The Role of Spousal Income in the Wife’s Happiness

Abstract

Few studies have examined spousal income in the context of happiness. This paper analyzes the Indonesia family life survey and finds a positive relationship between the husband’s income and his wife’s happiness. Specifically, a 100 % increase in the husband’s income is related to a 0.72 % point increase in his wife expressing very happy, which is about 11 % of the proportion expressing that response. Surprisingly, among the husband’s characteristics, only his income (along with health) is statistically significantly related to his wife’s happiness. This positive relationship is particularly strong among old, educated, and poor (in absolute and relative terms) urban residents.

Appendices

Appendix 1: Descriptions of Covariates

The variable of years of schooling is constructed using two variables: the highest education level attended and the highest grade completed at that school. Only regular education is considered, so a small number of respondents are excluded from analysis whose highest education refers to adult education, Open University, or Islamic schools. Cognitive capacity is measured as follows. Interviewers read 10 words slowly to the respondents and then asked them to repeat the list twice: once immediately after the list was read and some minutes later for a second time. The sum of words correctly recalled in the two sessions serves as a measure of cognitive capacity.

Height is a summary measure of the pre-adulthood environment because given one’s genetic factors, more nutrition and less disease make one taller. Height in IFLS4 is measured, instead of self-reported. Obesity is measured by body mass index. The World Health Organization (2000) suggested a body mass index (BMI) of 25 or greater to define obesity for adult Asians. This paper follows this suggestion to define obesity. The final measure of health is self-reported health status. Because few respondents said, “very unhealthy,” we combine the categories of very unhealthy and somewhat unhealthy.

In addition, a perceived income ladder on a scale of one (the poorest) to six (the richest) is included. For this variable, respondents answered the following question: “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 combine the fifth and sixth steps.

A final covariate—number of children—is shared by the husband and his wife. This variable is considered because there is an extensive literature on the relationship between parenthood and happiness (Hansen 2012, for a review). In this paper, children are defined by their relationship with the household head in the household roster. In preliminary analyses, ages of children are adjusted to be under 3, 5, 7, and 15. Regardless of cut-off points, the variable is little related to happiness. Thus, we report the coefficient on number of children without age restrictions.

Table 4 Descriptive statistics

Appendix 2: Cardinal Measures of Happiness

Some researchers (e.g., Ng 1997, 2008) favored cardinalism. In this case, OLS is a better choice. Furthermore, Ferrer-i-Carbonell and Frijters (2004) demonstrated specifically for longitudinal data that using a fixed effects model could change the results to some extent, but for cross-sectional data, using an ordered probit, ordered logit, or OLS did not alter the results. In this appendix, we show that cardinalism based on OLS produces the materially same results as those produced by ordinalism based on ordered probit models.

Cardinal measures of happiness are estimated in OLS with the same covariates as those in Column 4 of Table 1. In Column 1 of Table 5, the scale of happiness ranging from one to three is placed on the left hand side of the specification. This scale is the same as that used in the ordered probit models. The result suggests that an additional percentage increase in the husband’s income is related to a 0.016 increase in this happiness scale. Recall that very unhappy is combined with unhappy because few respondents opted very unhappy. In Column 2, these two categories are separated; now, one is assigned to very unhappy, two to unhappy, three to happy, and four to very happy. Nevertheless, α ₁ is almost the same as that for the 1–3 scale. In addition, Ng (1997, 2008) preferred cardinal measures of happiness to ordinal measures, arguing that people care about net happiness (i.e., enjoyment minus suffering including spiritual and sensuous). Hence, he proposed that negative values be assigned to unhappiness and positive values to happiness. Following this proposition, minus two is assigned to very unhappy, minus one to unhappy, positive one to happy, and positive two to very happy. This scale of happiness is considered in Column 3, and the result also indicates a positive relationship between the husband’s income and the wife’s happiness. Thus, the results in the table demonstrate that this positive relationship is robust to cardinal transformations of an ordinal measure of happiness.

Table 5 Cardinal Measures of Happiness

Testing the Proportional Odds Assumption

One crucial assumption for an ordered probit is that α ₁, α _h , and α _w are the same for each value of y, referred to as the proportional odds assumption. This assumption is testable. We perform this test with covariates identical to those in Column 4 of Table 1, i.e., with the full set of covariates. Specifically, we take this assumption, which is equivalent to a standard ordered probit, and then relax the assumption for the husband’s income; the first model is nested in the second model. Then, we run a likelihood ratio test, and the result suggests that the assumption is not violated.