Who Benefits? Race, Immigration, and Assumptions About Policy

Abstract

Existing scholarship suggests that attitudes about the real or imagined beneficiaries or targets of public policies shape public opinion about those policies, with racial and ethnic stereotypes driving policy evaluations for many Americans. Despite the importance of these assumptions, we lack strong evidence about how and why people form such assumptions in the first place. In a pre-registered survey experiment, I demonstrate that elements of policy design (e.g., a work requirement) significantly affect the assumptions that individuals make about policy beneficiaries (their race and national origin). These assumptions shape individuals’ evaluations of the policy, conditional on existing attitudes (e.g., racial resentment). Importantly, existing attitudes do not condition the effects at the assumption stage: even those who profess not to believe in racial stereotypes about work ethic still assume that the absence of a work requirement makes a policy more likely to benefit blacks and immigrants.

Appendices

Appendix 1. Questionnaire

• 1. We would like your opinion on a hypothetical federal government policy. Under this policy, people with lower incomes would receive assistance in the form of cash aid/a tax credit. [Also vary:] Only individuals that earn income through work would be eligible for this program. The government estimates that this policy would cost the U.S. Treasury about $73 billion per year.

  Would you approve or disapprove of this program?

  • (a) Strongly approve

  • (b) Approve

  • (c) Approve somewhat

  • (d) Neither approve nor disapprove

  • (e) Disapprove somewhat

  • (f) Disapprove

  • (g) Strongly disapprove

• 2. Here is a list of different groups in American society. Which of these groups do you think are most likely to benefit from this program? You may select multiple groups.

  • Poor people

  • The unemployed

  • Working-class people

  • Middle-class people

  • Wealthy people

  • Big business

  • Small business

  • Labor unions

  • Whites

  • Blacks or African-Americans

  • Latino or Hispanic Americans

  • Asian Americans

  • Immigrants

  • Americans born in the United States

  • Men

  • Women

• 3. Which political party do you believe would support this program?

  • (a) Democrats

  • (b) Republicans

  • (c) Neither Democrats nor Republicans

  • (d) Both Democrats and Republicans

  • (e) Don’t know

• 4. Do you think of yourself as a Democrat, a Republican, and Independent, or what?

  • (a) Democrat

  • (b) Republican

  • (c) Independent

  • (d) Other

  • (e) No preference

• 5. [If answered a or b to question 6] Would you consider yourself a strong Democrat/Republican, or a not very strong Democrat/Republican?

  • (a) Strong

  • (b) Not very strong

• 6. [If answered c, d, or e to question 6] Do you think of yourself as closer to the Republican Party or the Democratic Party?

  • (a) Closer to Republican

  • (b) Closer to Democratic

  • (c) Neither

• 7. Where would you place yourself on this scale?

  • (a) Very liberal

  • (b) Liberal

  • (c) Slightly liberal

  • (d) Moderate; middle of the road

  • (e) Slightly conservative

  • (f) Conservative

  • (g) Very conservative

• 8. What is your age?

  • (a) 18 to 24

  • (b) 25 to 34

  • (c) 35 to 44

  • (d) 45 to 54

  • (e) 55 to 64

  • (f) 65 to 74

  • (g) 75 or older

• 9. What racial or ethnic group best describes you?

  • (a) White

  • (b) Black

  • (c) Hispanic

  • (d) Asian

  • (e) Native American

  • (f) Mixed

  • (g) Middle Eastern

  • (h) Other

• 10. What is your gender?

  • (a) Male

  • (b) Female

• 11. Thinking back over the last year, what was your family’s income? [16-point ordinal scale]

• 12. What is the highest level of education you have completed?

  • (a) No high school

  • (b) High school graduate

  • (c) Some college

  • (d) 2-year degree

  • (e) 4-year degree

  • (f) Post-graduate degree

• 13. Do you agree or disagree with the following statement? “Over the past few years, blacks have gotten less than they deserve.”

  • (a) Agree strongly

  • (b) Agree somewhat

  • (c) Neither agree nor disagree

  • (d) Disagree somewhat

  • (e) Disagree strongly

• 14. Do you agree or disagree with the following statement? “Irish, Italian, Jewish, and many other minorities overcame prejudice and worked their way up. Blacks should do the same without any special favors.” [same 5-point agree–disagree scale]

• 15. Do you agree or disagree with the following statement? “It’s really a matter of some people not trying hard enough; if blacks would only try harder they could be just as well off as whites.” [same 5-point agree–disagree scale]

• 16. Do you agree or disagree with the following statement? “Generations of slavery and discrimination have created conditions that make it difficult for blacks to work their way out of the lower class.” [same 5-point agree–disagree scale]

• 17. Which comes closest to your view about what government policy should be toward unauthorized immigrants now living in the United States?

  • (a) Make all unauthorized immigrants felons and send them back to their home country.

  • (b) Have a guest worker program that allows unauthorized immigrants to remain.

  • (c) Allow unauthorized immigrants to remain in the United States and eventually qualify for U.S. citizenship, but only if they meet certain requirements like paying back taxes and fines, learning English, and passing background checks.

  • (d) Allow unauthorized immigrants to remain in the United States and eventually qualify for U.S. citizenship, without penalties.

• 18. There is a proposal to allow people who were illegally brought into the U.S. as children to become permanent U.S. residents under some circumstances. Specifically, citizens of other countries who illegally entered the U.S. before age 16, who have lived in the U.S. 5 years or longer, and who graduated high school would be allowed to stay in the U.S. as permanent residents if they attend college or serve in the military. From what you have heard, do you favor, oppose, or neither favor nor oppose this proposal?

  • (a) Favor

  • (b) Oppose

  • (c) Neither favor or oppose

• 19. Some states have passed a law that will require state and local police to determine the immigration status of a person if they find that there is a reasonable suspicion that he or she is an undocumented immigrant. Those found to be in the U.S. without permission will have broken state law. From what you have heard, do you favor, oppose, or neither favor nor oppose these immigration laws?

  • (a) Favor

  • (b) Oppose

  • (c) Neither favor or oppose

• 20. Do you think the number of immigrants from foreign countries who are permitted to come to the United States to live should be [increased a lot, increased a little, left the same as it is now, decreased a little, or decreased a lot/decreased a lot, decreased a little, left the same as it is now, increased a little, or increased a lot]? [Randomly reverse order]

  • (a) Increased a lot

  • (b) Increased a little

  • (c) Left the same as it is now

  • (d) Decreased a little

  • (e) Decreased a lot

• 21. Now we’d like to ask you about immigration in recent years. How likely is it that recent immigration levels will take jobs away from people already here -- [extremely likely, very likely, somewhat likely, or not at all likely/not at all likely, somewhat likely, very likely, or extremely likely]? [Randomly reverse order]

  • (a) Extremely

  • (b) Very

  • (c) Somewhat

  • (d) Not at all

Appendix 2. Descriptive Statistics

See Tables 4, 5, 6, 7, 8.

Table 4 Descriptive statistics for selected covariates

Table 5 Proportion of respondents making assumptions about the described policy

Table 6 Cross-tabulation of respondent assumptions about race of beneficiaries (proportions in parentheses)

Table 7 Cross-tabulation of respondent assumptions about nationality of beneficiaries (proportions in parentheses)

Table 8 Cross-tabulation of respondent assumptions that only Blacks and only immigrants would benefit (proportions in parentheses)

See Fig. 4.

Fig. 4

Mean approval rating of described policies, by experimental condition (with 95% confidence intervals)

Appendix 3. Representativeness of Sample

Table 9 displays summary measures for the Mechanical Turk sample on key demographic and political variables as compared to estimates for the US population. US Census Bureau estimates are calculated from the Census “QuickFacts” page (2016 estimates) and the annual population estimates by single year of age and sex for 2016. American National Election Studies (ANES) estimates are 95% confidence intervals calculated using full sample survey weights. Since two of the five questions used to construct the anti-immigration scale are missing from the 2016 ANES, I calculate a three-item version for comparison with the more recent benchmark (also rescaled to range from 0-1). A comparison with the 2012 benchmark for the full five-item version is also included.

Table 9 Demographic and political characteristics of mechanical turk sample compared to estimates of the US population

Appendix 4. Alternative Models of Race and Nationality Assumptions

This appendix displays the results of a series of alternative specifications of the “assumption” models presented in Tables 1 and 2 in the main text. Table 10 adds interaction terms of the treatments with education. Tables 11 and 12 add interaction terms of the treatments with ideology and party identification, respectively. Tables 13, 14, 15, 16 and 17 show the results of multinomial logit models that treat assumptions about each group as separate outcomes.

Table 10 Perceptions that minorities and not majorities benefit from the policy, with education interactions

Table 11 Perceptions that minority and not majority groups benefit from the policy, with ideology interactions

Table 12 Perceptions that minority and not majority groups benefit from the policy, with party identification interactions

Table 13 Experimental effects on race assumptions, multinomial specification (neither as base outcome)

Table 14 Experimental effects on race assumptions with interactions, multinomial specification (neither as base outcome)

Table 15 Experimental effects on nationality assumptions, multinomial specification (neither as base outcome)

Table 16 Experimental effects on nationality assumptions with interactions, multinomial specification (neither as base outcome)

Table 17 Experimental effects on both sets of group assumptions, multinomial specification (neither as base outcome)

Appendix 5. Alternative Specifications of Policy Approval Models

This appendix displays the results of alternative specifications of the OLS regressions of policy approval displayed in Table 3 in the main text.

Tables 18 and 19 displays the results of models that treat the assumptions that minority groups will benefit and that majority groups will benefit as separate variables. Figures 5 and 6 display these results graphically.

Tables 20 and 21 add interactions of the minority group assumptions with ideology and party identification, respectively.

Table 22 incorporates a dummy variable for the assumption that only the Democratic party supports the policy, with interaction terms with the relevant attitudinal variables.

Table 18 Model of policy approval with separate racial group assumptions and interactions

Fig. 5

Predicted policy approval by assumptions about beneficiaries and symbolic racism

Table 19 Model of policy approval with separate nationality assumptions and interactions

Fig. 6

Predicted policy approval by assumptions about beneficiaries and anti-immigration attitudes

Table 20 Policy approval, beneficiary group assumptions, and group attitudes with ideology interactions

Table 21 Policy approval, beneficiary group assumptions, and group attitudes with party identification interactions

Table 22 Model of policy approval incorporating assumptions of party support with interactions

Appendix 6. Analysis of Effects Across Overrepresented and Underrepresented Subgroups

This section graphically presents the results of models using interaction terms of the independent variables of interest in the main analysis with indicators identifying subgroups of respondents that are underrepresented in the sample. These results suggest how the central findings of the study might change if a more representative sample were used. Separate analyses were conducted for age (18 to 34 vs. 35 and up), ideology (liberals vs. moderates and conservatives), party identification (Democrats vs. Republicans and independents), gender (male vs. female), race (non-Hispanic whites vs. nonwhites and Hispanics), education (4-year degree and higher vs. no 4-year degree), and income (less than $50,000 vs. $50,000 and up). For each variable, categories were divided to maximize the size of the group that is underrepresented in the sample.

Effects of the work requirement treatment (collapsing across the tax vs. cash payment conditions) on the assumption that blacks will benefit to the exclusion of whites, and that immigrants will benefit to the exclusion of native-born Americans, are explored in Figs. 7 and 8, respectively. The effects of these assumptions on policy approval in interaction with symbolic racism and the anti-immigration scale are explored in Figs. 9 and 10.

Fig. 7

Effect of work requirement treatment on the assumption that Blacks (not Whites) will benefit from the policy, across subgroups (with 95% confidence intervals)

Fig. 8

Effect of work requirement treatment on the assumption that immigrants (not people born in the US) will benefit from the policy, across subgroups (with 95% confidence intervals)

Fig. 9

Marginal effect of the assumption that Blacks (not Whites) will benefit from the policy on policy approval across the range of the symbolic racism scale and subgroups (with 95% confidence intervals)

Fig. 10

Marginal effect of the assumption that immigrants (not people born in the US) will benefit from the policy on policy approval across the range of the the anti-immigration scale and subgroups (with 95% confidence intervals)

Appendix 7. Replication of Analyses using Survey Weights

Survey weights were generated using the ipfweight command in Stata (Bergmann 2011), based on American Community Survey estimates for race and ethnicity, age, gender, and income of the adult population for 2016. Specifically, the sample was weighted based on the following percentages for specific groups that are overrepresented or underrepresented in the data:

• Age 18 to 34: 30.4%

• Age 55 and up: 35.2%

• Female: 51%

• White, non-Hispanic: 67%

• African-American: 13%

• White Latino/Hispanic (proxy for people who would identify as “Hispanic” in the question): 14.5%

• Four-year degree or higher: 30%

• Income $75,000 and up: 36.8%

As recommended by Bergmann , I limit survey weights to a maximum of 5 for each respondent (Tables 23, 24, and 25).

Table 23 Perceptions that Blacks and not Whites benefit from the policy, with survey weights

Table 24 Perceptions that immigrants and not Americans born in the U.S. benefit from the policy, with survey weights

Table 25 Policy approval, beneficiary group assumptions, and group attitudes, with survey weights

Appendix 8. Mediated Moderation Analyses

The plots displayed in Fig. 3 are derived from a structural equation model and subsequent calculations recommended by Preacher, Rucker and Hayes (2007, see, in particular, Model 3 and Equation 20). These models take the following form:

$$\begin{aligned} assumption= & {} a_{0} + a_{1}workreq\\ approval= & {} b_{0} + b_{1}assumption + b_{2}attitude + b_{3}assumption \times attitude + c' workreq \end{aligned}$$

The first model is a linear probability model, which facilitates the calculation of mediated effects. The indirect or mediated effect of workreq is $\hat{a}_{1}(\hat{b}_{1} + \hat{b}_{3}attitude)$, and is distinct from the direct effect of the treatment on policy approval represented by $\hat{c'}$. The standard error of the mediated effect is calculated with the following equation:

$$\begin{aligned} SE = \sqrt{(\hat{b}_{1} + \hat{b}_{3}attitude)^{2}s_{\hat{a}_{1}}^{2} + (\hat{a}_{1}^{2} + s_{\hat{a}_{1}}^{2})(s_{\hat{b}_{1}}^{2} + s_{\hat{b}_{3},\hat{b}_{3}}+ s_{\hat{b}_{3}}^{2}attitude^{2})} \end{aligned}$$

These are “normal theory” standard errors that rely on an assumption that the product of $\hat{a}_{1}$ $\hat{b}_{1}$ is normally distributed, though this can be relaxed for large samples. Preacher et al. (2007) recommend bootstrapped standard errors to obviate the need for this assumption. As a robustness check, I also ran bootstrapped structural equation models with 10,000 repetitions and compared the standard errors for the effects of the work requirement treatment at the mean values of the attitude variables and one standard deviation above and below the mean. In all cases, the bootstrapped standard errors are virtually identical to those calculated with the normal theory method. The results of the two structural equation models are displayed in Tables 26 and 27.

Table 26 Mediated moderation model of policy approval (work requirement treatment effect mediated by assumption that Blacks, not Whites, will benefit)

Table 27 Mediated moderation model of policy approval (work requirement treatment effect mediated by assumption that immigrants, not people born in the US, will benefit)