Testing Association between racist beliefs and support for immigration

The ESS asked several European citizens whether they thought there were genetic race differences in intelligence and conscientiousness between races as well as their positions on immigration.

Overall, the effect seems to exist (who would’ve thought?) but it’s fairly small in terms of magnitude.

Data cleaning:

There are various questions on immigration: imsmetn - Allow many/few immigrants of same race/ethnic group as majority imdfetn - Allow many/few immigrants of different race/ethnic group from majority impcntr - Allow many/few immigrants from poorer countries outside Europe imbgeco - Immigration bad or good for country’s economy imueclt - Country’s cultural life undermined or enriched by immigrants imwbcnt - Immigrants make country worse or better place to live

The first general factor explained 64% of the variance in scores, and immigrationcomp contains the first principal component of the six questions; higher means less support for immigration.

pca(ess %>% select(imsmetn, imdfetn, impcntr, imbgeco, imueclt, imwbcnt))

I tried four different regression models with different control variables. Age was controlled for with a spline, as the association between age and beliefs about immigration could be nonlinear. Adding control variables did not change the association between believing some races are born more hard working and supporting immigration, but adding country-level controls decreased the relationship between believing some races are born more intelligent than others and support for immigration.

The size of the effects (d = .15 for believing races differ in inborn conscientiousness, (d = .4 for believing races differ in inborn intelligence) is rather small, and suggests that many more factors contribute to beliefs about immigration than just race.

lr <- lm(data=ess, immigrationcomp ~ chered + iqhered)
summary(lr)

Call:
lm(formula = immigrationcomp ~ chered + iqhered, data = ess)

Residuals:
     Min       1Q   Median       3Q      Max 
-2.68118 -0.60925 -0.04438  0.60942  2.60744 

Coefficients:
             Estimate Std. Error t value            Pr(>|t|)    
(Intercept) -0.216908   0.006463  -33.56 <0.0000000000000002 ***
chered       0.156307   0.011128   14.05 <0.0000000000000002 ***
iqhered      0.558152   0.014257   39.15 <0.0000000000000002 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.9229 on 34041 degrees of freedom
  (456511 observations deleted due to missingness)
Multiple R-squared:  0.06954,   Adjusted R-squared:  0.06949 
F-statistic:  1272 on 2 and 34041 DF,  p-value: < 0.00000000000000022

lr <- lm(data=ess, immigrationcomp ~ chered + iqhered + rcs(ag, 6) + gender)
summary(lr)

Call:
lm(formula = immigrationcomp ~ chered + iqhered + rcs(ag, 6) + 
    gender, data = ess)

Residuals:
     Min       1Q   Median       3Q      Max 
-2.83473 -0.61736 -0.05574  0.59719  2.74985 

Coefficients:
                  Estimate Std. Error t value             Pr(>|t|)    
(Intercept)      -0.427218   0.055918  -7.640   0.0000000000000223 ***
chered            0.144499   0.011097  13.021 < 0.0000000000000002 ***
iqhered           0.545094   0.014214  38.348 < 0.0000000000000002 ***
rcs(ag, 6)ag      0.003994   0.002342   1.705             0.088165 .  
rcs(ag, 6)ag'     0.003874   0.021188   0.183             0.854919    
rcs(ag, 6)ag''   -0.054467   0.076091  -0.716             0.474116    
rcs(ag, 6)ag'''   0.208147   0.110639   1.881             0.059937 .  
rcs(ag, 6)ag'''' -0.345748   0.091739  -3.769             0.000164 ***
gender            0.031965   0.009970   3.206             0.001346 ** 
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.9183 on 34022 degrees of freedom
  (456524 observations deleted due to missingness)
Multiple R-squared:  0.07903,   Adjusted R-squared:  0.07881 
F-statistic: 364.9 on 8 and 34022 DF,  p-value: < 0.00000000000000022

lr <- lm(data=ess, immigrationcomp ~ chered + iqhered + cntry)
summary(lr)

Call:
lm(formula = immigrationcomp ~ chered + iqhered + cntry, data = ess)

Residuals:
    Min      1Q  Median      3Q     Max 
-3.2041 -0.5967 -0.0487  0.5657  3.3106 

Coefficients:
             Estimate Std. Error t value             Pr(>|t|)    
(Intercept)  0.005221   0.023445   0.223             0.823786    
chered       0.160134   0.010905  14.685 < 0.0000000000000002 ***
iqhered      0.391966   0.014078  27.842 < 0.0000000000000002 ***
cntryBE     -0.110787   0.031385  -3.530             0.000416 ***
cntryCH     -0.452396   0.033150 -13.647 < 0.0000000000000002 ***
cntryCZ      0.441266   0.031385  14.060 < 0.0000000000000002 ***
cntryDE     -0.552139   0.028375 -19.458 < 0.0000000000000002 ***
cntryDK     -0.318061   0.033163  -9.591 < 0.0000000000000002 ***
cntryEE     -0.134618   0.031562  -4.265 0.000020022971433255 ***
cntryES     -0.283543   0.032265  -8.788 < 0.0000000000000002 ***
cntryFI     -0.300626   0.030610  -9.821 < 0.0000000000000002 ***
cntryFR     -0.134189   0.031198  -4.301 0.000017034245079123 ***
cntryGB     -0.004134   0.030099  -0.137             0.890753    
cntryHU      0.463142   0.034097  13.583 < 0.0000000000000002 ***
cntryIE     -0.147314   0.030525  -4.826 0.000001398744217534 ***
cntryIL     -0.055600   0.030526  -1.821             0.068558 .  
cntryLT     -0.085056   0.032382  -2.627             0.008626 ** 
cntryNL     -0.200274   0.031344  -6.389 0.000000000168627686 ***
cntryNO     -0.448276   0.033101 -13.543 < 0.0000000000000002 ***
cntryPL     -0.285127   0.034717  -8.213 0.000000000000000224 ***
cntryPT     -0.194751   0.035264  -5.523 0.000000033639296901 ***
cntrySE     -0.925287   0.031831 -29.069 < 0.0000000000000002 ***
cntrySI     -0.157808   0.036765  -4.292 0.000017725465750439 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.8773 on 34021 degrees of freedom
  (456511 observations deleted due to missingness)
Multiple R-squared:  0.1597,    Adjusted R-squared:  0.1592 
F-statistic:   294 on 22 and 34021 DF,  p-value: < 0.00000000000000022

lr <- lm(data=ess, immigrationcomp ~ chered + iqhered + rcs(ag, 6) + gender + cntry)
summary(lr)

Call:
lm(formula = immigrationcomp ~ chered + iqhered + rcs(ag, 6) + 
    gender + cntry, data = ess)

Residuals:
    Min      1Q  Median      3Q     Max 
-3.3533 -0.5900 -0.0533  0.5543  3.2272 

Coefficients:
                   Estimate Std. Error t value             Pr(>|t|)    
(Intercept)      -0.1105104  0.0582321  -1.898              0.05774 .  
chered            0.1472817  0.0108632  13.558 < 0.0000000000000002 ***
iqhered           0.3762649  0.0140307  26.817 < 0.0000000000000002 ***
rcs(ag, 6)ag      0.0005651  0.0022287   0.254              0.79983    
rcs(ag, 6)ag'     0.0226291  0.0201491   1.123              0.26141    
rcs(ag, 6)ag''   -0.0970479  0.0723378  -1.342              0.17974    
rcs(ag, 6)ag'''   0.2351898  0.1051538   2.237              0.02532 *  
rcs(ag, 6)ag'''' -0.3564956  0.0871901  -4.089  0.00004347598826855 ***
gender            0.0086630  0.0094924   0.913              0.36144    
cntryBE          -0.1013503  0.0312214  -3.246              0.00117 ** 
cntryCH          -0.4436614  0.0329681 -13.457 < 0.0000000000000002 ***
cntryCZ           0.4552008  0.0312647  14.560 < 0.0000000000000002 ***
cntryDE          -0.5575167  0.0282213 -19.755 < 0.0000000000000002 ***
cntryDK          -0.3136684  0.0329807  -9.511 < 0.0000000000000002 ***
cntryEE          -0.1328800  0.0313852  -4.234  0.00002303367414700 ***
cntryES          -0.2695716  0.0320850  -8.402 < 0.0000000000000002 ***
cntryFI          -0.3092109  0.0304441 -10.157 < 0.0000000000000002 ***
cntryFR          -0.1369424  0.0310146  -4.415  0.00001011279206558 ***
cntryGB          -0.0203214  0.0299347  -0.679              0.49723    
cntryHU           0.4599401  0.0339026  13.567 < 0.0000000000000002 ***
cntryIE          -0.1495776  0.0303438  -4.929  0.00000082855433542 ***
cntryIL          -0.0513958  0.0303663  -1.693              0.09055 .  
cntryLT          -0.0847840  0.0322155  -2.632              0.00850 ** 
cntryNL          -0.2145085  0.0311715  -6.882  0.00000000000602233 ***
cntryNO          -0.4433238  0.0329289 -13.463 < 0.0000000000000002 ***
cntryPL          -0.2703179  0.0345282  -7.829  0.00000000000000506 ***
cntryPT          -0.2051049  0.0350646  -5.849  0.00000000498030941 ***
cntrySE          -0.9336404  0.0316575 -29.492 < 0.0000000000000002 ***
cntrySI          -0.1530790  0.0365585  -4.187  0.00002830827338719 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.872 on 34002 degrees of freedom
  (456524 observations deleted due to missingness)
Multiple R-squared:  0.1699,    Adjusted R-squared:  0.1692 
F-statistic: 248.5 on 28 and 34002 DF,  p-value: < 0.00000000000000022

The relationship, even when no control variables are added, is rather weak; there’s much more overlap between the two groups than

GG_denhist(ess, var='immigrationcomp', group='iqhered')

Warning: Grouping variable contained missing values. These were removed. If you want an NA group, convert to explicit value.Warning: There were groups without any data. These were removed

One could argue that the relationship is biased downwards by acquiesence; some people may still think that races differ in genetic conscientiousness/intelligence but not admit to it on a survey. While it’s unrealistic to expect survey respondents to tell the truth 100% of the time, I don’t think this is a massive issue here. The percentage of people who claim that races differ genetically in intelligence is 18%, while the percentage who claim that races differ genetically in conscientiousness is 40%. Both of the averages seem plausible to me.