority of state-years experienced positive annual changes for most of the years in our analysis, we see a definite dip toward negative changes in 2009 (when all changes were below zero).

Percent annual change in state economic coincident index by year.
Findings
The results of the three multivariate negative binomial regression models are presented in Table 1; Model 1 presents results without interaction terms, and Models 2 and 3 include results with interaction terms to determine whether our predictors of interest are interacting to impact the outcome. In the first model without the interaction variable, we see that citizen political ideology, state economic coincident index change, and Hispanic population change since 2000 are all significantly predictive of punitive state immigration policy. Notably, neither of the two Asian population variables included in Model 1, percent Asian population and Asian population change since 2000, are significantly predictive of total punitive immigration policies; thus, they are not included in subsequent models. 13
Multivariate Analyses of Total Punitive Immigration Policies Enacted in State-Years, 2005–2012.
Note. Standard errors in parentheses. AIC = Akaike information criterion.
p < .06. *p < .05. **p < .01. ***p < .001.
In Model 2, we introduce white population variables, both percent of population and change since 2000, as well as an interaction term between Hispanic and white population change since 2000. We see that although both state economic coincident index change and Hispanic population change since 2000 remain significantly associated with the number of punitive immigration policies enacted in state-years in the same direction as they had in Model 1, with the introduction of the new variables we see that the coefficients of both state economic coincident index change and Hispanic population change decrease. Both white population change since 2000 and the interaction term are significantly associated with total punitive immigration policies enacted at the state level.
In Model 3, we introduce a second interaction term, interacting state coincident index change and Hispanic population change, two variables that are significantly predictive of the outcome in Model 2. Contrary to our expectations, we see that the new interaction term is not significantly predictive of total punitive immigration policies, indicating that the effects of economic conditions and changing Hispanic populations are operating independent of one another in terms of impact on state immigration policy making and not in a conditional manner as we hypothesized. Because the addition of the final interaction term in Model 3 does not provide a substantive improvement over Model 2, and Model 2 has slightly better fit indicators, we conclude that Model 2 is the best fit for the data.
Because direct interpretation of results of negative binomial models can be problematic, we integrate the results of postestimation with model results to provide insight into the effect size of the significant predictors included in Model 2. We calculate the marginal effects using an observed-value approach (Hanmer and Kalkan 2013), that is, setting the covariates to their observed values when predicting marginal effects. Thus, all marginal predicted values reported below are obtained holding all covariates at their observed values.
We interpret the marginal effects of each of the significant predictor variables in our best-fitting model, Model 2, assuming all other predictors held constant. In Model 2 then, as the state economic coincident index increases by one unit (indicating a stronger state economy than the previous year), the log of the count of punitive immigration policies in state-years decreases by 3.493. In terms of substantive marginal effects, as the annual change in state economic coincident index moves from its lowest value in the data set (−0.1478 observed in Nevada in 2009) to its highest value (0.1167 observed in North Dakota in 2012), we see a marginal decrease of almost one punitive policy enacted 14 (see Figure 3).

Predicted number of restrictive/punitive policies by percent change in annual state economic coincident index.
As the change in the percent Hispanic population in the state increases by one unit (100% increase), the log of the count of punitive immigration policies in state-years increases by 0.796. In terms of substantive marginal effects, as the change in the state Hispanic population since 2000 moves from its lowest value in the data set (−0.174 observed in West Virginia in 2005) to its highest value (1.627 observed in South Carolina in 2012), we see an increase of more than one punitive immigration policy enacted 15 (see Figure 4).

Predicted number of restrictive/punitive immigration policies by Hispanic population change since 2000.
As the change in the percent white population in the state increases by one unit (100% increase), the log of the count of punitive immigration policies in state-years increases by 6.976. In terms of substantive marginal effects, as the change in the white population since 2000 moves from its lowest value in the data set (−0.335 observed in New Mexico in 2005) to its highest value (0.143 observed in Utah in 2012), we see an increase of more than three punitive immigration policies enacted 16 (see Figure 5).

Predicted number of restrictive/punitive immigration policies by white population change since 2000.
Finally, as the interaction between the change in Hispanic population and change in white population since 2000 increases by one unit, the log of the count of punitive immigration policies in state-years decreases by 6.169. Because all but two state-years in the data set experienced positive Hispanic population change during the period under examination, 17 a negative value on the interaction term is indicative of state-years in which the Hispanic population is increasing and the white population is decreasing. The interaction term would be positive only in those state-years in which either both white and Hispanic population are increasing or both are decreasing. 18 Thus, we interpret the negative and strong coefficient on the interaction term to indicate that states in which the Hispanic population is increasing and the white population is decreasing are significantly more likely to enact punitive immigration policy, whereas states in which both the Hispanic and white populations are increasing are less likely to enact punitive immigration policy. This indicates that in addition to the significant direct impact Hispanic population growth is exerting, it is also exerting a conditional impact as the Hispanic population increases and the white population decreases. This is, in fact, what we observe in interpreting the substantive marginal effects of the interaction term—as the change in the Hispanic and white population change interaction term moves from its lowest value in the data set (−0.108 observed in Texas in 2012) to its highest value (0.194 observed in South Carolina in 2012), we see a decrease of almost two punitive immigration policies enacted. 19
Conclusion
The Great Recession of late 2007 through 2009 had a profound economic impact on U.S. states, substantially reducing general fund tax revenue that state policy makers had available in FY 2009 and FY 2010. In addition to dealing with the adverse impact of the economic downturn, during the years of the Great Recession, state legislatures were also busy enacting a record number of state laws addressing immigrants and immigration. We make an important contribution to the extant literature by directly testing the impact of the recession on immigration policy enactment across the U.S. states. We found a clear relationship here, as the financial stress to state economies brought about by the Great Recession is in fact correlated with increased restrictive/punitive immigration policy enactment across the American states. This relationship holds even after controlling for political factors, including citizen ideology, that were largely irrelevant predictors of immigration policy enactment in the fully specified model.
Our findings also reveal that changing demographics across the American states have had tremendous implications for immigration policy activity during this period of heightened economic stress. We theorized that increases in both Asian and Hispanic populations would be associated with increased restrictive/punitive policy making, as both are racialized immigrant groups that may potentially trigger anti-immigrant anxieties. We found only partial support for our theory, as increases in Hispanic population yields increased restrictive/punitive policy enactment while Asian population increase has no impact. This is an important finding that we believe speaks to the racialization of immigrants and the cognitive connection that many Americans, as well as policy makers, are making between immigration and the Latino population. The nonfinding for Asian Americans, although not supportive of our hypothesis, is not that surprising given that Asian Americans are perceived more positively relative to Latinos. Although Asian Americans undoubtedly have and continue to face racial discrimination, the “model-minority” stereotype associated with this population has led to the dominant population attributing many positive characteristics to this community (see Wu 2015), including being more quickly to assimilate to the dominant culture than Hispanics (Huntington 2004). The strong substantive effect of the Hispanic population variables in our models leaves little doubt that a more prominent presence of Latinos has led to more punitive immigration policies across the U.S. states.
We contrast the powerful role that Hispanic population has on immigration policy enactment with the failure of the presence of non-English-speaking, foreign-born, and undocumented populations to impact enactment of restrictive/punitive state immigration policies. Although these models are not included in Table 1 in the interest of space, we did test a series of alternative specifications using non-English-speaking, foreign-born, and undocumented populations both as a percent of population and as population change over time. Each alternative specification was tested in turn, including the alternative predictors with the two Hispanic population variables in the model without the Asian population variables. In no case were any of the six alternative predictors significant, suggesting that although there is some disagreement in the literature as to which immigrants may be the target of restrictive/punitive state immigration policy, policy enactment is associated with Hispanic population changes and not changes in non-English-speaking, foreign-born, or undocumented populations.
Our ability to test each of these potential contributing factors against each other provides some clarity to this debate in the literature. Although there remains room for future research in this area, we believe our analysis suggests policy makers (and arguably the larger population of constituents) is not clarifying whether the new residents of their state are immigrants, undocumented, or with documentation. And finally, we would like to caution researchers wishing to do work with this publicly available state-level immigration law data to consider the sensitivity these data may have to coding approaches. We have suggested in this article that it is important to treat punitive and positive laws as separate measures when they serve as outcomes or dependent variables. In our case that means focusing only on the punitive policies contained in laws passed by a state. However, we are of the opinion that capturing the more complete picture of immigration action in the states through an index combining punitive and positive laws in one variable may be the best approach when treating these laws as explanatory variables, as is the case in some of our other work.
Taken as a whole, these findings further suggest that immigration continues to be a racialized policy issue at the state level, as it has been at the federal level since at least the late 1800s (Ngai 2004; Tichenor 2002). Thus, although factors such as economic pressures do play a role in the politics of the legislative enactment of negative immigration policies, the impact of economic pressures must be considered alongside the presence and growth of racialized Hispanic populations and accompanying dynamics of the state non-Hispanic white population. In the end, over the course of the time period of our examination, the racial/ethnic population variables have a greater substantive impact on the outcome of restrictive/punitive immigration policy adoption than any of the other predictors. These findings should be of interest to a wide range of scholars, including those who do work in immigration politics and policy, racial and ethnic politics scholars, and those interested in state politics more broadly.
Footnotes
Appendix A
Appendix B
Appendix C
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) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors received financial support for the research from the Russell Sage Foundation under their Great Recession research program, Grant #92-12-16, 2012, Principal Investigator Gabriel R. Sanchez. The authors also received support from the RWJF Center for Health Policy at UNM.
