Information and the beauty premium in political elections.
Individuals are influenced every day by a variety of biases. These biases affect a range of decisions and may be explicit, where the person chooses to discriminate, or implicit, where there is no conscious intent (Bertrand, Chugh, and Mullainathan 2005). Biases may lead to unfair treatment of minority groups and can result in suboptimal outcomes. Political elections are an important setting where biases based on personal characteristics including gender, ethnicity, and beauty may influence the outcome. The effect of small biases on election outcomes is driven largely by the fact that many voters are uninformed about the candidates or issues (Battels 1996) forcing them to vote based on external cues and biases (McDermott 1997).
One particular bias in elections is a beauty bias where attractive candidates receive more votes. Past studies have documented a beauty premium with magnitudes differing based on the type of election and the degree to which voters are informed. Berggren, Jordahl, and Poutvaara (2015) find that the beauty premium is significant for both national and low-profile elections in Finland, but that it matters more for particular candidates (right-leaning candidates) in particular elections (low-profile elections). Berggren, Jordahl, and Poutvaara (2010) finds that the beauty premium is larger for nonincumbent candidates.
Extending this analysis to the United States, we examine the degree to which the beauty premium is smaller when the voters are likely to be better informed about the candidates. First, we test whether the beauty premium differs between elections for U.S. congress, where there is considerably more media attention, and elections for the House and Senate of individual states. (1) Second, we test whether the beauty premium is larger for incumbent or nonincumbent candidates, because there is likely to be much less information about nonincumbent candidates. Third, we test whether the beauty premium is diminished in elections with increased campaign spending, since higher spending is likely to be associated with more information available to voters. We recognize that more electable candidates may be able to raise more funds, which thereby induces endogeneity. Thus, the estimates regarding this issue should be interpreted as correlational rather than causal.
In all three cases, we find that the beauty gap is smallest in those settings when voters are most likely to be informed, suggesting that biases such as beauty bias are malleable to the environment in which the evaluation occurs. These results add to a growing literature about ways in which biases that affect how individuals evaluate others can be reduced (e.g., Marmaros and Sacerdote 2006; Parsons et al. 2011).
The paper is organized as follows. Section II discusses previous literature on the beauty premium in general and in politics specifically. Section III details the data and methodology. We present the results in Section IV and conclude in Section V.
Discrimination remains a large concern in labor economics. Beauty precipitates discrimination against individuals perceived as less attractive. The seminal paper of Hamermesh and Biddle (1994) was among the first to document a wage premium for individuals who are more attractive. Subsequently, there have been many studies which explore the beauty premium in a variety of contexts, including education and professional sports (Berri et al. 2011; Hamermesh and Parker 2005). Deryugina and Shurchkov (2015) study female college students and find no relationship between college grades and attractiveness, but do find substantial beauty-based sorting into areas of study. Ravina (2012) finds that beautiful applicants have a 1.59% higher probability of getting loans and pay 60 basis points less, despite having no difference in default rates than average looking people. The income premium of attractiveness in high school is found to persist through a person's early-50s, being robust to intelligence quotient (IQ), proxies for confidence and personality, family background, and high school experiences (Scholz and Sicinski 2015). In a laboratory setting, Wilson and Eckel (2006) find that attractiveness initially makes individuals more trusted by strangers, however they later face a beauty penalty by failing to live up to expectations.
Beauty and appearance have widely been examined in the context of elections, with significant consequences on both the election outcome and vote share. Hamermesh (2006) finds that beauty predicts a significant amount of variation in American Economic Association elections. King and Leigh (2009) and Berggren, Jordahl, and Poutvaara (2010) find evidence of the beauty premium in Australia and Finland, with an increase in beauty associated with a significant increase in vote share. Berggren, Jordahl, and Poutvaara (2015) observe that in Finland, right-wing candidates look better than left-wing candidates, and candidates on the right experience a larger beauty premium relative to those on the left, but only in low-profile elections. Todorov et al. (2005) find that judgments of competency based on 1 -second exposures to photos of two U.S. Congressional candidates predict nearly 70% of elections. Similar to beauty, Benjamin and Shapiro (2009) determine that voters make election forecasts based on candidates' personal characteristics, such as likeability, rather than inferences concerning candidates' policy positions.
We expect the beauty premium to differ between high- and low-profile elections due to varying levels of voter knowledge. Specifically, we expect the beauty premium to be lower for elections where voters are more highly informed about candidates. Recent literature corroborates these expectations. Riggle et al. (1992) show that absent all other information about candidates, attractiveness significantly influences voters' choices; however, when given other information, beauty becomes less important and possibly insignificant. Johns and Shephard (2011) additionally find that photographs of politicians have higher effects on uninformed voters than on informed voters. Todorov et al. (2005) posit that first impressions, including beauty impressions, may have a significant and lasting effect on voter choices.
Voters may choose to obtain more or less information based on their judgment of the costs and benefits of obtaining the information. For example, the costs of obtaining information are often lower in national elections, with a larger amount of advertisement in the media and a higher level of news coverage. Benefits may also differ; for example, a voter may derive more utility by becoming informed about national elections and being able to express political viewpoints to peer voters who may place a higher value on national elections.
III. DATA AND METHODOLOGY
We collected subjective data on perceived beauty of 800 different candidates from 400 randomly selected elections that took place in 2012. We randomly sampled 200 elections for the U.S. Senate and U.S. House as well as 200 elections for the State Senate and State House and focus on just the Republican and Democrat candidate for each election. The photos for each candidate come from votesmart.org, a nonpartisan, nonprofit political website. For the 2012 election year, this site contains a photo for over 1,200 candidates from 470 high-profile elections and a photo of over 8,000 candidates from over 5,600 low-profile elections. Votesmart uses the official picture from candidate campaign websites, when available. Other photographs are found via candidates' social media pages, such as Facebook or Twitter. Pictures are cropped in a fixed 1:1 ratio and generally include each candidate's head and shoulders only. The size of the photos used in the study is small (110 x 135 pixels), but uniform across photos. Our sample uses only color photographs of candidates and excludes elections where either the Democrat of Republican candidates has only a black and white photo on votesmart.org.
There are two potential problems with estimating a beauty premium using candidate photos. A reverse causality problem arises if more successful take better, more professional photos. In a robustness check, we attempt to control for this effect by using two subjective measures of photo professionalism. These are binary indicators for whether the photo appears to be professionally produced and an indicator for the quality of the photo. We coded the picture as being low quality if the picture appeared more pixelated and we coded whether the picture was professionally done based on the dress, style, and background. We find that 741 of the 800 photos were high quality and 619 professionally done. High-profile elections were more likely to have high-quality photos but less likely to have photos that appeared to be professionally done. We include these additional controls as part of a robustness check in our analysis. In addition, we hired three research assistants (RAs) to code the professionalism and quality of each photo on a 1-3 scale in order to give us an alternate measure of photo professionalism and quality not created by the authors. We took the average (mean) and mode of the three ratings (using 2 for the mode when all three ratings differed).
In the literature regarding the beauty premium, age is often controlled for as it is a confounding factor which is likely correlated with beauty as well as vote share and winning an election. In another robustness check, we estimate the main regressions using controls for age and an interaction term of age with beauty. We were able to identify the ages for 387 high-profile and 213 low-profile candidates and include these 600 observations in a robustness check with an age control.
Second, there may be omitted variable bias if beauty is correlated with other things that could lead to votes, such as oral skills, competence, or the number of people that campaigning candidates meet. We do not have a way to measure and or control for these other factors; therefore, results should be interpreted accordingly. Finally, Atkinson, Enos, and Hill (2009) suggest that a bias exists when a political party chooses a more beautiful candidate for an election it expects to be close; we do not address this particular issue in this paper.
We use an online survey with 991 participants from Amazon Mechanical Turk to obtain subjective ratings of beauty for the 800 candidate photos. Workers were required to be at least 18 years old, in the United States, and to have a project acceptance percentage of 90% or greater. (2) Each participant rated the attractiveness of 20 photographs that were selected and ordered randomly from the sample of 800 candidates. (3) Answers to the beauty question are coded on the same 1 -5 scale used by Berggren, Jordahl, and Poutvaara (2015). (4) Participants also indicated whether they recognized the persons they rated; we exclude recognized photo observations from our analysis. Participants were not informed that the photos were of political candidates in order to prevent qualities that may be associated with political ability, but not with beauty, from interfering with participants' evaluations. On average, each photo received about 25 ratings; the photo with the least number of ratings received 11 ratings. We take out rater fixed effects by regressing all scores on photo fixed effects and rater fixed effects and use the coefficients on the photo fixed effects as the beauty score, which are then normalize to have the same mean and standard deviation as the original data. Table 1 provides the average beauty rating across a set of different candidate characteristics. Female politicians were rated considerably higher than men in terms of beauty. Low-profile candidates who won had higher beauty scores on average, but there was no difference in high-profile elections. Surprisingly, incumbents had a slightly lower beauty score, perhaps due to age differences.
The data on the elections come from several sources. We use Carl Klarner's "State Legislative Election Returns Data, 2011-2012" for the state-level elections. These data are similar to that found in Klarner et al. (2013), but has been updated to include more recent years. U.S. House election data are collected from Gary Jacobson; U.S. Senate elections data were collated using data from Carl Klarner and NBC.com. OpenSecrets.org, run by the Center for Responsive Politics, supplies data on campaign expenditures.
For our analysis, the unit of observation is a candidate. Standard errors are clustered at the election level, with two observations per election. Our two primary dependent variables are the fraction of the vote that the candidate received (among those votes cast for one of the two major parties) and an indicator for whether or not the candidate won. Logistic regression reporting marginal effects at the mean is used when the outcome variable is winning and ordinary least squares is used for regressions with vote share as the outcome variable. We include a basic set of controls in each regression that include the candidate's incumbency status, gender, and race/ethnicity. Beauty has been standardized over all 800 candidates. As a set of robustness checks we include additional controls for the photo quality and photo professionalism.
In Table 2, we provide the results of our analysis when we pool the data from the state and national elections together. The results in this table indicate that a one standard deviation increase in a candidate's beauty is associated with a 1.1 percentage point increase in the fraction of votes received and a 6.0 percentage point increase in the probability of winning the election. (5) In columns 2 and 4, we include an interaction term between beauty and high-profile election and another interaction between beauty and incumbency status. The coefficient for the measure of beauty now represents the premium for nonincumbents in low-profile elections. For these candidates, we find that a one standard deviation increase in beauty is associated with a 2.5 percentage point increase in the vote share and a 10.2 percentage point increase in the probability of winning (both statistically significant at the 1% level). In contrast, the interaction term between beauty and high-profile election is -1.4 percentage points for vote share and -6.3 percentage points for winning (not significant), indicating that the beauty premium is much smaller for highprofile elections. The interaction term between beauty and incumbency status is also negative, with a coefficient of -2.2 percentage points for vote share and -9.0 percentage points in terms of winning the election, although this latter number is not significant. This result matches those of Berggren, Jordahl, and Poutvaara (2010), who also found that the effects of beauty were concentrated among nonincumbent candidates.
We further test the relationship between the beauty premium and voter information by considering the expenditures of a candidate during an election. Higher spending is correlated with higher voter information (Potters, Sloof, and Van Winden 1997). We thus expect elections with higher spending to have a smaller beauty premium, as voters will be better informed. In Table 3, we provide regression results, which include an interaction term between beauty and expenditures by the candidate in that election. We divide expenditures by the total number of votes cast in the election and then standardize the variable separately for low-profile and high-profile elections. (6) We consider national and state elections separately because spending and high-profile status are so strongly correlated.
In the first column of Table 3, we find that there is no beauty premium, on average, in elections for the U.S. House and Senate elections. However, in the fourth column when we include the interaction with spending, we do find a negative and statistically significant interaction term. The coefficients on the main effect for beauty and the interaction term suggest that for each standard deviation a candidate is above the beauty mean the candidate loses a beauty premium of 1.6 percentage points in vote share for every standard deviation they spend above the sample mean. Generally, the positive 3.3% direct effect of spending outweighs the added beauty premium but it does leave the possibility that spending more could outweigh the beauty premium for candidates more than two standard deviations below the sample mean. For the state-level elections, we also find a negative interaction term between beauty and campaign spending that is slightly smaller and not statistically significant. We recognize that the funding a candidate receives may be correlated with traits that attract votes. Owing to this endogeneity the relationship cannot be interpreted as causal. Campaign spending is often viewed as being an expense that is socially wasteful. Our results suggest that increased campaign spending may be socially beneficial by reducing biases that affect how individuals vote.
As robustness checks, in Table 4, we reestimate the specification of Table 2 and include subjective indicator variables for photo quality and for photo professionalism. We also estimate specifications with a control for candidate's age and an interaction of age with beauty. In the first two columns of Table 4 we find that the estimated beauty coefficient is about 25% smaller when we include controls for photo quality and professionalism. This suggests that a part of the beauty premium may operate through a candidate's presentation of themselves to the public. The final two columns of Table 4 include age controls; the marginal effect for beauty on winning is slightly smaller as compared to that in Table 2, but is slightly larger when vote share is the outcome variable. We also find that including these quality and professionalism controls reduces the magnitude of the interaction terms in Table 2. These changes in results are observed when professionalism and quality are included (shown using the mean of RA coding), but are not sensitive to how these variables were coded: the author rating, mean of RA coding, or mode of RA coding.
In terms of the direct effects of these other two measures, we find that photo quality is positively related to election outcomes, but not significant in any specification, while photo professionalism is positive, significant, and quite large. These results indicate that candidates who take more professionally-appearing photos achieve a higher vote share. This may mean that the measured beauty premium may, to some degree, be a side effect of better candidates--perhaps better candidates take more professional photos or respondents to the survey rated professional-looking photos higher because they perceived the candidate to be more beautiful due to the professionalism of the photo. (7)
Corroborating current literature, we find that beauty matters in elections. These results are robust to inclusion of controls for gender, race/ethnicity, age, and incumbency status. We find that candidates with higher levels of beauty experience electoral success. This effect is particularly large for nonincumbents in low-profile elections, a situation in which the voters are likely to have the least amount of information in the candidate. In contrast, we find no beauty premium for candidates in high-profile elections. This implies that voters use beauty as a substitute for information about candidates' ability. This finding may provide some amount of justification for the large amounts of resources invested in election campaigns as they may help eliminate the beauty bias of voters, again keeping in mind the endogeneity of the spending variable. In general, increased access to information can play a role in reducing various biases such as those based on beauty, gender, or race.
If beauty causes less-qualified or less-capable candidates to be elected due to attractiveness, policymakers who desire to prevent such outcomes will be motivated to explore ways of making beauty less of a factor in elections. It may be that beauty accurately signals to voters about the competency of a candidate and that the beauty premium may not be detrimental. Mobius and Rosenblat (2006) find that beauty is correlated with confidence and attractive workers have greater oral and social skills; however, they also find that for a given confidence level, physically attractive workers are (wrongly) considered more able by employers. We recognize our results depend on beauty not signaling competence or ability, and acknowledge this is an assumption that is difficult to empirically test. As Todorov et al. (2005) suggest, the beauty premium may be a symptom of a deeper inclination voters possess to vote according to their initial impression of a candidate, regardless of how this first impression may form. This analysis suggests that potential political candidates considering an investment of time and money to run for office will also be better able to estimate their chances for success by accounting for beauty effects. Finally, voters equipped with knowledge of the beauty premium may guard against its detrimental effects by increasing their knowledge about political candidates in order to cast less biased votes.
Atkinson, M. D., R. D. Enos, and S. J. Hill. "Candidate Faces and Election Outcomes: Is the Face-Vote Correlation Caused by Candidate Selection?" International Quarterly Journal of Political Science, 4(3), 2009, 229-49.
Bartels, L. M. "Uninformed Votes: Information Effects in Presidential Elections." American Journal of Political Science, 40(1), 1996, 194-230.
Benjamin, D. J., and J. M. Shapiro. "Thin-Slice Forecasts of Gubernatorial Elections." Review of Economics and Statistics, 91(3), 2009, 523-36.
Berggren, N., H. Jordahl, and P. Poutvaara. "The Looks of a Winner: Beauty and Electoral Success." Journal of Public Economics, 94(1-2), 2010, 8-15.
--. "The Right Look: Conservative Politicians Look Better and Voters Reward It." Munich Discussion Papers in Economics, 2015-10, 2015.
Berri, D. J., R. Simmons, J. Van Gilder, and L. O'Neill. "What Does It Mean to Find the Face of the Franchise? Physical Attractiveness and the Evaluation of Athletic Performance." Economics Letters, 111(3), 2011, 200-2.
Bertrand, M., D. Chugh, and S. Mullainathan. "Implicit Discrimination." American Economic Review, 95(2), 2005, 94-98.
Deryugina, T, and O. Shurchkov. "Does Beauty Matter in Undergraduate Education?" Economic Inquiry, 53(2), 2015,940-61.
Hamermesh, D. S. "Changing Looks and Changing 'Discrimination': The Beauty of Economists." Economics Letters, 93(3), 2006,405-12.
Hamermesh, D. S., and J. E. Biddle. "Beauty and the Labor Market." American Economic Review, 84(5), 1994, 1174-94.
Hamermesh, D. S., and A. Parker. "Beauty in the Classroom: Instructors' Pulchritude and Putative Pedagogical Productivity." Economics of Education Review, 24(4), 2005, 369-76.
Ipeirotis, P. G. "Demographics of Mechanical Turk." Working Paper CeDER-10-01, NYU Center for Digital Economy Research, 2010.
Johns, R., and M. Shephard. "Facing the Voters: The Potential Impact of Ballot Paper Photographs in British Elections." Political Studies, 59(3), 2011, 636-58.
King, A., and A. Leigh. "Beautiful Politicians." Kyklos, 62(4), 2009, 579-93.
Klarner, C, W. Berry, T. Carsey, M. Jewell, R. Niemi, L. Powell, and J. Snyder. "State Legislative Election Returns Data, 1967-2010," hdl: 1902.1/20401, Harvard Data-verse, VI, UNF:5:q/n5C9RQGfjy6AjbLG6JWQ==; 2013.
Marmaros, D., and B. Sacerdote. "How Do Friendships Form?" Quarterly Journal of Economics, 121(1), 2006, 79-119.
McDermott, M. L. "Voting Cues in Low-Information Elections: Candidate Gender as a Social Information Variable in Contemporary United States Elections." American Journal of Political Science, 41(1), 1997, 270-83.
Mobius, M., and T Rosenblat. "Why Beauty Matters." American Economic Review, 96(1), 2006, 222-35.
Parsons, C. A., J. Sulaeman, M. C. Yates, and D. S. Hamermesh. "Strike Three: Discrimination, Incentives, and Evaluation." American Economic Review, 101(4), 2011, 1410-35.
Potters, J., R. Sloof, and F. Van Winden. "Campaign Expenditures, Contributions and Direct Endorsements: The Strategic Use of Information and Money to Influence Voter Behavior." European Journal of Political Economy, 13(1), 1997, 1-31.
Ravina, E. "Love & Loans: The Effect of Beauty and Personal Characteristics in Credit Markets." Working Paper, 2012. http://papers.ssrn.com/sol3/papers.cfm7abstract_id=1107307
Riggle, E., V Ottati, R. Wyer, J. Kuklinski, and N. Schwarz. "Bases of Political Judgments: The Role of Stereotypic and Nonstereotypic Information." Political Behavior, 14(1), 1992,67-87.
Scholz, J. K., and K. Sicinski. "Facial Attractiveness and Lifetime Earnings: Evidence from a Cohort Study." Review of Economics and Statistics, 97(1), 2015, 14-28.
Todorov, A., A. N. Mandisodza, A. Goren, and C. C. Hall. "Inferences of Competence from Faces Predict Election Outcomes." Science, 308(5728), 2005, 1623-6.
Wilson, R. K., and C. C. Eckel. "Judging a Book by Its Cover: Beauty and Expectations in a Trust Game." Political Research Quarterly, 59(2), 2006, 189-202.
TODD R. JONES and JOSEPH PRICE (*)
(*) We would like to thank Carl Klarner and Gary Jacobson for providing data. We also thank Daniel Hamermesh, Levon Barseghyan, Daniel Benjamin, Cassandra Benson, Francine Blau, Lawrence Blume, Christine Coyer, Ronald Ehrenberg, and Lawrence Khan for providing helpful comments.
Jones: Ph.D. Candidate, Field of Economics, Cornell University, Ithaca, NY 14853. Phone (607) 255-4254, Fax (607) 255-2818, E-mail email@example.com
Price: Associate Professor, Department of Economics, Brigham Young University, Provo, UT 84602. Phone 801-422-5296, Fax 801-422-0194, E-mail firstname.lastname@example.org
1Q: Intelligence Quotient
RA: Research Assistant
(1.) Throughout, we use the terms "high-profile elections" and "national elections" interchangeably to refer to elections for the U.S. House and Senate. Similarly, we use "low-profile elections" and "state elections" interchangeably to refer to elections to the House and Senate of individual states.
(2.) Ipeirotis (2010) provides some demographic information about Mechanical Turk works and finds that 65% are female, 75% were born prior to 1985, almost 50% have at least a Bachelor's degree, and the median income is approximately $50,000.
(3.) The wording of the question and possible answers to the beauty question are similar to that used by Berggren, Jordahl, and Poutvaara (2015). Our survey also asked the respondent to rate the competence of the person in the photo; we do not consider competence in this analysis.
(4.) To provide a measure of inter-rater agreement among the survey-takers, we group ratings of 1 and 2 into a low category and 4 and 5 into a high category. We calculate a Kappa coefficient between these two groups of 0.523, a highly significant result that parallels that found by Berggren, Jordahl, and Poutvaara (2010).
(5.) We note that in regressions excluding incumbency status, the high-profile coefficient is close to 0; including incumbent causes the coefficient to increase in magnitude.
(6.) When we do not divide by the votes, the spending-beauty interaction terms at the national level are not significant.
(7.) Across all observations, the correlation of standardized beauty and photo professionalism is 0.11. Incumbent status is also strongly predictive of photo professionalism.
TABLE 1 Mean Beauty by Candidate Characteristics Low-Profile High-Profile Election Election Gender Male 2.287 2.396 (0.503) (0.481) Female 2.556 2.619 (0.600) (0.556) Incumbency Status Not incumbent 2.377 2.452 (0.539) (0.515) Incumbent 2.301 2.428 (0.541) (0.492) Candidate Result Lost 2.310 2.443 (0.522) (0.530) Won 2.396 2.440 (0.556) (0.480) Party Democrat 2.321 2.375 (0.553) (0.490) Republican 2.385 2.508 (0.527) (0.512) Race White 2.343 2.431 (0.540) (0.511) Not White 2.465 2.514 (0.534) (0.457) N 400 400 Notes: All of the statistics are calculated with Democrats and Republicans together with the exception of results by party, which are broken up by Democrat and Republican. TABLE 2 Correlation between Beauty and Election Outcomes Won Vote Share (1) (2) (3) (4) Beauty 0.060 (***) 0.102 (***) 0.011 (***) 0.025 (***) (0.023) (0.033) (0.004) (0.006) lncumbent *Beauty -0.090 -0.022 (***) (0.065) (0.007) High -profile *Beuuty -0.063 -0.014 (*) (0.045) (0.008) Incumbent 0.692 (***) 0.691 (***) 0.195 (***) 0.195 (***) (0.028) (0.028) (0.009) (0.009) High profile -0.144 (***) -0.133 (***) -0.024 (***) -0.024 (***) (0.031) (0.031) (0.005) (0.005) Female -0.023 -0.022 0.010 0.010 (0.061) (0.060) (0.009) (0.009) Black -0.035 -0.038 0.027 0.026 (0.090) (0.090) (0.022) (0.022) Hispanic 0.079 0.085 0.014 0.017 (0.111) (0.109) (0.023) (0.023) Other race/ -0.292 (***) -0.263 (***) -0.017 -0.010 Ethnicity (0.065) (0.083) (0.021) (0.018) N 800 800 800 800 R-squared .358 .362 0.437 0.445 Notes: Vote share is based on the two major party candidates. Mean beauty has been normalized across all observations by subtracting the mean and dividing by the standard deviation. We use logistic regression when winner is the outcome and report the marginal effects at the mean. Robust standard errors in parentheses. TABLE 3 Difference in Beauty Premium Based on Election Expenditures National Winner Vote Share (1) (2) (3) (4) Beauty -0.007 -0.011 -0.001 -0.004 (0.037) (0.042) (0.005) (0.005) Expenditures 0.279 (***) 0.291 (***) 0.029 (***) 0.033 (***) (0.059) (0.069) (0.009) (0.008) Expend* Beauty -0.061 -0.016 (*) (0.077) (0.008) Observations 399 399 400 400 R-squared .571 .573 0.578 0.584 State Winner Vote Share (5) (6) (7) (8) Beauty 0.059 (*) 0.052 0.012 (**) 0.012 (**) (0.031) (0.034) (0.006) (0.006) Expenditures 0.132 (**) 0.161 (**) 0.018 (***) 0.022 (***) (0.060) (0.071) (0.007) (0.008) Expend* Beauty -0.063 -0.009 (0.052) (0.006) Observations 400 400 400 400 R-squared .259 .264 0.363 0.368 Notes: Vote share is based solely on the two major party candidates. Expenditures was divided by total votes for the two major parties, then normalized separately for national and state elections. Observations are 399 in National with winner as the outcome because other race/ethnicity predicts the outcome perfectly. Mean beauty has been normalized across all 800 observations by subtracting the mean and dividing by the standard deviation. Each regression includes controls for incumbency, race/ethnicity, and gender. We use logistic regression when "winner" is the outcome and report the marginal effects at the mean. Robust standard errors in parentheses. (***) P< 01, (**)p<.05, (*) p<1. TABLE 4 Correlation between Beauty and Election Outcomes Accounting for Additional Controls Winner Vote Share Winner Beauty 0.078 (**) 0.020 (***) 0.090 (**) (0.034) (0.006) (0.044) Incumbent (*) Beauty -0.084 -0.019 (**) -0.066 (0.066) (0.007) (0.063) High profile (*) Beauty -0.050 -0.012 -0.061 (0.047) (0.008) (0.052) Incumbent 0.658 (***) 0.172 (***) 0.660 (***) (0.032) (0.010) (0.032) High profile -0.124 (***) -0.022 (***) -0.205 (***) (0.033) (0.005) (0.033) Photo professionalism 0.289 (***) 0.049 (***) (0.052) (0.008) Photo quality 0.010 0.005 (0.052) (0.007) Age -0.024 (0.028) Age (*) Beauty 0.026 (0.025) Observations 800 800 600 R-squared .395 0.474 .373 Vote Share Beauty 0.036 (***) (0.009) Incumbent (*) Beauty -0.026 (***) (0.008) High profile (*) Beauty -0.019 (**) (0.009) Incumbent 0.186 (***) (0.010) High profile -0.029 (***) (0.007) Photo professionalism Photo quality Age 0.003 (0.005) Age (*) Beauty 0.000 (0.005) Observations 600 R-squared 0.444 Notes: Vote share is based on the two major party candidates. Mean beauty has been normalized across all observations by subtracting the mean and dividing by the standard deviation. We use logistic regression when winner is the outcome and report the marginal effects at the mean. Age is in standard deviation units. Photo quality and professionalism use the average of RA coding. Robust standard errors in parentheses. (***) p<.01, (**)P<0.05, (*) p<.1.
|Printer friendly Cite/link Email Feedback|
|Author:||Jones, Todd R.; Price, Joseph|
|Publication:||Contemporary Economic Policy|
|Date:||Oct 1, 2017|
|Previous Article:||Does athletic success influence persistence at higher education institutions? New evidence using panel data.|
|Next Article:||The effect of casino proximity on lottery sales: Evidence from Maryland.|