Near-rational behavior in New Keynesian models

New Keynesian models, stressing nominal inertia in wages and prices, often assume economic agents are fully rational. The strong assumptions of full rationality suggesting all agents can solve complex dynamic problems are relaxed here. Using both empirical and theoretical approaches, we find that allowing for less-than-fully rational behavior by a fraction of agents proves to be useful in explaining the actual costs to disinflation in these models.; Chapter II investigates inflation dynamics implied by a class of Phillips Curves proposed by Gali and Gertler. They introduce ‘rule-of-thumb’ agents into an otherwise rational expectations micro-founded model, and find that though rule-of-thumb behavior is statistically significant, it is not important to aggregate dynamics. We use U.S. data to test this assertion and find the opposite: that rule-of-thumb behavior, though small in proportion, helps better explain aggregate dynamics. Furthermore, we argue that the measure of real aggregate activity used, real marginal cost, might not be appropriate in capturing the cyclical component of aggregate activity.; Under a continuous time staggered price-setting framework it has been shown that a credible disinflation announcement can actually produce a boom. Chapter III proposes a fraction of agents who use rule-of-thumb behavior in setting prices under this framework. The resulting output paths are more consistent with empirical evidence suggesting costly disinflation. More importantly, we show that only a relatively small proportion of rule-of-thumb behavior is necessary to produce recessions. Finally, with very high proportions of rule-of-thumb behavior, a cold-turkey disinflation maximizes output over time, but as the degree of rationality increases the best monetary policy shifts to a gradualist disinflation.; The last chapter discusses two separate extensions of the previous results. In the first, we re-examine the results of the Ball model with rule-of-thumb agents, including additional persistence in the form of sluggish labor adjustment, finding that adding sluggish labor adjustment does not help explain the costs to disinflation. We then reconsider the New Keynesian feature that discrete time models with a credibly committed Fed can result in costless disinflation, finding independent support for others who show that with more microfoundations, the solution dynamics imply costly disinflation.