Amanda M Michaud

Why do financial markets react to central bank communication when it mainly discusses publicly available data? We argue that the selection of which indicators to discuss is a choice that reveals how the central bank reads the state of the economy, and therefore signals its reaction function. Using Federal Reserve (Fed) statements, minutes, speeches, and transcripts, we document which labor market and inflation indicators the Fed selects over time and construct a measure we call spin: the reading of the state conveyed in communication relative to the full menu of available indicators. For individual communication, we show that spin is predicted by a speaker's policy preference. To avoid a mechanical correlation, we create a new, state-independent measure of policy stance with a multistage large-language-model procedure, which we validate against SEP projections and macroeconomic indicators. In Fed transcripts, hawks are more likely to cite hot labor market and inflation indicators, while doves are more likely to cite cool ones. Finally, in a short event window around FOMC policy statements, we find that a change in statement spin towards hotter signals (higher inflation and stronger labor markets) shifts up the Treasury yield curve, even when controlling for the policy rate change, market expectations, and other communication factors. In sum, indicator selection is thus a distinct channel of monetary policy communication.

To challenge and improve business cycle models of labor markets, we use data on separations into non-employment by destination and reason. Standard models where layoffs are random and quits are selective send too many laid-off workers to unemployment– empirically, a third leave the labor force– and they get the cyclical direction of labor force attachment backward. Adding selective layoffs and random quits resolves these issues and clarifies two channels that affect how labor markets evolve during recessions: (i) labor supply increases as marginal workers hoard their jobs and the displaced keep searching, and (ii) selection in layoffs shifts from marginal toward attached workers. Consequently, output per worker falls through composition rather than TFP, and the welfare cost of recessions is less than half that of the standard model. Through these channels, the volatility of the jobless who want to work is further disconnected from measured unemployment, a prediction we confirm in the data.

We construct monthly Current Population Survey series that classify every separation into non-employment by both its reason (quit or layoff) and its destination (unemployment or non-participation) from 1978 on. Standard destination flows mischaracterize churn: over a third of laid-off workers exit the labor force, leading employment-to-unemployment (EU) flows to understate true layoffs by 20% to mask a stabilizing negative correlation between quits and layoffs. Properly measured, layoffs account for 62% of employment-population volatility and 35% more of unemployment volatility than EU flows, amplified by a new channel: laid-off workers’ rising propensity to flow to unemployment during recessions.

There is substantial variation in Social Security Disability Insurance (DI) across the United States, with DI beneficiaries accounting from less than one percent to more than one-fifth of a county's working-age residents. We combine county-level data on DI applications and awards with measures of labor market characteristics, living costs and population health to model the geographic dispersion of DI. We find that local differences in health and income levels are about equally important in determining geographic differences, with living-cost differences also playing an important role. The DI program redistributes across space, delivering ex ante welfare gains to residents counties at the 90th percentile in terms of DI receipt that are more than twice as large as in counties at the 10th percentile. The place-based effects of DI are larger than most public policy initiatives specifically designed to support local economic activity. NBER SI 2023 Slides

As married women's labor force participation has increased in the United States, the cyclical volatility of their employment has also increased. We provide a unified theory that can reconcile these facts. Lower volatility of married women's employment over the business cycle is driven by counter-cyclical motives to remain in the labor force to provide insurance against their spouse's income risk. Changes in fundamentals that increase attachment subsequently lower the ability of wives to adjust labor supply to provide insurance in recessions-- they are more likely to have permanently high labor supply anyway. The model predicts that some forces driving the growth in female participation-- increasing returns to tenure and decreasing fixed cost of work-- increase attachment and cyclical volatility. The closing gender wage gap, by contrast, reduces both. A quantitative evaluation predicts that the former two forces have dominated in the United States with some ebbing for recent cohorts. Microeconomic evidence support both this prediction and the specific mechanism of precautionary spousal labor supply. Implications for welfare and intra-household insurance are discussed.

I develop a quantitative framework to study the impact of Unemployment Insurance (UI) expansions to workers earning below eligibility thresholds. A model of UI is calibrated to replicate experiences of the eligible and ineligible, including consumption after job loss. Eligibility rules distort labor supply in costly ways and removing them would benefit workers across the income spectrum, even with consideration of funding costs. The model captures job-finding rates by eligibility that I document during the GFC and COVID-19 recessions. Although expanded eligibility during COVID-19 lowered job-finding rates, other changes in UI policy were more impactful.

In recent decades, Social Security Disability Insurance (SSDI) claims have risen rapidly. We evaluate the importance of changing macroeconomic conditions in shaping this trend. Our quantitative framework considers that economic conditions interact with individuals' health status in their decisions to apply for SSDI. Crucially, these factors are correlated through the nature of work: multiple sectors differentially expose workers to health and economic risks. Decomposing factors driving SSDI growth in a calibrated model, we find the secular deterioration of economic conditions concentrated in populations with high health risks accounts for about half of the increase in SSDI claims predicted by the model, about a third overall.

Involuntary job separation causes earnings to fall 15.4% initially and remain depressed two decades later. We quantify how much these scars shape the income distribution: incorporating the empirical separation hazard and scar pro le into a standard income process, we nd that job loss reduces average life-cycle income growth by 14.7% and raises cross-sectional dispersion by 17.8% relative to a counterfactual without separations. We then highlight a simple but underappreciated implication of life-cycle concavity for reducedform income processes: any speci cation in which displaced workers restart along the same concave path mechanically delivers rapid recovery, missing the persistent scar in the data. We use a parsimonious learning-by-doing framework to show that this tension cannot be resolved by selection on wages or serially correlated separations alone; some channel that yields lower post-separation income growth is required. We provide a calibrated process matching both scars and life-cycle moments, intended for researchers who need a portable income process with realistic job-loss risk.

Only 14% of U.S. online job posts contain any wage information, and the firms that do post wages are not representative of the broader economy. Posted wages also exceed those reported in CPS and OES data, especially at the bottom of the wage distribution, so wage data from job postings should be used with caution.

Because past choices about crime and employment have lasting consequences, individuals respond slowly to changes in punitive policy, and the full deterrent effect arrives only as new cohorts replace older ones. A quantitative model calibrated to administrative data reproduces the path of U.S. crime and incarceration after 1980 and shows that the accompanying rise in inequality and decline in employment fall unequally across generations.

We show that occupational exposure significantly affects disability risk and study social disability insurance in a general equilibrium model with occupational choice. Insurance improves welfare through risk sharing and by reallocating workers across occupations, but at the optimum the two channels conflict, as additional risk sharing draws too many workers into risky occupations.

Government spending is procyclical in emerging economies and countercyclical in developed ones, a difference driven mainly by social transfers rather than spending on goods and services. In a small open economy model, differences in transfer policy account for a large share of the excess volatility of consumption relative to output in emerging economies.

I show that what employers learn about worker productivity connects wage and employment dynamics, since workers dismissed after being revealed as less productive go on to earn lower wages. A calibrated model attributes to employer learning 78 percent of post-unemployment wage losses, 24 percent of life-cycle wage growth, and 13 percent of cross-sectional wage dispersion.

Motivated by China's rapid growth under capital controls, we show that borrowing constraints on households can correct a learning-by-doing externality by raising labor supply and shifting labor toward traded goods. We compute the optimal constraints along the growth path and their implications for wages, the current account, and the real exchange rate.

SSDI awards based on vocational factors such as age, education, and work experience have tripled since 1985. Using state-level data, we find that rising educational attainment should have offset the effect of population aging on SSDI claims, particularly claims decided on vocational grounds.