Forecasting Macroeconomic Risk in Real Time: Great and Covid-19 Recessions
Roberto A. De Santis (European Central Bank)
We show that financial variables contribute to the forecast of GDP growth during the Great Recession, providing additional insights on both first and higher moments of the GDP growth distribution. If a recession is due to an unforeseen shock (such as the Covid-19 recession), financial variables serve policymakers in providing timely warnings about the severity of the crisis and the macroeconomic risk involved, because downside risks increase as financial stress and corporate spreads become tighter. We use quantile regression and the skewed t-distribution and evaluate the forecasting properties of models using out-of-sample metrics with real-time vintages.
Machine Learning, the Treasury Yield Curve and Recession Forecasting
Michael Puglia and Adam Tucker (Federal Reserve)
We use machine learning methods to examine the power of Treasury term spreads and other financial market and macroeconomic variables to forecast US recessions, vis-à-vis probit regression. In particular we propose a novel strategy for conducting cross-validation on classifiers trained with macro/financial panel data of low frequency and compare the results to those obtained from standard k-folds cross-validation. Consistent with the existing literature we find that, in the time series setting, forecast accuracy estimates derived from k-folds are biased optimistically, and cross-validation strategies which eliminate data “peeking” produce lower, and perhaps more realistic, estimates of forecast accuracy. More strikingly, we also document rank reversal of probit, Random Forest, XGBoost, LightGBM, neural network and support-vector machine classifier forecast performance over the two cross-validation methodologies. That is, while a k-folds cross-validation indicates tha t the forecast accuracy of tree methods dominates that of neural networks, which in turn dominates that of probit regression, the more conservative cross-validation strategy we propose indicates the exact opposite, and that probit regression should be preferred over machine learning methods, at least in the context of the present problem. This latter result stands in contrast to a growing body of literature demonstrating that machine learning methods outperform many alternative classification algorithms and we discuss some possible reasons for our result. We also discuss techniques for conducting statistical inference on machine learning classifiers using Cochrane’s Q and McNemar’s tests; and use the SHapley Additive exPlanations (SHAP) framework to decompose US recession forecasts and analyze feature importance across business cycles.
COVID-Induced Economic Uncertainty
Scott R. Baker (Northwestern University), et al.
Assessing the economic impact of the COVID-19 pandemic is essential for policymakers, but challenging because the crisis has unfolded with extreme speed. We identify three indicators – stock market volatility, newspaper-based economic uncertainty, and subjective uncertainty in business expectation surveys – that provide real-time forward-looking uncertainty measures. We use these indicators to document and quantify the enormous increase in economic uncertainty in the past several weeks. We also illustrate how these forward-looking measures can be used to assess the macroeconomic impact of the COVID-19 crisis. Specifically, we feed COVID-induced first-moment and uncertainty shocks into an estimated model of disaster effects developed by Baker, Bloom and Terry (2020). Our illustrative exercise implies a year-on-year contraction in U.S. real GDP of nearly 11 percent as of 2020 Q4, with a 90 percent confidence interval extending to a nearly 20 percent contraction. The exercise says that about half of the forecasted output contraction reflects a negative effect of COVID-induced uncertainty.
COVID-induced Recession Began in March 2020
Haixi Li (Freddie Mac) and Xuguang Simon Sheng (American University)
21 May 2020
The COVID-induced recession began in March 2020 for the United States. We identify this turning point by applying a sequential quickest detection method to a real-time index of economic activity. Supporting evidence is also found from macroeconomic data releases and stock markets.
Why Has the US Economy Recovered So Consistently from Every Recession in the Past 70 Years?
Robert E. Hall (Stanford University and NBER) and Marianna Kudlyak (San Francisco Fed)
It is a remarkable fact about the historical US business cycle that, after unemployment reached its peak in a recession, and a recovery began, the annual reduction in the unemployment rate was stable at around 0.55 percentage points per year. The economy seems to have had an irresistible force toward restoring full employment. There was high variation in monetary and fiscal policy, and in productivity and labor-force growth, but little variation in the rate of decline of unemployment. We explore models of the labor market’s self-recovery that imply gradual working off of unemployment following a recession shock. These models explain why the recovery of market-wide unemployment is so much slower than the rate at which individual unemployed workers find new jobs. The reasons include the fact that the path that individual job-losers follow back to stable employment often includes several brief interim jobs, sometimes separated by time out of the labor force. We show that the evolution of the labor market involves more than the direct effect of persistent unemployment of job-losers from the recession shock—unemployment during the recovery is elevated for people who did not lose jobs during the recession.
What’s Up with the Phillips Curve?
Del Del Negro (NY Federal Reserve), et al.
The business cycle is alive and well, and real variables respond to it more or less as they always did. Witness the Great Recession. Inflation, in contrast, has gone quiescent. This paper studies the sources of this disconnect using VARs and an estimated DSGE model. It finds that the disconnect is due primarily to the muted reaction of inflation to cost pressures, regardless of how they are measured—a flat aggregate supply curve. A shift in policy towards more forceful inflation stabilization also appears to have played some role by reducing the impact of demand shocks on the real economy. The evidence rules out stories centered around changes in the structure of the labor market or in how we should measure its tightness.
Measuring the Labor Market at the Onset of the COVID-19 Crisis
Alexander Bartik (University of Illinois at Urbana-Champaign), et al.
We use traditional and non-traditional data sources to measure the collapse and subsequent partial recovery of the U.S. labor market in Spring 2020. Using daily data on hourly workers in small businesses, we show that the collapse was extremely sudden – nearly all of the decline in hours of work occurred between March 14 and March 28. Both traditional and non-traditional data show that, in contrast to past recessions, this recession was driven by low-wage services, particularly the retail and leisure and hospitality sectors. A large share of the job loss in small businesses reflected firms that closed entirely. Nevertheless, the vast majority of laid off workers expected, at least early in the crisis, to be recalled, and indeed many of the businesses have reopened and rehired their former employees. There was a reallocation component to the firm closures, with elevated risk of closure at firms that were already unhealthy, and more reopening of the healthier firms. At the worker-level, more disadvantaged workers (less educated, non-white) were more likely to be laid off and less likely to be rehired. Worker expectations were strongly predictive of rehiring probabilities. Turning to policies, shelter-in-place orders drove some job losses but only a small share: many of the losses had already occurred when the orders went into effect. Last, we find that states that received more small business loans from the Paycheck Protection Program and states with more generous unemployment insurance benefits had milder declines and faster recoveries. We find no evidence so far in support of the view that high UI replacement rates drove job losses or slowed rehiring substantially.