Here's part of the abstract from Rios Morrison and Chung 2011, published in the Journal of Experimental Social Psychology:

In both studies, nonminority participants were randomly assigned to mark their race/ethnicity as either "White" or "European American" on a demographic survey, before answering questions about their interethnic attitudes. Results demonstrated that nonminorities primed to think of themselves as White (versus European American) were subsequently less supportive of multiculturalism and more racially prejudiced, due to decreases in identification with ethnic minorities.

So asking white respondents to select their race/ethnicity as "European American" instead of "White" influenced whites' attitudes toward and about ethnic minorities. The final sample for study 1 was a convenience sample of 77 self-identified whites and 52 non-whites, and the final sample for study 2 was 111 white undergraduates.

Like I wrote before, if you're thinking that it would be interesting to see whether results hold in a nationally representative sample with a large sample size, well, that was tried, with a survey experiment as part of the Time Sharing Experiments in the Social Sciences. Here are the results:

mc2011reanalysis

I'm mentioning these results again because in October 2014 the journal that published Rios Morrison and Chung 2011 desk rejected the manuscript that I submitted describing these results. So you can read in the Journal of Experimental Social Psychology about results for the low-powered test on convenience samples for the "European American" versus "White" self-identification hypothesis, but you won't be able to read in the JESP about results when that hypothesis was tested with a higher-powered test on a nationally-representative sample with data collected by a disinterested third party.

I submitted a revision of the manuscript to Social Psychological and Personality Science, which extended a revise-and-resubmit offer conditional on inclusion of a replication of the TESS experiment. I planned to conduct an experiment with an MTurk sample, but I eventually declined the revise-and-resubmit opportunity for various reasons.

The most recent version of the manuscript is here. Links to data and code.

Tagged with: , , , , , ,

In the Political Behavior article, "The Public's Anger: White Racial Attitudes and Opinions Toward Health Care Reform", Antoine J. Banks presented evidence that "anger uniquely pushes racial conservatives to be more opposing of health care reform while it triggers more support among racial liberals" (p. 493). Here is how the outcome variable was measured in the article's reported analysis (p. 511):

Health Care Reform is a dummy variable recoded 0-1 with 1 equals opposition to reform. The specific item is "As of right now, do you favor or oppose Barack Obama and the Democrats' Health Care reform bill". The response options were yes = I favor the health care bill or no = I oppose the health care bill.

However, the questionnaire for the study indicates that there were multiple items used to measure opinions of health care reform:

W2_1. Do you approve or disapprove of the way Barack Obama is handling Health Care? Please indicate whether you approve strongly, approve somewhat, neither approve nor disapprove, disapprove somewhat, or disapprove strongly.

W2_2. As of right now, do you favor or oppose Barack Obama and the Democrats' Health Care reform bill?

[if "favor" on W2_2] W2_2a. Do you favor Barack Obama and the Democrats' Health Care reform bill very strongly, or not so strongly?

[if "oppose" on W2_2] W2_2b. Do you oppose Barack Obama and the Democrats' Health Care reform bill very strongly, or not so strongly?

The bold item above is the only item reported on as an outcome variable in the article. The reported analysis omitted results for one outcome variable (W2_1) and reported dichotomous results for the other outcome variable (W2_2) for which the apparent intention was to have a four-pronged outcome variable from oppose strongly to favor strongly.

---

Here is the manuscript that I submitted to Political Behavior in March 2015 describing the results using the presumed intended outcome variables and a straightforward research design (e.g., no political discussion control, no exclusion of cases, cases from all conditions analyzed at the same time). Here's the main part of the main figure:

Banks2014Reproduction

The takeaway is that, with regard to opposition to health care reform, the effect of the fear condition on symbolic racism differed at a statistically significant level from the effect of the baseline relaxed condition on symbolic racism; however, contra Banks 2014, the effect of anger on symbolic racism did not differ at a statistically significant level from the effect of the relaxed condition on symbolic racism. The anger condition had a positive effect on symbolic racism, but it was not a unique influence.

The submission to Political Behavior was rejected after peer review. Comments suggested analyzing the presumed intended outcome variables while using the research design choices in Banks 2014. Using the model in Table 2 column 1 of Banks 2014, the fear interaction term and the fear condition term are statistically significant at p<0.05 for predicting the two previously-unreported non-dichotomous outcome variables and for predicting the scale of these two variables; the anger interaction term and the anger condition term are statistically significant at p<0.05 for predicting two of these three outcome variables, with p-values for the residual "Obama handling" outcome variable at roughly 0.10. The revised manuscript describing these results is here.

---

Data are here, and code for the initial submission is here.

---

Antoine Banks has published several studies on anger and racial politics (here, for example) that should be considered when making inferences about the substance of the effect of anger on racial attitudes. Banks had a similar article published in the AJPS, with Nicholas Valentino. Data for that article are here. I did not see any problems with that analysis, but I didn't look very hard, because the posted data were not the raw data: the posted data that I checked omitted, for example, the variables used to construct the outcome variable.

Tagged with: , , , , , , ,

Timofey Pnin linked to an Alice Eagly article that mentioned these two meta-analyses:

  • van Dijk et al. 2012 "Defying Conventional Wisdom: A Meta-Analytical Examination of the Differences between Demographic and Job-Related Diversity Relationships with Performance"
  • Post and Bryon 2015 "Women on Boards and Firm Financial Performance: A Meta-Analysis"

I wanted to check for funnel plot asymmetry in the set of studies in these meta-analyses, so I emailed coauthors of the articles. Hans van Dijk and Kris Byron were kind enough to send data.

The funnel plot for the 612 effect sizes in the van Dijk et al. 2012 meta-analysis is below. The second funnel plot below is a close-up of the bottom of the full funnel plot, limited to studies with fewer than 600 teams. The funnel plot is remarkably symmetric.

FP1

FP2

The funnel plots below are for the Post and Byron 2015 meta-analysis, with the full set of studies in the top funnel plot and, below the full funnel plot, a close-up of the studies with a standard error less than 0.4. The funnel plot is reasonably symmetric.

FP3

FP4

UPDATE (Apr 13, 2016):

More funnel plots from van Dijk et al. 2012.

Sample restricted to age diversity (DIV TYPE=1):

vDe - Age Diversity (1)

Sample restricted to race and ethnic diversity (DIV TYPE=2):

vDe - Race Ethnic Diversity (2)

Sample restricted to sex diversity (DIV TYPE=5):

vDe - Sex Diversity (5)

Sample restricted to education diversity (DIV TYPE=6):

vDe - Education Diversity (6)

Tagged with: , ,

The post is here.

Data are here for the 2016 ANES pilot study and here for the 2012 ANES time series study.

Stata code for the 2016 ANES pilot study analysis is here.

Stata code for the 2012 ANES time series study analysis is here.

Note that the use of blacks, Hispanics, gays and lesbians, feminists, transgender persons, and Muslims as the reference groups in the pilot study follows the use in the previous Monkey Cage post in white ethnocentrism.

---

To try to untangle the influence of attitudes about whites from attitudes about blacks, Hispanics, and Asians in predicting policy preferences, models were estimated for white respondents with the feeling thermometers for whites, blacks, Hispanics, and Asians entered separately, along with controls for sex, marital status, age, education, and household income. The feeling thermometers and the outcome variables were kept on an interval scale and were standardized. The data were weighted and were from the 2012 ANES time series study.

For support for more immigration, the point estimate for the standardized correlation of the Hispanic feeling thermometer was 0.22, and the point estimate for the standardized correlation of the white feeling thermometer was -0.24, indicating that attitudes about whites had roughly the same correlation as did attitudes about Hispanics: the more warm a white person feels toward Hispanics, the more immigration that person supports on average; and the more warm a white person feels toward whites, the less immigration that person supports on average.

For support for affirmative action for black students in university admissions, the point estimate for the standardized correlation of the black feeling thermometer was 0.28, and the point estimate for the standardized correlation of the white feeling thermometer was -0.12, indicating that attitudes about whites had roughly half of the correlation as did attitudes about blacks: the more warm a white person feels toward blacks, the more that person supports affirmative action in university admissions on average; and the more warm a white person feels toward whites, the less that person supports affirmative action in university admissions on average.

Stata code for the above analysis is here.

---

UPDATE (March 8, 2016)

Nathan Kalmoe asked about the results without the atypical groups for the white ethnocentrism measure. Results are below. Note that the phrase "oikophobic" refers to scores less than zero on an ethnocentrism scale, after the usage of Roger Scruton.

WHITES COMPARED TO ONLY BLACKS, HISPANICS, AND MUSLIMS

The percentage of white and nonwhite respondents on the negative side of the ethnocentrism scale is 21% and 48% (compared to 22% and 39% using the full set of six reference groups).

The range of candidate feeling thermometer scores for oikophobic whites is: Trump 22, Cruz 27, Carson 34, Fiorina 32, J Bush 33, Rubio 36, H Clinton 49, Obama 61, and Sanders 67.

The mean candidate feeling thermometer scores for nonwhites scoring less than zero are: Trump 24, Cruz 32, Carson 31, Fiorina 31, J Bush 32, Rubio 37, H Clinton 62, Obama 75, and Sanders 54.

The confidence intervals for oikophobic and ethnocentric whites on the black affirmative action item do not overlap and have a standardized difference of 0.55. The confidence intervals for oikophobic and ethnocentric whites on the immigration item do not overlap and have a standardized difference of 0.67.

WHITES COMPARED TO ONLY BLACKS AND HISPANICS

The percentage of white and nonwhite respondents on the negative end of the ethnocentrism scale is 29% and 60% (compared to 22% and 39% using the full set of six reference groups).

The range of candidate feeling thermometer scores for oikophobic whites is more muted for whites compared to using the full six reference groups or the black-Hispanic-Muslim set of reference groups: 32 for Trump, 36 for Cruz, 42 for Carson, 38 for Fiorina, 35 for J Bush, 42 for Rubio, 40 for H Clinton, 48 for Obama, and 54 for Sanders. The 95% confidence interval for Trump is [26, 37], which doesn't overlap with Obama or Sanders but overlaps with H Clinton [34, 45].

The range of candidate feeling thermometer scores is still fairly large for nonwhites scoring less than zero: 24 for Trump, 35 for Cruz, 34 for Fiorina, 33 for J Bush, 39 for Rubio, 62 for H Clinton, 73 for Obama, and 52 for Sanders. The 95% confidence interval for Trump is [18, 31], which doesn't overlap with Rubio, H Clinton, Obama, or Sanders, and is near the left edge of the confidence intervals for the remaining candidates (whose left edge confidence interval is either 29 or 30).

The difference between oikophobic and ethnocentric whites on the immigration item is still substantive: 0.65 standard deviations, with no confidence interval overlap.

The difference between oikophobic and ethnocentric whites on the item about affirmative action for blacks in university admissions is 0.56 on the 6-point scale (about 0.29 standard deviations), and the difference has a p-value of 0.04 even though the confidence intervals overlap.

Tagged with:

I tweeted about the possibility that whites are the proper target group for assessing racial bias among white liberals, so I thought I'd check available data to assess whether there is evidence for this. The recent Iyengar and Westwood 2015 AJPS article measuring different types of discrimination seemed a good place to look.

In the Iyengar and Westwood 2015 racial discrimination experiment, respondents were given a choice between two high school seniors competing for a scholarship, with names and clubs intended to signal race:

  • Arthur Wolfe, President of the Future Investment Banker Club
  • Jamal Washington, President of the African American Student Association

For some respondents, the two applicants had the same GPA (3.5 or 4.0), and, for other respondents, one of the applicants had a 3.5 GPA and the other had a 4.0 GPA.

Here are the results for white liberals and white conservatives:

EQUALLY QUALIFIED
Liberals: 73% selected the black target [n=34] CI: [58, 89]
Conservatives: 40% selected the black target [n=55] CI: [27, 53]
Difference between the 73% and the 40%: two-tailed p=0.001

BLACK TARGET MORE QUALIFIED
Liberals: 92% selected the black target [n=12] CI: [73, 110]
Conservatives: 56% selected the black target [n=23] CI: [35, 78]
Difference between the 92% and the 56%: two-tailed p=0.013

WHITE TARGET MORE QUALIFIED
Liberals: 44% selected the black target [n=18] CI: [19, 70]
Conservatives: 16% selected the black target [n=19] CI: [-2, 34]
Difference between the 44% and the 16%: two-tailed p=0.061

There were substantial differences in the estimates, with white liberals on average favoring the target with the black name when the targets were equally qualified.

Here are the results for white Democrats and white Republicans:

EQUALLY QUALIFIED
Democrats: 62% selected the black target [n=53] CI: [49, 76]
Republicans: 47% selected the black target [n=45] CI: [32, 62]
Difference between the 62% and the 47%: two-tailed p=0.125

BLACK TARGET MORE QUALIFIED
Democrats: 75% selected the black target [n=32] CI: [59, 91]
Republicans: 50% selected the black target [n=16] CI: [22, 78]
Difference between the 75% and the 50%: two-tailed p=0.104

WHITE TARGET MORE QUALIFIED
Democrats: 59% selected the black target [n=18] CI: [37, 81]
Republicans: 21% selected the black target [n=19] CI: [1, 41]
Difference between the 59% and the 21%: two-tailed p=0.012

Results indicated that white Democrats on average favored the target with the black name in all three scenarios, even when the white target was more qualified. The point estimate for white Republicans never crossed 50% in any scenario.

Data are here, and reproduction code is here.

Tagged with: