Home › Archive for Race › Page 6

Category: Race

By Posted on Posted in Race 4 Comments

The Chudy 2021 Journal of Politics article "Racial Sympathy and Its Political Consequences" concerns White racial sympathy for Blacks.

More than a decade ago, Hutchings 2009 reported evidence about White racial sympathy for Blacks. Below is a table from Hutchings 2009 indicating that, among White liberals and White conservatives, sympathy for Blacks predicted at p<0.05 support for government policies explicitly intended to benefit Blacks such as government aid to Blacks, controlling for factors such as anti-Black stereotypes:

Chudy 2021 thanked Vincent Hutchings in the acknowledgments, and Vincent Hutchings is listed as co-chair of Jennifer Chudy's "Racial Sympathy in American Politics" dissertation. But see whether you can find in the Chudy 2021 JOP article an indication that Hutchings 2009 had reported evidence that White racial sympathy for Blacks predicted support for government policies explicitly intended to benefit Blacks.

Here is a passage from Chudy 2021 referencing Hutchings 2009:

I start by examining white support for "government aid to blacks," a broad policy area that has appeared on the ANES since the 1970s. The question asks respondents to place themselves on a 7-point scale that ranges from "Blacks Should Help Themselves" to "Government Should Help Blacks." Previous research on this question has found that racial animus leads some whites to oppose government aid to African Americans (Hutchings 2009). This analysis examines whether racial sympathy leads some white Americans to offer support for this contentious policy area.

I think that the above passages can be reasonably read as suggesting an incorrect claim that the Hutchings 2009 "previous research on this question" did not examine "whether racial sympathy leads some white Americans to offer support for this contentious policy area [of government aid to African Americans]".

---

NOTES:

1. Chudy 2021 reported results from an experiment that varied the race of a target culprit and asked participants to recommend a punishment. Chudy 2021 Figure 2 plotted estimates of recommended punishments at different levels of racial sympathy.

The Chudy 2021 analysis used a linear regression, which produced an estimated difference by race on a 0-to-100 scale of -22 at the lowest level of racial sympathy and of 41 at the highest level of racial sympathy. These differences can be seen in my plot below to the left, with a racial sympathy index coded from 0 through 16.

However, a linear relationship might not be a correct presumption. The plot to the right reports estimates calculated at each level of the racial sympathy index, so that the estimate at the highest level of racial sympathy is not influenced by cases at other levels of racial sympathy.

2. Chudy 2021 Figure 2 plots results from Chudy 2021 Table 5, but using a reversed outcome variable for some reason.

3. Chudy 2021 used the term "predicted probability" to discuss the Figure 2 / Table 5 results, but these results are predicted levels of an outcome variable that had eight levels, from "0-10 hours" to "over 70 hours" (see the bottom of the final page in the Chudy 2021 supplemental web appendix).

4. The bias detected in this experiment across all levels of racial sympathy was 13 units on a 0-to-100 scale, disfavoring the White culprit relative to the Black culprit (p=0.01) [svy: reg commservice whiteblackculprit].

5. Code for my analyses.

Tagged with:

By Posted on Posted in Race No Comments

The Journal of Race, Ethnicity, and Politics published Buyuker et al 2020: "Race politics research and the American presidency: thinking about white attitudes, identities and vote choice in the Trump era and beyond".

Table 2 of Buyuker et al 2020 reported regressions predicting Whites' projected and recalled vote for Donald Trump over Hillary Clinton in the 2016 U.S. presidential election, using predictors such as White identity, racial resentment, xenophobia, and sexism. Xenophobia placed into the top tier of predictors, with an estimated maximum effect of 88 percentage points going from the lowest to the highest value of the predictor, and racial resentment placed into the second tier, with an estimated maximum effect of 58 percentage points.

I was interested in whether this difference is at least partly due to how well each predictor was measured. Here are characteristics of the predictors among Whites, which indicate that xenophobia was measured at a much more granular level than racial resentment was:

RACIAL RESENTMENT
4 items
White participants fell into 22 unique levels
4% of Whites at the lowest level of racial resentment
9% of Whites at the highest level of racial resentment

XENOPHOBIA
10 items
White participants fell into 1,096 unique levels
1% of Whites at the lowest level of xenophobia
1% of Whites at the highest level of xenophobia

So it's at least plausible from the above results that xenophobia might have outperformed racial resentment merely because the measurement of xenophobia was better than the measurement of racial resentment.

---

Racial resentment was measured with four items that each had five response options, so I created a reduced xenophobia predictor using the four xenophobia items that each had exactly five response options; these items were about desired immigration levels and agreement or disagreement with statements that "Immigrants are generally good for America's economy", "America's culture is generally harmed by immigrants", and "Immigrants increase crime rates in the United States".

I re-estimated the Buyuker et al 2020 Table 2 model replacing the original xenophobia predictor with the reduced xenophobia predictor: the maximum effect for xenophobia (66 percentage points) was similar to the maximum effect for racial resentment (66 percentage points).

---

Among Whites, vote choice correlated between r=0.50 and r=0.58 with each of the four racial resentment items and between r=0.39 and r=0.56 with nine of the ten xenophobia items. The exception was the seven-point item that measured attitudes about building a wall on the U.S. border with Mexico, which correlated with vote choice at r=0.72.

Replacing the desired immigration levels item in the reduced xenophobia predictor with the border wall item produced a larger estimated maximum effect for xenophobia (85 percentage points) than for racial resentment (60 percentage points). Removing all predictors from the model except for xenophobia and racial resentment, the reduced xenophobia predictor with the border wall item still produced a larger estimated maximum effect than did racial resentment: 90 percentage points, compared to 74 percentage points.

But the larger effect for xenophobia is not completely attributable to the border wall item: using a predictor that combined the other nine xenophobia items produced a maximum effect for xenophobia (80 percentage points) that was larger than the maximum effect for racial resentment (63 percentage points).

---

I think that the main takeaway from this post is that, when comparing the estimated effect of predictors, inferences can depend on how well each predictor is measured, so such analyses should discuss the quality of the predictors. Imbalances in which participants fall into 22 levels for one predictor and 1,096 levels for another predictor seem to be biased in favor of the more granular predictor, all else equal.

Moreover, I think that, for predicting 2016 U.S. presidential vote choice, it's at least debatable whether a xenophobia predictor should include an item about a border wall with Mexico, because including that item means that, instead of xenophobia measuring attitudes about immigrants per se, the xenophobia predictor conflates these attitudes with attitudes about a policy proposal that is very closely connected with Donald Trump.

---

It's not ideal to use regression to predict maximum effects, so I estimated a model using only the racial resentment predictor and the reduced four-item xenophobia predictor with the border wall item, but including a predictor for each level of the predictors. That model predicted failure perfectly for some levels of the predictors, so I recoded the predictors until those errors were eliminated, which involved combining the three lowest racial resentment levels (so that racial resentment ran from 2 through 16) and combining the 21st and 22nd levels of the xenophobia predictor (so that xenophobia ran from 0 through 23). In a model with only those two recoded predictors, the estimated maximum effects were 81 percentage points for xenophobia and 76 percentage points for racial resentment. Using all Buyuker et al 2020 predictors, the respective percentage points were 65 and 63.

---

I then predicted Trump/Clinton vote choice using only the 22-level racial resentment predictor and the full 1,096-level xenophobia predictor, but placing the values of the predictors into ten levels; the original scale for the predictors ran from 0 through 1, and, for the 10-level predictors, the first level for each predictor was from 0 to 0.1, a second level was from above 0.1 to 0.2, and a tenth level was from above 0.9 to 1. Using these predictors as regular predictors without "factor" notation, the gap in maximum effects was about 24 percentage points, favoring xenophobia. But using these predictors with "factor" notation, the gap favoring xenophobia fell to about 9.5 percentage points.

Plots below illustrate the difference in predictions for xenophobia: the left panel uses a regular 10-level xenophobia predictor, and the right panel uses each of the 10 levels of that predictor as a separate predictor.

---

So I'm not sure that these data support the inference that xenophobia is in a higher tier than racial resentment, for predicting Trump/Clinton vote in 2016. The above analyses seem to suggest that much or all of the advantage for xenophobia over racial resentment in the Buyuker et al 2020 analyses was due to model assumptions and/or better measurement of xenophobia.

---

Another concern about Buyuker et al 2020 is with the measurement of predictors such as xenophobia. The xenophobia predictor is more accurately described as something such as attitudes about immigrants. If some participants are more favorable toward immigrants than toward natives, and if these participants locate themselves at low levels of the xenophobia predictor, then the effect of xenophilia among these participants is possibly being added to the effect of xenophobia.

Concerns are similar for predictors such as racial resentment and sexism. See here and here for evidence that low levels of similar predictors associate with bias in the opposite direction.

---

NOTES

1. Thanks to Beyza Buyuker for sending me replication materials for Buyuker et al 2020.

2. Stata code for my analyses. Stata output for my analyses.

3. ANES 2016 citations:

The American National Election Studies (ANES). 2016. ANES 2012 Time Series Study. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2016-05-17. https://doi.org/10.3886/ICPSR35157.v1.

ANES. 2017. "User's Guide and Codebook for the ANES 2016 Time Series Study". Ann Arbor, MI, and Palo Alto, CA: The University of Michigan and Stanford University.

Tagged with: , , , ,

By Posted on Posted in Race No Comments

Political Research Quarterly published Garcia and Stout 2020 "Responding to Racial Resentment: How Racial Resentment Influences Legislative Behavior". The article abstract indicates (emphasis added):

Through an automated content analysis of more than fifty four thousand press releases from almost four hundred U.S. House members in the 114th Congress (2015–2017), we show that Republicans from districts with high levels of racial resentment are more likely to issue press releases that attack President Barack Obama. In contrast, we find no evidence of racial resentment being positively associated with another prominent Democratic white elected official, Hillary Clinton. Our results suggest that one reason Congress may remain racially conservative even as representatives' cycle out of office may be attributed to the electoral process.

---

Racial resentment conflates racial attitudes and political ideology, apparently even when controlling for factors such as partisanship and political ideology, so comparing how district racial resentment predicts the percentage of press releases attacking Barack Obama to how district racial resentment predicts the percentage of press releases attacking Hillary Clinton is a useful way to assess whether any association of racial resentment is due to the racial component of district racial resentment. But that comparison should involve a statistical test of whether the coefficient for district racial resentment in the Obama models differs from the coefficient for district racial resentment in the Clinton models. And my analyses indicate that, for results reported in the article, these coefficients don't differ at p<0.20.

My analyses indicated that the p-value is p=0.23 for a test of whether the coefficient on district racial resentment in an Obama model differs from the coefficient on district racial resentment in a Clinton model, using only a predictor of weighted district racial resentment and limiting the sample to Republican representatives. The p-value is about p=0.33 for a test comparing the key interaction coefficients in Models C and D in Garcia and Stout 2020 Table 1 (see the plot below).

---

The Garcia and Stout 2020 abstract's claim that "…we find no evidence of racial resentment being positively associated with another prominent Democratic white elected official, Hillary Clinton" (p. 812) is contradicted in the main text of Garcia and Stout 2020:

Pearson's R for the relationship between the unweighted (.16) and the weighted (.14) district's racial resentment score and Republicans issuing of negative-Clinton press releases are statistically significant at .05.

I think that the "no evidence" claim refers to the lack of statistical significance for the Clinton models in Table 1 when adding statistical control, but the coefficient/standard error ratio is about 1.3 for the key coefficient for Clinton in Table 1 Model D, so that's some evidence. Adding "cluster(robust)" to the regression specification increases this t-statistic to 1.78, which is not no evidence. And then removing the control for candidate margin of victory gets the p-value under p=0.05.

---

For the outcome variable codings of the percentage of press releases attacking Obama and the percentage of press releases attacking Clinton, 53% and 76% of the observations are zero, respectively. The outcome is a percentage, so I re-estimated the models using fractional logistic regression. As indicated in the output, the p-value for the interaction coefficient did not fall under p=0.15 in the Clinton models mentioned above or in the Obama models mentioned above.

---

The key Table 1 coefficient is an interaction term that involves district racial resentment and the political party of the representative, but the abstract claims are limited to Republican representatives. I estimated the Table 1 models limited to Republican representatives: the p-value for racial resentment did not fall under p=0.80 for the Obama models. The p-value for racial resentment did not fall under p=0.40 for the Obama models limited to non-Republican representatives.

---

So, in the fractional regression models discussed above, the key Table 1 interaction coefficient did not have a p-value under p=0.15; in the linear regression models discussed above with statistical control, district racial resentment did not predict at p<0.80 among Republican representatives the percentage of press releases attacking Obama; and in the linear regression models discussed above, the association of district racial resentment and the percentage of press releases attacking Obama did not differ at p<0.20 from the association of district racial resentment and the percentage of press releases attacking Clinton.

---

NOTES

1. Thanks to Jennifer R. Garcia for sending me data for the article.

2. Results reported in the post are for the weighted models, but the Stata output contains results for unweighted models, in which the inferences or a lack of inferences are the same or similar.

3. Stata code. Stata output. R code for the plot.

4. For what it's worth, Republican members of Congress from districts with relatively low levels of racial resentment were more likely to issue press releases that attacked Obama than to issue press releases that attacked Clinton, measuring low district racial resentment as the bottom 10% of GOP districts by racial resentment; the same pattern held for Republican members of Congress from districts with relatively high levels of racial resentment, measured as the top 10% of GOP districts by racial resentment. Stata code for this analysis. Stata output for this analysis.

Tagged with: , ,

By Posted on Posted in Race, Sex No Comments

Participants in studies reported on in Regina Bateson's 2020 Perspectives on Politics article "Strategic Discrimination" were asked to indicate the percentage of other Americans that the participant thought would not vote for a woman for president and the percentage of other Americans that the participant thought would not vote for a black person for president.

Bateson 2020 Figure 1 reports that, in the nationally representative Study 1 sample, mean participant estimates were that 47% of other Americans would not vote for a woman for president and that 42% of other Americans would not vote for a black person for president. I was interested in the distribution of responses, so I plotted in the histograms below participant estimates to these items, using the Bateson 2020 data for Study 1.

This first set of histograms is for all participants:

This second set of histograms is for only participants who passed the attention check:

---

I was also interested in estimates from participsnts with a graduate degree, given that so many people in political science have a graduate degree. Bateson 2020 Appendix Table 1.33 indicates that, among participants with a graduate degree, estimates were that 58.3% of other Americans would not vote for a woman for president and that 56.6% of other Americans would not vote for a black person for president.

But these estimates differ depending on whether the participant correctly responded to the attention check item: for the item about the percentage of other Americans who would not vote for a woman for president, the mean estimate was 47% [42, 52] for the 84 graduate degree participants who correctly responded to the attention check and was 68% [63, 73] for the 97 graduate degree participants who did not correctly respond to the attention check; for the item about the percentage of other Americans who would not vote for a black person for president, respective estimates were 44% [39, 49] and 67% [62, 73].

Participants who reported having a graduate degree were 20 percentage points more likely to fail the attention check than participants who did not report having a graduate degree, p<0.001.

---

These data were collected in May 2019, after Barack Obama had been elected president twice and after Hillary Clinton won the popular vote for president, and each aforementioned mean estimate seems to be a substantial overestimate of discrimination against women presidential candidates and Black presidential candidates, compared to point estimates from relevant list experiments reported in Carmines and Schmidt 2020 and compared to point estimates from list experiments and direct questions cited in Bateson 2020 footnote 8.

---

NOTES

1. Stata code for my analysis.

2. R code for the first histogram.

Tagged with: , ,

By Posted on Posted in Race No Comments

Study 1 of Strickler and Lawson 2020 "Racial conservatism, self-monitoring, and perceptions of police violence" in Politics, Groups, and Identities was an experiment in which participants rated how justified a police shooting was. The experiment had a control condition, a "stereotype" condition in which the officer was White and the suspect Black, and a "counterstereotype" condition in which the officer was Black and the suspect White.

The article indicates that:

And while racial resentment did not moderate how whites responded to treatment in the White Officer/Black Victim condition, it did impact response to treatment in the Black Officer/White Victim condition. As Table 3 and Figure 4 demonstrate, for whites, those with higher levels of racial resentment are significantly less likely to view shooting as justified if it involves a black officer and a white victim.

However, the 95% confidence interval in the aforementioned Figure 4 crosses zero at high levels of racial resentment. I emailed lead author Ryan Strickler for the data and code, which he provided.

---

Instead of using a regression to estimate the outcome at higher levels of racial resentment, I'll estimate the outcome for only participants at given ranges of racial resentment (see Hainmueller et al. 2019). This way, inferences about particular groups are based on data for only those groups.

Plots below report point estimates and 95% confidence intervals from tests comparing the outcome across conditions, at various ranges of racial resentment, among all White respondents or among Whites who responded correctly to manipulation checks about the officer's race and the suspect's race. Racial resentment was coded from 1 through 17.

The outcome for the first four plots was whether the participant indicated that the officer's actions were justified.

In the top left plot, the top estimate is for White participants at the highest observed level of racial resentment. The estimate is positive 0.06, which indicates that high racial resentment participants in the stereotypic condition were 6 percentage points more likely to rate the shooting as justified, compared to high racial resentment participants in the counterstereotypic condition; however, the 95% confidence interval crosses zero. The next lower estimate compared outcomes for White participants at a racial resentment of 16 and 17. The bottom estimate (RR>=1) is for all White participants, and the negative point estimate for this bottom estimate indicates that White participants in the counterstereotypic shooting condition were more likely to rate the shooting as justified, compared to White participants in the stereotypic shooting condition.

The evidence for bias among Whites high in racial resentment is a bit stronger in the right panels, which compared the counterstereotypic condition to the control condition, but the 95% confidence intervals still overlap zero. There is an exception among White participants who scored 14 or higher on the racial resentment scale, when excluding participants who did not pass the post-treatment manipulation check, but it's not a good idea to exclude participants after the treatment.

---

Tables in the main text of Strickler and Lawson 2020 reported results for a dichotomous outcome coded 1 if, for the first item of the branching, the respondent indicated that the officer's actions were justified. But tables in the appendix used ratings of the extent to which the shooting was justified, measured using branched items that placed respondents into nine levels, from "a great deal certain" that the shooting was not justified to "a great deal certain" that the shooting was justified.

The plots below report results from tests that compared conditions for this ordinal measure of justification, placed on a 0-to-1 scale. Evidence in the right panel is a bit stronger using this outcome, compared to the dichotomous outcome. Like before, the top estimate is for White participants at the highest observed level of racial resentment. Middle estimates (RR>=1 and RR<=17) are for all Whites; below that, estimates are for more extreme levels of low racial resentment, ending with RR==1, for White participants at the lowest observed level of racial resentment.

Results for Whites who passed the manipulation checks are in the output file.

---

NOTES

1. Thanks to Ryan Strickler for sending me data and code for the article.

2. Stata code and R code for my analyses. Data for the first four plots. Data for the final two plots.

Tagged with: ,

By Posted on Posted in Race No Comments

The American National Election Studies Time Series Cumulative Data File (1948-2016) contains data for feeling thermometer measures for Whites and for Blacks, collected in face-to-face or telephone interviews, for each U.S. presidential election year from 1964 to 2016.

Feeling thermometers range from 0 to 100, with higher values indicating warmer or more favorable feelings about a group. The ANES Cumulative Data File and some early individual year ANES Time Series files collapse responses of 97 through 100 into a response of 97. This means that a respondent who selected 97 for Whites and 100 for Blacks would have the same "difference" value as a respondent who selected 100 for Whites and 97 for Blacks. Therefore, I placed respondents with a substantive value for the feeling thermometer about Whites and the feeling thermometer about Blacks into one of three categories:

  • rated Whites more than 3 units above Blacks
  • rated Whites within 3 units of Blacks, and
  • rated Blacks more than 3 units above Whites.

Abrajano and Alvarez (2019) reported evidence from ANES Time Series Studies that responses to racial feeling thermometers differed between the non-internet mode and the internet mode, so my reported results do not include results from the internet mode, which do not go back to 1964.

Below is a plot of how Whites Americans (left) and Black Americans (right) fell into each of the three categories, not including the respondents in the cumulative data file who did not report a substantive response to the items, which ranged from 1% to 8% (see the Notes). Documentation for the cumulative data file indicated that in 1964 and 1968 a response was recorded as 50 for a "don't know" response or if the participant indicated that the participant did not know too much about a group.

---

The plot below indicates how these thermometer ratings associated with two-party vote choice, among White participants:

The right panel indicates a steep drop in two-party vote for the Republican presidential candidate among Whites who rated Blacks more than 3 units higher than Whites, which seems to be consistent with evidence of a "Great Awokening" (see, e.g., Yglesias 2019 and Goldberg 2019, and this image linked to in Goldberg 2019).

---

The plot below is the plot above, but with columns grouped by year:

---

NOTES

1. Percentage non-responses to one or both thermometer items, by year: 3% (1964), 4% (1968), 8% (1972), 5% (1976), 5% (1980), 7% (1984), 5% (1988), 4% (1992), 4% (1996), 8% (2000), 3% (2004), 3% (2008), 1% (2012), 2% (2016).

2. Code for my analyses and black-and-white plots.

3. Feeling thermometer ratings about Chicanos/Hispanics and about Asians are not available in ANES Time Series Cumulative Data File until 1976 and 1992, respectively.

4. A color version of the first plot, for comparison:

5. A color version with a black line divider:

Tagged with:

By Posted on Posted in Race No Comments

The Adida et al. 2020 PS: Political Science & Politics article "Broadening the PhD Pipeline: A Summer Research Program for HBCU Students" claimed that (p. 727):

The US academy today is overwhelmingly white, with only 8% to 9% of full-time science and engineering faculty as underrepresented minorities (DePass and Chubin 2008, 6).

This evidence to support the claim that the "US academy" is "overwhelmingly white" is the percentage of a *subset* of the U.S. academy (science and engineering) that is not White *and not Asian*, given that Asians were not considered underrepresented minorities in the calculation of the percentage. Moreover, the cited publication is more than a decade old, and the data might be even older than that.

Below is a plot of data from 2018, for the U.S. academy as a whole, of data from the National Center for Educational Statistics. The light areas indicate the percentage White for each rank and overall, compared to the total White, Black, Hispanic, Asian, Pacific Islander, and persons of two or more races; the percentage does not include persons with an unknown race/ethnicity and does not include non-resident aliens.

Overall, in Fall 2018, about 76% of U.S. full time faculty at U.S. degree-granting postsecondary institutions were White, which matches a calculation in this Pew study or Fall 2017. So, if you randomly selected four of these faculty, one of them would be expected to be non-White. I'm not sure whether that counts as being "overwhelmingly white".

---

NOTES

1. R code for the plot.

Tagged with: ,