Federal Reserve Board of Governors

ABSTRACT This paper leverages free‐form textual responses from a key manufacturing survey to create sentiment indexes that mirror categorical measures from the same survey and also contain predictive content—both in and out‐of‐sample—for manufacturing output. We use textual data from the Institute for Supply Management to compare sentiment metrics based on dictionary and deep learning natural language processing methods. The best performing sentiment measures classify comments based on fine‐tuned deep learning models. To add interpretability, we apply Shapley decompositions to show that a relatively small number of words—associated with very positive and very negative sentiment—account for much of the variation in the aggregate sentiment index.

We provide evidence that banks use loan covenants to prepare for future monetary policy tightening, thereby facilitating the bank lending channel of monetary policy transmission. Specifically, banks with greater monetary policy exposure—those whose lending capacity contracts more as the federal funds rate increases—include stricter financial covenants in loan contracts, granting them flexibility to reduce existing loan commitments during monetary policy tightening when firms breach covenants. The resulting credit reductions to covenant violators by high-exposure banks account for over one-third of the total decline in credit during recent federal funds rate hikes.

Abstract Allocative wages—the labor costs considered when deciding to form or dissolve a long-term employment relationship—are more sensitive to cyclical conditions for more educated workers. Specifically, college-educated workers’ allocative wages are highly pro-cyclical, while high school dropouts’ wages exhibit only moderate cyclicality. Further, as education increases, an increasing share of the sensitivity of allocative wages is driven by the persistent scarring effects of the cyclical position at the time of hiring on the wages associated with higher levels of tenure, amounting to more than a third of the overall sensitivity for the college educated. The greater job stability of the more educated—and therefore the exposure to scarring—contributes to these differences. In addition, more significant scarring at each horizon of tenure amplifies the effect. In service of documenting these facts, I develop new methods for inferring the sensitivity of labor costs to shocks when agents are forward-looking and wages may be intertemporally smoothed.

We estimate the time-varying distribution of aggregate supply (AS) and aggregate demand (AD) shocks. We distinguish between traditional Gaussian uncertainty and “bad” uncertainty, associated with negative skewness. The Great Moderation is driven by a reduction in the volatility of AS shocks and the Gaussian component of AD shocks. The increased role of “bad” demand uncertainty implies that the conditional skewness of GDP growth and inflation has decreased over time. The correlation between AS/AD shocks and shocks to their conditional volatilities is generally strongly negative. The correlation between inflation and growth shocks has increased due to a decrease in AS volatility. (JEL C52, D81, E12, E23, E31, E32)