Varadarajan V. Chari, Principal Investigator

Computation and Estimation of Dynamic General Equilibrium Model with
Microeconomic Heterogeneity

  The goal of this project was to compute and estimate a dynamic, general equilibrium model of life-cycle consumption with endogenous borrowing constraints. This project attempted to advance the understanding of computation, estimation, and economic behavior of dynamic general equilibrium models with microeconomic heterogeneity.

  Standard, complete markets models of consumption cannot explain the observed hump-shape pattern of consumption over the life-cycle. The research addressed this puzzle using a life-cycle model with durable goods and endogenous debt constraints. The key idea is that households' desire to accumulate durable goods at the beginning of their life cycle is limited by the enforceability of debt contracts. In case of default, durable goods used as collateral do not generally cover all outstanding payments because of their low liquidation value. Then the threat of exclusion from intertemporal trade must be used as an additional enforcement mechanism. These constraints change endogenously over the life-cycle, forcing households to deviate from consumption smoothing and to delay investment in liquid assests, which generates in that way an endogenous distribution of agents. The main consequence is that, if the income profile follows a hump-shape, consumption will also follow this pattern.

  After a characterization of the equilibrium path, the model was estimated and simulated using data from the Consumer Expenditure Survey. The results will be compared with other recent papers that have tried to explain this puzzle using models where exogenous (and time invariant) borrowing constraints coexist with income uncertainty.

  The goal of a further project undertaken by this group was to estimate a model of the career decisions of U.S. politicians, focusing on Congressional careers. In this model, a politician first decides whether or not to run for the House of Representatives. If the person is elected, then in the next period he or she must decide whether to:

• run for reelection,

• run for election to higher office (e.g., the Senate),

• exit politics for a private sector job, or

• retire from politics.

  If reelected, the politician faces the same set of choices in the next period.

  The key feature of this model is that the researchers recognized that the choice problem faced by politicians is "dynamic." By this is meant that each choice option not only generates current period rewards, but also future rewards. Specifically, the payoff to winning a seat in Congress is not only the wage and prestige that a person recieves from being in Congress for one term. It also includes the increment in expected future payoffs that accrue as a result of election. For example, election to Congress may enhance a person's chance of obtaining lucrative consulting opportunities once he or she exits Congress and enters the private sector. In such a situation, it is reasonable to model individuals as if they solve a discrete stochastic dynamic programming (DS-DP) problem in order to make optimal current period decisions-that is, to model agents' behavior as if they forecast how their current decisions will affect their future payoffs.

  Unfortunately, although most economists would agree that a DS-DP problem is the proper way to model choice behavior in such dynamic environments, implementation of such models has been severely limited by the computational burdens involved. Estimation of this type of model requires the use of high performance computer facilities like the IBM SP.

  Congressional seat and on the career decisions of politicians. Such policies include the introduction of term limits, changes in the pension regime, changes in the wages of Congress, changes in the seniority rule for committee appointments, as well as other actions.

  Another research focus concentrated on developing new methods to solve and estimate descrete stochastic dynamic programming (DS-SP) models and use these to study occupational and educational choices. In recent years it has become common in economics to model individuals who are making choices in dynamic environments as if they were solving a DS-SP problem to determine their optimal decisions. But empirical implementation of such models has been hampered because their solution and estimation requires that very high order numerical integrations be performed.

  This research investigated the use of simulation methods to circumvent these integration problems. These methods were employed in two substantive projects. The first was a model of life cycle decisions of young women; the focus was on their human capital investment (i.e., school attendance and work), fertility, and marriage decisions, and how these are affected by public welfare programs. After this model was completed, it was found to be useful for predicting the effects of changes in welfare rules on teenage childbearing and dropout rates, among other things. The second project was a model of decision-making under uncertainty about future prices, as applied to marketing problems.