I study how downsizing establishments choose which workers to lay off and how the cost of job loss varies by the type of layoff rule using administrative employer-employee data for West Germany. I estimate establishment-level layoff rules for over 4,400 mass layoffs from 1980 to 2009 and explore how layoff rules have changed over time and the business cycle. The probability of being laid off is decreasing in seniority, wage, age, and (slightly in) education, with most mass layoffs involving a significant level of discretion. Layoff decisions have become more selective and less seniority-based over time, consistent with Germany’s decreasing firing costs and collective bargaining coverage. I also find significant business cycle effects: during recessions establishments become less likely to lay off their low-seniority workers and more likely to lay off their workers with low wage residuals. I develop an asymmetric employer learning model with heterogeneous firing costs to explain these findings and and to test whether workers’ current employer has superior information about their productivity than outside employers. I show that the cost of job loss for workers should correlate with the level of discretion used in the layoff rule and find evidence for asymmetric employer learning: event studies show that workers who are laid off using seniority layoff rules experience significantly smaller earnings losses, since these layoffs do not serve as a negative signal of workers’ productivity. Finally, asymmetric employer learning appears to be more important for workers with low education and workers in high-skilled service occupations.