I investigate the contribution of hand-to-mouth (HtM) households to aggregate consumption fluctuations and the channels through which these effects operate. To do so, I estimate a two-agent New Keynesian (TANK) model using a spike-and-slab prior on the share of HtM households. This approach nests the representative-agent (RANK) model within the TANK framework, accounts for specification uncertainty between the two nested models, and delivers the posterior probability assigned to each specification. The data assign greater posterior probability to the TANK model and identify labor income as the primary channel through which HtM households affect aggreate consumption following monetary policy and government spending shocks.

We develop a novel identification strategy that combines sign and policy coefficient restrictions to estimate the exchange rate response to U.S. monetary policy shocks. Central to our approach is a restriction grounded in historical narratives: monetary policy systematically tightens following a depreciation of the U.S. dollar. This assumption, particularly relevant during the Volcker era, yields results consistent with exchange rate overshooting and a conditional form of uncovered interest parity. Unlike standard sign-restriction methods, our strategy delivers robust impulse responses across regimes, including both the Volcker and post-Volcker periods.