This study aims to test whether unitary models are consistent with Japanese household behavior. The methodology uses information regarding household decision-making systems to test unitary models. The method used in this study differs from those used in previous literatures. Unitary models are categorized into three types. First, a dictatorial family member in a household decides the household resource allocation based on his/her utility function. Second, all the family members have the same utility functions. Third, all the family members have different utility functions but they approve of a household utility function that represents the sum of their respective utility functions using the exogenous welfare weight. Some Japanese surveys show that the first category is not realistic. Therefore, this study considers the second and third types. This method aims to interpret the type of decision-making system selected by households as a proxy variable for the welfare weight. When controlling for individual and household characteristics, if the differences in the decision-making systems between households affect the distributions in household resource allocation, it can be concluded that one family member's preference is not similar to another's. Therefore, the second type is rejected. If the decision-making systems are endogenously determined by the extra-household environmental parameters (EEPs), it is valid to interpret that the distribution of the decision-making power is a proxy variable for the endogenous welfare weight. Therefore, the third type is rejected. The estimation strategy is used to consider the selectivity bias. In case individuals with a specific characteristic all tend to select a specific decision-making system, the effect of this decision-making system on the household resource allocation should be estimated independently from the effect of the characteristic on the allocation. If the effect of the decision-making system is not independently estimated, the estimation will include selectivity bias. In order to exclude the bias, this study employs the treatment effect model provided by Barnow, Cain, and Goldberger (1980). The analysis uses micro data from the 2000 and 2001 Japanese General Social Survey (JGSS), which is a nationwide cross-sectional survey; the analysis focuses on the data of two earner couples. The estimation results reveal that in a household wherein the wife is the principal decision-maker, the husband works 15% more and the wife works 59% less than is the case in a household that adopts a different decision-making system. The effect of the household decision-making system on labor supply is consistent with the implications derived from both the collective and Nash bargaining models. Further, the variables regarded as the EEPs (husband's and wife's wage rates, the proxy variable for nonwage income, health conditions) affect the type of decision-making system that the household selects. In particular, it is found that a spouse who is healthy and has the ability to earn a high income tends to have strong decision-making power in the household. These two results show that the unitary model is rejected, and the family members' decision-making power shows the property of endogenous welfare weight assumed by both the collective and Nash bargaining models.