We study the economic incentives for automation when labor and machines are perfect substitutes. Labor may still be employed in production, even when it is a costlier input than robots on a productivity-adjusted basis. This occurs if firms face idiosyncratic risk, adjusting the stock of machines is costly, and workers can be hired and fired quickly enough. Even though labor survives, jobs become less stable, as workers are hired in short-lived bursts to cope with shocks. We calibrate a general equilibrium, multi-industry version of our model to match data on robot adoption in US manufacturing sectors, and use it to compute the employment and labor share consequences of progress in automation technology. A fall in the relative price of robots leads to relatively few jobs losses, while reductions in adjustment costs, or improvements in relative robot productivity, can be far more disruptive. The model-implied semi-elasticity of aggregate employment to robot penetration (number of robots per thousand employees) ranges between 0.01% and 0.12%, depending on the underlying source of increased robot adoption, consistent with findings in the empirical literature. In an extension, we show that reduced-form hiring and firing costs unambiguously depress long-run employment.

我们研究了当劳动力和机器是完美替代品时自动化的经济激励。即使在生产力调整的基础上，劳动力的投入成本高于机器人，但仍可能在生产中使用劳动力。如果公司面临特殊风险，调整机器库存的成本很高，并且可以足够快地雇用和解雇工人，就会发生这种情况。尽管劳动力幸存下来，但工作变得不那么稳定，因为工人是在短期内被雇佣以应对冲击。我们校准了模型的一般均衡、多行业版本，以匹配美国制造业机器人采用的数据，并使用它来计算自动化技术进步的就业和劳动力份额结果。机器人相对价格的下降导致相对较少的工作岗位流失，同时降低了调整成本，或相对机器人生产力的提高，可能更具破坏性。模型隐含的总就业对机器人渗透率（每千名员工的机器人数量）的半弹性介于 0.01% 和 0.12% 之间，这取决于机器人采用率增加的潜在来源，这与实证文献中的发现一致。在扩展中，我们表明减少形式的招聘和解雇成本明确地抑制了长期就业。