Dartmouth Political Survey Leftovers: Part 3

This is the third of three blog posts in which I dive deeper into some interesting results from surveys I’ve ran at my school. The first part can be found here and the second one here.

The Roommate Question

The original results article for a survey I ran back in April sparked some interest for one specific finding: when asked about how comfortable they would be with having a roommate with opposing political views to their own, Democratic students at Dartmouth were significantly more likely to say they were uncomfortable than Independent and Republican students were. I’ve briefly addressed what this result could suggest, but I wanted to examine something outside of the partisan response differences here.

For responses to each substantive survey question, I always break them down by demographic (race, gender, etc.) and political (party) variables to check for significant subgroup differences. That’s where the partisan divide in roommate comfort level came from. Interestingly, that proved to be just one of five significant subgroup differences. The other four arose when splitting responses up by gender, race, family income, and student Greek affiliation status.

More male Dartmouth students (60 percent) said they would be comfortable with a roommate with opposing political views than female students did (38 percent). Fewer non-white students at 39 percent said they were comfortable than white students at 58 percent did. Income backgrounds split responses as well: 60 percent of students from higher family income brackets (>$200,000) said they were comfortable while only 39 percent of students from lower income backgrounds (<$200,000) did so.

Subgroup differences also formed for Greek affiliation status. As I’ve explained before, the Greek system at Dartmouth means students can fall into three buckets: non-freshmen students who are affiliated with Greek life, non-freshmen who aren’t affiliated with Greek life (whom I’ll just refer to as “unaffiliated students”), and freshmen, who cannot enter the Greek system in their first year. Unaffiliated non-students (38 percent) were less comfortable compared to affiliated students (53 percent) and freshmen (52 percent). Moreover, fewer international students expressed comfort with this type of roommate than non-international ones, but a small sample for the former group did not allow this difference to pass conventional levels of statistical significance.

Data and Methods

Simply put, while partisanship proved a big factor for shaping responses to this question on roommate preference based on politics, similar-sized differences at significant levels appeared across various personal attributes. To isolate the influence of these many variables, I ran a multivariate linear model predicting roommate political discomfort. This dependent variable was built from the five-point Likert scale from the roommate question. Comfort levels ranged from very comfortable, somewhat comfortable, neither comfortable nor uncomfortable, somewhat uncomfortable, to very uncomfortable; values of 1, 2, 3, 4, and 5, respectively, were assigned to these responses to create a continuous discomfort measure for the outcome in the model. Five independent variables were ultimately of interest here:

  1. Partisanship; a categorical variable was used, with Democrats (the largest of three groups) as the base in the regressions
  2. Gender; men were used as the base (two groups; the group of non-binary gender identifiers was not large enough to include as a category in the model)
  3. Race; whites were used as the base (two groups)
  4. Family income; students from families making more than $200,000 were used as the base (two groups)
  5. Greek affiliation status: students affiliated with Greek life were used as the base (three groups)


I include three different specifications, with the third one containing all five variables of interest. The results of each are shown in the below table.

regression output 6-9-17

The first model regresses the discomfort measure on what ultimately were the two strongest predictors: party and gender. As seen in the first column, relative to Democrats, Independents were 0.653 points more comfortable with a roommate with opposing political views to their own on this 1-5 discomfort scale. Republican students showed lower signs of discomfort with this situation as well, being 0.747 points more comfortable than Democratic students were. Even with this important variable in partisanship accounted for, the statistically significant difference between genders held: female students at Dartmouth were 0.591 points more uncomfortable with a political opponent for a roommate than male students were. I’ll return to this important result a bit later.

Crucially, as other “demographic” variables are introduced in models 2 and 3, the effect sizes and significance levels of these two predictors–partisanship and gender–for roommate discomfort do not change. In the model with every variable included seen in column 3, party and gender are the strongest predictors of roommate discomfort. Other demographics prove not as strong as I take into account more factors. The second model in column 2 adds just race and income into the mix. While significant at the p<0.05 level and confirming the relationship at the bivariate level (lower income and non-white students express more discomfort), these two predictors prove weaker than gender and party identification in shaping roommate discomfort levels. In the last specification in column 3, taking into account Greek affiliation status leaves the coefficient for family income less significant. Unaffiliated students are more uncomfortable relative to affiliated students, but only at the p<0.10 level.

Explaining Race, Income, and Greek Affiliation Effects

The mechanism behind the somewhat significant effects of race and to a lesser extent income and Greek affiliation status seems to stem from (possibly negative) inter-group sentiments and perceptions specific to Dartmouth. For example, several high-profile and controversial race-related issues and protests have occurred on campus over the last few years, with communities of color long calling to address marginalization of non-white students. These incidents and more subtle, common forms of racial marginalization on campus could have heightened racial tensions and relations–specifically for non-white students, who most feel the brunt of all this. That could conceivably shape relational aspects of student life to the point where non-white students feel less comfortable interacting and living with others outside their group. This mechanism is limited, as the roommate discomfort question was about a roommate from another political group–not another racial one. However, given how closely racial and political identities intertwine, this relational aspect on campus could begin to shed light on the significance of the race variable in the models.

I speculate that somewhat similar processes could be occurring in regards to income backgrounds and Greek affiliation effects. For example, in another recent survey I’ve done (results here, under the “The New Story of Dartmouth” article), I found that 81 percent of all Dartmouth students believe there is a socioeconomic divide on campus, and 79 percent say economic factors shape relationships on campus. I’ve also found students unaffiliated with the Greek system have significantly different experiences and perceptions of social life on campus, often expressing views more critical of the Greek system. Perhaps asking about relations with a student of a different political identity activated not just views toward outgroup political identities but also outgroup economic and social identities as well. Lower income background students and those outside the Greek system might’ve been more sensitive to accepting interactions and living with outgroup members compared to higher income and affiliated students. However, this of course veers into a lot of speculation, and given it’s concerning some of the least powerful predictors of roommate discomfort, this should not be should not be viewed as too convincing.

Why is Gender Such a Strong Predictor?

That still leaves one important and very puzzling question unanswered: why are female students–with all other factors accounted for–less comfortable with roommates of opposing political views than male students are? One explanation for the partisan divide in roommate discomfort centered on framing Democrats as less tolerant of students of opposing political views. Could this supposed political intolerance be gender-based too, in that female students are more intolerant than male ones? Overall, the possibility of political intolerance ignores other key factors, but does have at least some merit for the partisanship effects. The notion that female students are more intolerant, however, seems much more unlikely. Rather, there’s a strand of political and social science research regarding gender-based differences in socialization that I find most instructive for understanding this result.

In examining the roots of the gender political knowledge gap, Wolak and McDevitt (2011) find that adolescent women learn more about politics in environments defined by greater political consensus than political conflict, which they measure through types of political discussion and home county partisan competitiveness (i.e. partisan division). In other words, conflict plays a role in shaping women’s relationship with politics whereby women find social contexts with less conflict more agreeable. Hooghe and Stolle (2004) also study gender differences in politics using data on people at an early age and find that boys favor participating in confrontational and thus more conflict-oriented forms of political engagement than girls do. Along similar lines, Niederle and Vesterlund (2007) find that women are less likely to embrace competition than men are–another sign of possible conflict avoidance. Older evidence, such as from Lever (1976 and 1978), speaks to a similar story, and importantly confirms an early age process and thus a tie to as powerful a force as socialization: boys engage in more conflict-based games and relationships than women do at an early age.

I believe this socialized personality difference regarding conflict, and particularly in the context of one’s orientation towards politics, provides a compelling explanation for why female students at Dartmouth express greater discomfort at having a political opponent for a roommate. This situation of living with someone with opposing political views clearly implies some level of conflicting opinion, if not greater disagreement and dislike between roommates. At the very least, it stands firmly in contrast to the environment marked by consensus that Wolak and McDevitt (2011) emphasize in describing how women prefer to interact with their political surroundings. This conflict avoidance, rooted in socialization processes that span both political and broader social realms, could therefore lead women to report less comfort with the prospect of having a roommate of opposing political views. The current political era–defined by strong and growing hostile outgroup sentiment and negative partisanship–could especially heighten this consideration of conflict avoidance among women. Most importantly, though, I consider the socialization-based development of conflict avoidance among women and its application to political relations as best explaining this gender difference in roommate comfort.

Dartmouth Political Survey Leftovers: Part 3

One thought on “Dartmouth Political Survey Leftovers: Part 3

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s