Should Dynamic Scoring be done with Heterogeneous Agent-Based Models? Challenging the Conventional Wisdom

Traditionally, Dynamic Scoring calculations experiements are carried out using representative agent based macroeconomic models. Existing literature does not provide any objection to this approach. In this paper, I develop a heterogeneous agent model similar to the SaverSpenders model of Mankiw(2000). But spenders in my model are merely credit constrained and not Rule of thumb consumers. Both groups are intertemporal optimizers because of the existence of Internal Habit Persistence. Transition path of most of the macro and …scal variables for various tax cuts under alternative …nancing scheme shows pattern which are signi…cantly di¤erent and sometimes contrasting to the representative agent model. Dynamic scoring calculations reveal a downward bias of the representative agent model. Underestimation of the dynamic response could be as large as 45%. Finally, steady state results indicate smaller impact of contractionary policies on major …scal variables such as net tax revenue and tax base. Over all, the paper argues that the need to use heterogeneous agent based model in dynamic …scal calculations is not only desirable but also essential. Key Words: Savers-spenders model, rule-of thumb consumer, intertemporal optimizers, dynamic scoring, habit persistence, alternative …nancing, debt …nancing, …scal policy JEL code: E62, H2, H3, H6 1 Introduction Should we use heterogeneous agent based model to analyze aggregate change rather than representative agent based model? More precisely, should we use models that have rich heterogeneity across their agents in terms of market participation, preference and labor supply decisions to analyze policy experiments such as dynamic scoring that focus on the aggregate implications of alternative …scal policies? The conventional wisdom among policy making agencies and to some extent, in the literature is the following; although heterogeneous agent based models are crucial for analyzing distributional policies, for standard dynamic …scal policy analysis where the main focus is the aggregate e¤ect such as debt …nancing and dynamic scoring, the representative agent based model, although subject to minor measurement error, could provide a reasonable approximation to macroeconomic response to various policy changes, and therefore could be used as I would like to thank my Advisor , Eric Leeper, for his valuable suggesstions during my research. I would like to specially thank Nora Traum for her valuable comments and also for sharing matlab codes for Leeper and Yang(2006). I would also like to thank Susan Yang for providing code for the steady state calculations of Leeper and Yang(2006) and helping me to understand the fundamentals of dynamic scoring. Finally, I thank all the participants of the Fall 2007 Macroeconomics Brown Bag workshop. All errors are mine