Does a taller husband make his wife happier?

Personality and Individual Differences 91 (2016) 14–21

Contents lists available at ScienceDirect

Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid

Does a taller husband make his wife happier? Kitae Sohn Department of Economics, Konkuk University, 120 Neungdong-ro, Gwangjin-gu, Seoul 05029, South Korea

a r t i c l e

i n f o

Article history: Received 19 August 2015 Received in revised form 17 November 2015 Accepted 18 November 2015 Available online xxxx Keywords: Height Happiness Marriage Evolution Indonesia

a b s t r a c t Although it has been known that women prefer tall men in mating for evolutionary reasons, no study has investigated whether a taller husband makes his wife happier. We analyzed two datasets (N = 7850) that are, together, representative of the Indonesian population to determine whether this is true. A greater height difference in a couple was positively related to the wife's happiness. This relationship gradually weakened over time and entirely dissipated by 18 years of marital duration. The husband's resourcefulness was a minor mediator in the relationship. We thus argue that the husband's height and its correlates made his wife initially happy, but their influence waned over time. Nevertheless, the long period of the dissipation indicates a powerful impact of male height on women's psychology, probably prepared by evolution. © 2015 Elsevier Ltd. All rights reserved.

1. Introduction Many women say that they prefer tall men to short men. Vaillant and Wolff (2010) documented from French data collected in 1993–1999 that 41% of women specifically mentioned tallness as a desirable trait, whereas only 7.5% of men did. It could be that women only say this but do not act upon it. Pawlowski and Koziel (2002), however, analyzed the lonely hearts advertisements that appeared in Poland between 1994 and 1996, and found that taller men received more responses. This was the same for online dating in Boston and San Diego (Hitsch, Hortaçsu, & Ariely, 2010). One could still object that responses to advertisements, whether online or offline, might be just for fun with no actual consequences. Kurzban and Weeden (2005) employed data provided by HurryDate, which is a commercial dating service aimed at adult singles living in major metropolitan areas of the US; their dataset revealed the participants' choices with actual consequences. They continued to find that women chose taller men more frequently for a date. Dating does not require serious commitment, so women may behave differently in the case of marriage. Even for marriage, however, taller men are more likely to get married. Although evidence supporting this is not unanimous (Hacker, 2008), a sufficient amount of evidence has been advanced by Fu and Goldman (1996) for the US in 1979–1991, by Harper (2000) for the UK in 1991, by Murray (2000) for the 1884–1899 classes of Amherst College in the US, by Herpin (2005) for France in 2001, by Belot and Fidrmuc (2010) for UK interethnic marriages, and by Manfredini, Breschi, Fornasin, and Seghieri (2013) for two Italian communities at the turn of the 20th century.

E-mail address: [email protected].

http://dx.doi.org/10.1016/j.paid.2015.11.039 0191-8869/© 2015 Elsevier Ltd. All rights reserved.

One can attribute the origin of the female preference for male height to evolutionary processes. Taller men were perceived to be stronger, and women presumed that taller men could provide more resources and protection for them and their offspring; as a consequence, ancestral women tended to select tall men for mating (Buss, 2003; Courtiol, Raymond, Godelle, & Ferdy, 2010). Such men would also beget male offspring who in turn would be reproductively successful and carry copies of the women's genes—the sexy son hypothesis (Weatherhead & Robertson, 1979). In fact, height is correlated with many positive attributes such as physical strength, cognitive and noncognitive skills, and socio-economic status in contemporary Western countries (Case & Paxson, 2008; Lundborg, Nystedt, & Rooth, 2014; Persico, Postlewaite, & Silverman, 2004) and Indonesia (Sohn, 2015a, 2015e). Women thus perceive male height, correctly or not, as a marker of a good provider. Considering the female preference for male height, we hypothesized that a woman who marries a taller man is happier. Two points are worth recalling at this point. First, women prefer not just tall men but men taller than themselves, which is known as the male-taller norm (Pierce, 1996). Pawlowski (2003) used six pairs of human outlines with different levels of sexual dimorphism in stature (SDS) and demonstrated that Polish people adjusted their preferences for SDS in relation to their own height. Fink, Neave, Brewer, and Pawlowski (2007) used the same strategy and found the same results for people in Germany, Austria, and the UK. The norm is not universal since it was not clearly observed in some traditional societies (Sorokowski & Butovskaya, 2012; Sorokowski, Sorokowska, Fink, & Mberira, 2011; Sorokowski et al., 2015). However, Indonesia is no longer a traditional society where people hunt, gather, herd, or forage for a living. Indonesian couples exhibit an SDS of 1.07 (Sohn, 2015d, 2015e) and assortative mating (later explained in Fig. 1); an SDS of 1.07 in couples is typical among humans (Gaulin & Boster, 1992; Sohn, in press-c). Indonesians thus

K. Sohn / Personality and Individual Differences 91 (2016) 14–21

Fig. 1. Assortative mating by height.

conform to the norm in actual mating although it is another question whether Indonesians consciously consider the norm in mating (Sohn, 2015d). These findings urged us to consider not only a husband's height but also the height difference in a couple. Second, marriage is based on commitment, so marriage typically lasts for a certain period. It is thus necessary to consider marital duration. One can consider two scenarios of marital happiness over marital duration. A wife may enjoy happiness to the same degree over time. Alternatively, the sources of initial (un)happiness gradually lose their influence. The wife might get used to her husband's height and its correlates such as physical attractiveness and strength, income, wealth, health, and education. She might lose her characteristics that enabled her to marry her tall husband, such as beauty; the loss could cause her unhappiness directly or indirectly by changing her husband's behavior such as showing less affection to her, more affection to other women, and providing less childcare. We can determine the trend in the wife's happiness by interacting the height difference in a couple with their marital duration, while taking the wife's happiness as the dependent variable. If the interaction term is negative and statistically significant, the weakening hypothesis is preferred; otherwise, the constant hypothesis is preferred. Aside from the trend in happiness, one may also want to identify the sources of her happiness. There are many sources, and it is impossible to accurately measure all. We thus could not identify all sources, but two potential sources are worth considering. One is the intrinsic value of height; that is, women simply like tall men, while unable to say why. This is similar to people favoring fatty, salty, and sugary foods without knowing exactly why: such foods are essential to survival but were scarce as humans evolved; hence craving such foods increased reproductive fitness in the past (Lindeberg, 2010). Similarly, the female preference for male height increased women's reproductive fitness. We could not directly measure this source, and it was included in the “everything else” category. The other is male resourcefulness, which we could measure, although crudely. The female preference for male resourcefulness in mating is prevalent and related to height. If a husband's height acts as a marker of his resourcefulness, controlling for his resourcefulness would substantially weaken the relationship between his height and his wife's happiness. Otherwise, we have to leave the question unanswered. We tried to answer these two questions (i.e., happiness trends and sources) by using two datasets that are, together, representative of the Indonesian population. To the best of our knowledge, this study is the first to tackle this issue in the happiness literature. This study is also of

15

importance because only a few studies have discussed the relationships of spousal characteristics to happiness. For example, Groot and Van Den Brink (2002) explained how age and educational differences within couples were related to their happiness. Relatedly, García, Molina, and Navarro (2010) showed that the husband's education level was positively related to his wife's satisfaction with income. Powdthavee (2009) elucidated that one's happiness was positively related to the spouse's happiness. Guven, Senik, and Stichnoth (2012) treated happiness as an independent variable and argued that a gap between happiness levels in a couple was a good predictor for their future divorce. In all, height is missing. Moreover, this study contributes to regional studies as the country of interest is Indonesia. A series of studies by Sohn (2013a, 2013b, in press-a, in press-b) considered several aspects of happiness in Indonesia, but none of them considered height and happiness in marriage. When height and happiness in marriage are a research topic, the Indonesian population provides a straightforward case because almost all women there marry and divorce is rare (Sohn, 2015d, 2015e,in press-b). Therefore, bias resulting from selection into and out of marriage is minimal. Moreover, when height is considered, Indonesia presents an interesting case because it belonged to the region where the mean height was the shortest in the world over the past two centuries (Baten & Blum, 2012); the population remains one of the shortest populations in the world at present (Sohn, 2014, 2015a, 2015b, 2015c, 2015e, in press-a). When tallness is scarce, women may enjoy more happiness from tallness than otherwise, and consequently, we would find strong evidence of the relationship between height difference in a couple and the wife's happiness in Indonesia. 2. Data We analyzed two datasets: the Indonesian Family Life Survey (IFLS) and the Indonesian Family Life Survey East (IFLS East). The IFLS, an ongoing longitudinal survey, started collecting data on more than 22,000 individuals in 7224 households in 13 provinces in 1993 (IFLS1); the population of the provinces represented by IFLS1 accounted for 83% of the Indonesian population in the year. Four follow-ups ensued in 1997 (IFLS2), 1998 (IFLS2 +), 2000 (IFLS3), and 2007 (IFLS4). Although the survey is a longitudinal survey, we employed IFLS4 because only this follow-up contains the variable of happiness. Of the 10,994 target households, IFLS4 re-contacted 90.6% (6596 original IFLS1 households and 3366 old split-off households), and an additional 4033 new splitoff households were contacted. The IFLS has excluded most of the eastern part of Indonesia for cost and security considerations. There has been, however, a growing interest in promoting a more balanced development and in extending development benefits to the less developed eastern part. In 2012, IFLS East collected data from 7 provinces not covered by the IFLS. We combined IFLS4 and IFLS East and constructed a nationally representative dataset. Of 3159 households selected for IFLS East, 2547 (80.6%) households provided at least a partial interview. These households jointly had 10,887 household members, of which 10,759 (98.8%) provided at least a partial interview, but partial interviews were rare. In addition, 9929 (91.2%) were measured in the biomarker module (including height). These response rates led to the high quality of the data. Moreover, the questionnaires of IFLS East are almost the same as those of IFLS4, so we faced few of the problems that usually arise from merging multiple datasets of different structures. The dependent variable of interest was happiness, which was measured by the respondent's answer to the following question: “Taken all things together how would you say things are these days—would you say you were very happy, pretty happy, or not too happy?” This question listed three levels of happiness in itself (i.e., very happy, pretty happy, and not too happy), but the respondent was presented with four options: very unhappy, unhappy, happy, and very happy. This question is identical to that of the US General Social Survey and is nearly identical

16

K. Sohn / Personality and Individual Differences 91 (2016) 14–21

to that of the Euro Barometer Survey Series, and both surveys offer three levels of happiness. In addition, the response of very unhappy was provided by 87 of 29,059 respondents in the raw IFLS4 data and 42 of 5910 in the raw IFLS East data. Furthermore, the survey question listed three responses. For these reasons, we combined the responses of very unhappy and unhappy. Below, we showed that the results remained the same regardless of this combination. Kahneman and Deaton (2010) highlighted the distinction between evaluative and hedonic subjective wellbeing; the former is related to long-term wellbeing, and the latter to short-term wellbeing. Our happiness question is not a pure form of either evaluative or hedonic subjective wellbeing: “Taken all things together” invites respondents to evaluate their lives, but “happy” invites contamination by the respondent's current hedonic state. This mixed nature of the question is helpful for this study since we were concerned with current happiness, while taking into account marital duration. Although happiness levels are generally stable over time, it is not unusual to see them change (Diener, Lucas, & Scollon, 2006). Therefore, the happiness question is crude but effective in appreciating happiness over time. One of the two independent variables of interest was height difference in a couple, which was defined by the husband's height minus his wife's height. We excluded heights outside of the range 120–200 cm because they were probably recording errors.1 Specially trained nurses measured height, so there was little concern of measurement error and bias resulting from reported height. If measurement error was present in the variable, it was probably of the classical type, which would underestimate the coefficient on height difference (i.e., attenuation bias) and reduce its estimation precision. Because height difference was created by two height variables, if measurement error in height was severe, measurement error in height difference would be more so. Potential attenuation bias implies that, if the coefficient on height difference was statistically significant, this only reinforces our argument that the coefficient was indeed statistically significant; the coefficient was a lower bound. Marital duration, the other independent variable of interest, was available for ever married women aged 15–49, and for IFLS4, it was also available for women who completed the same module in IFLS3. Because women self-select into marriage, selection bias was of concern. For example, relative to an unhappy woman, if a happy woman married more often and her height difference was greater, the coefficient on height difference would be biased upward. We checked the degree of this bias. For similar reasons, survival bias was a potential threat; that is, relative to an unhappy woman, if a happy woman lived longer and her height difference was greater, the coefficient on height difference would also be biased upward. Because 90% of women in the sample belonged to the age range 15–49, survival bias was probably minimal. Nevertheless, we also checked the degree of this bias. Our focal group was married women, and it is worth noting that divorce is rare in Indonesia. For example, of men and women aged 30–65 in IFLS4, divorced (but not remarried yet) men and women accounted for only 1.5% and 3.5%, respectively. This fact is important because if happiness and divorce were closely related, divorce would cause another type of selection bias if unhappy couples tend to get divorced. In general, people are happy, married individuals are happier than unmarried individuals, and women are happier than men (Diener & Diener, 1996). This is also true in Indonesia (Sohn, 2013a, 2013b, in pres-a, in press-b). Consequently, our focal group was happier than the average Indonesian population. By construction, the happiness measure in the data had a ceiling; no one could choose an option higher than

1 Almost no people were affected by this rule, so the rule was immaterial. In IFLS4, of 22,758 individuals aged 20–60 with non-missing heights, only 174 were outside this height range. In IFLS-East, the corresponding figures were 7 of 4731.

very happy, no matter how happy she was. Hence our sample was already a happier group than the average population, who were in turn generally happy, but there was a ceiling in the response. As a result, there was not much room for improvement in happiness in our sample; that is, the variation in happiness was small, and therefore, estimation precision decreased. Our sample was extracted in a way that would make it difficult to find a statistically significant relationship between height difference and the wife's happiness. If the relationship was statistically significant, this only reinforces our argument that the relationship indeed existed and was not a false positive. We chose covariates that have been argued to exert great influence on happiness: age, years of schooling, self-reported healthy status, perceived income ladder, urban (vs. rural) residence, husband's income in the previous year, a dummy indicating zero income, and county fixed effects. To save space, we provided descriptions of the variables in the appendix. We also experimented with many other covariates used by previous studies, but the substance of the results remained the same (not shown). Upon excluding women with missing values, we were left with 7850 wives. 3. Empirical strategy Happiness was measured to be ordinal instead of cardinal, so we cannot say that the difference between three and one is somehow twice as important as the difference between two and one or three and two. Respondents used unknown cutoff points to indicate their happiness levels. An ordered probit model takes these into account and estimates a relationship between the dependent variable and an independent variable. Most happiness studies have employed this model for analyzing cross-sectional data (Ferrer-i-Carbonell & Frijters, 2004), and so did we. Nevertheless, we later performed robustness checks by using other methods. We based our empirical specification on the following simple conceptual framework (the appendix lists the empirical specification). We regarded that the wife's happiness (U) is a function of height difference, marital duration, and other covariates: U ¼ U ðHD; MD; HD  MD; X h ; X w Þ;

where HD refers to height difference, MD to marital duration, HD × MD to the interaction between the two, and Xh and Xw to sets of the husband's and wife's characteristics, respectively. We controlled for county fixed effects, county being the smallest administrative unit available. This strategy is particularly apt since Indonesia, with its 13,000 islands, is the world's largest archipelagic state and one of the most spatially diverse nations on earth in its economic development, resource endowments, population settlements, location of economic activity, ecology, and ethnicity (Hill, Resosudarmo, & Vidyattama, 2008). In addition, it is possible that some places have better resources improving the residents' happiness (e.g., better infrastructure and lower crime rates). Controlling for county fixed effects allowed us to examine the relationship between the independent variables and happiness within, instead of between, counties. We applied crosssection person weights with attrition correction to make estimations representative and clustered standard errors at the county level to account for potential correlation within counties. For ease of interpretation, we reported the mean marginal effect of each independent variable on being very happy. Considering these empirical strategies together with the nature of the sample, our results were conservative in the sense that we did our best not to cause a false positive. We did not extend the same exercise to husbands because although there is compelling and consistent evidence of the female preference for male height, the opposite is not the case. We mainly used commands of oprobit and margins in STATA ver. 13.

K. Sohn / Personality and Individual Differences 91 (2016) 14–21 Table 1 Descriptive statistics.

Continuous variable Height (cm) Height difference (cm) Marital duration (year) Age (year) Years of schooling Husband's height (cm) Ln(husband's yearly earnings)

Mean

SD

Range

151.1 10.9 16.0 36.0 7.7 162.0 15.92

5.4 7.4 10.7 9.8 4.3 6.1 1.16

125.5–184.6 −36.6 to 47.1 0–62 14–62 0–18 120–192.6 9.49–18.9 %

Discrete variable Very unhappy Unhappy Happy Very happy Unhealthy/somewhat unhealthy Somewhat healthy Very healthy Perceived income ladder 1 Perceived income ladder 2 Perceived income ladder 3 Perceived income ladder 4 Perceived income ladder 5/6 Rural residence Urban residence Ln(husband's earnings) ≠ 0 Ln(husband's earnings) = 0 N

0.27 5.76 86.25 7.72 16.06 73.68 10.25 7.20 23.02 51.87 16.33 1.58 51.81 48.18 91.73 8.27 7850

4. Results 4.1. Descriptive statistics Table 1 presents descriptive statistics. For brevity, we explained only variables closely related to the subsequent results. The mean height was 151.1 cm; this short height reflects the low income level of the population and is an extension of the historical trend in height. The mean height difference in couples was 10.9 cm. The mean marital duration was 16.0 years, which makes sense considering the mean age of wives (35.9 years). The mean age also suggests that the absolute majority was not old; recall that 90% of the sample was under age 50. Like other populations, most Indonesian wives were happy in general; 86.3% of the sample said happy, but there was enough variation in the variable because 7.7% said very happy, while 6.0% said unhappy or very unhappy. Fig. 1 presents a scatterplot of couples' heights. The plot was oval-shaped, exhibiting a positive association with a correlation coefficient of 0.18. 4.2. Effect of height difference and marital duration on happiness In Column 1 of Table 2, we controlled for the independent variables of interest (height difference and marital duration) along with age and the wife's height to estimate their total effects on happiness; subsequently, we checked their robustness to specification changes. The marginal effect of height difference suggests that a 10 cm increase in height difference was related to a 3.92 percentage point increase in the probability of saying very happy. Given that 7.72% of the sample said very happy, the size of the association (50.8%) was large.2 Other covariates could mediate the relationship between height difference and happiness. We thus further controlled for years of schooling, self-reported health status, perceived income ladder, and urban (vs. rural) residence (Colum 2). The coefficient on height difference lost statistical 2 Since the coefficient on an independent variable decreases with the dependent variable, one can roughly assess the size of the association by dividing the coefficient by the mean of the dependent variable.

17

significance, but the coefficient on marital duration turned statistically significant. It could be that time-invariant differences across counties drove the results in Column 2. We thus added county fixed effects (Column 3) and found that the coefficient on height difference increased to some degree. This increase suggests that county fixed effects did not drive the relationship between height difference and happiness. Although weakly significant, its estimation precision also improved. Moreover, the marginal effects of marital duration and age decreased, and the marginal effect of education lost statistical significance. All these changes demonstrate that time-invariant county characteristics exerted considerable effects on the relationship between the independent variables and happiness. However, the relationship between perceived income ladder and happiness hardly changed, which is consistent with the powerful influence of subjective income on happiness (Clark, Frijters, & Shields, 2008). It could be that the wife interpreted her husband's height as a marker of his resourcefulness. To check this, it was necessary to control for his resourcefulness and see whether the coefficient on height difference would considerably decrease. Albeit imperfect, we entered the natural log of the husband's earnings in the past year (along with a dummy indicating no earnings) and found a positive relationship between the husband's earnings and his wife's happiness (Column 4). A 100% increase in the husband's earnings was related to a 2.1 percentage point (27% relative to the mean) increase in his wife saying very happy. This strong relationship suggests that the husband's earnings had sufficient influence on his wife's happiness. The coefficient on height difference indeed became smaller, but not considerably. If height difference had acted only as a marker of his resourcefulness, the reduction would have been more pronounced. It appears that his resourcefulness does not much account for the relationship between height difference and the wife's happiness. That said, the overall results in the table indicate that height difference was somehow related to happiness, but its weak statistical significance implies that the relationship was not as hypothesized in the current form. When we added the squared term of height difference, neither the linear nor squared term was statistically significant (not shown). Furthermore, the considerable changes in the relationship between marital duration and happiness across columns suggest that marital duration was influenced by and influencing many factors related to happiness. As a result, we paid more attention to marital duration. 4.3. Interaction effect of height difference and marital duration on happiness The results in the previous subsection motivated us to hypothesize that height difference is positively related to happiness beyond as a marker of male resourcefulness, and the relationship weakens as the sources of initial (un)happiness lose their influence over time. We tested the hypothesis by entering the interaction of height difference and marital duration into the specification. Table 3 lists the results, and each column of the table includes the same covariates as those in the corresponding column of Table 2, but we did not list them for brevity. All coefficients on the three variables of interest were statistically significant. While height difference and marital duration were positively related to happiness, their interaction was negatively related. When we added more covariates (Column 2), unlike those in Table 2, the marginal effects of both height difference and marital duration remained statistically significant. When we added county fixed effects, just like before, the relationship between height difference and happiness increased, whereas that between marital duration and happiness decreased to some extent. Regardless, the main message continued to hold. The results in Columns 1–3 are consistent with the weakening hypothesis, but they do not distinguish whether happiness generated by height difference resulted from the husband's resourcefulness or others. To determine which was more likely, we added the natural log of the

18

K. Sohn / Personality and Individual Differences 91 (2016) 14–21

Table 2 Effect of height difference and marital duration on happiness: ordered probit. 1

2

3

4

Height difference (/100) Marital duration (/100) Height (/100) Age (/100) Years of schooling (/100) Somewhat healthy Very healthy Perceived income ladder 2 Perceived income ladder 3 Perceived income ladder 4 Perceived income ladder 5/6 Urban residence Ln(husband's earnings) Ln(husband's earnings) = 0

0.392 (0.151)⁎⁎⁎ 0.148 (0.148) 0.316 (0.224) −0.389 (0.185)⁎⁎

0.189 (0.126) 0.432 (0.160)⁎⁎⁎ 0.001 (0.202) −0.691 (0.178)⁎⁎⁎ 0.755 (0.254)⁎⁎⁎ −0.016 (0.028) 0.009 (0.035) 0.069 (0.037)⁎ 0.108 (0.034)⁎⁎⁎ 0.174 (0.041)⁎⁎⁎ 0.229 (0.063)⁎⁎⁎

0.231 (0.118)⁎ 0.311 (0.147)⁎⁎ 0.028 (0.195) −0.594 (0.171)⁎⁎⁎ 0.337 (0.245) 0.003 (0.028) 0.035 (0.031) 0.069 (0.034)⁎⁎ 0.105 (0.035)⁎⁎⁎ 0.173 (0.038)⁎⁎⁎ 0.209 (0.062)⁎⁎⁎

0.202 (0.116)⁎ 0.246 (0.143)⁎ b0.000 (0.190) −0.560 (0.167)⁎⁎⁎ 0.241 (0.241) 0.001 (0.028) 0.038 (0.031) 0.064 (0.035)⁎ 0.092 (0.035)⁎⁎⁎ 0.156 (0.039)⁎⁎⁎ 0.191 (0.061)⁎⁎⁎

0.028 (0.024)

0.027 (0.031)

0.011 (0.030) 0.021 (0.007)⁎⁎⁎ 0.025 (0.046)

County fixed effects Pseudo R squared

No 0.013

No 0.089

Yes 0.148

Yes 0.155

Notes: The sample size was 7850. We applied cross-section sample weights with attrition correction. Standard errors clustered at the county level are in parentheses. ⁎ p-Value b 0.10. ⁎⁎ p-Value b 0.05. ⁎⁎⁎ p-Value b 0.01.

husband's earnings in the previous year and a dummy indicating no earnings (Column 4). The marginal effect of height difference decreased, but the decrease was not large enough to assert that happiness owing to height difference primarily resulted from his resourcefulness. Overall, the wife was happier as her husband was taller, implying that his height correlated with many sources of her happiness. More importantly, her happiness associated with their height difference waned over time: according to Column 4, the marginal effect of height difference on the wife saying very happy completely dissipates by 18 (=0.00692/0.000386) years of marital duration.

Another concern was survival bias. Given that the life expectancy at birth was 69 years for Indonesian men and 73 years for Indonesian women in 2012, survival bias for women aged 49 and under should be small. This subgroup produced similar results for the pooled sample (Column 2). The main findings held even when we restricted the sample to ages 30–49, thereby alleviating both selection and survival bias (Column 3). Some may be concerned about the empirical model (i.e., ordered probit model). We thus employed an ordered logit model and found that this model change did not affect the main message (Column 4) in any meaningful way. We also checked whether a cardinal (instead of ordinal) level of happiness affected the result by running OLS. Although the magnitude of each coefficient was not comparable because the dependent variable was cardinal now, the sign of each coefficient confirmed that the result was robust (Column 5). It was the same when we distinguished very unhappy from unhappy and ran OLS (not shown).

4.4. Robustness checks Table 2 suggests that height difference and marital duration were positively related to happiness, but the relationships were not simply linear. Table 3 then demonstrates that marital duration mediated the relationship between height difference and happiness. Only when this interaction was taken into account, each of height difference and marital duration was statistically significantly and positively related to happiness. Incrementally adding more covariates did not change the substance of this main finding. To make the case more convincing, we performed further robustness checks after controlling for the full set of covariates (i.e., those in Column 4 of Table 2) and listed the results in Table 4. First, we checked if selection into marriage influenced the main results. We attempted to address this concern by exploiting the fact that almost all Indonesian women had ever been married by age 30 (Sohn, 2015d, 2015e, in press-b). Roughly speaking, if every woman is married, there is no selection into marriage by definition. When we analyzed this subsample (Column 1), all coefficients on height difference, marital duration, and their interaction were similar to those of the pooled sample.

5. Discussion This study was motivated by a casual observation that women value male height in mating. We tested whether a woman was happier if the height difference with her husband was greater and, if this was the case, whether her happiness owing to the height difference diminished over marital duration. We employed two datasets collected from a population who probably enjoyed much happiness from height. The datasets posed fewer concerns regarding bias resulting from selection in and out of marriage and survival. We found that height difference in a couple increased the wife's happiness, which is consistent with the casual observation. At the same time, this happiness diminished over time. Despite the decrease in her

Table 3 Interaction effect of height difference and marital duration on happiness: ordered probit.

Height difference (/100) Marital duration (/100) Height difference × duration (/1000) Pseudo R squared

1

2

3

4

0.813 (0.295)⁎⁎⁎ 0.493 (0.222)⁎⁎ −0.331 (0.163)⁎⁎

0.633 (0.249)⁎⁎ 0.809 (0.214)⁎⁎⁎ −0.350 (0.156)⁎⁎

0.728 (0.241)⁎⁎⁎ 0.728 (0.181)⁎⁎⁎ −0.392 (0.153)⁎⁎

0.692 (0.241)⁎⁎⁎ 0.658 (0.190)⁎⁎⁎ −0.386 (0.156)⁎⁎

0.018

0.095

0.156

0.163

Notes: We controlled for but not listed covariates identical to those in the corresponding column of Table 2. The sample size was 7850. We applied cross-section sample weights with attrition correction. Standard errors clustered at the county level are in parentheses. ⁎⁎ p-Value b 0.05. ⁎⁎⁎ p-Value b 0.01.

K. Sohn / Personality and Individual Differences 91 (2016) 14–21

19

Table 4 Robustness checks. 1

2

4

5

30 ≤ age ≤ 62

15 ≤ age ≤ 49

30 ≤ age ≤ 49

Ordered logit

OLS

Height difference (/100) Marital duration (/100) Height difference × duration (/1000)

0.681 (0.294)⁎⁎ 0.604 (0.242)⁎⁎ −0.397 (0.190)⁎⁎

0.693 (0.241)⁎⁎⁎ 0.660 (0.191)⁎⁎⁎ −0.387 (0.156)⁎⁎

0.683 (0.296)⁎⁎ 0.606 (0.243)⁎⁎ −0.399 (0.191)⁎⁎

0.698 (0.254)⁎⁎⁎ 0.694 (0.205)⁎⁎⁎ −0.407 (0.170)⁎⁎

1.168 (0.437)⁎⁎⁎ 1.120 (0.336)⁎⁎⁎ −0.660 (0.277)⁎⁎

Covariates same as those in Col. 4 of Table 2 N Adjusted or pseudo R squared

Yes 5414 0.219

Yes 7041 0.163

Yes 4606 0.220

Yes 7850 0.160

Yes 7850 0.180

Notes: We applied cross-section sample weights with attrition correction. Standard errors clustered at the county level are in parentheses. ⁎⁎ p-Value b 0.05. ⁎⁎⁎ p-Value b 0.01.

happiness, it is remarkable that the complete dissipation took about 18 years. Whatever the reasons for her happiness, her husband's height and its correlates are long-lasting sources of her happiness. Evidence suggests that some interventions and life events change happiness for a certain period, but it is rare to see the changes last for such a long period (Diener et al., 2006). It is no wonder that women want to marry tall men. In fact, Indonesian women value male height and its correlates in mating so much that the value of 1 cm height difference in a couple was about 3% of the husband's earnings (Sohn, 2015e). When an extensive set of covariates were taken into account, the value decreased to 1%, but the size is not negligible. We also tried to determine whether the husband's resourcefulness played a substantial role in his wife's initial happiness and its waning. We found a negative answer. One reason is that our measure of resourcefulness was too crude. A more important reason is probably that height is correlated with many characteristics of husbands and wives, and they change over time. To make matters more complicated, they change while interacting with each other. It is a limitation of this study, but it is empirically impossible to disentangle them and to estimate the influence of one source on her happiness. For example, she was happy to be married with a tall and rich man. Over time, he may lose interest in her; she then becomes less happy. Since she is less happy, he is also less happy, further losing interest in her. Even in this single case, we cannot empirically trace the origin of the vicious circle, not to mention cut the circle and estimate its influence on her happiness. This illustrates the reflective nature of marital happiness (Powdthavee, 2009). A taller husband is happier, which in turn makes his wife happier, which in turn makes her husband happier, and so on. Manski (1993) recognized this reflection problem in the distribution of behavior in a population and explained that it is possible to address only under unlikely conditions. We look forward to seeing some researchers tackle this problem, but we doubt that it is possible in the near future. We acknowledge another limitation, which is related to the reflection problem. We did not estimate any putative causality running from height difference to happiness. This is possible if we could experiment with marriages, which is out of the question, or have plausible instrumental variables (i.e., variables correlated with height difference and, as later revealed, marital duration, but uncorrelated with happiness). No one has found such variables. As far as happiness is concerned, however, Easterlin (2003) presented evidence that reverse causality was not a great concern for happiness in the marriage domain. One may wonder whether our results are generalizeable to other populations. Much evidence supports this possibility. For example, most Indonesian women were happy (Table 1), which is the same as for other populations (Diener & Diener, 1996). Indonesians exhibited assortative mating by height (Fig. 1), which is also the same for other populations (Silventoinen, 2003); its strength, measured by correlation coefficient, is the same as that for the UK (Stulp, Buunk, Pollet, Nettle, & Verhulst, 2013). Age was negatively related to happiness (Table 2), which is, at first glance, inconsistent with the U-shaped relationship

(Blanchflower & Oswald, 2008). Upon scrutiny, however, our finding is consistent because the trough of the U-shape usually took place in the 50s and 90% of our sample was under 50. Thus the age range of our sample corresponded to the left side of the U-shape. The positive relationship between education and happiness is also consistent with the literature (Michalos, 2008) and makes sense since education improves material and nonmaterial resources for happiness (Sohn, 2013a). Wives who perceived that they were richer than others were happier, which is consistent with the importance of relative rather than objective income (Clark et al., 2008). Pseudo R squared statistics in our results were small, which agrees with the literature, too (Diener, 1994, p. 115). The low R statistics supports the consistency of our results and should not discount the results; the statistics only indicate that happiness was determined by many factors other than our independent variables. Our primary interest lay in marginal effects. Future research with better data can improve upon this study. Recall that our study population is unusually short, so they are probably more sensitive to height. From this discussion, it would be of interest to check whether tall populations (e.g., North Europeans) exhibit similar patterns. Because women's preference for male height and resourcefulness in mating seems to be cross-cultural, we believe that similar patterns would be found, but to a smaller degree possibly because of satiation. Additionally, if one insists that height has no value other than as a marker of male resourcefulness, he can consider male resourcefulness in more detail to check this possibility. Acknowledgments I am grateful to the two anonymous reviewers for helpful comments and suggestions. Appendix A A.1. Model specification With an ordered probit model with linearization, our specification becomes: y ¼ α 1 HD þ α 2 MD þ α 3 HD  MD þ X h α h þ X w α w þ e; and ejHD; MD; HD  MD; X h ; X w  Normal ð0; 1Þ; where y⁎ is the latent variable of happiness, the α are (vectors of) coefficients to be estimated, and e indicates the error term. Happiness levels are defined as follows: y = 1 if y⁎ ≤ μ1 for very unhappy or unhappy y = 2 if μ1 b y⁎ ≤ μ2 for happy. y = 3 if y⁎ N μ2 for very happy, where μ1 and μ2 are unknown cutoff points, to be estimated by maximum likelihood, along with the α. Strictly speaking, the α are (vectors

20

K. Sohn / Personality and Individual Differences 91 (2016) 14–21

of) coefficients, but we also treated them as their marginal effects to save notations. A.2. Descriptions of variables Age was self-reported in years. The variable of years of schooling was constructed using two variables: the highest education level attended and the highest grade completed at that school. We excluded a small number of women whose highest education level attended was adult education, open university, pesantren (Islamic school), or school for the disabled. Self-reported health status was measured by an answer to “In general, how is your health?” The respondent answered very healthy, somewhat healthy, somewhat unhealthy, or unhealthy. Because few respondents said very unhealthy, we combined the categories of very unhealthy and somewhat unhealthy. The perceived income ladder refers to the answer to “Please imagine a six-step ladder where on the bottom (the first step) stand the poorest people, and on the highest step (the sixth step) stand the richest people. On which step are you today?” Because few respondents chose the sixth step, we combined the fifth and sixth steps. Husband's income refers to income earned during the year prior to the interview. Paid employees answered “Approximately what was your salary/wage during the last year (including the value of all benefits)?” and the self-employed answered “Approximately how much net profit did you gain last year, after taking out all your business expenses?” We inflated incomes in IFLS4 by 1.3217 to match the money value (i.e., take into account inflation rates) between the two survey years, 2007 and 2012. Income for the month prior to the interview was available, but job insecurity, economic shocks, and seasonality greatly affect workers in Indonesia. Since this variable was intended to capture the resourcefulness of the husband, a monthly time span was too short a duration to effectively capture this aspect. Although income during the past year is not the same as the husband's resourcefulness, it was the closest available. As is often the case, we took the natural logarithm of income, and we constructed a dummy indicating zero income. References Baten, J., & Blum, M. (2012). Growing tall but unequal: New findings and new background evidence on anthropometric welfare in 156 countries, 1810–1989. Economic History of Developing Regions, 27(Suppl. 1), S66–S85. http://dx.doi.org/10.1080/20780389. 2012.657489. Belot, M., & Fidrmuc, J. (2010). Anthropometry of love: Height and gender asymmetries in interethnic marriages. Economics & Human Biology, 8(3), 361–372. http://dx.doi.org/ 10.1016/j.ehb.2010.09.004. Blanchflower, D. G., & Oswald, A. J. (2008). Is well-being U-shaped over the life cycle? Social Science & Medicine, 66(8), 1733–1749. http://dx.doi.org/10.1016/j.socscimed. 2008.01.030. Buss, D. M. (2003). The evolution of desire: Strategies of human mating. New York: Basic Books. Case, A., & Paxson, C. (2008). Stature and status: Height, ability, and labor market outcomes. Journal of Political Economy, 116(3), 499–532. Clark, A. E., Frijters, P., & Shields, M. A. (2008). Relative income, happiness, and utility: An explanation for the Easterlin Paradox and other puzzles. Journal of Economic Literature, 46(1), 95–144. http://dx.doi.org/10.1257/jel.46.1.95. Courtiol, A., Raymond, M., Godelle, B., & Ferdy, J. B. (2010). Mate choice and human stature: Homogamy as a unified framework for understanding mating preferences. Evolution, 64(8), 2189–2203. http://dx.doi.org/10.1111/j.1558-5646.2010.00985.x. Diener, E. (1994). Assessing subjective well-being: Progress and opportunities. Social Indicators Research, 31(2), 103–157. Diener, E., & Diener, C. (1996). Most people are happy. Psychological Science, 7(3), 181–185. Diener, E., Lucas, R. E., & Scollon, C. N. (2006). Beyond the hedonic treadmill: Revising the adaptation theory of well-being. American Psychologist, 61(4), 305–314. http://dx.doi. org/10.1037/0003-066X.61.4.305. Easterlin, R. A. (2003). Explaining happiness. PNAS, 100(19), 11176–11183. http://dx.doi. org/10.1073/pnas.1633144100. Ferrer-i-Carbonell, A., & Frijters, P. (2004). How important is methodology for the estimates of the determinants of happiness? Economic Journal, 114(497), 641–659. Fink, B., Neave, N., Brewer, G., & Pawlowski, B. (2007). Variable preferences for sexual dimorphism in stature (SDS): Further evidence for an adjustment in relation to own height. Personality and Individual Differences, 43(8), 2249–2257. http://dx.doi.org/ 10.1016/j.paid.2007.07.014.

Fu, H., & Goldman, N. (1996). Incorporating health into models of marriage choice: Demographic and sociological perspectives. Journal of Marriage and the Family, 58(3), 740–758. García, I., Molina, J. -A., & Navarro, M. (2010). The effects of education on spouses' satisfaction in Europe. Applied Economics, 42(28), 3607–3618. http://dx.doi.org/10.1080/ 00036840802314572. Gaulin, S. J. C., & Boster, J. S. (1992). Human marriage systems and sexual dimorphism in stature. American Journal of Physical Anthropology, 89(4), 467–475. Groot, W., & Van Den Brink, H. M. (2002). Age and education differences in marriages and their effects on life satisfaction. Journal of Happiness Studies, 3(2), 153–165. Guven, C., Senik, C., & Stichnoth, H. (2012). You can't be happier than your wife. Happiness gaps and divorce. Journal of Economic Behavior & Organization, 82(1), 110–130. Hacker, J. D. (2008). Economic, demographic, and anthropometric correlates of first marriage in the Mid-Nineteenth-Century United States. Social Science History, 32(3), 307–345. http://dx.doi.org/10.1215/01455532-2008-001. Harper, B. (2000). Beauty, stature and the labour market: A British cohort study. Oxford Bulletin of Economics and Statistics, 62(S1), 771–800. Herpin, N. (2005). Love, careers, and heights in France, 2001. Economics & Human Biologys, 3(3), 420–449. http://dx.doi.org/10.1016/j.ehb.2005.04.004. Hill, H., Resosudarmo, B. P., & Vidyattama, Y. (2008). Indonesia's changing economic geography. Bulletin of Indonesian Economic Studies, 44(3), 407–435. Hitsch, G. J., Hortaçsu, A., & Ariely, D. (2010). What makes you click?—Mate preferences in online dating. Quantitative Marketing and Economics, 8(4), 393–427. http://dx.doi.org/ 10.1007/s11129-010-9088-6. Kahneman, D., & Deaton, A. (2010). High income improves evaluation of life but not emotional well-being. PNAS, 107(38), 16489–16493. http://dx.doi.org/10.1073/pnas. 1011492107. Kurzban, R., & Weeden, J. (2005). HurryDate: Mate preferences in action. Evolution and Human Behavior, 26(3), 227–244. http://dx.doi.org/10.1016/j.evolhumbehav.2004. 08.012. Lindeberg, S. (2010). Food and western disease: Health and nutrition from an evolutionary perspective. Oxford: Wiley-Blackwell. Lundborg, P., Nystedt, P., & Rooth, D. -O. (2014). Height and earnings: The role of cognitive and noncognitive skills. Journal of Human Resources, 49(1), 141–166. Manfredini, M., Breschi, M., Fornasin, A., & Seghieri, C. (2013). Height, socioeconomic status and marriage in Italy around 1900. Economics & Human Biology, 11(4), 465–473. http://dx.doi.org/10.1016/j.ehb.2012.06.004. Manski, C. F. (1993). Identification of endogenous social effects: The reflection problem. Review of Economic Studies, 60(3), 531–542. Michalos, A. C. (2008). Education, happiness and wellbeing. Social Indicators Research, 87(3), 347–366. Murray, J. E. (2000). Marital protection and marital selection: Evidence from a historicalprospective sample of American men. Demography, 37(4), 511–521. Pawlowski, B. (2003). Variable preferences for sexual dimorphism in height as a strategy for increasing the pool of potential partners in humans. Proceedings of the Royal Society of London B: Biological Sciences, 270(1516), 709–712. Pawlowski, B., & Koziel, S. (2002). The impact of traits offered in personal advertisements on response rates. Evolution and Human Behavior, 23(2), 139–149. Persico, N., Postlewaite, A., & Silverman, D. (2004). The effect of adolescent experience on labor market outcomes: The case of height. Journal of Political Economy, 112(5), 1019–1053. Pierce, C. A. (1996). Body height and romantic attraction: A meta-analytic test of the male-taller norm. Social Behavior and Personality, 24(2), 143–149. Powdthavee, N. (2009). I can't smile without you: Spousal correlation in life satisfaction. Journal of Economic Psychology, 30(4), 675–689. Silventoinen, K. (2003). Determinants of variation in adult body height. Journal of Biosocial Science, 35(2), 263–285. http://dx.doi.org/10.1017/s0021932003002633. Sohn, K. (2013a). Monetary and nonmonetary returns to education in Indonesia. The Developing Economies, 51(1), 34–59. http://dx.doi.org/10.1111/deve.12001. Sohn, K. (2013b). Sources of happiness in Indonesia. Singapore Economic Review, 58(2), 1350014. http://dx.doi.org/10.1142/s0217590813500148. Sohn, K. (2014). Age and size at maturity in Indonesian women: A norm of reaction? American Journal of Human Biology, 26(5), 713–715. http://dx.doi.org/10.1002/ajhb.22571. Sohn, K. (2015a). Height and happiness in a developing country. Journal of Happiness Studies. http://dx.doi.org/10.1007/s10902-014-9566-8 (in press-a). Sohn, K. (2015b). The role of spousal income in the wife's happiness. Social Indicators Research. http://dx.doi.org/10.1007/s11205-015-0934-5 (in press-b). Sohn, K. (2015c). Sexual stature dimorphism as an indicator of living standards? Annals of Human Biology (in press-c). Sohn, K. (2015d). The height premium in Indonesia. Economics and Human Biology, 16, 1–15. http://dx.doi.org/10.1016/j.ehb.2013.12.011. Sohn, K. (2015e). The influence of birth season on height: Evidence from Indonesia. American Journal of Physical Anthropology, 157(4), 659–665. Sohn, K. (2015f). Is leg length a biomarker of early life conditions? Evidence from a historically short population. American Journal of Human Biology, 27(4), 538–545. http://dx. doi.org/10.1002/ajhb.22682. Sohn, K. (2015g). The male-taller norm: Lack of evidence from a developing country. HOMO — Journal of Comparative Human Biology, 66(4), 369–378. http://dx.doi.org/ 10.1016/j.jchb.2015.02.006. Sohn, K. (2015h). The value of male height in the marriage market. Economics & Human Biology, 18, 110–124. Sorokowski, P., & Butovskaya, M. L. (2012). Height preferences in humans may not be universal: Evidence from the Datoga people of Tanzania. Body Image, 9(4), 510–516. http://dx.doi.org/10.1016/j.bodyim.2012.07.002. Sorokowski, P., Sorokowska, A., Butovskaya, M., Stulp, G., Huanca, T., & Fink, B. (2015). Body height preferences and actual dimorphism in stature between partners in two

K. Sohn / Personality and Individual Differences 91 (2016) 14–21 non-western societies (Hadza and Tsimane'). Evolutionary Psychology, 13(2), 455–469. Sorokowski, P., Sorokowska, A., Fink, B., & Mberira, M. (2011). Variable preferences for sexual dimorphism in stature (SDS) might not be universal: Data from a seminomad population (Himba) in Namibia. Journal of Cross-Cultural Psychology, 43(1), 32–37. http://dx.doi.org/10.1177/0022022110395140. Stulp, G., Buunk, A. P., Pollet, T. V., Nettle, D., & Verhulst, S. (2013). Are human mating preferences with respect to height reflected in actual pairings? PloS One, 8(1), e54186. http://dx.doi.org/10.1371/journal.pone.0054186.

21

Vaillant, N. G., & Wolff, F. -C. (2010). Positive and negative preferences in human mate selection. Review of Economics of the Household, 9(2), 273–291. http://dx.doi.org/10. 1007/s11150-010-9089-y. Weatherhead, P. J., & Robertson, R. J. (1979). Offspring quality and the polygyny threshold: “The sexy son hypothesis”. American Naturalist, 113(2), 201–208.