Testing Nash Bargaining Household Models with Time-series Data

Economists who analyze household decisionmaking allocation have traditionally assumed that the household acts as a single unit. They assume that there exists one decisionmaker whose preferences form the basis of household welfare and that all household resources are effectively pooled. This approach is known as the "unitary model," the "common preference model," or the "joint family utility model," depending on the study consulted.

In recent years, the assumptions behind this approach have been questioned. An alternative framework, the "collective model," has been explored. This model takes as its starting point the possibility that different household members have different preference orderings and that the resolution of these differences is a nontrivial problem. A plethora of theoretical models have built on this starting point, but several scholars have noted that attempts to differentiate unitary from collective models suffer from several failings. One is the problem of "observational equivalence." Certain collective models yield predictions identical to those derived from a unitary model. Another problem is that regressors that reflect bargaining power, such as an individual's income, can plausibly be regarded as endogenous, thus introducing bias into the estimated parameters of the regressors. Finally, even if a "bargaining variable" is truly exogenous, it may be correlated with some unobserved characteristic of the individual, again producing biased parameter estimates.

This paper provides an empirical test that discriminates between the unitary model and one type of collective model, the Nash bar-gaining model. The core argument of the Nash bargaining model is that a change in women's fallback position outside of the household improves their position within the household. The empirical challenge is to find such a change that is not subject to the critiques noted above.

A Turning Point in Canadian Divorce Law
The study argues that changes in provincial divorce law in Canada provide precisely such an example of an exogenous effect supporting the assumptions of the collective model. Before proceeding to the conclusions of the study, an overview of Canadian divorce law is offered to give a sense of the empirical issues at stake.

Under early common law, the husband and wife became one person by marriage, and common law added the rider that the husband became that person. The husband acquired control over all property owned by his wife at the time of marriage or acquired by her during marriage, except for minor, personal items such as jewelry and clothing. In the late 19th century, the enactment of the "Married Woman's Property Act" in the United Kingdom, and its subsequent replication in all Canadian provinces except Quebec, gave women the right to hold property in their own name, including property acquired both before and during marriage. However, the acquisition and ownership of property was tied to direct financial contribution. If a husband was the household breadwinner while his spouse remained at home, the wife had no legal claim to assets acquired in her husband's name.

...the study tests the Nash cooperative bargaining model by examining the impact of the change in divorce law on the suicide rates of Canadian women 25 and older.

A case before the Supreme Court of Canada in 1973, Murdoch v. Murdoch, was the catalyst for change in this aspect of family law. The case involved a contested divorce between a Mr. and Mrs. Murdoch. For the first four years of their marriage, the Murdochs worked on ranches as a hired couple, with their pay being given to Mr. Murdoch. These funds were used, in part, to purchase a ranch and homestead. Over the next 20 years, Mrs. Murdoch made a substantial contribution to the operation and management of the ranch. When the marriage broke down, she sought a judicial separation and claimed she was entitled to a one-half share, not only in the home-stead, but also in the ranch. However, in the absence of a direct, financial contribution or an extraordinary financial contribution, the court held that Mrs. Murdoch's actions were "just about what the ordinary rancher's wife does." As there was no explicit agreement linking Mrs. Murdoch's labor efforts to an entitlement to a share of the ranch, she was deemed to have no interest in the ranch. She thus lost her case and faced financial destitution.

Almost immediately, remedial legislative action was put into place. The Government of Ontario passed the Family Law Reform Act, 1975. It provided that "no spouse should be dis-entitled to an interest in property simply because his or her efforts were no more than might be expected from a reasonable spouse of the kind." Nonmonetary contributions to the household, such as child care, gave the contributory spouse an interest in assets acquired during marriage. It is important to note that the law applied to all marriages in existence at that time. The act was subsequently amended by the Family Law Act, 1978, which went on to "provide that upon marriage breakdown each spouse was entitled to an equal interest in the ‘family assets' regardless of who owned the assets." Other provinces followed suit, with the exception of Quebec, which had inherited and modified French civil law.

Testing the Nash Model
The study tests the Nash cooperative bargaining model by examining the impact of the change in divorce law on the suicide rates of Canadian women 25 and older (a proxy for married women). It draws on formal economic analyses of suicide to derive testable implications for the theory of household models. Using statistics on suicide rates in Ontario from 1960 to 1994, the study groups the data according to four age cohorts—excluding children and people over 65—and uses a multivariate analysis that controls for causes of suicide other than divorce law.

The Nash-bargained household model is tested with three hypotheses: (1) the change in divorce law should only reduce suicides among married women; (2) it should have a larger effect on the suicide rates of married, older women; and (3) its impact is asymmetric, that is, it should not affect the suicide rates of adult men. The results of the tests conform to the hypotheses. The data show a marked decline in female suicide rates in Ontario for the two older age cohorts from the mid-1970s onward. By contrast, there is no obvious change in the rates for the younger cohorts. Furthermore, the two older cohorts share the timing and sign of the structural break in the suicide rate. An examination of male suicide rates in Ontario also confirms the effects of change in the divorce law on women. Male suicide rates shift several years after female rates shift, suggesting that factors other than divorce law reform caused a break in the trend of men's rates.

The tests are also extended to two other provinces. In Quebec, where no legislative change took place, no structural break in the data is observed. And the results for British Columbia lend some support to the idea that the possibility that "something will improve" (in this case one's legal status) will reduce suicide. But the results for British Columbia are less than conclusive. This may be a consequence of greater measurement error in the recording of suicides and the presence of a much larger Aboriginal population in that province. Accordingly, the authors conclude that their results favor a variant of the collective model over the unitary model.

To order a discussion paper, please fill out an online order form, email IFPRI-FCN@cgiar.org or send requests to Food Consumption and Nutrition Division, IFPRI, 2033 K Street, N.W., Washington, D.C. 20006, U.S.A.