UC Santa Cruz

This dissertation consists of three essays in macroeconomics. The first essay examines labor-market efficiency by incorporating part-time employment. The second and third essays study real-time recession detection, using labor market data in Japan and real-time GDP data in the United States. Chapter 1 extends the sufficient-statistics formula for efficient unemployment developed by Michaillat and Saez (2021a) to account for part-time employment. I introduce two additional sufficient statistics that measure the share of part-time employment and part-time hours relative to full-time hours. Applying the framework to the United States (1951–2026) and Japan (1970–2025), I compare the effects of total part-time employment and involuntary part-time employment on efficient unemployment. Total part-time employment has substantially larger effects than involuntary part-time employment. While involuntary part-time employment provides information about labor-market slack, the main change in efficient unemployment comes from part-time work itself because part-time workers supply fewer market hours than full-time workers. Under the total part-time calibration, efficient unemployment averages 4.7% in the United States before COVID and 4.2% after COVID. In the Japanese application, the full-sample average is 2.7%. The distinction is especially important in Japan, where part-time employment is widespread and often reflects flexible work arrangements. These findings suggest that aggregate labor input, rather than involuntary part-time employment alone, is an important determinant of labor-market efficiency. Chapter 2, a joint work with Neha Sikand, examines whether labor market–based recession indicators developed for the United States can be adapted to Japan. Recession indicators are often viewed as U.S.-specific, raising the question of whether labor market–based rules such as the Sahm Rule and the Michez Rule can reliably detect recessions in other countries. To answer this, we evaluate whether such rules can be adapted to Japan by calibrating thresholds and smoothing parameters to Japanese labor market data. We construct a large set of 95,832 recession indicators combining unemployment and vacancy data. The selected classifiers are statistically perfect as they identify all 11 historical recessions in the 1970–2021 training period without generating any false positives. Among these, 193 classifiers lie on the anticipation–precision frontier. Restricting attention to the high-precision segment yields six classifiers with a standard deviation of detection errors below 3 months. The selected classifier ensemble signals recessions, on average, 0.06 months after their true onset. Overall, these findings suggest that slack-based labor market rules provide a general framework for improving real-time recession detection across countries. Chapter 3, a joint work with Neha Sikand, examines whether real-time GDP announcements can reliably identify business-cycle turning points. Using U.S. real-time GDP vintages from 1947 to 2021, we construct 4,356 recession indicators based on alternative smoothing methods and scaling variations. We then combine these recession indicators with different thresholds to obtain 137,457 perfect recession classifiers. The selected classifiers identify all 12 historical recessions without generating false positives. Restricting attention to the high-precision segment yields two classifiers with a standard deviation of detection errors below three months, while the selected ensemble signals recessions, on average, 3.04 months after official onset. Our framework accurately identifies recession episodes across vintages, suggesting that real-time data revisions alone cannot fully account for discrepancies reported in prior work. These results suggest that real-time GDP announcements provide a practical real-time proxy for NBER-style recession dating.