Reporting on the Race vs. Gender of a Suspect (Survey Results)

1. Overview

This is an informal survey-based study of desire for police and media to take a policy of not including race or gender in their description of the suspect of a crime. Two-hundred-and-one Americans, balanced for gender (50:50 men and women) and party affiliation (50:50 Republican and Democrat, used as a proxy for the broader “Red Tribe” and “Blue Tribe” cultures in the United States that differ in politics, religion, gender roles, geography, urbanization, etc.), indicated their approval level (on a 1–7 scale) of one of the two following policy statements:

  1. When a crime is committed, police and media should avoid mentioning race in their description of the suspect or wanted person, to avoid fueling stereotypes about race and crime.
  2. When a crime is committed, police and media should avoid mentioning gender in their description of the suspect or wanted person, to avoid fueling stereotypes about gender and crime.

Results indicate a greater approval rating for the first policy (of not mentioning race) than the second policy (of not mentioning gender).


Note that I do not ask this with the assumption that either the policy of mentioning less or more about the suspect is the right one. Regardless of my or your overall leanings here, I want to raise the question of whether race and gender should be treated differently.

2. Main findings

The average approval rating when asked about gender was 2.5 out of 7, compared to 3.5 out of 7 for race (SDs = 2.0 and 2.4). This is a difference of 1 point, or about half a standard deviation (Cohen’s d = 0.45), and is statistically significant (see section 3).

Democrats had overall higher approval responses across both questions (average of 3.6/7) than Republicans (2.3). There was additionally a trend in the data towards Democrats having a larger disparity between their responses for race and gender, but this was not statistically significant (meaning that this is something that might be worth looking into more in the future, but at this point no concrete conclusions should be drawn).

3. Additional details

The 201 respondents were the same as in the previous survey study comparing attitudes on charging men more for car insurance with attitudes on charging women more for health insurance. Respondents were assigned to the question about race or gender based on identifying whether the last digit in their day of birth was an even number or an odd number, which divided them approximately in half. Respondents also provided their ages, which are reported in the other study. Republicans were a bit older than Democrats.

Statistical analysis was done using a linear regression (lm in R). The response variable was the 1–7 rating of acceptability; the predictor variables were question (i.e., which question was asked: #1 about race or #1 about gender), gender of the respondent (man, woman), and party affiliation of the respondent (Republican or Democrat). The ANOVA table is provided below. The most relevant result is the significant main effect of group (people had higher approval ratings for a policy of not mentioning race than a policy of not mentioning gender).

Analysis of Variance Table

Response: crimeresp
                       Df Sum Sq Mean Sq F value    Pr(>F)    
question                1  47.84  47.839 10.9997  0.001089 ** 
gender                  1   1.39   1.394  0.3206  0.571895    
party                   1  88.71  88.707 20.3965 1.094e-05 ***
question:gender         1   0.02   0.016  0.0037  0.951409    
question:party          1   7.44   7.437  1.7100  0.192537    
gender:party            1   0.07   0.075  0.0172  0.895867    
question:gender:party   1   1.15   1.150  0.2644  0.607679    
Residuals             193 839.38   4.349                      
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Charging Men More for Car Insurance vs. Women More for Health Insurance (Survey Results)

1. Overview

This is an informal survey-based study on people’s views of gender-based pricing in insurance. Two-hundred-and-one Americans, balanced for gender (men and women) and party affiliation (Republican and Democrat, although my interest isn’t party politics itself so much as the broader “Red Tribe” and “Blue Tribe” cultures in the United States that differ in politics, religion, gender roles, geography, urbanization, etc.), indicated their approval level (on a 1–7 scale) of one of the two following policy statements:

  1. Insurance companies should be allowed to charge women more for health insurance, if they find that women on average access health services more often.
  2. Insurance companies should be allowed to charge men more for car insurance, if they find that men on average get in car accidents more often.

This survey asks about car insurance for men and health insurance for women because those forms of gender-based pricing have received attention in the media (e.g., “Alberta man changes gender on government IDs for cheaper car insurance” from CBC and “Why Republicans Want to Make Women Pay More Than Men for Health Insurance” from NYMag), making these the most realistic or familiar scenarios.


Results indicate greater approval for charging men more for car insurance than for charging women more for health insurance. Respondent gender and party affiliation were relevant—women exhibited a larger difference than men in approval ratings, and the “Blue Tribe” (Democrats) had a larger difference than the “Red Tribe” (Republicans).

Some might question whether car insurance and health insurance are comparable. I think they are comparable, although there are obviously some differences. Health insurance is more essential, although car insurance is not far behind in the notoriously car-dependent United States (77% of Americans drive to work, not counting carpools). Health insurance is also more expensive to buy, but more likely to be provided by someone else (a majority of Americans have coverage through their employer, Medicare, or Medicaid). I don’t see these details giving an obvious justification for why gender pricing would be acceptable in one type of insurance and not the other, but that’s an open possibility.

2. Main findings

The overall average approval rating for gender-based pricing in insurance was 3.3 out of 7 (SD = 2.1). The average for charging men more for car insurance was 3.8, compared to 2.8 for charging women more for health insurance. This is a difference of 1 point or about half a standard deviation (Cohen’s d = 0.49), and is statistically significant (see section 3).

Figure 1

Both gender and party of respondent had a significant effect on the difference in ratings between scenarios. Men exhibited a smaller difference in their approval ratings than women did, and Republicans exhibited a smaller difference than Democrats. As a result of the effects of party and gender, Democrat women exhibit numerically the largest disparity (2.5 points) and Republican men the smallest (0.3 points in the other direction), with Democrat men (1.2) and Republican women (1.1) in the middle.

Figure 2

3. Additional details

The 201 participants were recruited from an online research platform and paid a small sum to complete the survey, which took under a minute (and included one other policy question). Respondents were assigned to the question about health insurance or car insurance based on identifying whether the last digit in their day of birth was an even number or an odd number, which divided them approximately in half.

Table 1: Question answered, by gender and party affiliation

Total respondentsAnswered question on men & car insuranceAnswered question on women & health insurance
Republican men502327
Republican women502822
Democrat men502525
Democrat women512823

Having participants answer one question, but not both, has the advantage that seeing one question doesn’t influence their answer to the second question, but the disadvantage of lower statistical power. Respondents also provided their ages, which I’ve summarized below. Republicans were a bit older than Democrats.

Table 2: Age of sample, by gender and party affiliation

MenWomen
Democrat28.029.3
Republican34.039.7

Statistical analysis was done using a linear regression (lm in R). The response variable was the 1–7 rating of acceptability; the predictor variables were scenario (men on car insurance, women on health insurance), gender of the respondent (man, woman), and party affiliation of the respondent (Republican or Democrat).

The ANOVA table of the output is provided below. The most relevant results are the significant main effect of scenario (charging men more for car insurance was rated as more acceptable than charging women more for health insurance) and the significant scenario:gender and scenario:party interactions (both party and gender affect the disparity in ratings, as explained above). The non-significant scenario:gender:party interaction is also relevant; it means that the disparity exhibited by each party + gender combination is predicted by the separate effects of party and gender. In other words, the scenario ratings difference of Democrat women is largely predicted by the separate effects of “Democrat” and “woman” rather than something special about Democrat women (and similarly for Republican women, Republican men, and Democrat men).

Analysis of Variance Table

Response: insuranceresp
                       Df Sum Sq Mean Sq F value    Pr(>F)    
scenario                1  52.43  52.432 15.0413 0.0001442 ***
gender                  1  45.14  45.144 12.9504 0.0004065 ***
party                   1  48.27  48.273 13.8481 0.0002599 ***
scenario:gender         1  23.43  23.434  6.7226 0.0102488 *  
scenario:party          1  30.44  30.441  8.7325 0.0035152 ** 
gender:party            1  28.59  28.593  8.2026 0.0046460 ** 
scenario:gender:party   1   0.21   0.215  0.0617 0.8041652    
Residuals             193 672.78   3.486                      
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

A complete table of the results is provided below, with effect sizes (Cohen’s d, the effect size measured in standard deviations, calculated using the effsize package in R).

Table 3: Full results

Charge men more
Charge women moreDifferenceCohen’s d
Full sample (n = 201)3.842.811.030.49
Men (n = 100)3.983.600.380.18
Women (n = 101)3.711.911.801.02
Republicans (n = 100)3.943.710.230.10
Democrats (n = 101)3.741.901.841.12
Republican men (n = 50)4.484.81-0.330.17
Republican women (n = 50)3.502.361.240.54
Democrat men (n = 50)2.522.281.140.68
Democrat women (n = 51)3.931.482.451.85

Do People Give More Money to Female Panhandlers? (Survey Results)

This is original data, written up in the general style of a scientific paper (albeit somewhat more informally). It should be readable regardless of background, although for a summary you can skip to the discussion or conclusion. I encourage comments, criticism, and suggestions, especially (but not only) from people with backgrounds in research.

Note: Since posting, this page has undergone nontrivial revisions. See § 7 for details.

Sections:

  1. Introduction
  2. Methodology
  3. Results
  4. Discussion
  5. Conclusion
  6. Appendix
  7. Revisions

(Length: 1,500 words.)

Continue reading “Do People Give More Money to Female Panhandlers? (Survey Results)”

Attitudes on Profiling Men vs. Profiling Minorities (Survey Results)

This is original data, written up in the general style of a scientific paper (albeit somewhat more informally). It should be readable regardless of background, although for a summary you can skip to the discussion or conclusion. I encourage comments, criticism, and suggestions, especially (but not only) from people with backgrounds in research.

Note: Since posting, this page has undergone nontrivial revisions. See § 7 for details.

Sections:

  1. Introduction
  2. Methodology
  3. Results
  4. Discussion
  5. Conclusion
  6. Appendix
  7. Revisions

(Length: 2,500 words.)

Continue reading “Attitudes on Profiling Men vs. Profiling Minorities (Survey Results)”

Men’s Issues Reading List (Book Recommendations)

If your exposure to men’s issues comes from online resources then you might be surprised to find out that there is actually a pretty strong body of offline literature too. The first two books work well as an overview or introduction to men’s issues (one is more philosophical and the other more personal, so they complement each other well). The other books cover other relevant topics (specific men’s issues, gender roles, feminism, etc.) in more detail.

Continue reading “Men’s Issues Reading List (Book Recommendations)”