Testing the Backlash Argument: Voter Responses to (Pro-)Immigration Reforms

Authors: Alexander Kustov

Journal: Journal of European Public Policy (2023)

DOI: 10.1080/13501763.2022.2061034

Resources

Figures & Tables

Abstract

Do significant pro-immigration reforms—that open legal pathways for labor and family immigration—increase populist voting? Despite the common assumption that such reforms would lead to counter-productive voter backlash informed by the literature on immigrant group threat, the extent to which immigration policy itself influences voters has been unclear. To address this question, this paper estimates the impact of immigration policies on (right-wing) populist voting and immigration attitudes by exploiting the timing of major changes to immigration legislation in a new dataset linking the best available public opinion and policy data across the last forty years in 24 European countries. My analysis shows that, while the absolute levels of immigration policy openness are associated with slightly higher populist voting across countries in a naive cross-sectional analysis, pro-immigration (or anti-immigration) policy changes do not affect populist voting or immigration concerns within countries. This suggests pro-immigration reforms do not backfire due to voter backlash.

Keywords: Immigration policy; populism; public opinion, TSCS data, voting behavior

Introduction

In one of her recent interviews, the former US presidential candidate Hillary Clinton—a pro-immigration politician who had previously remarked about her “open borders dream”—recommended curbing the existing “generous and compassionate” immigration policies which were “inflaming voters and contributed to the election of Donald Trump and Britain’s vote to leave the EU,” according to the Guardian from November 22, 2018. Similarly, a great number of otherwise immigration-friendly politicians and scholars across the political spectrum have repeatedly attributed the rise of populism to voter backlash to immigration and the corresponding political failure to sufficiently restrict it. Despite a lack of solid empirical evidence, these popular arguments against pro-immigration reforms based on particular instances of alleged immigration backlash have been extremely influential among policymakers. After all, since most voters across high-income countries oppose increasing immigration, policymakers often tend to dismiss any possibility of significant pro-immigration reforms as “politically unfeasible.”

In line with these arguments, there has been a large literature documenting the “group threat” that immigrants pose to natives and the related voter response to immigration shocks in terms of the rising salience of the issue, populist voting, or even xenophobia [@Hopkins2010, Kaufmann2019, Dennison2019b, Claassen2021]. The extent to which immigration policy itself directly influences political behavior or causes backlash beyond its effects on demographic change, however, is still unclear. This important omission is unfortunate since, outside of the immigration realm, scholars have widely recognized that government policies do not just reflect but also significantly shape voter preferences [@Ura2014, Bishin2015, Tankard2016].

Do significant pro-immigration reforms that increase legal pathways for labor and family immigration (Note: Despite their complexity, most immigration policies regulate the long-term admission of non-citizens into the country. While governments also have distinct policies concerning irregular and humanitarian migration, as well integration and enforcement, these policies are largely contingent on the allowed number of legal immigrants [@Ruhs2013] so I do not focus on them here in detail due to space limitations.)—especially in the contexts where most people have negative immigration attitudes—cause voter backlash? Can these reforms even be counterproductive to what they are trying to achieve by increasing xenophobia and right-wing populist vote (Note: While populism—usually defined as a type of antagonistic political rhetoric that emphasizes the general will of “the people”—does not have to be anti-immigration in principle, it predominantly is so in practice (especially in its European right-wing variant). I will thus use right-wing populist voting as a proxy for anti-immigration political behavior for the purposes of this paper, which is consistent with the other literature on the topic [@Dennison2019c].)?

To address these questions, I estimate the potential impact of pro-immigration (and anti-immigration) reforms on voting behavior and public attitudes by exploiting the variation in timing of select (labor and family) immigration policy changes over the last four decades across Europe. While there are valid theoretical reasons to expect that voters retaliate against the liberalization of international mobility by voting for populist and anti-immigration parties, it is also possible that many people can accept such liberalization as legitimate or even grow to embrace immigration further.

Overall, I show that the absolute levels of immigration policy openness are associated with slightly higher levels of right-wing populist voting across countries in a naive cross-sectional analysis of panel data. However, I also demonstrate that pro-immigration (or anti-immigration) policy changes do not robustly affect populist voting or immigration concerns within countries when time-invariant and standard time-varying confounders are accounted for in panel fixed effects models. If anything, pro-immigration policies may be related to more positive voter attitudes both within and between countries. Taken together, this evidence suggests that meaningful policy reforms that liberalize legal immigration are unlikely to backfire due to voter backlash.

The Backlash Argument and Immigration

Most generally, the notion of public backlash refers to an adverse reaction to the advancements of (disadvantaged) social groups and causes by a sufficiently large segment of the population that may ultimately be counterproductive to these advancements. The adverse reaction may relate to changes in public attitudes toward the disadvantaged groups, causes, and related policies in question, the intensity of these attitudes, or their behavioral manifestations. While these changes do not have to be political, the public backlash to various advancements in today’s democracies is often channeled electorally where people can and do vote for politicians, parties and programs with an aim to reverse these advancements.

While a variety of scholars across different disciplines employ the concept of backlash, there is little agreement about its precise definition. Nonetheless, the underlying argument is usually about more then just the mere existence of a negative reaction to a certain change by some people [@Bishin2015, Norris2019]. After all, as long as voters have heterogeneous preferences, no change can possibly yield a unanimous public support in a large polity. Any non-trivial and analytically fruitful backlash argument necessarily explicates that a significant adverse reaction against a particular advancement can be counterproductive to the advancement’s goals in the long run. Importantly, when the backlash argument is invoked, the advancement is counterproductive exactly due to the direct adverse popular reaction to it, independent of any actual economic or other social effects of such advancement. Therefore, most backlash arguments can be viewed as empirically testable claims about voter responses to political change of the following kind: “regardless of its merits, if you cause advances too much now, you may eventually get less than otherwise would be the case due to the more active resistance of those who disagree with your cause.”

Throughout history, backlash arguments along these lines have been applied to a variety of disadvantaged groups and causes, including slavery abolition, women suffrage, racial equality, and same-sex marriage. Of course, backlash arguments can be and are often factually mistaken and they can also be strategically employed by the opponents of a certain cause to undermine it [@Keck2009]. One of the reasons for that is, though falsifiable in principle, the “excessiveness” of any advancement (or adverse reaction to it) is context-dependent and largely subjective. (Note: In the case of social status or other zero-sum rivalries, any advancement of the lower-status groups over the status quo can be perceived as excessive by the higher-status groups [@Kustov2019a].) Furthermore, the inherent ambiguity of what constitutes the right time frame makes it difficult to specify appropriate counterfactuals and come to an agreement about whether the backlash has occurred. When it comes to the harsh government restrictions on international labor mobility in today’s high-income democracies, backlash arguments usually try to convey that any significant pro-immigration advancement over the status quo has a potential to encourage more people to vote for populist or anti-immigration politicians in the short term, which can undermine or even reverse this very advancement in the long term.

Conceived this way, there can be both the “weaker” and the “stronger” versions of the immigration backlash argument. According to the weaker (or less extreme) backlash argument, pro-immigration advancements merely increase opportunities for political entrepreneurs to rally against immigration due to rising salience (or relative importance) of the issue among those who oppose these advances, which in turn increases the probability of anti-immigration voting and subsequent immigration restrictions. According to the stronger (or more extreme) backlash argument, pro-immigration advancements do not just increase the probability of rising issue salience and anti-immigration voting but also change the underlying preferences of the electorate by making people more anti-immigration (or “xenophobic”). (Note: Note that this distinction parallels the ongoing debate in political psychology between galvanizing and mobilizing effects of (immigration) threat [@Mader2018, Kustov2019b].)

Voter Backlash to Immigrants versus Immigration Policy

What constitutes pro-immigration advancements that voters lash back against? Most prominently, scholars have conceptualized immigration backlash as an adverse voter response to the rising physical presence of immigrants in terms of ethnic demographic change. (Note: Some political economy scholars may also view immigration backlash as an instance of voter responses to globalization shocks alongside international trade and foreign capital flows [@Naoi2020].) While this rather intuitive idea—often dubbed as “group threat”—has been extensively studied in the social science literature, the existing observational evidence is far from being conclusive [@Pottie-Sherman2017, Kaufmann2018, Claassen2021]. At the same time, (quasi-)experimental studies which (as if) randomly assign demographic composition are difficult and rare.

More generally, there is a growing longitudinal evidence that questions the very possibility of the strong backlash argument as a response to demographic change due to the relative stability of immigration preferences or the small explanatory power of geographic context [@Kustov2019b, Maxwell2019]. Nonetheless, there has been relatively more evidence in support of the weak backlash argument in terms of increasing salience of immigration and populist voting due to rapid demographic change [@Dennison2019b, Kaufmann2019]. (Note: For additional supportive quasi-experimental studies exploiting the 2015 Refugee Crisis, see @Hangartner2019 [@Schaub2020]. It is important to note, however, that even the best quality evidence on popular backlash to substantial refugee shocks and the related policies cannot be easily extrapolated to the regulation of regular immigration flows explored here.) Still, some scholars point out that even these effects are likely significantly moderated by the media environment and parties’ competition for whom immigration shocks merely present one additional opportunity to shift their agenda focus [@Hopkins2010, Mader2018, Dai2022]. (Note: Other scholars have also explored the voter responses to more or less pro-immigration political rhetoric or media discourse as analytically distinct from both immigration demographic and policy changes [@Flores2018a]. While an examination of the potential voter backlash to pro-immigration rhetoric in itself is warranted, it is beyond the scope of this paper.)

Indeed, the focus on the immigration-induced demographic change arguably misses a significant part of the issue since immigrant presence (both actual or anticipated) is evidently neither necessary nor sufficient for voter backlash [@Solodoch2020]. Even if one fully grants the possibility that ceteris paribus rapid immigration flows increase the salience of immigration among voters and the probability of populist voting across contexts, this can only explain a small part of variation in these outcomes. Furthermore, from the perspective of policymakers, the evidence regarding immigrant presence is not very helpful inasmuch as it has multiple complex causes beyond migration policy [@Czaika2013]. The extent to which immigration policy itself influences political behavior beyond its effects on immigration levels and flows, however, is still unclear [@Flores2017, Abou-Chadi2018d, Vranceanu2018, Solodoch2020].

To illustrate this point, consider the full implications when someone argues that immigration backlash contributed to Brexit. Even if one explicitly focuses on the effects of immigration-related demographic change, such a causal argument also necessarily contains an implicit counterfactual claim that if the UK’s immigration policy had been sufficiently more restrictive, then the UK would have fewer immigrants and thus Brexit would not have happened. In other words, to the extent anti-immigration voters negatively react to the increases of immigrant presence, they should also negatively react to any pro-immigration policy change that facilitates these increases within the immediate electoral context. (Note: Of course, voters may fail to react if they underestimate the significance of a certain policy (as it was prominently the case with the 1965 Immigration Act in the United States). Such critique, however, is arguably applicable to any political stimuli in the world of uncertainty and bounded rationality.) To give even a more extreme example, if the UK government had suddenly declared the end of all border enforcement, this could plausibly cause significant backlash among voters in an upcoming election regardless of how many people actually migrated (or wanted to migrate) there simply due to the existence of some categorical opposition to loosing control over immigration in the British public [@Kustov2020, Solodoch2020].

So how do voters respond to immigration reforms? To answer this question, I build on the related literature exploring the possibility of voter backlash to other contentious laws and policies such as same-sex legislation. Overall, this literature recognizes that, in addition to simply reflecting public preferences, government policies and laws can significantly affect voter attitudes and behavior by inducing widespread resistance or legitimation [@Ura2014]. In the case of LGBT issues, while many authors have speculated to the possible backlash to the proliferation of same-sex legislation, most studies have been unable to detect these effects. Furthermore, across various contexts and empirical strategies (from survey experiments and event studies to difference-in-differences approaches), scholars have found that, if anything, legal LGBT advances have increased their public acceptance [@Bishin2015, Flores2016, Abou-Chadi2019a, Aksoy2020]. (Note: Interestingly, unlike in the case of immigration, scholars rarely studied backlash to increasing or more visible LGBT population where backlash is more likely to be conceptualized in terms of policy change.)

Specifying Empirical Expectations

As emphasized above, it is analytically fruitful to consider the hypothesized popular backlash to pro-immigration policies as being independent of their long-term effects as long as immigration is generally disliked by a significant share of voters (due to its perceived negative effects, ethnic prejudice or other causes). Any consistent and non-trivial backlash argument is thus not about the consequences of pro-immigration policy changes per se but rather about the net negative attitudinal or behavioral voter responses to these changes. To that end, here I focus on the more proximate and immediate manifestations of immigration backlash related to the attitudinal and electoral change in the aggregate, rather than the reactions of particular voter subgroups or other possible broad and slow-evolving effects (e.g., related to the increase in ethnic violence, decrease in social cohesion, or more restrictive immigration policy itself in the long term).

Departing from the studies of policy impact on political behavior mentioned above, I have formulated the following empirical expectations with regard to the “backlash” effects of immigration policy reforms as opposed to immigration flows or changes in media discourse explored elsewhere (see Figure fig:backlash). First, and most important, both the stronger and the weaker backlash arguments stipulate that pro-immigration reforms probabilistically increase the short-term share of right-wing populist vote in the electorate (baseline Hypothesis 1).

Second, the stronger backlash argument stipulates that pro-immigration reforms can increase anti-immigration preferences in the electorate within the course of the electoral cycle (Hypothesis 2a). This idea has been present both in the media and academic literature, and it is also in line with the evidence documenting attitudinal change on LGBT issues. Unlike LGBT issues, however, public immigration preferences have been rather stable since the very start of opinion polling across the immigrant-receiving countries [@Kustov2019b]. At the same time, most of the recent immigration reforms have also been relatively minor and it is hard to find an equivalent immigration advancement to the legalization of same-sex marriage. Consequently, an alternative plausible expectation is that, unless the change is truly radical, immigration policy cannot significantly affect immigration preferences which are normally very stable. To the extent significant policy change does occur sometimes, it is also possible that pro-immigration reforms can alter the underlying social norms and legitimize these advancement across the electorate in the short term—countervailing the backlash-induced increases in anti-immigration preferences or even further decreasing them (Hypothesis 2b).

Third, regardless of immigration preferences themselves or their change, however, important political outcomes such as populist voting may still be dependent on the salience (or personal issue importance) of immigration among voters. Unlike the case of preference stability, there is substantial evidence linking contextual factors to changes in the salience of immigration attitudes. Accordingly, the weaker backlash argument stipulates that pro-immigration reforms can at least increase the issue salience of immigration even if does not immediately trigger changes in immigration levels (Hypothesis 3). Unlike the stronger version, this expectation is also more plausible econometrically simply due to the greater volatility of immigration salience [@Dennison2019b].

Although it is usually assumed that the relationship between pro-immigration policy and populism is mediated through either immigration preferences or salience, there can be other possible causal pathways such as related to various ethnic and anti-elitist preferences [@Kaufmann2019]. Consequently, at least for the purposes of this paper, I assume that all sets of hypotheses are independent of each other (i.e., immigration reforms can affect populist vote even when they are not related to immigration attitudes).

Figure. The Immigration Backlash Argument: Causal Pathways and Hypotheses

While my substantive focus here is on the effects of pro-immigration reforms, it is important to consider the flip side of the common backlash argument or what it implies for how voters respond to (almost equally common) anti-immigration reforms. Most prominently, @Kaufmann2019 puts forth a symmetric argument that anti-immigration reforms should conversely appease existing restrictionist sentiments, lower the salience of the issue, and undermine populist support in the electorate. Most other backlash arguments mentioned above do not discuss this issue explicitly, but they seem to similarly assume that voter responses to positive and negative changes are symmetric (in line with Hypotheses 1-3 above).

Nonetheless, one could also argue that voter responses to pro-immigration and anti-immigration reforms might operate under distinct causal mechanisms. Although a comprehensive treatment of the issue is beyond the scope of this paper, one important alternative empirical expectation to consider is that anti-immigration reforms simply should not affect populist voting or immigration attitudes (Hypotheses 1-3c). If most voters already have stable anti-immigration preferences, the passage of anti-immigration reforms is congruent with such preferences and thus can be seen as nothing more than an adequate public policy response by the public.

Data and Methods

Unlike many other government policies, the causal identification of immigration reform effects are complicated by the fact that they are predominantly “assigned” at the national level and thus cannot be easily randomized. At the same time, since immigration policy changes also likely reflect the preferences of voters and parties in power alongside other unobserved factors, any cross-sectional associations (or the lack thereof) are likely subject to reverse causality and omitted variable bias. The standard econometric solution to these problems is to produce plausible causal estimates of policy effects by utilizing longitudinal or panel data from as many countries and years as possible.

To test for the voter backlash to immigration reforms, I have thus gathered a TSCS dataset linking the best available voter behavior, public opinion, and policy data at the country-year level for 1980-2014 across 24 European countries. (Note: Unlike related research on policy backlash [@Vranceanu2018, Abou-Chadi2019a], I do not use individual-level public opinion data since such data have a much more limited coverage. While useful for examining potential heterogeneous effects by voter characteristics (which is beyond this paper’s scope), individual-level data do not provide any advantage over aggregate-level data in the immigration policy setting where all “treatments” are assigned at the country-year level.) The most prominent set of statistical techniques for this purpose is related to either unit or two-way fixed effects (FEs) regression models, which can account for unobserved confounders under a number of more or less realistic assumptions. The main identifying assumption of such approach here is that populist vote or immigration attitudes would have developed similarly in countries with and without the observed immigration reforms had they not been adopted.

While the inclusion of unit FEs in a linear model adjusts for all unit-specific and time-invariant confounders, the inclusion of time FEs adjusts for common trends. Accordingly, two-way (unit and time) FE models have been especially popular in the policy literature as the (as-if) generalization of the design-based “difference-in-difference” technique, which identifies treatment effects by simply comparing the average change over time in the outcome variable for the treatment group with the one in the control group (assuming parallel trends). This practice, however, has been recently challenged by some scholars who point out to the unclear counterfactual interpretation of two-way FEs and the impossibility of simultaneously adjusting for unit-specific and time-specific confounders in a nonparametric way [@Kropko2020, Imai2021b].

I thus use the simpler unit FEs as my starting specification to answer the following question: “as immigration policy changes within a certain country, how does it relate to the changes in aggregate voter behavior (or attitudes) over time?” Most of the following step-by-step specifications can be summarized as follows (for country $c$ and year $t$ where $y$ stands for populist vote or attitudes, $X$ stands for a set of lagged control variables, $C$ stands for country-fixed effects, and $T$ stands for time-fixed effects):

y_ct = _0 + _1Immigration Reform_ct + _2C_c + _3T_t + _4X_ct-1 + _5C_cTrend + _6y_ct-1 + _ct.

Besides FEs and standard control variables, some models include linear country-specific time trends to account for the differential trajectories in voter behavior potentially confounded with immigration reforms [@Aksoy2020]. Given that past anti-immigration voter preferences and populist electoral victories likely decrease the probability of pro-immigration reforms, I also include lagged outcomes in some of my specifications to account for this issue. Importantly, all coefficient estimates are reported using heteroskedasticity-robust and (given that the policies are “assigned” at the level of countries) country-clustered standard errors. (Note: Since there is no consensus in the literature regarding whether the usually more conservative (and econometrically justifiable) country clustering should be preferred to the heteroskedasticity robustness in the relatively small-N samples like the one here, I report both types of standard errors for the sake of transparency.)

As for the main dependent variable, I rely on the Timbro Authoritarian Populism Index which provides a comprehensive measure (1980-2020) of the share of votes cast for “right-wing populist” parties in European national legislatures (in each country-year since the last elections) as coded by the expert Chapel Hill Expert Survey (CHES). These are exactly the parties that have been either explicitly anti-immigration or otherwise attractive to the anti-immigration electorate in line with the backlash argument explicated above (for the list of parties, see Appendix). Given that populist vote is not observed every year, some of my models also estimate a version of the equation above at the more empirically appropriate country-election level. (Note: I also corroborate my analysis using Timbro index for all right-wing and left-wing populist parties, as well as using the alternative coding of “far-right populist parties” from the Popu-List project with no change in the substantive results (not shown).)

To complement this behavioral variable with the more fine-grained, attitudinal measures of voter preferences, I also rely on the country-year “immigration conservatism” index compiled by @Caughey2019 as my second dependent variable. Based on the aggregation of major public opinion survey questions regarding immigration (e.g., “Do you think immigration should be decreased?”), this index indicates the general restrictiveness of the electorate in a given country/year. As another attitudinal dependent variable, which has been increasingly emphasized in the literature, I also rely on “immigration salience” as measured by the responses to the “most important issue” question that mention immigration in Eurobarometer [@Claassen2021]. (Note: Unfortunately, the immigration preference and especially salience measures are available for a more limited set of countries and years (1990-2014 and 2002-2014 respectively).)

As for the main independent variables, I rely on two major immigration policy datasets: The Immigration Policies in Comparison (IMPIC, 1980-2010) [@Helbling2016] and Determinants of International Migration (DEMIG POLICY, 1980-2014) [@DeHaas2015]. While IMPIC allows comparing the absolute levels of immigration policy openness across countries and years (used as the first explanatory variable in my analysis), DEMIG codes the relative changes to immigration policy by its valence and magnitude in each particular country and year. (Note: To summarize all DEMIG policy changes by country-year and country-election, I follow @Haas2019 and calculate the weighted average of reforms by their valence and magnitude. While DEMIG data is generally not suited for comparison between countries, to the extent the paper is focused on the relationship between immigration policy changes and voter behavior within countries, it should be appropriate as a treatment variable in unit and and two-way FEs models.) To make these datasets comparable, I also calculate IMPIC’s annual first difference or the change in the IMPIC score from the previous year (or the previous election when appropriate). Given the common symmetry assumption described above, these two further explanatory variables based on either the annual summary of DEMIG or the annual change in IMPIC treat immigration reforms as a continuum from anti-immigration to pro-immigration policy changes. (Note: To allow for the possibility of asymmetric responses, however, I also examine the possible effects of pro-immigration and anti-immigration reforms based on these measures separately in Tables tab:TIMBROrobust2, tab:voterscorerobust, and tab:immsaliencerobust with no change in the substantive results.)

Although I treat the measures of immigration policy change based on either IMPIC and DEMIG as conceptually related, the resulting correlation coefficient between their symmetric versions is only marginally positive ($r = 0.2$). Most changes to immigration law include both pro-and anti-immigration provisions, which often makes the assessment of their restrictiveness and comparison difficult (for both scholars and voters alike). To address the possible ambiguity issues, I also create a composite binary measure of unambiguous pro-immigration policy shocks which simply takes a value of 1 if the change is coded as positive by both datasets (for the list of identified major pro-immigration changes, see Appendix). (Note: For a discussion of important differences between the two datasets, see @Schmid2016.)

As emphasized earlier, since my primary focus is on the possible voter backlash to reforming legal immigration pathways, my main empirical specifications only include the “`Labor migration” and “Family reunification” components of IMPIC and the comparable “Legal entry and stay” component of DEMIG (while excluding humanitarian and irregular migration policies from these indices). (Note: I do not analyze the “Co-ethnics” component of IMPIC or the “Integration” and “Exit” components of DEMIG due to the unclear theoretical expectations regarding the potential voter backlash to these policies.) While my main results do replicate for the “Asylum and refugees” and “Control” components of IMPIC and the “Border and land control” component of DEMIG (Table tab:TIMBROhumcon), it is worth noting that the coverage of both datasets ends prior to the peak of Europe’s refugee crisis, which limits the external validity of these analyses. (Note: I also use the separate IMPIC scores for labor and family migration (see Tables tab:TIMBROrobust2, tab:voterscorerobust, and tab:immsaliencerobust), as well as nation-level reforms as coded in DEMIG only (not shown) with no change in the substantive results.)

Some specifications include the (lagged) control variables common to this literature: the share of immigrant population, unemployment rate (log), and GDP per capita (log). Since the way how governments implement [@Kolbe2020] and how voters respond to immigration policies [@Chou2021] may depend on the identity of the incumbent, I also control for the left-right ideology of the government coalition since the latest election as measured by the Schmidt index. All variables have been standardized to vary from 0 to 1.

Descriptive Analysis

Do pro-immigration reforms fuel populism? I start with a (naive) descriptive analysis documenting the basic trajectories of populist voting and immigration policy across time and countries. As can be seen from Figure fig:TIMBROtimelineG, both the share of right-wing populist voting and the openness of immigration policy have gradually increased in Europe over the last forty years. Although some version of this stylized fact has been often used by political commentators to draw the connection between these phenomena, this does not mean that there is any causal relationship or even that the same correlated trend has held within each particular country.

Indeed, there is a lot of between-country heterogeneity in the trajectories of immigration policy and populism (see Figure fig:TIMBROtimeline). While countries like Germany and Switzerland notably show a trend similar to Europe as a whole (more open immigration and rising populism), one may find quite a few counter-examples. Slovakia, for instance, has experienced both the slow liberalization of its immigration system and the fall of populist vote while Greece and Finland have further restricted immigration alongside rising populism. At the same time, countries like Portugal and Spain have prominently managed to liberalize its immigration system without experiencing any significant wave of right-wing populism. Finally, despite the understandably slow change of immigration policy, Figure fig:TIMBROtimeline generally indicates a wide range of both between-country and within-country variation in both dependent and independent variables, which is crucial for effect estimation under TSCS methods. (Note: For a summary of the trajectories of immigration policy and preferences, see Figure fig:ATTITUDEStimeline. In line with the previous literature [@Dennison2019c, Kustov2019b], Figure fig:ATTITUDEStimeline documents the relative stability of immigration preferences and the relative volatility of immigration salience.)

Setting aside the variation across time, it appears that the positive relationship between immigration policy openness and populism also holds cross-nationally. In particular, according to Figure fig:TimbroIMPIC, few countries with restricted immigration have experienced any significant right-wing populist vote, while there is a lot of variation in populism across countries that are relatively open to immigration. Interestingly, however, immigration policy openness is negatively related to restrictive immigration preferences of the public.

To get at the contemporaneous relationship between the changes in immigration policy and populist voting, I also examine the first differences of these variables. As can be seen from Figure fig:dIMPICshort, there does not seem to be any relationship between the change in immigration policy and right-wing populist voting between elections. When it comes to immigration preferences, however, it is negative as in the case of absolute levels.

At the same time, when one compares the amount of accumulated pro-immigration policy change and populist vote (or anti-immigration preferences) over the last forty years, this relationship is also negative across countries for both variables (see Figure fig:dIMPICLong). Although such analysis is evidently underpowered and it is still subject to endogeneity concerns, it is descriptively true that in the long run those countries that liberalized their immigration system the most did not experience more populist voting or restrictive voter attitudes. Importantly, none of these bivariate relationships explored so far necessarily imply that changes in immigration policy cause (or do not cause) changes in populist vote or immigration attitudes within particular countries.

Panel Data Regression Analysis

After establishing the theoretical and empirical possibility of backlash (and possible counterevidence) in the descriptive analysis, I take advantage of the time-series cross-sectional nature of my dataset and fit a set of fixed-effects linear regressions to estimate the effects of pro-immigration policy changes within countries (that are free from between-unit time-invariant confounders).

Table tab:TIMBRO summarizes the results of these empirical models. As can be seen, while the general openness of immigration system (IMPIC score) has a consistent positive relationship with populist vote, most of the coefficients are substantively small and none of them are statistically significant after accounting for time-invariant differences between countries and clustering even in the baseline model. Even if one takes some of the more conservative estimates at their face value, Table tab:TIMBRO (3-5) implies that going from the most restrictive to the most open immigration system is related to only 2-3 percentage points of populist vote. Alternatively, this means that a one standard deviation increase in immigration openness is at best related to a 0.05-0.08 standard deviation increase in populist vote.

Since the general immigration openness is a slow moving variable which may be partially subsumed by including fixed effects, I further consider the possible effects of the much more volatile shocks to the immigration system as measured by either the annual changes to the immigration system (IMPIC score), the number/direction of immigration reforms in a particular year (DEMIG score), or the binary combination of the two. As before, none of these results are substantively or statistically significant.

As an important robustness check, in Table tab:TIMBROBE I also consider a set of models look at the between-election variation only, which yields no change in the substantive results. At the same time, Table tab:vturn indicates that none of the immigration policy changes are related to the overall turnout in the electorate. Table tab:TIMBROrobust2 further shows the (null) effects of disaggregated policy indices. Finally, it is not the case that the null effect of pro-immigration policies on populism is conditional on the underlying preference restrictiveness on the electorate or the ideology of the government coalition in power (i.e., there is no interaction effect between these sets of variables, not shown).

It is worth noting that, across all models, the very direction of policy coefficients strongly depends on a particular measure or specification in use. Given that I ran 80 different model specifications with right-wing populism as an outcome and found statistically significant coefficients (in either direction) for policy in only 3 of those models, this is about what one would expect based on random chance. Even if one considers the most consistently positive coefficient of immigration openness under the less conservative model and standard error specifications, it is still not substantively significant in terms of its magnitude.

Additional Evidence on Immigration Attitudes

Although immigration policies do not seem to be related to right-wing populist vote (or overall turnout), it is still possible that they can affect a much more proximate outcome of immigration attitudes, including both voter preferences and issue salience. While previous research has documented the general stability of preferences and volatility of salience [@Kustov2019b, Dennison2019c], an important advantage of using either measure over populist voting is that it changes—whenever they exist—are not confined to election years. I thus further replicate the models from Table tab:TIMBRO on these aggregate attitudinal outcomes.

Overall, as Tables tab:voterscore and tab:immsalience indicate, voters do not change their immigration attitudes in the more salient and/or restrictive direction in response to pro-immigration reforms. If anything, it appears to be the opposite—immigration openness and pro-immigration reforms are related to more positive immigration preferences and less salience of the issue. The positive relationship between immigration policy openness and voter preferences is particularly remarkable since it holds even after accounting for fixed effects and other controls. For instance, Table tab:voterscore (3) implies that a one standard deviation increase in immigration openness is related to a 0.4 standard deviation decrease in anti-immigration preferences.

Nonetheless, it would be premature to claim any causal policy effects since the potential for reverse causality and omitted variable bias here is arguably even more present than in the case of populist vote. For instance, one can argue and show that improvements in immigration preferences in public opinion polls push (or simply better allow) policymakers to pass pro-immigration reforms. Still, it is notable that the positive relationship between pro-immigration policy and preferences is very consistent across all model specifications. Furthermore, there appears to be no relationship between immigration openness and (placebo) economic issues as measured by @Caughey2019 (see Table tab:voterscoreEcon).

At the same time, the positive correlation between the openness of immigration policy and voter preferences is not statistically significant after including country-specific trends and it appears to be particularly driven by labor migration (see Table tab:voterscorerobust). Furthermore, none of the more volatile “reform” variables are significantly related to immigration attitudes. Overall, while the robustness of the positive “legitimizing” relationship between policy and preferences may be reasonably disputed, it is notable that none of the results are consistent with the idea of counterproductive attitudinal backlash to immigration policy.

Table. (Pro-)Immigration Policy Changes and Right-Wing Populist Vote

Conclusion

According to most evidence, the enormous benefits of more open global labor mobility outweigh its many possible costs [@Clemens2011, Caplan2019]. Still, it is also the case that most voters in rich countries—even those who are educated and unprejudiced—oppose increasing immigration, and often (com)passionately so [@Kustov2020a]. Many scholars and policymakers reasonably worry that any significant relaxation of existing restrictions would face voter resistance and ultimately be counterproductive. Building on the best available public opinion and policy data from the last forty years in Europe, this paper provides a data-driven assessment of such concerns, which is of both theoretical and practical interest. Overall, my results indicate that existing pro-immigration reforms have not been counterproductive in terms of increasing populism or immigration concerns.

Specifically, despite the gradual increase of both populism and pathways for legal immigration over the last decades, the timing of existing pro-immigration reforms is not related to the increases in populist vote in subsequent elections. Similarly, pro-immigration reforms are not related to any negative changes in immigration preferences or their salience to voters. If anything, it appears that most pro-immigration reforms may further legitimize freer immigration in the electorate. To the extent most immigration is as of now already restricted, the evidence thus suggests that even significant relaxations of these restrictions do not cause popular backlash. At the same time, additional immigration restrictions are not related to the reduction of populist vote or xenophobia either.

Of course, my analysis is not without limitations. First, one can still reasonably worry about reverse causality. To the extent that populist parties may restrict immigration right after they come to power, for instance, this could arguably bias the regression estimates against finding any contemporaneous backlash effects of pro-immigration reforms. Even more broadly, if some limited electoral success of populists may influence the subsequent positions of mainstream parties by lowering their resolve to pass pro-immigration reforms, this can also bias the results downward.

According to my estimates, however, most immigration reforms (as measured by DEMIG) are not more likely to happen during election years, and those that do, overwhelmingly happen prior to the national elections (not shown). After all, many of the significant changes to the immigration system are usually planned in advance and sometimes takes years to negotiate. At the same time, the most recent evidence indicates that populist parties have largely failed to shift the positions of center-left and center-right parties on the issue [@Dancygier2020] or change the immigration policy itself when in power [@Lutz2019].

Second, it is still possible that there are shorter-term effects of immigration reforms on voter behavior and attitudes that simply cannot be captured in the country-year data. Relatedly, some reforms may have an impact only on some voters, and some of these impacts may even cancel each other out in the aggregate (e.g., when reforms increase or decrease populist voting among anti-immigration or pro-immigration voters). To the extent these effects exist, however, they do not seem to be long-lasting enough to be counter-productive, i.e., have a meaningful impact on the electoral fortunes of populist parties.

Still, future research may benefit from expanding the data coverage to more countries and years, as well as exploring the potential heterogeneous effects of different policy types on different types of voters across a variety of time periods, especially using individual-level longitudinal data. For example, selective pro-immigration reforms that explicitly focus on advancing national interest (e.g., increasing high-skilled immigration) may have the most legitimizing impacts among those who currently oppose immigration [@Kustov2020a]. At the same time, some of the more controversial pro-immigration law enforcement policies that instead draw public attention to the lack of border control [@Harell2017] may conversely generate voter backlash even among those who are not particularly hostile to immigration.

Third, while backlash is usually conceptualized as a response to pro-immigration advancements, one can also imagine the “reverse backlash” to anti-immigration advancements, especially given the increasing voter polarization on the issue. One prominent example is the pro-immigration reaction of the public in the aftermath of Brexit in the UK [@Schwartz2020]. More generally, since the US and the UK attitudes toward immigration have never been more positive than they are now, (assuming these trends persist) the notorious Trump or the Brexit backlashes may not be counterproductive to pro-immigration advancements after all.

Finally, immigration reforms are rarely fully exogenous to voter behavior. Consequently, one can argue that the reason why policy backlash has not been empirically present is that even the most significant pro-immigration reforms of the last forty years are conservative enough for voters not to be considered “excessive” within the existing political equilibrium [@Chou2021]. Even with this limitation in mind, the evidence presented here certainly challenges the common claim that existing pro-immigration policy advancements—at least when it comes to opening pathways for legal labor and family immigration—may have been counterproductive or conducive to the rise of populism.

Acknowledgements

I would like to thank my colleagues who have read and commented on previous drafts of this article. I am especially grateful to Samuel Diaz, Elif Naz Kayran Meier, Michael Kenwick, Omer Solodoch and anonymous reviewers for their helpful suggestions. Different parts of the project were also presented at the APSA, SPSA, MPSA, and ISA annual meetings, as well as the Yale, EUI, Ryerson, UPenn, and World Bank research seminars.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes on contributor

Alexander Kustov is an assistant professor of political science at the University of North Carolina at Charlotte.

Correspondence address: University of North Carolina at Charlotte, Charlotte, NC 28223

Email: akustov@uncc.edu

Supplemental materials and replication data for this article can be accessed at https://doi.org/10.1080/XXX.

ORCID

Alexander Kustov https://orcid.org/0000-0001-9361-8739

References

See source bibliography.

Supplementary Materials (Online Appendix)\

Testing the Backlash Argument: Voter Responses to (Pro-)Immigration Reforms

roman

Tables and Figures

Atable

Table. (Pro-)immigration Policy Changes and Right-Wing Populist Vote Across Elections

Atable

Table. (Pro-)Immigration Policy Changes and (Anti-)immigration Voter Preferences

Atable

Table. (Pro-)Immigration Policy Changes and Voter Immigration Salience

Afigure

Table. (Pro-)immigration Policy Changes and Voter Turnout Across Elections

Afigure

Table. (Pro-)Immigration Policy Changes and Right-Wing Populist Vote (Alternative Policies)

Atable

Table. (Pro-)Immigration Policy Changes and Voter Economic Preferences (Placebo Test)

Atable

Table. Immigration Policy Changes and Right-Wing Populist Vote (Alternative Operationalizations)

Afigure

Table. Immigration Policy Changes and (Anti-)Immigration Voter Preferences (Alternative Operationalizations)

Afigure

Table. (Pro-)Immigration Policy Changes and Voter Immigration Salience (Alternative Operationalizations)

Afigure

Figure. Immigration Openness and Right-Wing Populist Vote in Europe (1980-2014)

Afigure

Figure. Immigration Openness, Preferences, and Populist Vote across Countries

Afigure

Figure. Short-term Change in Immigration Openness, Preferences, and Populist Vote

Afigure

Figure. Long-term Change in Immigration Openness, Preferences, and Populist Vote

Afigure

Figure. The Trajectories of Immigration Policy and Right-Wing Populist Vote Across Europe (1980-2014)

Afigure

Figure. The Trajectories of Immigration Policy and Attitudes Across Europe (1990-2014)