Estimation of asset demands by heterogeneous agents

We develop optimization models to analyze the demand for financial assets by heterogeneous agents. The models extend Frankel's earlier approach, and relax the assumption of normality of asset returns. Instead, we assume that investors maximize an expected utility of terminal wealth based on heterogeneous attitudes toward risk. Solving a bi-level optimization program, we endogenously estimate the risk aversion parameters and derive the optimal asset holdings for each agent. The models are tested on United States market data, explaining the market structure better than previously postulated models.