Unemployment Risk

Abstract

Fluctuations in the risk of a large increase in unemployment are examined. The analysis compares medium‐term risks—that is, risks at a 3‐year horizon—to those over a 1‐year horizon. The primary approach involves quantile regressions. Robustness exercises examine risks using a logistic regression to model the probability of a large increase in the unemployment rate. U.S. experience reveals an elevated risk of large increases in unemployment over the medium term when credit growth is high and when the unemployment rate is low. Near‐term risks to unemployment are closely tied to changes in corporate bond spreads or the slope of the yield curve, consistent with research on recession prediction, but these factors do not play a sizable role in medium‐term risks. These results highlight the value of considering different near‐ and medium‐term risk factors.

Specification Choices in Quantile Regression for Empirical Macroeconomics

“…To have a relatively large model in this case, we pull in indicators from a range of studies, some that have considered tail risks to economic activity measures other than GDP growth. More specifically, in the large GDP growth application, our choice of broad indicators is informed by the results of applications in Caldara, et al (2021) and Plagborg-Moller, et al (2020), who find that broad factor indexes of economic and financial conditions have predictive content for growth-at-risk; in a range of studies in the forecasting literature that generally find credit spreads to be helpful for macroeconomic forecasting (e.g., Faust, et al (2013)); and in Kiley (2022), who finds that a credit spread and medium-term changes in the credit-to-GDP ratio have predictive content for tail risks to economic activity as measured by the unemployment rate. Drawing on specifications from these studies, the broad set of indicators used for tail risk prediction in this large GDP growth application consists of the four NFCI subindexes for leverage, non-financial leverage, credit, and risk; the Baa corporate/10-year Treasury bond spread; the Aaa corporate/10-year Treasury bond spread; the four-year growth rate of the credit-to-nominal GDP ratio; and the Chicago Fed's national activity index (CFNAI) of the business cycle.…”

“…Following Kiley (2022), our fourth application examines tail risks to unemployment with a model that relates the change in the unemployment rate to the lagged level of the unemployment rate, the Baa corporate/10-year Treasury bond spread, the 10-year Treasury bond/federal funds rate term spread, the four-year growth rate of the credit-to-nominal GDP ratio, and the four-quarter rate of inflation in PCE prices excluding food and energy (henceforth, core PCE prices). In these results, our specification is patterned along the lines of Kiley (2022), to predict the change in unemployment rather than the level.…”

“…Following Kiley (2022), our fourth application examines tail risks to unemployment with a model that relates the change in the unemployment rate to the lagged level of the unemployment rate, the Baa corporate/10-year Treasury bond spread, the 10-year Treasury bond/federal funds rate term spread, the four-year growth rate of the credit-to-nominal GDP ratio, and the four-quarter rate of inflation in PCE prices excluding food and energy (henceforth, core PCE prices). In these results, our specification is patterned along the lines of Kiley (2022), to predict the change in unemployment rather than the level. In this case, the variable being predicted is y an unemployment rate gap measured as the actual rate less the Congressional Budget Office's (CBO's) estimate of the non-cyclical rate of unemployment; relative import price inflation; and the four-year growth rate of the credit-to-nominal GDP ratio.…”

“…Quantile regression has become widely used in empirical macroeconomics, in particular for estimating and forecasting tail risks to macroeconomic indicators. Several examples in a voluminous literature include Giglio, Kelly, and Pruitt (2016) and Adrian, Boyarchenko, and Giannone (2019) for output growth; Kiley (2022) for unemployment; Gaglianone and Lima (2012), Korobilis (2017), Lopez-Salido and Loria (2022), and Manzan and Zerom (2013) for inflation; and Manzan (2015) for a range of variables.…”

“…Accordingly, in this paper we examine various choices in the specification of quantile regressions used for modeling and forecasting macroeconomic tail risks. Our analysis is based on applications to US data on GDP growth, the unemployment rate, and inflation, patterned on empirical work by Adrian, Boyarchenko, and Giannone (2019), Caldara, et al (2021), and Plagborg-Moller, et al (2020) for output growth; Kiley (2022) for unemployment; and Korobilis, et al (2021) and Lopez-Salido and Loria (2022) for inflation. For each application, we compare the accuracy of quantile forecasts obtained from simple quantile regression, averages of quantile regression forecasts obtained with one indicator at a time, partial quantile regression, quantile regression with ridge penalty, and a few different specifications of priors for Bayesian quantile regression.…”