Numerous studies have today shown that beautiful people earn more than ugly people. Not just among women, but also among men. A short summary of the current research is provided by Landsburg (2001) and Engemann & Owyang (2005). The correlation seems stronger among men and still hold even after adjustment for education, experience, and job categories. Taller people also earn more than shorter people. It is also said that women suffer more for being fat. A widely spread belief is that beauty correlates with higher wages because employers discriminate ugly people. The assumption is probably wrong, as we shall see.
Here’s the detailed summary.
Roszell et al. (1989) demonstrate that beauty is associated with higher earnings in both women and men, for which the unstandardized beta regressions were 0.688 and 1.266, given a 5-point scale in the attractiveness variable and a unit equivalent of 1000 dollars in the dependent variable (income1981), indicating a gain of 688 and 1266 dollars per year per unit increase in beauty variable. In other words, 4-unit increase in beauty produces a gain of 688*4=2752 et 1266*4=5064 dollars per year. That means the relationship is stronger among men. This weakens considerably the idea that wage premium for beauty is a psychological effect, although there exists some evidence of blond effect that works for blond women only and when the employer is a man, not a woman (Johnston, 2010; Guéguen, 2012). It is important to control for job category because we can assume that beauty can be important in some jobs (those involving more human contact) but not others. Thus is the necessity to prove the wage-beauty correlation stands at all levels/types of jobs. That correlation seems to hold even when year of education and occupation status (manual work versus non-manual work) are held constant. Of great interest here is that the (dependent) variable income1981 was predicted simultaneously by beauty, income1979, and gender job composition. In this way, previous differences in income are statistically controlled. This means that beauty increases wages rank-ordering from 1979 and 1981, i.e., in a longitudinal way. Given that IQ also increases earning longitudinally, that is, IQ exerts its effect on earnings through pay trajectory but not entry pay (Ganzach, 2011), there is a possibility that the beauty effect on wage longitudinal increases is due to its synergy with IQ, on the condition of course that IQ and beauty are correlated. The possibility of a synergy is suggested also by the fact that both IQ and attractiveness are heritable (Kanazawa & Kovar, 2004). But if stereotype effect exerts its impact on entry pay (when the employer has fewer information on the candidates’ capabilities) rather than on pay trajectory (when the employer knows much more about the actual employees’ capabilities), the evidence for the latter suggests the beauty effect has nothing to do with some kind of stereotypes.
Frieze et al. (1991) found, among MBA graduates, that beautiful men earn higher starting salaries and this difference increases over time, adjusted for inflation started salaries, whereas for women there was no starting salaries differences but the more beautiful women had experienced higher wage growth, consistent with what is found among men. This, again, can be put in parallel with the fact that IQ seems to have an impact mainly on growth salaries but not on starting salaries (Ganzach, 2011). The very fact that the impact of beauty on starting salary is meaningless for women while being large for men is further proof that it is due to psychological effects, such as, e.g., the employer has been enchanted by the beautiful woman. Height is not associated with better starting salaries for both sexes whereas weight is associated with worse starting salaries for both sexes. The effect of height on later wages is large for both sexes, and weight on later wages is large for men but very small for women. Here’s the table.
Although not explicitly stated by the authors, the coefficients are obviously unstandardized. The constant at 81.14 and 69.34 (thousands of dollars per year) for men and women denotes the estimated salary when all of the independent variables have a value of zero (i.e., b0). Thus, as per the authors’ description, a coefficient of 1.00 and 0.13 in beauty for men and women would mean that each unit-increase in the 5-point scale beauty variable increases starting salaries by 1000 dollars and 130 dollars, or 4000 and 520 dollars for 4-unit increase in beauty. The latter is obviously a weak effect. However, when we look at the third panel, we see that the same coefficients were 1.47 and 1.91, for men and women. In other words, one-unit increase in beauty predicts an additional 1470 and 1910 dollars per year, or 5880 and 7640 dollars per year for 4-unit increase in beauty. The effect is huge, if only attributed to beauty and not to confounding, omitted variables. The coefficients for height in inches are 0.42 and 0.14, or 420 and 140 dollars per year. Small effects, but this also means that a gain of 10 inches (~25 centimeters) in height produces an additional 4200 and 1400 dollars per year, which is not trivial. Coefficients for weight (dichotomy) are -2.38 and -0.40, which means being overweight reduces later wages by 2380 and 400 dollars per year. The effect is strong among men, small among women. The absence of a strong effect of BMI for women is quite inconsistent with later research, as we will see.
Hamermesh & Biddle (1994, Tables 3, 4, 5) studied 3 different survey data, QES, QAL, QOL, using regression with, for instance, inclusion of health, marital status, educational attainment, tenure with the firm, as additional covariates. The measure of attractiveness, as usually the case, consists of a 5-point scale, with the middle being average, and the other scores below- and above-average for ugliness and beauty, respectively. Among men, the impact of being ugly appears more important than the impact of being beautiful in the 3 surveys. However, among women, ugliness is much more associated with earnings changes than is attractiveness in the QES, whereas their contribution were almost equal in the QOL, and both ugliness and attractiveness are associated with better earnings in the QAL. Perhaps a plausible explanation is the suspected high measurement errors associated with attractiveness ratings. This possibility receives some support from the QOL survey which has the great advantage of having 3 beauty assessment. We see, Table 5, that the impact of beauty is modest in each separate measures but when these measures are aggregated, the impact of beauty is strong. The interesting fact about the seems to be the SD of beauty ratings, “What is most interesting is that the ratings of women are more dispersed around the middle category. This is a common finding in the social-psychological literature: women’s appearances evoke stronger reactions, both positive and negative, than men’s (Hatfield and Sprecher, 1986).” (p. 1179). This, in itself, can be taken as evidence and explanation for the differential effect of attractiveness of social outcome across the sexes. But the simple fact that the wage-beauty relationship appears stronger among men seems to indicate that physical appearance among men evoke stronger reactions and feelings as reflected by higher penalty among men, even when the SD of beauty scores is larger among women. If women appearance really evokes stronger reactions, at the very least, its impact on wages is less than what would have been predicted. Whatever the case, the impact of beauty on wages is not affected so much by the inclusion of height or weight variables (when available). Taken as a whole, the effect of ugliness is to reduce wages by 9% among men and 5% among women.
The same authors also found, in the QES, that obesity increases wages among men but reduces it among women. Either this means that beauty has a very different meaning among the sexes or that outcome must be explained by some omitted variables. In both QES and QAL, the effect of being tall or short for both men and women shows no consistent pattern at all.
Landsburg (2001) said that the ugliest married women are 8% less likely to look for a job than married women in general, and that this effect causes those people to disappear in the statistics because they are not counted. The consequence is to lower the SD of beauty scores among women and thus a range restriction will cause a decline in the correlations for women. But he mentioned that Hamermesh & Biddle already addressed this point, as they wrote “There is thus some evidence that women select themselves out of the labor force if they are particularly unattractive. However, this selectivity has no important impact on the basic estimates of the effects of looks on earnings [in column (iv) of Table 4 and column (v) of Table 5]. Correcting for selectivity in the QAL changes the estimated premium associated with above-average looks from 0.128 to 0.130. Accounting for this form of selectivity does not alter the premium in the QOL and changes the earnings penalty from – 0.058 to – 0.036.” (p. 1188). That the effect is stronger among men may deserve explanation. Maybe it elicits stronger feelings.
Using 3 different measures of “occupational looks” or more precisely the ratings of beauty premium by occupation, Hamermesh & Biddle (1994, Tables 8 & 9) found evidence of self-selection, i.e., beautiful people are more likely to work on jobs for which physical appearance is considered more important. Table 9, the authors present regressions with inclusion of an ugliness*occupation and beauty*occupation interaction, that must be interpreted as the increasing or diminishing return of ugliness/beauty when occupational looks increase. In the QES (negative main effect of ugliness and positive main effect of beauty in both sexes), among men, ugliness*occupation tends to be negative whereas beauty*occupation is positive, which means both directions were in accordance with each other. That indicates ugliness penalized more and more whereas beauty earned more and more when the importance of look increases. For women, however, both interactions showed negative signs, i.e., the negative impact of ugliness increases whereas the positive impact of beauty diminishes when occupational looks increase, suggesting that women lose more for being ugly than they earn for being beautiful. In the QAL, there is no clear tendency of increasing or diminishing return among men (inconsistent negative signs in main effect of ugliness but consistent positive signs in main effect of beauty) whereas the interaction tends to be negative for ugliness (with positive main effect in both ugliness and beauty) among women but no clear trend for beauty*occupation among women. Again, this conflicting signs in correlation is probably due to measurement errors. Taken as a whole, the negative/positive main effect of ugliness/beauty indicates the pervasiveness of beauty premium in all levels of occupational looks. Because if the interactions accounted for the totality of the beauty premium, the regression coefficients would fall to zero when controlling for the effect of these interactions.
One possible hypothesis on the beauty-wage correlation is that income itself buys beauty. Landsburg argue that there is a limit to how much one can accomplish with cosmetics. And Hamermesh & Biddle (1994) write the following :
Three pieces of evidence suggest that these simultaneity problems are not crucial here. First, the social-psychological evidence we mentioned in Section I showed how little individuals’ relative physical appearances change during adulthood. This implies that there is limited scope for using unexplained earnings differences to “buy” differences in beauty. Second, if differences in unexplained earnings were used to affect beauty, their persistence over a working life should lead to a greater simultaneity bias among older workers than among younger workers, and thus smaller apparent penalties and premia if we restrict the samples in Tables 3-5 to workers aged 18-30. In fact, all beauty premia and penalties in the QES are larger in this subsample than in the basic estimates in Table 3. In the other two samples, half the estimates increase in absolute value, while half decrease. There is no evidence of a weaker relation between earnings and beauty among younger workers.
Later, Biddle & Hamermesh (1998) studied a longitudinal sample of lawyers, between the 5th year and 15th year of job practice. Appearance is rated through photographs taken before the subjects enter the labor market, which means the “money-buys-beauty” theory is not supported here again. Their expectation was that the earnings of attorneys in the public sector will be higher than in the private sector if employers demand better-looking attorneys. Instead, the self-employed attorneys earn more money than the employees when studied longitudinally. This probably is indicative of the fact that the client/customer itself prefers attractive attorneys. It’s probably for this reason that private attorneys tend to be more attractive than public attorneys due to private demand. The difference in beauty (and earnings) between public and private attorneys (Table 8) increases with years of practice (or job). In fact, the beauty score was stronger (for private attorneys) at year 15 but not at year 5 perhaps because the more attractive ones moved from the public to the private sector where they can expect higher earnings advantage for more beautiful attorneys. For instance, the authors write “The results in the second panel of table 7 show that private stayers (the comparison group) are more attractive than public stayers. Furthermore, lawyers who left the private sector between years 5 and 15 are less attractive than those who remained, while lawyers who switched from the public sector are more attractive than those who stayed” (p. 192).
Harper (2000, Tables 6 & 7) attempts to control for occupation type as well as work experience, years of job tenure, and other variables, in a large longitudinal sample. He finds that attractiveness (rated by teachers on people when they were children) is not meaningfully associated with earnings (at age 33) in both sexes, but being unattractive is negatively associated with earnings in both sexes. The effect of ugliness is again stronger for men. Interestingly, the (non)attractiveness measures at either age 7 or 11 show a very weak correlation but the average score shows a stronger correlation, an indication of the importance to minimize as much as possible the insidious effect of measurement errors. Additionally, the coefficients increase when occupational status is statistically controlled, which means, as expected, that job heterogeneity decreases the beauty-wage correlation, due to the omitted confounding factors. This aside, being tall is also associated with higher earnings in both sexes, and especially in jobs involving direct personal contact. Obesity was associated with 5% wage penalty, but for women only. Harper included several interaction terms between attractiveness/unattractiveness and different job categories, so that the main effect of appearance-wage correlation must be interpreted as to say that the beauty effect exists independently of job categories. In table 7, columns 1-2 and 5-6 show an increase in ugliness-wage correlations when the interaction terms have been controlled, going from -0.057 to -0.111 and from 0.006 to 0.052 for men and women respectively. The positive effect of ugliness for women is of course unexpected, but may be due to unreliability. Indeed, they use the age-11 ratings, and not the age 7-11 aggregated. Also, one would believe that the result could have been slightly different if adult physical appearance had been used instead.
Harper (2000, Table 7) found that being short (between 0 and 9th percentile) reduces wages in both sexes but that being tall (between 80 and 89th percentile) increases wage meaningfully only among men. Obesity (higher than 90th percentile) at age 23 has more negative effect among women. In Table 6 (columns 3 & 6), with no interactions, but controlling for occupation status and other socio-economic background dummy variables, Harper found that, among men, the effect of obesity tends toward zero when controlling for the confoundings whereas, among women, the effect is still negative (coefficients around -0.050) and that the effect of being short/tall is negative/positive among men, as expected, but the signs appeared to be inconsistent among women. Harper (2000, Table 8) also investigates whether physical appearance (at age-11) affects the probability of being unemployed, with coefficients of 0.019, 0.019 and 0.038 for “average”, “attractive”, “unattractive”, among men, i.e., 0.038-0.019=0.019 or 1.9% more probability of being unemployed for ugly men. The respective numbers for women were 0.020, 0.009, 0.019, i.e., 0.020-0.009=0.011 or 1.1% less probability of being unemployed for beautiful women. Thus, the probability for ugliness is higher for men. This, in itself, is interesting if we remember Hamermesh & Biddle (1994) discussion mentioned above.
In itself, the fact that height is associated with higher earnings could be mediated through IQ, because there is a correlation between height and IQ but it may be due as well to the historical belief that taller people are more able. As Kanazawa & Kovar (2004) put it, “For instance, one of the stereotypes that people have is that high-status people are taller. In one experiment (Dannenmaier & Thumin, 1964), 46 freshmen in a nursing school estimated the height of four people they knew well who differed in their status (assistant director of the school, instructor at the school, their class president, and one specific fellow student). The students consistently overestimated the height of two high-status people (assistant director and instructor), and underestimated the height of two low-status people (class president and fellow student). In another experiment (Wilson, 1968), the same man was introduced to five different groups of students. In each group, he was introduced as someone with a different academic status (student, demonstrator, lecturer, senior lecturer, and professor). After the man left the room, the students were asked to estimate his height. The status of the stimulus person had a positive and monotonic effect on his estimated height. Participants who thought he was a student estimated him to be less than 5 ft 10 in.; those who thought he was a professor estimated him to be more than 6 ft. … It may be because higher-status persons tended to be taller than lower-status people throughout the evolutionary history. In the ancestral environment, many (if not most) competitions for status were physical, although alliances and coalitions were also important (de Waal, 1982). Our ancestors physically fought each other, and those who won the physical battle came out on top to occupy high status. In the ancestral environment, taller and bigger people therefore had an advantage over shorter, smaller competitors, and they often occupied high status. … Our stereotype that higher-status people are taller thus accurately reflects how things tended to be in the ancestral environment and, to a lesser extent, how things still tend to be in the current environment. We contend that our perception that beautiful people are more intelligent has a similar origin. Individuals believe that more attractive people are more intelligent today, because such a correlation existed in the ancestral environment, and may have survived to the current environment (like the correlation between height and status).” (pp. 231-232). If true, that means the greater reward for being beautiful and tall was merely due to unconscious thoughts. One could even assimilate it to some sort of superstitions, except that the latter has no empirical basis. In any case, the hypothesis of stereotype as a mere social construction is considerably weakened.
French (2002) replicated the earlier studies. The regression included ethnicity, marital status, occupation (e.g., professional, technical, service, transportation), experience, experience^2, working full-time, beauty (below- and above-average). The effect of appearance on wage was similar in both gender groups, but ugliness was more deleterious for men whereas beauty yielded more reward for womens. Unlike the previous studies where the majority used appearance ratings based on photographs, appearance here is a self-reported or self-assessed rating. Since that specific data is not longitudinal, it wasn’t possible to evaluate its reliability.
A criticism sometimes made against self-reported attractiveness or attractiveness reported by, say, an examiner, is that such a score cannot be accurate because beauty is “subjective”. This is true in a sense. However, the argument implies that beauty assessment becomes not very accurate, because of higher disagreement causing lower reliability. In this case, it will simply lower the correlations between attractiveness and the other variables, on the assumption of random measurement errors; reliability, in fact, can be improved by additional raters or multiple self-ratings at different points in time. Besides, wage itself is not expected to have a great reliability, especially if self-reported. This factor lowers the observed correlations even more. Furthermore, it is likely that when more people find a given person beautiful, the probability of a random person to rate the studied person as handsome is suspected to be higher, and vice versa. Likewise, if a certain person hears from someone that he is ugly as many times he hears from someone that is beautiful, the average score would be at the medium level, for example. Thus, even if self-reported attractiveness can be confounded with levels of self-confidence or self-esteem, these factors are not independent of what people can hear and can see. The weak point would be made apparent, however, when the scores are too heterogeneous. In such cases, we can admit that (non)attractiveness may be difficult to assess. But contra what the critics claimed, the accuracy of attractiveness ratings can surely be studied statistically. The reason for the common (mis)belief that beauty cannot be measured (because of it being overly subjective) is because people rely on anecdote, not data. Anecdote relies on small samples e.g., less than 5 persons, rated by perhaps 1 or 2 persons. Due to the small number of persons and raters, the amount of measurement errors is very large and so is the proportion of disagreement among the individuals. However, one detail not usually addressed is the plausible bias with beauty measured and rated by others through photographs of the subjects, i.e., when errors were not random but systematic. For example, it could be that more intelligent and educated persons know how to make and present a good photo whereas others don’t know that. It could be, also, that less intelligent and educated persons care less about their physical appearance and how they wear clothing. It remains to be seen, however, whether these factors can meaningfully bias the predicted earnings. Nonetheless, the very fact that most studies adjust for educational levels and other background variables suggests that the potential bias due to confoundings is somewhat minimized.
Cawley (2004, Tables 1 & 2) found that obese white women earn less, and this, whether or not a (7-year) lagged measure of weight is used instead of a contemporaneous value. When the potential confounding factors were included in the regression (e.g., IQ, respondents’ grade, parents’ grade, job tenure, years of actual work experience, etc.) the correlation for black women and hispanic women became meaningless, with the obesity parameter being 0.002 and -0.02, respectively, when for white women it was -0.061. When controlling for these same parameters, if anything, obese men tend to earn more, with coefficients as much as 0.013, 0.031, 0.023, for whites, blacks, hispanics, respectively. Why this differential effect ? One can believe that weight has a different meaning for men and women. Being fat for a women could be interpreted by others as they are careless about themselves, which would send a bad signal to the employers. Cawley (2004, Table 4) found that the IQ difference (ASVAB) in the NLSY1979 data is too small but that the level of attained education is different for fat/slim women of all races. However, it must be recalled that multiple regressions in principle hold constant these parameters. And yet, the null effect of obesity in hispanic and black women suggests again that some omitted parameters can play a substantial role in explaining these numbers.
Gregory & Ruhm (2009) reach more or less the same conclusion when using PSID and NLSY1979 data with a lagged or contemporaneous BMI measure. Rather unexpectedly, the BMI function for men shows an inversed U-shaped curve, low wages at low and high BMI. For white women, the wage penalty begins before obesity threshold whereas for black women the wage penalty begins at higher BMI, but still before obesity threshold. They argue the pattern does not square well with the assumption that the negative effect of women BMI on wages is mediated by health problems or medical costs (partly paid by the employer in the US through the US health insurance system) because the wage penalty starts from “healthy” to “overweight” when, it fact, such hypothesis would predict further decline from “overweight” to “obese”. The data however shows a flat curve, i.e., no relationship at this stage. They write “We are doubtful of such a mechanism for the simple reason that the conditional wage function for women turns downwards so early – at a BMI of under 23 – far below either the obesity threshold or the level at which health costs might be expected to increase. … If such expenditures are the source of the fall-off in wages, we should therefore expect, ceteris paribus, a steeper BMI-wage gradient for older than younger persons. Instead, figure 9 shows that the conditional wage function declines from its peak much more rapidly for 35-44 than for 45-55 year old women.” (pp. 18-19). This pattern is striking because obesity should be reflective of a negative market signal, so that employers would care about health costs while instead it appears having no effect.
In the above studies linking beauty with wages, none of the studies had controlled for IQ. Even if education has been taken into account in most cases, IQ and education are not perfectly correlated because it lies somewhere around 0.70.
Now, one would ask. Is it really the case that beautiful people earn more just because employers have a marked preference for beautiful people, regardless of their productivity ? The evidence suggests otherwise. Beautiful people may have some characteristics that ugly people do not have. For instance, Satoshi Kanazawa (2011) has shown that high IQ people are more physically attractive than low IQ people.
Intelligence and Physical Attractiveness (Kanazawa, 2011)
A simple look at the unstandardized coefficients, in the UK, reveals that the IQ score (derived by factor analyzing several tests and by converting it into a standard IQ metric, mean=100, SD=15) among more beautiful people (where the beauty variable is dichotomous, i.e., coded 1 if attractive and 0 otherwise) is higher by 11.39 and 13.61 for men and women, without the adjustments. And adjustments don’t do much to minimize those IQ gaps. In the US, however, the PPVT (age-standardized with mean of 100, SD=15) advantage for beautiful people by each SD increase (because the beauty variables were factor analyzed and yielded a factor score with mean=0 and SD=1) is 1.95 and 2.27 IQ points for men and women, without adjustement. However, the PPVT is not really an IQ score, just a vocabulary score, which is, by the way, overly culture loaded.
Now that we know about the IQ-beauty correlation, we have to explain why IQ can be a strong mediator of the wage-beauty correlation. Perhaps the best indirect evidence of this is provided by Mobius & Rosenblat (2005). They conducted an experimental study which reproduces as much as possible the conditions of the labor market, in an attempt to decompose the beauty premium during a wage negotiation process, where they found that a wage premium for beautiful people still exist through oral interaction (telephone interview) between worker and employer. Obviously, beautiful people seem to have better oral skills, suggesting a higher productivity. This is easily understandable if one consider the link between intelligence and beauty. Furthermore, the finding shows that indeed something other than physical attractiveness accounts for the beauty premium. This is suggested by the absence of additive effects of the beauty premium in treatment VO (visual + oral) compared to treatment O (oral).
The visual and oral stereotype effects do not seem additive: in treatments VO and FTF where employers and workers interact both visually and orally the beauty premia are only marginally greater but not significantly so.
It seems, therefore, that “visual” and “oral” represent two faces of a same coin. When they are considered together, the individual effect of “visual” seems to be weakened. If beautiful people are more intelligent, this may explain why there seems to be a beauty premium through oral interaction : intelligent people send a signal that they are more productive workers. The authors also found that confidence explains 15-20 percent of beauty premium, suggesting that beautiful people are more confident or perhaps have better self-esteem.
To summarize, we find that about 15-20 percent of the beauty premium is transmitted through the confidence channel and about 40 percent each through the visual and oral interaction channels. However, this decomposition comes with the caveat that the visual and oral stereotype channels are not fully additive: the coefficient on BEAUTY*VISUAL*AUDIO is negative and weakly significant.
A question still remain. Why intelligence correlates with beauty ? In two articles, Kanazawa speculates that this link can be explained by homogamy (assortative mating). The same point has been made before (Kanazawa & Kovar, 2004). Kanazawa (& Kovar, 2004) believes that attractiveness-intelligence correlation is extrinsic, i.e., not causal, and mainly due to assortative matings. Intelligence, height and beauty are heritable. Intelligent people tend to have higher SES. Taller people are both more intelligent and have higher SES. Health is correlated with IQ, SES and height. The more beautiful women have more chance to end up married with high SES men and by the same token taller men. Given all these synergies, IQ, height, and beauty are expected to be correlated.
A question now is whether all of these associations also operate within other ethnicities. If not, it may rise some concerns about the hypotheses advanced in explaining this pattern. Kanazawa was clear that this relationship must hold regardless of races.
- Biddle Jeff E., Hamermesh Daniel S. (1998). Beauty, Productivity, and Discrimination: Lawyers’ Looks and Lucre.
- Cawley John (2004). The Impact of Obesity on Wages.
- Engemann Kristie M. & Owyang Michael T. (2005). The Link between Wages and Appearance.
- French Michael T. (2002). Physical appearance and earnings: Further evidence.
- Frieze Irene Hanson, Olson Josephine E., Russell June (1991). Attractiveness and income for men and women in management.
- Ganzach Yoav (2011). A dynamic analysis of the effects of intelligence and socioeconomic background on job-market success.
- Gregory Christian A., Ruhm Christopher J. (2009). Where Does the Wage Penalty Bite?.
- Guéguen Nicolas (2012). Hair color and wages: Waitresses with blond hair have more fun.
- Hamermesh Daniel S., Biddle Jeff E. (1994). Beauty and the Labor Market.
- Harper Barry (2000). Beauty, Stature, and the Labour Market: A British Cohort Study.
- Johnston David W. (2010). Physical appearance and wages: Do blondes have more fun?.
- Kanazawa Satoshi (2011). Intelligence and Physical Attractiveness.
- Kanazawa Satoshi, Kovar Jody L. (2004). Why beautiful people are more intelligent.
- Landsburg Steven E. (2001). Hey, Gorgeous, Here’s a Raise! As for you fatties, we’re cutting your salaries.
- Mobius Markus & Rosenblat Tanya S. (2005). Why Beauty Matters.
- Roszell Patricia, Kennedy David, Grabb Edward (1989). Physical Attractiveness and Income Attainment Among Canadians.