THROUGH cross-sectional analysis of the asset holdings of individual households, this study adds to the evidence available on relative risk aversion. It utilizes the National Longitudinal Surveys (NLS)1 which have advantages over the data bases used in previous studies. Assumptions about relative risk aversion are made in a number of theoretical economic and financial models.2 An empirical study by Cohn, Lewellen, Lease, and Schlarbaum (CLLS, 1975) concluded that relative risk aversion declined as wealth increased across households, while a second empirical study by Friend and Blume (F&B 1975) concluded from mixed evidence that relative risk aversion remained relatively constant as wealth increased. The present study finds that by restricting the sample to higher wealth households and by defining wealth narrowly, patterns consistent with decreasing or constant relative risk aversion emerge and these are compatible with the earlier studies. However, the use of a broader based sample and a more comprehensive measurement of wealth alters the conclusions and a pattern indicative of increasing relative risk aversion emerges. Thus, the paper cautions that constant or decreasing relative risk aversion assumptions in theoretic models may not be realistic descriptions of the risk attitudes of typical U.S. households. First, the literature on relative risk aversion will be reviewed. Second, the model used to estimate the coefficient of relative risk aversion will be discussed. The third section will present empirical results and contrast them with those of earlier studies. Finally, the material will be summarized and the conclusions indicated.