We elicit the empirical relevance of transmission channels for HANK models relying on MPC heterogeneity and the cyclicality of inequality in various measures of income and consumption. We use a unique dataset from Norway that links household-level transaction information on consumption to administrative data on income and wealth. Consumption and MPC differences are substantial even in a high-income and relatively equal society such as Norway. A key question is whether in recessions, income and consumption losses are disproportionately large for the high-MPC, and also for hand-to-mouth households. We compute two sufficient statistics. First, the aggregate MPC that takes into account individual MPCs, income shares, and cyclicalities. And second, the cyclicality of consumption inequality between two household groups, classified according to their hand-to-mouth status, i.e. whether their choices obey an Euler equation or not.
In a nutshell, we find that estimates based on labour earnings do imply amplification through the model’s lens: the cyclicality of labour earnings is higher for higher-MPC agents. However, we show that estimates based on disposable income overturn this: if anything, the cyclicalities of high-MPC agents’ disposable incomes is lower, implying that disposable income inequality (in the MPC dimension) is procyclical, and a dampening of aggregate-demand fluctuations through the model’s lens. We show that part of this mitigation is due to capital income (which is cyclical and accrues predominantly to low-MPC individuals) and part to the tax and transfer system in Norway. Nevertheless, we also show through the lens of a counterfactual using an estimated US tax-and-transfer system, that a similar mitigation, albeit of a smaller magnitude, would occur if such a system were in place in Norway instead.
Finally, we compute the cyclicalities of individual consumption and find that they are virtually flat in the MPC dimension, which is suggestive evidence for consumption insurance, or risk sharing. We group agents according to whether they are hand-to-mouth or not, using a wide range of definitions based on liquid wealth or net worth. Even according to this split which informs our second sufficient statistic, we find at best modest evidence for more consumption cyclicality of hand-to-mouth agents, implying a very small multiplier effect through the lens of the model, of the order of 1.05. Even when focusing on the Great Recession as a case study, and using a generous definition implying that 30% of the population is hand-to-mouth, the implied multiplier is at most around 1.1.