Personal Economic Shocks and Public Opposition to Unauthorized Immigration

Abstract Do negative economic shocks heighten public opposition to immigration, and through what mechanisms? Extant research suggests that economic circumstances and levels of labour market competition have little bearing on citizens' immigration attitudes. Yet personal economic shocks have the potential to trigger the threatened, anti-immigration responses – possibly through channels other than labour market competition – that prior cross-sectional research has been unable to detect. To examine these propositions, we used a unique panel study which tracked a large, population-based sample of Americans between 2007 and 2020. We found that adverse economic shocks, especially job losses, spurred opposition to unauthorized immigration. However, such effects are not concentrated among those most likely to face labour market competition from unauthorized immigrants. Instead, they are concentrated among white male Americans. This evidence suggests that the respondents' anti-immigration turn does not stem from economic concerns alone. Instead, personal experiences with the economy are refracted through salient socio-political lenses.

High rates of immigration to industrialized democracies, combined with electoral successes by far-right parties, have spurred substantial research on public attitudes towards immigration.Initially, a prominent argument centred on labour market competition as a key explanation for natives' anti-immigration sentiments.This view was empirically supported by evidence that natives with less formal education were more opposed to immigration.Researchers interpreted this pattern as evidence that those most directly competing with foreign labourersthat is, low-skilled nativesviewed immigration more negatively due to labour market threat (Mayda 2006;Scheve and Slaughter 2001).
Some subsequent research has cast doubt on this interpretation, showing that this correlation likely stems from another source: the less educated tend to be more averse to the presence of foreigners and their perceived cultural and sociotropic impacts (Sniderman, Hagendoorn, and Prior 2004).In this view, less-educated natives oppose immigration for reasons unrelated to their employment prospects.This interpretation was bolstered by studies finding weak empirical associations between the characteristics of natives' jobs, occupations, or employment status and their immigration attitudes (Bansak, Hainmueller, and Hangartner 2016;Card, Dustmann, and Preston 2012;Hainmueller, Hiscox, and Margalit 2015).These findings reinforce research contending that individuals' labour market circumstances have little bearing on their immigration attitudes (see Hainmueller and Hopkins (2014) for a summary).
While this view is backed by considerable evidence, we suspect certain caveats are warranted (Dancygier 2010;Dancygier and Donnelly 2013;Goldstein and Peters 2014;Malhotra, Margalit, and Mo 2013).First, prior evidence overwhelmingly rests on cross-sectional data (but see Kustov, Laaker, and Reller 2021) observational or experimentalthat shows associations between individuals' labour market standing and their immigration attitudes at a specific point in time.Research has not explored whether changes in individuals' labour market standing are associated with subsequent shifts in their immigration attitudes. 1xtant accounts may therefore overlook a dynamic effect that connects individuals' labour market experiences with their immigration views.For example, when natives experience a negative economic shock, such as losing a job, they may grow more hostile to immigration.However, evidence that labour market shocks influence immigration attitudes is not necessarily evidence that labour market competition is at work.Experiencing a labour market shock may lead people to seek a scapegoat for their hardships, with foreigners as one potential target.In that case, people who suffer a layoff may grow more hostile toward immigrants, even with little reason to blame their experiences on immigrants.
Moreover, recent research finds significant heterogeneity in responses to economic threats across different ethnic/racial groups and social circumstances (Baccini and Weymouth 2021).Such reactions are partly attributed to the perceived threat experienced by members of dominant social groups (for example, men and whites) as a result of their worsening economic standing (Cramer 2016;Kurer and Van Staalduinen 2022;Mutz 2018;Quillian 1995).Specifically, two structural developments in recent decades have contributed to the decline in white men's social status in high-income democracies.The first is the widespread movement of women into the labour market, especially into well-paid occupations.The second includes profound shifts that challenged the traditional social boundaries through which white men maintained their social status and selfesteem: the institutionalization of cultural frameworks pursuing racial and gender equality (Gidron and Hall 2017;Hochschild 2018).If correct, this logic suggests that economic shocks such as layoffs might spur anti-immigration attitudes among some social groups more than others.Personal experiences with the economy are refracted through salient socio-political lenses (Hopkins 2010;Sides, Tesler, and Vavreck 2019).This logic also explains why the effects of job loss may not be concentrated on attitudes toward those immigrants seen as direct labour market competitors.
To test these propositions, we analyzed a unique panel that tracked a large, population-based sample of Americans over fourteen years.Respondents reported information about their employment situation and attitudes toward unauthorized immigration.Most American workers were not in direct competition with unauthorized immigrants.By focusing on attitudes on unauthorized immigration, we thus considered an attitudinal domain in which labour market experiences, if they proved influential at all, were likely to operate through channels other than labour market threat.
Our results show that individuals who lose their jobs are substantially more likely to support the deportation of unauthorized immigrants, with a shift of 6.7 points, representing an increase of 16 per cent.Notably, this effect holds when accounting for respondent fixed effects and timevarying characteristics such as local labour market conditions.While this research emphasizes attitudes toward unauthorized immigrants, Supplemental Material (SM) Figure SM-2 shows generally strong associations between attitudes toward authorized and unauthorized immigrants, suggesting that these findings may be instructive about a more general shift in immigration attitudes (see also Wright, Levy, and Citrin 2016).
Furthermore, we find that the shift in unauthorized immigration attitudes predominantly occurs among white Americans who lose their jobs.As compared to minority Americans, whites tend to support deporting immigrants by six percentage points more, a gap that more than doubles if whites lose their jobs.We observe a similar pattern among males, particularly among white males, whose shift against unauthorized immigration after job loss is substantially greater than females'.
In contrast, we do not find that job loss leads to stronger anti-immigrant views among individuals employed in locales with bleaker employment prospects, or in places with larger immigrant populations.Considered as a whole, our evidence indicates that economic experiences can influence immigration attitudes, though not via the labour market competition channel.Moreover, economic shocks typically affect only a small number of Americans at a time and are not a prevalent source of public opposition to unauthorized immigration.
Together, these results suggest that the turn against unauthorized immigration following job loss is more likely due to the newly unemployed directing their frustrations at foreigners than reflecting a calculated response stemming from a desire to reduce the threat to one's employment prospects.Put differently, job loss can shape attitudes through channels other than labour market competition.Personal experiences and group-level threats are often competing explanations for immigration attitudes, but we show how they can interact.

Data
We employed a population-based panel of American respondents who were eighteen or older in 2008.Knowledge Networks (later GfK and then Ipsos) recruited panelists offline via addressbased sampling or random-digit dialing.The first wave was administered in October 2007, and the fifteenth wave in October 2020.2Surveys were administered primarily before and after presidential and mid-term elections.The sampling was weighted to generate marginal distributions on key variables matching those of the US population.See Table SM-1 for dates and sample sizes of each wave and an explanation of the sampling method.
Table 1 reports key demographics, showing that the sample mirrors the target population of English-speaking adults.For example, respondents in the October 2020 survey were 12 per cent black, 11 per cent Hispanic, and 71 per cent white, with 38 per cent reporting a college degree.
The key dependent variable used to capture public attitudes toward unauthorized immigration is based on the following question: 'On immigration, some people argue that U.S. policy should focus on returning illegal immigrants to their native countries.Others argue that U.S. policy should focus on creating a pathway to U.S. citizenship for illegal immigrants.Still, others are somewhere in between.Where would you place yourself on this scale, or haven't you thought much about this?' Responses are on a 7-point scale ranging from 1 ('Return illegal immigrants to their native countries') to 7 ('Create a pathway to U.S. citizenship for illegal immigrants'). 3 We then created an indicator variable representing support for deportation, which equals 1 if the answers are below the midpoint. 4 We examined the effects of two types of economic shocks.The first was job loss.This dichotomous variable equals 1 for unemployed respondents who were employed during the last time they were asked. 5The second shock is a sizable drop in income.Annual household income was measured using a 21-point scale, from 'Less than $5,000' to '$250,000 or more'. 6We consider a 'sizable drop' as a fall of at least two levels, representing a decrease of 25 to 50 per cent.A substantial drop in measured household income, when not the result of losing one's job, is likely to be the result of the respondent's spouse losing their job.Overall, the share of unemployed respondents in our sample is 3.5 per cent, the proportion of recent job losers is 1.5 per cent, and 7.7 per cent of the observations are of respondents who experienced a significant drop in income.See the appendix for full summary statistics.

Results
Table 2 examines how recent, short-term or persistent unemployment correlates with voters' opposition to unauthorized immigration.First, for each available wave, we estimate the correlation between unemployment and support for the deportation of unauthorized migrants using linear probability models (LPM), controlling for the respondent's age, gender, race, education, income, employment status, and partisanship. 7In the last two columns, using the pooled sample, we estimate the model with and without year and individual-level fixed effects.
Consistent with prior cross-sectional analyses, we find no evidence that unemployed voters are less supportive of unauthorized immigration than employed voters.However, the analysis underscores that when relying on cross-sectional data, the association between unemployment and opposition to unauthorized immigration is noisy and varies across waves in terms of sign, substantive magnitude, and statistical significance.Furthermore, when pooling the data over time The survey randomly presented a reversed version of this scale for half of the sample.

4
In Table SM-15 we use an alternative dependent variable and estimate the effect of economic shocks on respondents' support for 'increasing border security by building a fence along part of the U.S. border with Mexico.' (in columns 8-9), the analysis shows a negligible and statistically non-significant association between unemployment and opposition to unauthorized immigration.
While voters' employment status does not seem to be a meaningful source of influence on immigration attitudes, changes in such economic circumstances have a different and more meaningful impact.Table 3 assesses how a recent job loss or a drop in income affects opposition to unauthorized immigration.All models control for the respondents' education and employment status. 8 Across different specifications, job loss consistently and significantly increases voters' opposition to unauthorized immigration.Respondents who recently became unemployed are 6.7 percentage points more supportive of deporting unauthorized immigrants.This effect represents an increase of 16 per cent above the baseline rate of support for deportation (40.7 per cent) among employed respondents.Similarly, a major income drop (of at least 25 per cent) is associated with a 3.8-point increase in opposition to unauthorized immigration, representing an increase of 9.3 per cent above the baseline.
The analysis indicates that recent re-employment does not generate a counter-balancing (pro-immigration) effect.The coefficient for finding a new job in column 5 is substantively small (0.007) and statistically non-significant.This finding raises doubt about the notion that labour market competition is a central consideration shaping immigration attitudes; if so, we might expect some drop in opposition to unauthorized immigration following re-employment.

Effect Heterogeneity
To further investigate whether job loss's effects on immigration attitudes stem from labour market competition, we examine whether low-skilled respondents and labour union members are differentially affected by unemployment shocks. 9If labour market competition underlies the attitudinal impact of economic shocks, low-skilled individuals-i.e., workers for whom unauthorized immigrants are more likely labour market competitors-should grow more supportive of deportation following job loss relative to highly skilled individuals.10Conversely, labour union members, who are likely to have more generous compensation guarantees if laid off (Budd and McCall 1997), should be less affected by job losses than non-unionized workers.Yet, as We use respondents' educational attainment to proxy for professional skills.Low-skilled respondents are those with 12 or fewer years of education.columns 2 and 3 in Table 4 show, we find no evidence of statistically significant heterogeneous effects across respondents differentiated by skill level or union membership.
Next, we use county-level data to examine two additional observable implications of the labour market competition hypothesis.If job loss is generating an anti-immigrant effect because of concerns about economic competition, we would expect the effect to be stronger among natives residing in counties where the unemployment rate or the share of foreign-born residents is higher.
Yet, we again find no support for this conjecture.Residents of counties with above-median unemployment or shares of foreign-born residents are not more strongly affected by job loss despite greater labour market exposure to competition from immigrants.11Moreover, the signs of the (statistically non-significant) interaction effects are negative.This suggests that natives exposed to stronger county-level economic competition are, if anything, less likely to adopt anti-immigrant views after losing their job.
If labour market position doesn't amplify or buffer the impacts of unemployment, we might instead consider characteristics that moderate the political implications of job loss.Specifically, recent research shows that white Americans exhibit a stronger shift to conservative positions following exposure to the negative impacts of deindustrialization and the closures of local manufacturing plants (Baccini and Weymouth 2021).This is attributed to the anxiety experienced by a previously dominant group when its privileged social status becomes threatened (Gidron and Hall 2017).In the SM, we report survey evidence consistent with this claim, as Americans perceive white men to have lost considerable status between 1980 and 2020.
Indeed, we find that following a layoff, white respondents are a sizeable 17 percentage points more opposed to unauthorized immigration.In contrast, non-white respondents exhibit no significant change in their attitudes toward unauthorized immigration after experiencing job loss.12We observe a similar interaction effect among males.These results are consistent with the notion that personal experiences are refracted through cultural and racial lenses (Sides, Tesler, and Vavreck 2019).

Effect Durability
How long does the effect of economic shocks persist?The panel allows us to examine this question in two ways.First, the analysis in Table 3 shows that newly re-employed respondents are not more supportive of unauthorized immigration than the previously unemployed who did not acquire a job, suggesting that the effect of job loss endures beyond the resumption of employment.
Second, we examine whether respondents who experienced an economic shock in previous survey waves are still more likely to hold anti-immigrant attitudes.In Fig. 1, we estimate the effect of job loss and income drops on support for deportation using separate LPMs.We find that the effects of job losses and income drops are strongest when respondents are still experiencing the shock.This effect diminishes over time; three waves after exposure to the shock (forty-five months later, on average), it is no longer statistically distinguishable from zero.
However, a caveat is necessary.Unlike the robust null result of re-employment, the endurance of the lagged unemployment shock is sensitive to model specification.Specifically, using individual fixed or random effects instead of state-level fixed effects imprecisely estimates the lagged job loss coefficient (see Table SM -14).Thus, while it is clear that the effect of job loss endures after re-employment, the results are less clear-cut concerning the pattern of decline over time.

Conclusion
Public opinion research on immigration offers seemingly conflicting findings: whereas macro-level studies document that opposition to immigration rises in periods of economic crisis (Hatton 2016;Polavieja 2016), analyses of individual-level data reveal a weak relationship between people's standing in the labour market and their immigration attitudes (Hainmueller and Hopkins 2014).Our results help reconcile these findings: the experience of an economic shock such as job loss generates greater opposition to unauthorized immigration, but this attitudinal shift has little to do with labour market considerations.Instead, this shift appears to reflect, at least partly, concerns about status threat and people's tendency to assign blame to an out-group for their own hardships.These findings highlight the need for more research on the interaction between economic conditions and attitudinal shifts on non-economic issues.Research in political economy typically focuses on the link between changes in individuals' economic standing and their attitudes on directly related policy issues: for example, increased risk of unemployment and support for redistribution (Cusack, Iversen, and Rehm 2006;Rehm 2009).Yet, several recent studies provide evidence that exposure to economic hardship can also lead to shifts in other domains, be it an embrace of authoritarian dispositions (Ballard-Rosa et al. 2020) or nativist sentiments (Colantone and Stanig 2018).Findings in this vein, however, still serve much like a 'proof of concept', demonstrating that economic conditions can influence attitudes in non-economic domains.But basic aspects of this phenomenon remain unexplored: under what conditions does such cross-domain influence occur, and in what segments of the public is such influence more prevalent?These important questions warrant systematic theorizing and greater empirical attention.
In a recent article, Kustov, Laaker, and Reller (2021) report that immigration attitudes tend to be stable over time.Examining multiple panels, they note a high correlation (0.72) in immigration attitudes reported in the first and final waves.Nonetheless, our results show that individuals do become significantly more apprehensive about unauthorized immigration in the aftermath of a personal economic setback.Yet, as the incidence of these setbacks tends to be low in the overall population, this study has shown that these shock-induced changes in attitudes pertain to only a narrow segment of the electorate (Malhotra, Margalit, and Mo 2013).Hence, our results appear to be consistent with the characterization of overall stability in citizens' immigration views.
While our evidence comes only from the United States, the theoretical claims of our study could apply elsewhere.As shown with respect to the impact of job loss on social policy preferences, patterns of attitude change are similar even in countries with differing welfare state regimes (Margalit 2019), suggesting that behavioural responses to economic setbacks may not depend on country-specific characteristics.As long-running panel studies that include immigration items become available, replicating this research design in other high-income democracies will shed more light on the generalizability question.
Our research underscores the value of panel data for studying theories of change.Long limited by data availability, arguments hypothesizing shifts in political views as a function of changes in objective conditions are often supported by cross-sectional data, relying on cross-respondent variation in the independent and dependent variables.Yet this design can mask important dynamics, particularly when there is heterogeneity in individuals' responses to said change, as in our study.The growing availability of panel data, partly due to the greater ease of administering repeated surveys online to large samples, provides new opportunities to informatively re-examine theories of political change.

Figure 1 .
Figure 1.The effect of economic shocks over time.Note: N = 5,106.Dots and lines represent point estimates and 90-95 per cent confidence intervals drawn from LPM regressions controlling for respondents' race, age, income, employment status, union membership, partisanship, and state of residence.

Table 1 .
Demographics for waves seven (2012-13; n = 2,264) and fifteen(October 2020; n = 1,115) a Variables in years.b Variable in thousands of dollars.

Table 2 .
Contemporaneous unemployment and support for deportation of unauthorized migrants Outcome variable is a binary indicator that equals 1 if the respondent supports deporting unauthorized migrants, and 0 otherwise.All regressions control for respondents' age, race, gender, level of education, income employment status (retired, disabled or other), and partisanship (an indicator variable for Republicans).Robust standard errors in parentheses; **p < 0.05. Notes:

Table 3 .
The effect of economic shocks on voters' support for the deportation of unauthorized immigrants Outcome variable is a binary indicator that equals 1 if respondents support deporting unauthorized migrants and 0 otherwise.All regressions control for respondents' level of education, level of income (logged), and employment status.Robust standard errors in parentheses; **p < 0.05.
Notes:8The employment status variable is otherwise coded as employed, retired, disabled, or 'other.'

Table 4 .
Effect heterogeneity by respondent characteristics Outcome variable is a binary indicator that equals 1 if the respondent supports deporting unauthorized migrants, and 0 otherwise.The linear probability models in columns 2-7 interact with unemployment shocks with indicator variables denoting the different groups of respondents.All regressions control for the constitutive terms of the interaction with job loss and age, income, employment status (retired, disabled or other) and partisanship.Robust standard errors in parentheses; **p < 0.05, *p < 0.1. Notes: