A new model to predict the distributional effect of economic uncertainty

Massimiliano Marcellino (Bocconi University and Baffi Centre), Florian Huber (University of Salzburg) and Tommaso Tornese (Università Cattolica del Sacro Cuore) have studied the distributional implications of uncertainty shocks by developing a model that links macroeconomic aggregates to the US distribution of earnings and consumption. The model is described in detail in the paper “The Distributional Effects of Economic Uncertainty”, forthcoming in the “International Economic Review”.

With policy makers and economic analysts increasingly interested in the heterogeneity among economic agents, recent macroeconomic studies have focused more extensively on the relationship between macroeconomic events and distributional changes. 

The new model

In their paper, Marcellino, Huber and Tornese adopt a Functional Structural Vector Autoregression (F-SVAR), which treats the full income and consumption distributions as endogenous objects. The F-SVAR, therefore, generalizes the widely used SVAR model in empirical macroeconomics to the functional data setting. Despite their growing popularity in the statistical literature, functional data models are still rarely used in applied macroeconomics and have not yet been applied to study the distributional consequences of uncertainty shocks.

The new modeling approach of the authors builds on a three-step procedure

First, they interpolate the cross-sectional data for each time period using kernel estimators to obtain smooth earnings and consumption distributions. Second, they transform these distributions to remove unit-integration and non-negativity constraints, and approximate the transformed curves with a set of basis functions, using functional principal components (FPCs). Finally, they model the FPCs jointly with a set of macroeconomic and financial indicators using a Bayesian VAR. This framework allows the authors to trace out impulse response functions (IRFs) of the entire earnings and consumption distributions to structural uncertainty shocks. A key feature of the approach is that, by performing FPC Analysis (FPCA), they are able to condense the information embedded in the distributions in a small number of scores, which captures most of the relevant time variation while keeping the VAR dimension in the third step manageable. Moreover, the transformation they apply to the distribution provides an unconstrained object suitable for FPCA and ensures that the endogenous distribution remains proper following any type of shock.

The findings

The empirical findings of the paper reveal a two-phase propagation mechanism of uncertainty shocks. In the short run, as employment and output fall, a larger share of low-income workers lose their jobs. Among those who remain employed, wages tend to rise relative to GDP per capita, leading to a reduction in the mass of employed individuals with low income-to-GDP ratios. This likely reflects job losses concentrated among less specialized workers.

At the same time, the consumption distribution shifts markedly: the proportion of households with low consumption levels rises, while the mass in the middle of the distribution declines. This pattern suggests that affected households are unable to smooth consumption due to limited access to credit or insurance mechanisms.

In the medium run, as the macroeconomic effects of the uncertainty shock dissipate, the share of low-income workers increases beyond pre-shock levels — presumably due to rehiring of previously unemployed, low-skilled workers whose productivity remains low following reduced investment. While the income distribution becomes more unequal, the consumption distribution returns to its pre-shock shape, with a relevant distinction between durables and nondurables/services. In particular, the consumption of nondurables and services recovers more quickly than that of durable goods, which remains persistently depressed among lower- and middle-income households, possibly due to limited access to credit.

Overall, the short-run phase reduces earnings inequality among employed individuals but increases consumption inequality. In the medium run, income inequality rises substantially, while consumption inequality returns to baseline. One potential explanation for the diverging consumption dynamics lies in the persistence of tighter credit conditions for poorer households, limiting access to durable goods. At the same time, redistributive transfers or improved sentiment upon re-employment may temporarily boost spending on nondurables and services.