In Sickness and in Wealth

Abstract

As the percentage of wives outearning their husbands grows, the traditional social norm of the male breadwinner is challenged. The upward income comparison of the husband may cause psychological distress that affects partners’ mental and physical health in ways that affect decisions on marriage, divorce, and careers. This article studies this impact through sexual and mental health problems. Using wage and prescription medication data from Denmark, we implement a regression discontinuity design to show that men outearned by their wives are more likely to use erectile dysfunction medication than their male breadwinner counterparts, even when this inequality is small. Breadwinner wives suffer increased insomnia/anxiety medication usage, with similar effects for men. We find no effects for unmarried couples or for men who earned less than their fiancée prior to marriage. Our results suggest that social norms play important roles in dictating how individuals respond to upward social comparisons.

Keywords

social comparison marriage intimacy happiness family gender roles sexual identity sexuality emotion in relationships romantic relationships regression discontinuity

The percentage of American wives who outearn their husbands has grown from 4% to 22% in the last 20 years (Fry & Cohn, 2010), with similar patterns in other countries (Bloemen & Stancanelli, 2008). Although this trend may reflect better educational and career opportunities for women (Goldin, 2006), as well as changes in marriage patterns (Schwartz & Mare, 2005), it also has major implications for psychological and physical health because it dramatically changes income comparisons in marriage. Social psychology, economics, and sociology have demonstrated that spouses and romantic partners provide highly relevant social comparison points (Beach & Tesser, 1993; Pinkus, Lockwood, Schimmack, & Fournier, 2008) and that income is a particularly salient dimension for comparison (Clark & Oswald, 1996; Luttmer, 2005).

Existing research in social psychology suggests that income comparison in marriage might substantially affect both spouses. Much of this work has been based on Tesser’s (1988) self-evaluation maintenance (SEM) model, which he and colleagues (Beach et al., 1996) extended toward comparisons in romantic couples. In the SEM model, spouses might respond negatively to upward comparison in highly relevant tasks that are self-defining, but positively in tasks that are less important to the self. Romantic or emotional closeness further amplifies these effects, but may also produce empathetic effects, where the impact of comparison on the other reflects back to the partner (Beach & Tesser, 1993).

This theory, combined with survey-based and laboratory research (Beach et al., 1998; Mendolia, Beach, & Tesser, 1996), suggests that upward comparison, even on self-defining tasks, might have minimal negative consequences in marriage. Pinkus et al. (2008), through survey and laboratory studies, even argue that the collective identity of the couple may make upward comparisons positive for the low-performing partner. Yet, Beach et al. (1998) and Pinkus et al. (2008) find some evidence of a negative impact of upward social comparison for men on highly relevant tasks, including career concerns. What if the romantic relationship itself, through norms of gender roles in marriage (e.g., Eagly, 1987), renders upward comparison on career and income dimensions negative for men?

In fact, considerable evidence exists that upward comparisons by men on the dimension of income can yield significant costs in marriage because they violate traditional social norms of male breadwinners. When women outearn their husbands, it challenges the traditional social norms and roles of the man as economic provider and female income as supplemental (Eagly, Wood, & Diekman, 2000; Johannesen-Schmidt & Eagly, 2002; Zelizer, 1994). These couples may respond by disguising and pooling finances (Atkinson & Boles, 1984), suffer lower marital satisfaction (Gray-Little & Burks, 1983), and fail to balance housework and marital power to correspond to wives’ earnings (Pollak, 1988). Men have strong preferences for outearning their wives and less desire than women for high-earning partners (Hitsch, Hortacsu, & Ariely, 2010; Zuo, 1997). Surveys show men frequently suffer depression when outearned by their wives (Crowley, 1998), partially explaining why marriages with female breadwinners are correlated with increased marital disruption and spousal abuse (Atkinson, Greenstein, & Lang, 2005; Heckert, Nowak, & Snyder, 1998). These survey results are consistent with a much broader literature demonstrating the costs of upward comparisons (e.g., Tesser, Millar, & Moore, 1988; Wheeler & Miyake, 1992) and the intensity of such comparisons with socially close others (e.g., Tesser, 1988).

In this research, we focus on an important manifestation of the psychological costs of wage comparison in marriage: sexual health. Marital sex is a frequent medium for social exchange tied to financial or social resources (Baumeister, Catanese, & Vohs, 2001). Male sexual desire and behavior is tied to cultural and social factors such as patriarchy, money, and social networks (Cornwell & Laumann, 2011; Hill, 1997; Kitzinger, 1987), potentially causing men to suffer reduced sexual desire or dysfunction when perceiving their traditional provider role to be usurped. In fact, the medical literature has shown that anger and frustration can lead to serious sexual problems such as erectile dysfunction (ED; Levy, 1994), a problem also linked to unemployment and decreased household income (Laumann, Paik, & Rosen, 1999). Similarly, work in social psychology has shown the link between sexuality and gender role conformity (Rohlinger, 2002; Sanchez, Crocker, & Boike, 2005).

We also examine related manifestations of the psychological costs of wage comparison: insomnia, anxiety, and depression. Given existing work linking these problems to spousal incomes, and their role as potential mediators for sexual dysfunction, we examine whether increases in mental health issues are also observable when marital income norms are violated. While we examine the impact of upward income comparison on male mental health, an additional benefit is that we can study how this norm violation also affects female health. Given existing evidence on empathetic responses to spouses’ negative comparisons (e.g., Lockwood, Dolderman, Sadler, & Gerchak, 2004; Pinkus et al., 2008), women may also suffer psychologically from the negative effects of wage comparison on their husbands.

We examine the relationship between marital income comparison and sexual and mental health using income and prescription medication data for more than 200,000 married Danish couples from 1997 to 2006. We first use linear regression analysis to simply describe the relationship between relative income and ED medication usage, controlling for demographic, family, and medical factors that also might affect ED. Our results show a strong relationship between relative income among couples where the woman outearns her husband and ED medication usage, but little relationship in couples with a primary male breadwinner. Furthermore, we find initial evidence of a discontinuity at the income equality point, where ED medication is significantly higher for men who are slightly outearned by their wives than for those who slightly outearn them, and that this effect is smaller for unmarried cohabitating couples.

This initial analysis is limited by its inability to control for individuals intentionally choosing their partners and careers based on their attitudes toward female breadwinners. It is also limited by omitted variables that might affect relative wages and ED medication usage, such as unobserved medical issues or job-specific stressors. To address these issues, we use a quasi-experimental regression discontinuity (RD) design that allows us to compare control and treatment groups immediately on either side of the point of marital income equality (Thistlewaite & Campbell, 1960). This analysis allows for a flexible relationship between income inequality and ED medication, while examining the discrete increase in medication usage observed near the point where women begin to outearn their husbands. This approach minimizes the limitations of the initial analysis, as couples close to this income equality point are observed to be identical apart from small differences in relative wages. The results from RD analysis show that even small differences in relative income are associated with large changes in ED medication usage when they shift the marriage from a male to a female breadwinner.

Furthermore, we find that this effect does not exist for unmarried cohabitants, suggesting that the social construct of marriage plays a critical role in how men view wage comparison. We also find that the effect only exists for couples where the man outearned the woman prior to marriage, suggesting that men who knowingly marry high-earning females suffer no psychological costs from future income comparison. We also find some evidence that the impact of wage comparison on sexual dysfunction decreases with the age of the husband, despite increased base rates of ED medication usage. This result, presented in an online appendix, is consistent with existing literature on higher testosterone increasing threats to masculinity (e.g., Willer, Rogalin, Conlon, & Wojnowicz, in press). Furthermore, wives who outearn their husbands have an increased likelihood of using benzodiazepine medication for anxiety and insomnia, and husbands suffer increased benzodiazepine and antidepressant usage. These results suggest that upward comparisons in marriage and other relationships can directly affect psychological and physical health when they violate salient social norms.

Empirical Analyses

Data

We identify the sexual cost of marital income comparison using two data sets linked through individual social security numbers in Denmark. Denmark, which ranks 14th in the United Nations Development Programme’s Gender Empowerment Measure (GEM), is a conservative setting to study traditional gender roles because of its progressive gender attitudes. Despite this, prior studies in the Netherlands, Sweden, and Denmark suggest that the cultural norm of male breadwinners is still relevant to northern Europe (Sainsbury, 1996; Sorenson & Dahl, 2012). The major benefit of studying Denmark is that the nearly complete Danish social security system enables accurate matching of individual-level databases. The first data set, the Integrated Database for Labor Market Research (IDA), contains the Danish government’s annual demographic information from the third week of November for all individuals in Denmark from 1980 to 2007.

We match the IDA data set to a second data set, the Danish Register of Medicinal Product Statistics (RMPS), maintained by the Danish Medicines Agency. This database includes all medical prescriptions for the entire Danish population from 1995 to 2007. This combination of demographic and medical data has previously been used by Dahl, Nielsen, and Mojtabai (2010) and Dahl (2011). Denmark is ideal for studying the impact of relative income on health because employment and income changes will not greatly influence access to medical care. Medication requires some small payment from patients, but public support for low-income patients ensures that payment amounts decrease heavily with the number of prescriptions. Visits to general physicians are free of charge.

The linked IDA and RMPS databases allow us to identify annual wage-based income and drug prescriptions for all wife and husband pairs as well as for cohabitating heterosexual couples and homosexual domestic partners. Using the World Health Organization’s drug classification classes (Anatomical Therapeutic Chemical [ATC] Classification System), the RMPS database allows us to identify sexual dysfunction across time through the observation of prescriptions of phosphodiesterase type 5 (PDE-5) inhibitors, which is the first line treatment for ED (Sandner, Hütter, Tinel, Ziegelbauer, & Bischoff, 2007): sildenafil (ATC: G04BE03 [Viagra]), tadalafil (ATC: G04BE08 [Cialis]) or vardenafil (ATC: G04BE09 [Levitra]). For our dependent variable, we use a dummy variable that takes a value of one if a man has at least one prescription of the three ED medications in the year following income reporting. It is important to note that prescriptions reflect those who have sought treatment for sexual problems; we cannot observe those suffering but unwilling to seek treatment. Our use of PDE-5 inhibitors as a proxy for sexual dysfunction does not allow us to distinguish between newly occurring sexual dysfunction or preexisting conditions for which the patient is newly seeking medical treatment. PDE-5 inhibitors are in some cases used for pulmonary hypertension (Sandner et al., 2007), but this condition occurs predominantly in females and typically forces early retirement, limiting the impact on our study.

To more directly test psychological costs, we also examine the use of two different types of medication classified by ATC-classes, following Dahl et al. (2010) and Dahl (2011). Insomnia and anxiety are treated with benzodiazepine-related medications (ATC: N05CF and N05BA). Anxiety and depression are treated with selective serotonin reuptake inhibitors (SSRI; ATC: N06AB). For alternative dependent variables, we create dummy variables for each of these two medication classes, for both the husband and wife. Each variable takes a value of one if the spouse held a prescription for that drug class in the year following income reporting.

In Table 1, column 1, we present summary statistics for our main sample consisting of all married heterosexual couples where both spouses are between ages 25 and 49 and where both spouses have full-time jobs. We focus on this age group for two reasons. First, there is selection bias for younger men and women due to choices to attend university and for older couples to retire early. This selection bias means that younger or older couples who are both working full-time are unlikely to be representative of the broader population. Second, the base rate for ED medication is extremely low for younger couples and much higher for the older group.¹ We include all couples within this central age range for every year between 1997 and 2006, with 569,257 unique couples and 2,831,779 total observations. All wages are inflated to 2010 price levels using a consumer price index for Denmark. To remove extreme outliers, we exclude couples with wage differences lower than the first and higher than the 99th percentile.²

Table 1.

Descriptive Statistics for Married and Unmarried Cohabitant Samples

	Married		Unmarried
	M	SD	M	SD
Erectile dysfunction	0.007	0.083	0.005	0.069
Wage difference (1,000s)	129.05	163.41	99.35	144.57
Female breadwinner	0.184	0.388	0.227	0.419
Household income	671,737	213,658	594,792	199,583
Man income	337,854	136,227	347,071	136,135
Woman income	229,032	93,448	247,721	108,784
Man age	39.48	5.94	34.64	6.33
Woman age	37.61	5.92	32.96	6.14
Man education (months)	158.1	28.6	156.7	27.5
Woman education (months)	156.2	27.3	157.1	26.1
Children (0-5 years)	0.545	0.427	0.459	0.661
Children (6-13 years)	0.771	0.844	0.354	0.653
Children (14-17 years)	0.272	0.515	0.098	0.331
Dad erectile dysfunction	0.024	0.153	0.027	0.162
Antidepressants	0.019	0.136	0.008	0.091
Insomnia	0.029	0.168	0.022	0.148
Heart condition	0.019	0.136	0.011	0.105
Diabetes	0.012	0.107	0.008	0.091
Observations	2,831,779		1,064,502

In Figure 1, we plot the distribution of wage differences for our married sample using a kernel density function,³ which creates a smoothed histogram. In Figure 2, we present kernel density distributions for male and female wage income. One concern in studying income comparison in marriage is that wage difference, our independent variable, is a difference score. Although we will address this issue later, we first present Figure 3, which shows the relationship between marital wage difference and the wife’s and husband’s wages and household income. We create this figure by aggregating couples into bins of 10,000 Danish kroner (~US$1,750) based on marital wage difference and plotting the average wages of the wife, the husband, and the total household for each bin. The three points exactly on the income equality line, for example, represent all couples where the husband’s wages are between 0 and 10,000 kroner more than his wife’s wages. The solid dots, hollow dots, and x’s represent these couples’ average household, male, and female wages, respectively.

Figure 1.

The distribution of wage income differences in Danish marriages

Figure 2.

The distribution of Danish spousal wage income

Figure 3.

Continuous relationship between marital wage difference and spousal wages

We note several important observations about Figure 3. First, for wage differences between −200,000 and 200,000 kroner, household income is remarkably constant. This suggests that any results based on wage difference within this range are unlikely to be confounded by total household income. Second, and most importantly for our subsequent RD analysis, man and woman income are continuous through the point of income equality. This suggests that the point of marital income equality does not correspond with a discrete change in either the income of the husband or the wife. Figure 4 aggregates couples into bins of 10,000 kroner based on the wage income of the wife and the husband, then separately plots the average ED medication usage for each bin.⁴

Figure 4.

Relationship between spousal wages and ED medication usage

Descriptive Analysis of Relative Wages and ED Medication

We first provide a rough descriptive analysis of the relationship between relative wage income and ED medication usage across a broad range of income differences. Although this analysis cannot suggest any causal relationship and cannot control for other factors such as choice of spouse or career, it provides an initial description of the association between ED medication usage and marital income. We use linear probability models regressing the use of ED medication on relative income in marriage.⁵ We specify these models as linear splines, where the linear relationship between ED medication and relative income is estimated separately for marriages with female breadwinners and for those with male breadwinners. The linear spline specification also allows for a discontinuous change in this relationship at the income equality point. We note that our model restricts this relationship to be linear on each side of the equality point, which we will relax later in the article. Our fully specified model is:

E D_{it + 1} = β_{0} + β_{1} (W_{i t} - W_{j t}) + β_{2} L_{i j t} + β_{3} (W_{i j} - W_{j t}) * L_{i j t} + β_{3} Y_{t} + β_{4} D_{i j t} + β_{5} F_{i} + β_{6} H_{i t} + ε_{i j t}

where ED_it+1 is a dummy variable indicating the man i in year t has a prescription for sildenafil, tadalafil, or vardenafil the following year, (W_it − W_jt) is the difference between the wage incomes of the man i and woman j in year t, and L_ijt is a dummy variable indicating the man makes less than the woman. The remaining matrices (groups of variables) represent control variables for time (year dummies Y_t), husband and wife demographics (D_ijt), family ED predisposition (F_i), and related health problems (H_it).

Heterosexual married couples

We present our results for married couples in columns 1 to 3 of Table 2, with coefficients and p values for the wage variables in parentheses.⁶ Coefficients are multiplied by 100 for presentation purposes to reflect percentage changes of ED medication.⁷ Column 1 presents the model with only time controls. In column 2, we add demographic controls, including the age and education of both spouses, total household income, and the logged number of children for three categories: ages 0 to 5, 6 to 13, and 14 to 17. We also control for family disposition—whether the husband’s father has ever been prescribed medication for ED. Finally, in column 3, we add controls for medications that may either affect ED or else reflect an underlying medical condition that affects it. For these controls, we include dummy variables reflecting the use of drugs for heart conditions and diabetes.⁸ Diabetes is treated with insulin and insulin analogues (ATC: A10), whereas heart conditions are treated with beta blocking agents (ATC: C07) and statins (ATC: C10; McVary, 2007).

Table 2.

Linear Models of the Relationship Between Marital Wage Difference and ED Medication

	Dependent variable: man uses ED medication
	1	2	3	4
Sample	Married heterosexual	Married heterosexual	Married heterosexual	Nonmarried heterosexual
Explanatory variables
Man earns less	0.104*** (0.000)	0.071*** (0.005)	0.0648*** (0.009)	0.0242 (0.405)
Wage difference	0.0001*** (0.001)	0.0003*** (0.00006)	0.0004*** (0.000)	0.0004*** (0.000)
Wage Difference × Man Earns Less	−0.0013*** (0.000)	−0.00016*** (0.005)	−0.0015*** (0.000)	−0.0012*** (0.000)
Controls
Year dummies	Included	Included	Included	Included
Demographics	—	Included	Included	Included
Parent health	—	Included	Included	Included
Physical health	—	—	Included	Included
R²	0.0016	0.0056	0.0115	0.0086
Observations	2,831,779	2,831,779	2,831,779	1,064,502

Note: ED = erectile dysfunction. All models use linear regressions. Robust standard errors are clustered at the level of the couple, with p values in parentheses. All coefficients and standard errors are multiplied by 100 to represent percentages.

p < .10. **p < .05. ***p < .01.

The results in all three columns demonstrate two notable associations between wage differences and ED medication. First, although ED medication increases with the absolute value of the wage difference for female and male breadwinner households, the magnitude of this effect is much larger when the woman outearns her husband. Although each additional 1,000 kroner increases the probability of ED medication for male breadwinners by only 0.0004% in our fully controlled Model 3 (b = .0004, p < .001), the marginal effect when women earn relatively more is 0.011% (b = -.0015, p < .001).⁹ Even this marginal impact is small, but given the mean ED medication usage of 0.6%, it implies that a wage difference of 100,000 kroner (~US$16,000) is associated with a doubling of ED medication usage—a medically and socially significant relationship. We note that the model implies that men who greatly outearn their wives also have increased usage of ED medication. Although this could be due to many factors, the relationship between male income and ED medication in Figure 4 suggests it may be related to increased stress from higher earning jobs (Laumann et al., 1999).

The second, and perhaps more important, association is at the point of wage equality. The coefficient for the dummy variable indicating that the man earns less suggests that simply being left of the equality point is associated with increased ED medication of 0.06% (b = .065, p < .01), which is a 10% increase from the 0.6% population mean. This suggests that there is discretely higher ED medication usage when the man earns slightly less than his wife, compared with when he earned slightly more. An alternative specification addresses concerns that the wage-difference coefficient reflects underlying relationships between ED medication and the individual wages of either the husband or wife. The model, adapted from Edwards (1994), is presented in the online appendix and supports our preliminary findings of increasing ED medication usage in marriages with female breadwinners.

Unmarried heterosexual cohabitants

As a comparison point, we use an alternative sample of cohabitating heterosexual couples.¹⁰ The descriptive statistics for this sample are presented in column 2 of Table 1. We run our fully controlled model from column 3 of Table 2, and present this for cohabitating couples in column 4. The results show similar linear relationships between relative wages and ED medication on both sides of the equality point (b = .0004, p < .001; b = −.0012, p < .001), but no statistically significant jump in ED medication usage at the equality point (b = .024, p = .405). This suggests that although there may be some underlying relationship between income and ED medication, there is no discrete psychological shock to a man of being outearned by an unmarried female partner. These results suggest that either marriage makes men more sensitive to upward income comparisons with their partners, or alternatively that men who are comfortable with a breadwinning partner choose not to marry.

Regression Discontinuity Design

Identifying the direct psychological impact of female breadwinners on male sexual health is difficult, because any correlation between the two could be explained by spouse and career choices, low male earnings from other health problems affecting sexual health, or lifestyle choices predicting health and earnings. Although our earlier models tried to control for these factors through medical data and alternative samples, there are almost certainly still unobserved factors that are influencing our results. An experimental design randomly assigning men and women to marriages might resolve some of these issues but is infeasible for reasons of practicality and ethics.

Instead, we use a quasi-experimental sharp regression discontinuity (RD) design, which involves assigning individual observations to a treatment (e.g., female breadwinner) or control (e.g., male breadwinner) group based on a continuous assignment variable (e.g., marital wage difference; Berk & Rauma, 1983; Imbens & Lemieux, 2008). Those observations above (or below) some discrete threshold in the assignment variable are assigned to the treatment group, whereas the others are considered controls. Examples of such threshold assignments include vehicle emissions limits (Pierce & Snyder, 2012), minimum seniority levels during layoffs (Mark & Mellor, 1991), or a majority election vote share (Dal Bo, Dal Bo, & Snyder, 2009). In each case, being slightly above or below the threshold has a decidedly different impact on the outcome (e.g., passing a test, keeping a job, or winning an election).

Despite being developed in psychology (Thistlewaite & Campbell, 1960), the use of RD in the discipline has been relatively sparse,¹¹ with more widespread implementation in economics and political science.¹² It has frequently, however, been advocated in psychology as an effective alternative to randomized experiments (Rutter, 2007; Shadish, Cook, & Campbell, 2002; West, 2009; West, Biesanz, & Pitts, 2000).

The continuous assignment variable in our model will be marital income difference, with the dependent variable as ED medication usage. We define our threshold as the exact point of income equality between husband and wife, based on a standard categorization of couple types (Winkler, McBride, & Andrews, 2005). A sharp RD design assumes that populations immediately on either side of the income equality threshold are nearly identical, as they differ only by small magnitudes in the continuous variable that assigns them to either side. The two groups near the threshold form pseudocontrol (male breadwinner) and pseudotreatment (female breadwinner) groups defined by the discrete threshold, so long as the assignment of the subject to either side is based solely on this variable (income difference). We thereby examine whether the psychological “treatment” of being outearned by a wife predicts sexual problems represented by increased prescriptions of ED medication.

Although relative marital income may have a complicated relationship with ED for reasons mentioned above, these reasons are unlikely to have a discrete impact at the point of equality. Any impact at this point we therefore interpret as the psychological impact on the man of being outearned by his wife. An additional advantage of this design is that it addresses the potential interpretation problems with difference score measures mentioned earlier in the text. So long as spousal incomes are continuous at the point of wage equality (which Figure 3 demonstrates), any discontinuous impact in wage difference cannot be explained by the individual wage components.

Our implicit assumption is that the population of couples where the husband barely outearns his wife is equivalent on all other characteristics to the population where she barely outearns him. To test this assumption, we present Table 3, which represents those populations where each spouse earns within 30,000 Danish kroner (~US$5,000) of one another (n = 409,341). As expected, the prescription drug variables are all higher for the group where the wife earns more. We present the difference in ED medication usage for these two groups in Figure 5, with 95% confidence intervals. Medication usage is higher for couples where the wife earns more (M = 0.77%) than for those where the man earns more (M = 0.70%; p < .01). More importantly, nearly every other variable is identical for the two groups—an important assumption in RD models (Lee, 2008). The lone exception is wife education, which is slightly higher in the group where the wife outearns her husband.

Table 3.

Descriptive Statistics for Marriages Within 30,000 Kroner on Each Side of the Income Equality Threshold

	Wife earns more		Husband earns more
	M	SD	M	SD
Erectile dysfunction	0.0077	0.087	0.007	0.083
Man insomnia	0.033	0.179	0.03	0.171
Man depression	0.024	0.156	0.021	0.145
Woman insomnia	0.045	0.206	0.044	0.206
Woman depression	0.031	0.173	0.032	0.176
Wage difference (1,000s)	−13.92	8.64	15.35	8.67
Man age	39.82	6.05	39.71	6.03
Woman age	38.03	5.9	37.87	5.93
Months of education (man)	153.61	28.43	153.51	28.26
Months of education (woman)	158.45	27.35	155.87	27.4
Children (0-5 years)	0.458	0.686	0.469	0.694
Children (6-13 years)	0.738	0.832	0.744	0.834
Children (14-17 years)	0.285	0.523	0.286	0.523
Household income	627,723	167,462	620,786	155,996
Observations	172,761		236,470

Figure 5.

ED medication usage for heterosexual married couples with wage differences less than 30,000 kroner

We first study the psychological and sexual cost to men using the ED drug dummy as our dependent variable. Using logistic regression, we measure the discrete impact of the psychological treatment at the point of income equality on the likelihood of using ED medication. That is, we estimate the discrete increase in medication for men earning just less than their wives, compared with the control group of men earning just more than their wives. This approach is similar to the discrete increase in ED medication observed in our earlier linear spline models, but is more robust to any nonlinear relationship between wage difference and the ED medication. Similar to Table 2, we will identify this discrete increase as the coefficient for the man earns less dummy variable.

Our RD design models all variables other than our discrete treatment variable (man earns less) as part of a continuous function. As it is important to allow this function the flexibility to be nonlinear, we use fourth-order (quartic) polynomial smoothing for the wage difference, which involves including wage difference and its squared, cubed, and fourth-order values as independent variables. We include the full set of control variables from our earlier linear models, including age, education, the ages and number of children, father ED prescriptions, year and regional dummies, medications with known ED side effects, and comorbidity. As the quartic polynomial should provide sufficient functional flexibility, any statistically significant positive coefficient on man earns less reflects a discontinuity or jump in ED medication usage that suggests psychological and sexual costs from a man earning just less than his wife.

We present a graphical representation of this model in Figure 6, which plots wage differential (man–woman) against probability of ED medication. Like earlier figures, each point represents the average ED medication rate for all couples with wage differentials within 10,000 Danish kroner (~US$1,750). Lines represent predicted quartic polynomial relationships between wage differential and medication probability on either side of the income equality point, with dashed lines representing 95% confidence intervals.¹³ The discontinuity at income equality is clearly visible. Couples where the wife outearns her husband (to the left of equality) have considerably higher ED medication usage than those where the husband earns more (to the right of equality). Figure 7 repeats the process using smaller bins of 5,000 kroner. The fitted lines are more precisely estimated due to greater numbers of bins, although the plots are noisier due to fewer observations per bin. We again observe a discontinuity, where men earning just less than their wives (the treatment group) have higher usage of ED drugs than do the control group.

Figure 6.

Regression discontinuity model of the relationship between marital wage difference and ED medication usage

Figure 7.

Regression discontinuity model with 5,000 kroner bins

Male sexual health

We present the results from our logistic regression models for ED medication in columns 1 to 3 of Table 4 for married couples. Column 1 includes only year- and wage-difference polynomial controls, whereas columns 2 and 3 add additional controls similar to Table 2. The impact of the discrete treatment at wage equality, or man earns less, is statistically significant in our fully controlled Model 3 (b = .111, p = .002, OR = 1.12).¹⁴ The odds ratio indicates that the odds that men who are barely outearned by their wives use ED medication are 1.12 times larger than the odds for those who earn slightly more (an increased risk of approximately 10%). It is important to note that similar to our earlier models, we include controls for medications used to treat diabetes, cholesterol, and heart conditions, thereby ruling out these health conditions as driving our result.¹⁵ To test for differential effects across age groups, we repeat our fully controlled Model 3 separately for couples with husbands age 18 to 34, 35 to 49, and 50 to 65. These results, presented in the online appendix, suggest that the impact of upward wage comparison on male sexual health decreases as they age, which is consistent with higher testosterone levels increasing threats to masculinity (Willer et al., in press).

Table 4.

Sharp Regression Discontinuity Models of the Impact of Female Breadwinner on ED Medications for Married Couples

	Dependent variable: M uses ED medication
	1	2	3	4	5	6	7
Model	Logit	Logit	Logit	Logit Spline	Logit	OLS	Kernel
Treatment
Man earns less	0.084** (0.020)	0.114*** (0.002)	0.111*** (0.002)	0.122** (0.039)	0.072** (0.022)	0.066** (0.013)	0.051** (0.044)
	OR = 1.087	OR = 1.121	OR = 1.118	OR = 1.130	OR = 1.075
Control variables
Quartic polynomial (wage difference)	Included	Included	Included	—	—	Included	—
Quartic polynomial spline (wage difference)	—	—	—	Included	—	—	—
Quartic polynomial (spousal wages)	—	—	—	—	Included	—	—
Year controls	Included	Included	Included	Included	Included	Included	Included
Demographics	—	Included	Included	Included	Included	Included	Included
Parent health	—	Included	Included	Included	Included	Included	Included
Physical health	—	—	Included	Included	Included	Included	Included
Pseudo R²	0.021	0.069	0.091	0.091	0.092	0.012
Observations	2,883,968	2,831,779	2,831,779	2,831,779	2,831,779	2,831,779	2,831,779

Note: ED = erectile dysfunction; OR = odds ratio. Coefficients and odds ratios presented for all logit models. OLS and kernel models are marginal percentage effects (coefficient multiplied by 100). p values presented in parentheses, and are calculated based on heteroskedasticity-robust standard errors clustered at the level of the couple.

p < .10. **p < .05. ***p < .01.

For robustness, we alternatively use separate quartic polynomials on each side of the equality point, with no substantial change in the results presented in column 4 (b = .122, p = .039, OR = 1.13). Similarly, to account for concerns about the use of a difference score as a running variable, we use quartic polynomials for husband and wife wages as our assignment variables. The coefficient for the threshold (column 5) is smaller but is consistent with other models and remains statistically significant (b = .072, p = .022, OR = 1.08).¹⁶ Column 6 presents an alternative linear specification, which is more traditionally used in RD models and is consistent in magnitude and significance. Finally, column 7 presents a kernel regression model, where we also test the robustness of our model to narrowing the bandwidth (range) of data on which model is tested.¹⁷ For a bandwidth of −200,000 and 200,000 kroner wage difference, the results are very similar and statistically significant (b = .051, p = .044), whereas smaller bandwidths produce similar coefficients with slightly weaker statistical significance.

Our RD models consistently show that married men who are slightly outearned by their wives are on average 8% to 10% (e.g., 0.66% vs. 0.6%) more likely to use ED medication that those who slightly outearn their wives.

Alternative samples

We reran our RD design with our alternative sample of heterosexual unmarried cohabitants, using a logistic regression model with full controls. We present these results in Table 5, column 1. We again find no evidence of a discontinuity (b = −.032, p = .650, OR = .968), which is consistent with marriage being critical in defining the social norm of female breadwinner. We present the average ED medication usage for the primary married sample and the heterosexual cohabitant sample together in Figure 8, collapsing all couples into 10,000 kroner bins, and adding fitted quartic polynomials and 95% confidence intervals. The discontinuity from Figures 6 and 7 is evident for married couples, whereas there is no evident discontinuity for unmarried cohabitants. These results suggest that marriage plays some role in defining the social norm of male breadwinner, as we observe no apparent ED costs in unmarried men with higher earning cohabitants.

Table 5.

Sharp Regression Discontinuity Models for Alternative ED Samples and Psychotropic Dependent Variables

	1	2	3	4	5	6	7
Sample	Unmarried cohabitants	Fiancée makes more	Fiancé makes more	All	All	All	All
Model	Logit	Logit	Logit	Logit	Logit	Logit	Logit
Dependent variable	Man ED	Man ED	Man ED	Woman Benzo	Woman SSRI	Man Benzo	Man SSRI
Treatment
Man earns less	−0.032 (0.650)	−0.053 (0.682)	0.221** (0.027)	0.046*** (0.003)	−0.017 (0.359)	0.047*** (0.009)	0.055** (0.016)
	OR = 0.968	OR = 0.949	OR = 1.247	OR = 1.047	OR = 0.983	OR = 1.048	OR = 1.056
Control variables
Quartic polynomial	Included	Included	Included	Included	Included	Included	Included
Year controls	Included	Included	Included	Included	Included	Included	Included
Demographics	Included	Included	Included	Included	Included	Included	Included
Parent health	Included	Included	Included	Included	Included	Included	Included
Physical health	Included	Included	Included	Included	Included	Included	Included
Pseudo R²	0.109	0.123	0.103	0.032	0.016	0.024	0.016
Observations	1,064,502	142,945	479,907	2,831,779	2,831,779	2,831,779	2,831,779

Note: ED = erectile dysfunction; SSRI = selective serotonin reuptake inhibitors; OR = odds ratio. Coefficients and odds ratios presented for all logit models. p values presented in parentheses, and are calculated based on heteroskedasticity-robust standard errors clustered at the level of the couple.

p < .10. **p < .05. ***p < .01.

Figure 8.

Regression discontinuity models of the relationship between ED medication and wage difference for married and unmarried cohabitants

Selection models

Given that people do not randomly choose their spouse (Burdett & Coles, 1997; Eagly, Eastwick, & Johannesen-Schmidt, 2009), we separate our married sample into two groups: those where the fiancée (wife) outearned the fiancé (husband) prior to the marriage, and those where the man was the higher premarital earner. We identify these groups based on the year prior to marriage, which restricts our sample to only those couples who were not married to their current spouse in the 1st year of our data. One would expect that men who married high-earning fiancées would not be psychologically affected by later, unfavorable wage differences. We implemented our fully controlled RD models separately on these two samples and present the results in Table 5, columns 2 and 3. In column 2, where the woman was the higher premarital earner, the impact of being outearned post-marriage is not statistically significant (b = −.053, p = .682, OR = .949). In contrast, column 3, which presents the sample where the man was the higher premarital earner, the impact is statistically significant and twice as large as the effects observed in early models (b = .221, p = .027, OR = 1.247). The odds ratios indicate that being outearned by a spouse increases the likelihood of ED medication usage by approximately 20% near the income equality point. Together, these results suggest that breadwinner status is only associated with ED medication usage for those men who did not knowingly marry a female breadwinner. Men who knowingly married a female breadwinner appear to suffer no costs from being outearned.

Mental health

Our analysis suggests a distinct psychological cost to men from being outearned by their wives. Men appear to suffer increased ED at precisely the point where wives start outearning them. Yet, what is the impact on the wife, given existing evidence in social psychology on empathetic impacts on partners in romantic relationships (e.g., Lockwood et al., 2004; Pinkus et al., 2008)? This question is particular important given the observation of decreased happiness in American women despite growth in earnings and opportunity (Stevenson & Wolfers, 2009). Furthermore, recent work has shown links in spouses’ negative moods and hormones (Fletcher, 2009; Saxbe & Repetti, 2010). We test the impact on wives’ and husbands’ mental health by regressing anxiety/depression and insomnia medications using our earlier RD logit models.

Columns 4 and 5 of Table 5 present results from fully controlled logistic regression models predicting the impact of female breadwinner status at the point of wage equality on wife anxiety/depression and insomnia medications. Results for SSRI drugs are not statistically significant (b = −.017, p = .359, OR = .983), but benzodiazapine has a small but statistically significant increased likelihood for women who slightly outearn their husbands (b = .046, p = .003, OR = 1.047). We can only speculate whether this insomnia and anxiety come from direct psychological costs or as spillovers from the reduced psychological or sexual health of husbands, and what role it might play in the sexual health of both partners. We also present the same models for husbands’ mental health in columns 6 and 7, which show a similar increase in benzodiazepine and SSRI usage in men just below the point of income equality (b = .047, p = .009, OR = 1.048; b = .055, p = .016, OR = 1.056). These results, which show increased psychotropic drug usage among men who are outearned by their wives, are consistent with the psychological costs we previously observed manifested through ED.¹⁸

Robustness Tests

Fuzziness in the discontinuity

Our use of the treatment threshold as the exact point of income equality is referred to as a “sharp” RD design. In traditional sharp RD implementation, all individuals on one side of the threshold are believed to be treated, whereas all others remain untreated. There are several important reasons why we might believe this is not the case for our setting. The first is that not all husbands who are outearned will suffer psychological or sexual costs. Second, not all husbands who do suffer such costs will be “treated” at the same level of marital income difference. Some husbands may feel emasculated even if they earn slightly more than their wives, whereas others may be unaffected so long as they earn within a certain amount of their wife. Third, not all men will have exact information about the income of themselves and their spouses. Although information on wage income is readily observable through paychecks and tax filings in Denmark, we acknowledge that for the (likely few) marriages where the wife handles all the financial and tax documentation, husbands may be unaware of relative wage income.¹⁹

These reasons all suggest that the discontinuity may be more “fuzzy” than “sharp,” which would typically suggest the use of a fuzzy RD design. In a fuzzy RD design, the treatment effect is not assumed to change from 0 to 1 exactly at the threshold. Instead, the probability of treatment is assumed to increase across the continuous assignment variable (in this case wage difference) near the threshold, with a smaller discrete jump in this probability at the threshold. Thus, in estimating a fuzzy RD design, the sharp threshold is typically used as an instrumental variable for observed treatment in a two-stage least squares regression (2SLS; Hahn, Todd, & Van der Klaauw, 2001). Although this fuzzy model fits our setting much better than the sharp RD design, we are unable to implement this design because we cannot directly observe treatment, or which men are psychologically affected by upward wage comparison with this wives.

To better understand the nature of the “fuzziness” in our treatment, we rerun our fully controlled sharp RD model using linear probability models (Table 4, column 6) 101 times, increasing the threshold in each iteration by 1,000 kroner from a minimum of −50,000 to 50,000. Figure 9 presents the discontinuity estimates (coefficients on the threshold dummy) for each iteration with 95% confidence intervals. As the figure shows, a sharp discontinuity is statistically significant at the 5% level for a range of thresholds near income equality, with the magnitude (and statistical significance) highest near 9,000 kroner. We draw two conclusions from these results. First, the fuzziness of the discontinuity is indeed substantial, such that the severe drop in ED medication usage as the man outearns his wife occurs over a range of approximately 30,000 kroner. Second, the average wage difference at which men view themselves as losing breadwinner status appears to be slightly higher than equality, which suggests that as previous work has argued (Nock, 2001; Winkler et al., 2005), many men need to significantly outearn their wives to avoid psychological and sexual costs from violating marital norms. We believe future work, combining survey-based instruments (to measure actual treatment) with medical data, could allow a more effective implementation of the fuzzy RD design.

Figure 9.

Estimated marginal change in ED medication at different regression discontinuity threshold choices

Testing for sorting

Given the fuzziness observed in the discontinuity, and our inability to observe true treatment, we are concerned that couples near the wage equality threshold may be intentionally sort to one side or the other by basing income or career decisions on the relative income of the spouses. A wife, for example, might be less aggressive in pursuing a promotion or salary increase if it were to increase her wages to slightly more than her husband. Similarly, a husband might work harder or change jobs to regain his breadwinner status. This identification concern, noted in psychology and economics (Lee & Lemieux, 2010; Shadish et al., 2002; West et al., 2000), would weaken any causal inference in the quasi-experimental design. We address this concern in two ways. First, we present the frequency distribution of wage differences for our primary married sample in Figure 10, using bins of 1,000 kroner. If there were significant endogenous sorting, we would expect a discontinuity in frequencies at the point of equality as couples sorted to one side or the equality point. We note that the distribution is reasonably continuous, although there appears to be a slight bump in frequency just to the right of equality. This suggests that a few couples may be choosing to allow the husband to slightly outearn the wife. We note, however, that this selection bias, if it exists, would likely bias against our results, as we would expect the husbands most likely to be effected by being outearned to sort from the left to the right of the equality point.

Figure 10.

Distribution of wage differences for married couples

To more formally test this continuity, we follow the suggestion of McCrary (2008) and run a sharp RD design on the density of the running variable (wage difference), using a flexible linear regression with quartic polynomial smoothing (similar to Table 4, column 4) for all couples between −100,000 and 100,000 kroner in wage difference.²⁰ We find a small discontinuity, where the number of couples per 1,000 kroner interval jumps by approximately 188 at the point of income equality. This small increase, compared with an average bin density of 2,650, suggests that a small percentage of couples may be intentionally moving from the female to the male breadwinner side. Given the small size and direction of the discontinuity, we do not believe this can explain our results, but caution that this discontinuity, plus the high number of identical marital wages explained earlier, suggests that better understanding this sorting process is important future work.

Discussion and Conclusion

In this article, we provide evidence that there is a distinct psychological and sexual cost to upward income comparisons in marriage. We observe a sharp increase (approximately 10%) in the use of ED drugs when women slightly outearn their husbands, compared with when they are slightly outearned. Similarly, we see wives near the point of income equality having increased stress or insomnia when they are the primary breadwinner. Furthermore, the increase in ED medication usage continues as the gap between the wages of the wife and husband increases. These results are consistent with a broad literature suggesting psychological and sexual costs from men losing their traditional marital role of breadwinner. More broadly, they suggest that the impact of upward social comparison in marriage depends on which partner is looking upward, as social norms in marriage dictate the traditional role of the husband as financial provider. Although we cannot observe the exact psychological mechanisms that yield sexual dysfunction, anxiety, and insomnia, our use of medication to identify psychological costs from social comparison is unique. Future work must rely on complementary survey and experimental methods to further explore which mechanisms are driving this effect.

We do not believe our results to be inconsistent with recent work by Beach et al. (1998) or Pinkus et al. (2008), who found evidence for positive benefits from upward comparisons in couples. In fact, the measures of affect in Pinkus et al. (2008) following upward and downward career comparisons, although not statistically significant, are consistent with our results. Individuals in their first study had lower affect following upward career comparison than after downward comparisons. Similarly, the second study in Beach et al. (1998) shows that a man’s mood decreases when recalling a self-relevant task on which his wife outperformed him, even when that task was also relevant for his wife. Social norms of breadwinning make income a highly relevant dimension, which suggests consistency between the studies and our field evidence. We believe our study complements their work by showing that the specific impact of social comparison in relationships depends on the dimension of comparison. Couples may indeed benefit from upward comparisons on many dimensions due to empathy (Beach & Tesser, 1993) and shared outcomes or identity (Aron, Aron, Tudor, & Nelson, 1991), but when those upward comparisons conflict with prevailing social norms of gender roles, they are likely to produce negative psychological outcomes for the individual with lower performance and for their partner who empathizes with them.

It is important to note that our RD models necessarily identify this relationship at the point of income equality. However, the data in Figure 6 and the estimates in Table 2 suggest that this effect may continue to worsen as the income gap grows, although we must be cautious in our interpretation of the purely psychological impact at other points due to selection and omitted variable bias. We must also exercise caution in this broader relationship due to the difficulty in interpreting difference scores, and the strong relationship between male income and ED medication observed in Figure 4. Finally, we note that the fuzziness of the treatment observed in Figure 9, as well as the small level of sorting suggested by Figure 10, makes our identification less precise than many RD models and require us to use wider bandwidths than are typically recommended (Imbens & Lemieux, 2008). Although we believe the fuzziness and sorting problems bias against our findings, thus making the results more remarkable, we also caution the reader that causal inference is difficult in our setting.

It is also important to note that we expect this problem to be considerably larger in countries with stronger patriarchal norms. Denmark is one of the most progressive countries on women’s issues, yet we still observe a significant role of traditional gender norms, consistent with recent work by Sorenson and Dahl (2012). We can only imagine how much more severe these problems may be in countries where social norms of male breadwinners are uniform and primary to the family social construct.²¹

We in no way suggest that the trend toward female breadwinners is socially harmful; greater equality and opportunity for women present undeniable economic and social benefits and are part of long historical trend toward greater female involvement in the labor force (Goldin, 2006). Nor do we argue that all men will respond to upward income comparisons negatively; many husbands are proud of and attracted to high-earning wives, which is likely why we do not see a larger or sharper impact in our data. Yet recent evidence suggests that gender role attitudes have changed little in the past 20 years (Cotter, Hermsen, & Vanneman, 2011). If social norms against female breadwinners continue to be strong, increasing female income will produce real costs in marriage, including the anxiety, insomnia, and ED identified here. These costs may be understated in our study, given that many women may never pursue high-paying careers due to social pressure for them to either work in the home or serve as secondary earners. Finally, we emphasize that ED and sexual health more generally have many causes, and that our findings do not discount the importance of any of them.

Footnotes

Acknowledgements

The authors acknowledge the support of Statistics Denmark and the Harrington Foundation, as well as the helpful comments and suggestions of Nicholas Argyres, Adam Augustine, Cynthia Cryder, Associate Editor Brent Donnellan, Francesca Gino, Celia Moore, Robert Pollak, Jason Snyder, Janet Schwartz, Anne Winkler, two anonymous reviewers, and seminar participants at Washington University and London Business School.

Authors’ Note

Nielsen consulted on the prescriptions and treatments.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Aron

E. N.

Tudor

Nelson

(1991). Close relationships as including other in self. Journal of Personality and Social Psychology, 60, 241-253.

Atkinson

M. P.

Boles

(1984). WASP (wives as senior partners). Journal of Marriage and the Family, 46, 861-870.

Atkinson

M. P.

Greenstein

M. P.

Lang

M. M.

(2005). For women, breadwinning can be dangerous: Gendered resource theory and wife abuse. Journal of Marriage and Family, 67, 1137-1148.

Bailey

J. M.

Kim

P. Y.

Hills

Linsenmeier

J. A. W.

(1997). Butch, femme, or straight acting? Partner preferences of gay men and lesbians. Journal of Personality and Social Psychology, 73, 960-973.

Baumeister

R. F.

Catanese

K. R.

Vohs

K. D.

(2001). Is there a gender difference in strength of sex drive? Theoretical views, conceptual distinctions, and a review of relevant evidence. Personality and Social Psychology Review, 5, 242-273.

Beach

S. R. H.

Tesser

(1993). Decision making power and marital satisfaction: A self-evaluation maintenance perspective. Journal of Social and Clinical Psychology, 12, 471-494.

Beach

S. R. H.

Tesser

Fincham

F. D.

Jones

D. J.

Johnson

Whiticker

D. J.

(1998). Pleasure and pain in doing well, together: An investigation of performance-related affect in close relationships. Journal of Personality and Social Psychology, 74, 923-938.

Beach

S. R. H.

Tesser

Mendolia

Anderson

Crelia

Whitaker

Fincham

F. D.

(1996). Self-evaluation maintenance in marriage: Toward a performance ecology of the marital relationship. Journal of Family Psychology, 10, 379-396.

Berk

R. A.

Rauma

(1983). Capitalizing on nonrandom assignment to treatments: A regression-discontinuity evaluation of a crime-control program. Journal of the American Statistical Association, 78, 21-27.

10.

Bloemen

H. G.

Stancanelli

E. G. F.

(2008). Modeling the employment and wage outcomes of spouses: Is she outearning him? (IZI Working Paper No. 3455). Retrieved from http://www.iza.org/en/webcontent/publications/papers/viewAbstract?dp_id=3455

11.

Brandt

Henry

P. J.

(2012). Gender inequality and gender differences in authoritarianism. Personality and Social Psychology Bulletin, 38, 1301-1315.

12.

Burdett

Coles

(1997). Marriage and class. Quarterly Journal of Economics, 112, 141-168.

13.

Clark

A. E.

Oswald

A. J.

(1996). Satisfaction and comparison income. Journal of Public Economics, 61, 359-381.

14.

Cook

(2008). “Waiting for life to arrive”: A history of the regression-discontinuity design in psychology, statistics, and economics. Journal of Econometrics, 142, 636-654.

15.

Cornwell

Laumann

E. O.

(2011). Network position and sexual dysfunction: Implications of partner betweenness. American Journal of Sociology, 117, 172-178.

16.

Cotter

Hermsen

Vanneman

(2011). The end of the gender revolution? Gender role attitudes from 1977 to 2008. American Journal of Sociology, 117, 259-289.

17.

Crowley

M. S.

(1998). Men’s self-perceived adequacy as the family breadwinner: Implications for their psychological, marital, and work-family well-being. Journal of Family and Economic Issues, 19, 7-23.

18.

Dahl

M. S.

(2011). Organizational change and employee stress. Management Science, 57, 240-256.

19.

Dahl

M. S.

Nielsen

Mojtabai

(2010). The effects of becoming an entrepreneur on the use of psychotropics among entrepreneurs and their spouses. Scandinavian Journal of Public Health, 38, 857-863.

20.

Dal Bo

Snyder

J. A.

(2009). Political dynasties. Review of Economic Studies, 76, 115-142.

21.

Eagly

(1987). Sex differences in social behavior: A social role interpretation. Hillsdale, NJ: Lawrence Erlbaum.

22.

Eagly

Eastwick

Johannesen-Schmidt

(2009). Possible selves in marital roles: The impact of the anticipated division of labor on the mate preferences of women and men. Personality and Social Psychology Bulletin, 35, 403-414.

23.

Eagly

Wood

Diekman

(2000). Social role theory of sex differences and similarities: A current appraisal. In Eckes

Trautner

H. M.

(Eds.), The developmental social psychology of gender (pp. 123-174). Mahwah, NJ: Lawrence Erlbaum.

24.

Edwards

(1994). The study of congruence in organizational behavior research: Critique and a proposed alternative. Organizational Behavior and Human Decision Processes, 58, 51-100.

25.

Fletcher

(2009). All in the family: Mental health spillovers effects between working spouses. BE Journal of Economic Analysis & Policy, 9(1), Article 3.

26.

Fry

Cohn

(2010). New economics of marriage: The rise of wives. Pew Research Center. Retrieved from http://pewresearch.org/pubs/1466/economics-marriage-rise-of-wives

27.

Goldin

(2006). The quiet revolution that transformed women’s employment, education, and family. American Economic Review, 96, 1-21.

28.

Gray-Little

Burks

(1983). Power and satisfaction in marriage: A review and critique. Psychological Bulletin, 93, 513-538.

29.

Hahn

Todd

Van der Klaauw

(2001). Identification and estimation of treatment effects with a regression-discontinuity design. Econometrica, 69, 201-209.

30.

Hakuta

Bialystok

Wiley

(2003). A test of the critical-period hypothesis for second-language acquisition. Psychological Science, 14, 31-38.

31.

Heckert

T. A.

Nowak

T. A.

Snyder

K. A.

(1998). The impact of husbands’ and wives’ relative earnings on marital disruption. Journal of Marriage and the Family, 60, 690-703.

32.

Hill

C. A.

(1997). The distinctiveness of sexual motives in relation to sexual desire and desirable partner attributes. Journal of Sex Research, 34, 139-153.

33.

Hitsch

G. J.

Hortacsu

Ariely

(2010). Matching and sorting in online dating. American Economic Review, 100, 130-163.

34.

Imbens

Lemieux

(2008). Regression discontinuity designs: A guide to practice. Journal of Econometrics, 142, 615-635.

35.

Johannesen-Schmidt

Eagly

(2002). Diminishing returns: The effects of income on the content of stereotypes of wage earners. Personality and Social Psychology Bulletin, 28, 1538-1545.

36.

Katz

(2001). Bias in conditional and unconditional fixed effects logit estimation. Political Analysis, 9, 379-384.

37.

Kitzinger

(1987). The social construction of lesbianism. London, England: SAGE.

38.

Kleven

H. J.

Knudsen

M. B.

Kreiner

C. T.

Pedersen

Saez

(2011). Unwilling or unable to cheat? Evidence from a tax audit experiment in Denmark. Econometrica, 79, 651-692.

39.

Laumann

E. O.

Paik

Rosen

R. C.

(1999). Sexual dysfunction in the United States. Journal of the American Medical Association, 281, 537-544.

40.

Lee

D. S.

(2008). Randomized experiments from non-random selection in U.S. House elections. Journal of Econometrics, 142, 675-697.

41.

Lee

D. S.

Lemieux

(2010). Regression discontinuity designs in economics. Journal of Economic Literature, American Economic Association, 48, 281-355.

42.

Levy

(1994). Impotence and its medical and psychosocial correlates: Results of the Massachusetts male aging study. Journal of Urology, 151, 54-61.

43.

Lockwood

Dolderman

Sadler

Gerchak

(2004). Feeling better about doing worse: Social comparisons within romantic relationships. Journal of Personality and Social Psychology, 25, 152-161.

44.

Luttmer

E. F. P.

(2005). Neighbors as negatives: Relative earnings and well-being. Quarterly Journal of Economics, 120, 963-1002.

45.

Mark

M. M.

Mellor

(1991). Effect of self-relevance of an event on hindsight bias: The foreseeability of a layoff. Journal of Applied Psychology, 76, 569-577.

46.

McCrary

(2008). Manipulation of the running variable in the regression discontinuity design: A density test. Journal of Econometrics, 142, 698-714.

47.

McVary

K. T.

(2007). Erectile dysfunction. New England Journal of Medicine, 357, 2472-2481.

48.

Mendolia

Beach

S. R. H.

Tesser

(1996). The relationship between marital interaction behaviors and affective reactions to one’s own and one’s spouse’s self-evaluation needs. Personal Relationships, 3, 279-292.

49.

Nichols

(2011). RD 2.0: Revised Stata module for regression discontinuity estimation. Retrieved from http://ideas.repec.org/c/boc/bocode/s456888.html

50.

Nock

S. L.

(2001). The marriages of equally dependent spouses. Journal of Family Issues, 22, 755-775.

51.

Organisation for Economic Co-operation and Development. (2009). Benefits and wages: Denmark. Paris, France: Author.

52.

Pierce

Snyder

(2012). Discretion and manipulation by experts: Evidence from a vehicle emissions policy change. B.E. Journal of Economic Analysis & Policy, 13(3).

53.

Pinkus

R. T.

Lockwood

Schimmack

Fournier

M. A.

(2008). For better and for worse: Everyday social comparisons between romantic partners. Journal of Personality and Social Psychology, 95, 1180-1201.

54.

Pollak

R. A.

(1988). Tied transfers and paternalistic preferences. American Economic Review, 23, 240-244.

55.

Rohlinger

D. A.

(2002). Eroticizing men: Cultural influences on advertising and male objectification. Sex Roles, 46, 61-74.

56.

Rutter

(2007). Proceeding from observed correlation to causal inference: The use of natural experiments. Perspectives on Psychological Science, 2, 377-395.

57.

Sainsbury

(1996). Gender, equality, and welfare states. Cambridge, UK: University of Cambridge.

58.

Sanchez

Crocker

Boike

(2005). Doing gender in the bedroom: Investing in gender norms and the sexual experience. Personality and Social Psychology Bulletin, 3, 1445-1455.

59.

Sandner

Hütter

Tinel

Ziegelbauer

Bischoff

(2007). PDE5 inhibitors beyond erectile dysfunction. International Journal of Impotence Research, 19, 533-543.

60.

Saxbe

Repetti

R. L.

(2010). For better or worse? Coregulation of couples’ cortisollevels and mood states. Journal of Personality and Social Psychology, 98, 92-103.

61.

Schwartz

C. R.

Mare

R. D.

(2005). Trends in educational assortative marriage from 1940 to 2003. Demography, 42, 621-646.

62.

Shadish

W. R.

Cook

T. D.

Campbell

D. T.

(2002). Experimental and quasi-experimental designs for generalized causal inference. Boston, MA: Houghton Mifflin.

63.

Sorenson

Dahl

M. S.

(2012). Geography, joint choices and the reproduction of gender inequality (Working paper). Yale School of Management and Aalborg University. Retrieved from https://dx-doi-org.web.bisu.edu.cn/10.2139/ssrn.1968440

64.

Stevenson

Wolfers

(2009). The paradox of declining female happiness. American Economic Journal: Economic Policy, 1, 190-255.

65.

Tesser

(1988). Toward a self-evaluation maintenance model of social behavior. In Berkowitz

(Ed.), Advances in experimental social psychology (Vol. 21, pp. 181-227). New York, NY: Academic Press.

66.

Tesser

Millar

Moore

(1988). Some affective consequences of social comparison and reflection processes: The pain and pleasure of being close. Journal of Personality and Social Psychology, 54, 49-61.

67.

Thistlewaite

Campbell

(1960). Regression-discontinuity analysis: An alternative to the ex post facto experiment. Journal of Educational Psychology, 51, 309-317.

68.

West

S. G.

(2009). Alternatives to randomized experiments. Current Directions in Psychological Science, 18, 299-304.

69.

West

S. G.

Biesanz

J. C.

Pitts

S. C.

(2000). Causal inference and generalization in field settings: Experimental and quasi-experimental designs. In Reis

H. T.

Judd

C. M.

(Eds.), Handbook of research methods in social and personality psychology (pp. 40-84). New York, NY: Cambridge University Press.

70.

Wheeler

Miyake

(1992). Social comparison in everyday life. Journal of Personality and Social Psychology, 62, 760-773.

71.

Willer

Rogalin

Conlon

Wojnowicz

(in press). Overdoing gender: A test of the masculine overcompensation thesis. American Journal of Sociology.

72.

Winkler

McBride

T. D.

Andrews

(2005). Wives who outearn their husbands: A transitory or persistent phenomenon for couples? Demography, 42, 523-535.

73.

Wooldridge

J. M.

(2002). Econometric analysis of cross section and panel data. Cambridge, MA: MIT Press.

74.

Zelizer

V. A.

(1994). The social meaning of money. New York, NY: Basic Books.

75.

Zuo

(1997). The effect of men’s breadwinner status on their changing gender beliefs. Sex Roles, 37, 799-816.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.92 MB