Heterogeneous Agent models and Marginal Propensity to Consume (MPC)
Kaplan and Violante, in this paper argue that empirical evidence suggests that in response to an unexpected windfall, the MPC is large on average and displays a wide degree of heterogeneity across households. But canonical representative agent models fail in accounting for this evidence. The reason why heterogeneous agent (HA) models fit better the data is because they encompass two realistic features of MPCs: (i) the aggregate MPC can be much larger than the interest rate, and (ii) the distribution of MPCs can be very dispersed. This is because these models include poor hand-to-mouth households which consume the majority of any extra liquidity they receive. At low levels of wealth, precautionary saving motives are especially strong and consequently lower the level of consumption, but disappear as wealth grows sufficiently large. And finally, households have ex-ante differences such as impatience, high willingness to substitute, low returns on saving or behavioral biases which can all be sources of high MPCs. By solving different versions of HA models they conclude:
One-asset heterogeneous agent models with idiosyncratic income risk and incomplete markets generate an average MPC that is an order of magnitude bigger than in representative agent models, but still significantly smaller than in the data. Allowing for ex-ante heterogeneity across households or certain behavioral biases in preferences can raise the average MPC to be in line with the data. However, this comes at the cost of misrepresenting key features of the wealth distribution. Most importantly, these models vastly overstate the fraction of households with low wealth.
The Marginal Propensity to Consume in Heterogeneous Agent Models
Authors: Greg Kaplan, Giovanni L. Violante
From: University of Chicago, Princeton University
HANK, MPCs and stabilization policy
In this presentation, Greg Kaplan explains how in HANK models the ability of monetary policy to stabilize economic activity is weakened. Unlike in representative agent models where monetary policy can rely on the intertemporal substitution channel—lowering interest rates induces households to consume – in HANK models this channel is much weaker because a number of households have a hard time shifting consumption over time either because they have little wealth or because their wealth is illiquid. As a result, the main transmission channel of monetary policy in HANK models involves changing household disposable income, and in particular affecting income for households with higher marginal propensities to consume. An important lesson from the HANK literature, argues Kaplan, is that the redistributive effects of policy can no longer be ignored and that gains from alleviating individual hardship swamp gains from aggregate stabilization.
What Has The HANK Literature Taught Us About Monetary Policy?
Author: Greg Kaplan
From: University of Chicago
