We propose a three-step factor-flows simulation-based approach to forecast the duration distribution of unemployment. Step 1: estimate individual transition hazards across employment, temporary layoff, permanent layoff, quitter, entrant, and out of the labor force, with each hazard depending on an aggregate component as well as an individual’s labor force history. Step 2: relate the aggregate components to the overall unemployment rate using a factor model. Step 3: combine the individual duration dependence, factor structure, and an auxiliary forecast of the unemployment rate to simulate a panel of individual labor force histories. Applying our approach to the November Blue Chip forecast of the COVID-19 recession, we project that 750,000 workers laid off in April 2020 remain unemployed eight months later. Total long-term unemployment rises thereafter and eventually reaches 4.2 million individuals unemployed for more than 26 weeks and 1.4 million individuals unemployed for more than 46 weeks. Long-term unemployment rises even more in a more pessimistic recovery scenario, but remains below the level in the Great Recession due to a high amount of labor market churn.

我们提出了一种基于三步因素流模拟的方法来预测失业的持续时间分布。步骤1：估算整个就业，临时裁员，永久裁员，退出，进入和进入劳动力市场的个人过渡危害，每种危害取决于总体构成要素和个人的劳动力历史。步骤2：使用因素模型将总体构成要素与总体失业率相关联。步骤3：结合个人持续时间依赖性，要素结构和失业率的辅助预测，以模拟一组个人劳动力历​​史记录。将我们的方法应用于11月蓝筹股对COVID-19衰退的预测中，我们预计2020年4月下岗的750,000名工人将在八个月后失业。此后，长期总失业率上升，最终导致420万失业人数超过26周的失业者和140万失业人数超过46周的失业者。在更为悲观的复苏情况下，长期失业率甚至会进一步上升，但由于劳动力市场的大量动荡，长期失业率仍低于大萧条时期的水平。