Abstract: This paper studies the nature of countercyclical price change dispersion and revisits the real effects of monetary policy over the business cycle. Using Nielsen Retail Scanner data, I show that the increase in the dispersion of price changes during the Great Recession is largely driven by the diverging price adjustment between goods of different qualities. Prices of low-quality goods drop dramatically, while those of high-quality goods remain stable. Conversely, the contribution of increasing volatility of individual prices is negligible. By linking Nielsen Consumer Panel with Retail Scanner data, I find that households who purchase a larger fraction of low-quality goods during the recession become more price-sensitive, suggesting that sellers of low-quality goods face increasing demand elasticity. Motivated by the empirical findings, I develop a quantitative multi-sector menu cost model incorporating countercyclical elasticity of demand in the sector of low-quality goods. The model successfully replicates the countercyclical dispersion of price changes and the price change distribution of goods in different quality groups. In terms of the effectiveness of monetary policy, time-varying demand elasticity affects monetary non-neutrality mainly through the selection effect. Quantitatively, contrary to models driven by second-moment shocks, monetary policy is still very effective during recessions: the cumulative output response to a monetary policy shock is only 4% smaller in recessions.
Abstract: This paper investigates the optimal monetary policy in a two-sector model with imperfect common knowledge between firms due to limited attention. A welfare criterion based on the utility of consumers is derived to evaluate the real effects of monetary policy. When sectoral productivity shocks are symmetric, complete price stabilization is optimal. However, with asymmetric sectoral productivity shocks, price stabilization is no longer optimal. When responding to these shocks, central banks face trade-offs between output and price stabilization. The optimal monetary policy places more weight on price stabilization when firms have more information processing capacity.
Work in Progress
Heterogeneous Price Stickiness, the Keynesian Cross and the Effects of Monetary Policy
Abstract: I document using Nielsen Retail Scanner data that the prices of retail goods typically purchased by low-income households adjust more often. I also show that households’ Marginal Propensities to Consumer (MPCs), estimated from the 2008 Economic Stimulus Supplement Survey, are negatively correlated with the price stickiness of the goods they buy. To investigate the implications for the monetary transmission mechanism, I develop a Two-sector Two Agent New Keynesian model (T-TANK). In the model, Ricardian (saver) and Keynesian (spender) households have different MPCs and consumption bundles and thus face different frequencies of price changes. Keynesian households face higher inflation after a nominal stimulus. As a result, compared with a Two Agent New Keynesian model with homogeneous price stickiness, the Keynesian multiplier in my model is smaller, making monetary policy less effective.
Using Tax Rebate data to Estimate MPC: Fixed-Effects vs. Matching Methods
Abstract: Fixed-effects estimators are widely used with tax rebate data to estimate households’ marginal propensity to consume (MPCs). These estimators are supposed to estimate the treatment effects of tax rebate on consumption, by eliminating unobserved heterogeneity and common time factors. This paper shows that the fixed-effects estimator is biased: when subtracting the average spending across households within each period, it not only eliminates the time fixed effects but also subtracts some effects on spending from the tax rebate, which confounds the estimation. I develop a new matching method, by explicitly constructing treatment and control groups, to estimate households’ MPCs. Using the 2008 Economic Stimulus Supplement Survey, I show that the matching method consistently recovers the time profile of MPC.