«Trevor Breusch and Edith Gray (trevor.breusch The Australian National University Abstract It has often been observed that married men ...»
Does marriage improve the wages of men and women in Australia?
Trevor Breusch and Edith Gray (firstname.lastname@example.org)
The Australian National University
It has often been observed that married men have higher wages than unmarried men, in
what is described as a ‘marriage premium’. For women, the evidence is much less
conclusive, with various researchers reporting positive but small, or zero or even negative
effects of marriage on women’s remuneration. We explore these issues in Australia using data from the HILDA survey collected in 2001. Our finding for men is consistent with the bulk of other evidence, and for women we find a smaller but nevertheless substantial positive premium. We further examine the concept of ‘marriage’, which has broadened in recent years, and explore the differences in wage outcome between legal-, de facto- and ex-marriage states. There is never a significant difference in the premium between men in the various marriage states, while for women the differences between marriage states are numerically large and sometimes statistically significant.
Paper prepared for the 12th Biennial Conference of the Australian Population Association, 15-17 September 2004, Canberra.
1. Introduction It has often been observed that married men have higher wages than unmarried men, in what is commonly referred to as the ‘marriage premium’ (Korenman and Neumark, 1991; Gray, 1997).
There is a vast amount of literature on the marriage premium for men from research in the United States and Europe, but there is a surprisingly small amount of Australian research (one exception is Chalmers, 2002). While the research has concentrated on men and marriage, there has been less interest in a women’s premium, and what evidence there is on women is decidedly mixed. In recent decades, rising rates of cohabitation and marital breakdown have brought change to the concept of ‘marriage’, which further questions what is being rewarded and who receives the rewards.
So why might a married man earn more than a single man? There are two major theoretical explanations of the premium to a man’s wage from being married. The first is that marriage makes a man more productive through intrahousehold specialisation between husbands and wives (Becker, 1981; 1985; Korenman and Neumark, 1991). Men are freed from household labour by the traditional division of responsibilities, and hence allowed to pursue specialisation in the labour market. Now able to maximise their potential in the workplace, they obtain higher wages (and perhaps work longer hours) and thus bring additional income to the household. Wives are also better off in this model, not through higher wages of their own, but rather through access to their husbands’ higher incomes.
The second explanation describes a selection effect whereby a man who is favoured in the labour market is also favoured as a potential marriage partner (and vice versa) (Becker, 1981).1 Here it is argued that men with certain characteristics—notably qualities of their human capital and personality that may not be obvious to an outside observer—are more likely to marry and also more likely to be successful in employment. This explanation is not as widely supported in the literature as the specialisation mechanism, but it does emphasise the importance of controlling for other characteristics associated with earnings when comparing the wages of men in different relationship states (whether they be single, widowed or divorced, cohabiting or married).
While the intrahousehold specialisation argument is dominant in the literature, some research has cast doubts on this explanation. For example, Hersch and Stratton (2000) find the wage premium for being married is the same for men regardless of the amount of time they spend on A third theoretical proposition is found in the literature, suggesting positive discrimination toward married men as compared to unmarried men (see Hill, 1979).
housework. There have, however, been reports of recent declines in the marriage premium, which have been attributed to declining productivity effects as household specialisation decreases (Gray, 1997).
It is apparent that the male breadwinner model of the family is less applicable to modern marriages. The social changes in western societies since the middle of last century have brought declines in intrahousehold specialisation. The main marker of these changes is the strong increase in women’s labour force participation, particularly that of married women. This participation by women in the public sphere of paid work has had other consequences for intrahousehold dynamics.
As women have increased economic independence, the time spent on housework and family responsibilities by husbands and wives has converged. Men’s participation in domestic duties has increased—albeit it marginally2—although the convergence of time spent on domestic tasks is primarily driven by a decrease in the amount done by women (Gershuny and Robinson, 1988). The changes evident within marriage relationships are also reflected in society more generally.
Attitudinal research shows that over 65 per cent of 18 to 54 year olds in Australia disagree that the husband should be the principal breadwinner and the wife should have primary responsibility for the home and children (McDonald, et al. 1999).
If the premium that has been observed for men is the result of favourable selection, then married women should receive similar premia to their male counterparts. It is argued that if a ‘symmetric model’ of marriage is favoured over a ‘specialisation model’, then women are valued for their earnings potential rather than as simply a support person for a breadwinner (Cherlin, 2000).
Hence in societies where women and men are expected to contribute to both the household income and household tasks, women’s economic potential would be seen as an important factor in marriage formation. This would suggest that a marriage premium should be found for women.
Some of the decline in intrahousehold specialisation may be attributed to the large increases in the proportion of people cohabiting both before, and as an alternative to, legal marriage. In 2002 it was reported that 70 per cent of marriages were preceded by cohabitation (ABS, 2003), compared to 22 per cent in 1978 (ABS, 1999). This widespread change in family formation is also associated with revised responsibilities in the household. It is suggested that cohabiting couples share the division of household labour more than married couples do (South and Spitze, 1994), and that married couples who cohabit prior to marriage are less likely to embrace the traditional division of household labour upon marriage (Baxter, 2001). The increase in cohabitation and the different Hence the predictions by social scientists that as women’s labour force participation increased, so too would men’s involvement in domestic duties, have largely not been fulfilled (Hochschild, 1989).
behaviour evidenced in cohabiting relationships suggest that this relationship type should be identified separately in analyses of the marriage premium.
Cohen (2002) suggests there are two main reasons cohabiters have been treated as single or unmarried in previous research in the United States. First, cohabitation in the past was relatively rare and hence not considered to have a major influence on findings, and second, cohabitation information was just not available in the data on earnings. Cohen finds that the apparent decline in the male marriage premium in the US is not as prominent if cohabiters are not included as unmarried men.
Recent US research examines differences in the marriage and cohabitation income premium including women as well as men (Light, 2004). Although the focus of that paper is on changes in total family income according to movement between marital states, the author finds that there is no difference for women according to their own income in different marital states, while married men have substantially higher income than single or cohabiting men.
We should not assume that results in the dominant US literature will carry over to Australia.
There are substantial differences in the social and legal frameworks in these two countries, which may make for important variations in behaviour between cohabiting and married couples. Such differences in institutional settings have been found important in related research, such as Heimdal and Houseknecht (2003) who examine differences in cohabiting and married couples’ income organisation between Sweden and the US.
This paper examines the marriage premium in Australia. We consider women as well as men, and we examine the other marriage states of cohabitation and ex-married in addition to the traditional dichotomy of legally married and single. In support of the selection explanation for the premium, we find married men continue to enjoy a strong premium in their wages despite the reported decline in intrahousehold specialisation, and that women too have a substantial positive marriage premium. The women’s premium is smaller than for men, suggesting that some of the premium that is observed for men may be due to specialisation that is not available to working women.
We hypothesise that as intrahousehold specialisation is less pronounced in cohabiting couples than married couples, we should find a lesser premium for cohabitation than marriage. However, we suggest that the difference between the premium of marrieds and cohabiters may not be a strong effect as reported in US studies, because cohabitation is such a common and legally recognised institution in Australia. Indeed in the case of men, we find that cohabiters have a wage premium that is not significantly different from the legally married, although we find the premium for cohabiters is markedly less in the case of women.
We also find that ex-married men have a wage premium over those who are single and never married that is quite large, although it is considerably less than the premium of the currently partnered men (whether married or cohabiting). However, ex-married women have a premium over the singles that is trivially small and insignificant.
HILDA is a longitudinal survey of Australian households that has been funded by the Commonwealth Government. It investigates life in Australia, focussing on income, labour market, and family dynamics. We use the second release of the Wave 1 data, collected between August 2001 and January 2002. See Watson and Wooden (2002) for overview.
The survey used four questionnaires, including a household form, a household questionnaire, a person questionnaire for all household members aged 15 years and over, and a self-completion questionnaire. In Wave 1, all components were administered by personal interview except for the self-completion questionnaire.
The reference population is all members of private dwellings with some exceptions. A multistage approach is used to select households. There are 11,693 in-scope households of which 6,872 (59 per cent) fully-responded and 810 (7 per cent) partially-responded. Of potentially eligible adults in the households involved, 92 per cent of individuals responded (13,969 respondents). A smaller proportion of individuals completed the self-completion questionnaire (87 per cent).
HILDA is generally representative of the Australian population although there are some differences. Sydney residents are somewhat underrepresented, as are males, unmarried persons and immigrants from a non-English-speaking background.
We want to estimate the effects on a person’s potential wage of hypothetically changing some of their characteristics, where this person is a typical man or woman from the population, and not just someone who is observed to be working at the time of the survey. This is a version of the wellknown ‘selectivity’ problem, and brings an estimation problem associated with the zero wage rate recorded for anyone who is not working when sampled. The statistical model we use has the standard form of a Tobit Type II selection (or Heckman) model consisting of two equations, one for the wage rate and the other for selection into employment,
Here yi∗ and zi∗ are latent variables, the former representing ‘potential wage’ (actually the natural logarithm of the potential wage rate) and the latter representing something like ‘the propensity to be employed’. The indicator that the person is employed is the event zi∗ 0. Wage wi is formed by exponentiating log-wage and it is observed only when the person is employed. The two parts of the model are tied together through the joint error process, which is represented as a bivariate normal distribution in which the correlation is determined by the parameter ρ. For simplicity the explanatory variables xi are written the same in both equations, which implies that each of the coefficient vectors β and γ may need some restriction imposed on it for the other coefficient vector to be well identified. See Verbeek (2000, p. 207) for an exposition of this class of models.
There are many factors that determine a worker’s wage rate, just as there are many factors that influence whether a person is working or not. We largely follow the received literature when specifying the variables to control for in our joint model of employment and wage rates. Much of the observed variation in wages between people can be explained by differences in general education, job-specific training, experience in the workforce, and tenure in the same position (or with the same employer). These measures are usually grouped together as contributions to the person’s human capital, which in turn is interpreted as the productive resource that influences their rewards in employment. Several studies underscore the importance of distinguishing between a person’s age and their actual labour market experience: advancing age without addition to experience is a marker of the depreciation of human capital. The impact of a higher stock of human capital is to increase the productivity of employment, and hence to increase both the person’s wage rate and the chances of finding them in paid employment.
For women, a factor associated with wage rates is whether they have ever had a child or not.