Econ Papers Daily / 工作论文
我们刚刚创建了超过1,000个合成控制组来研究加州20美元快餐最低工资
We Just Created More Than 1,000 Synthetic Controls To Study California's $20 Fast Food Minimum Wage
摘要
Synthetic control methods hold the promise of identifying a credible counterfactual by closely reproducing pretreatment outcome paths. Their reliability, however, depends in part on how effectively they constrain researchers’ less-visible degrees of freedom. We explore these methods’ sensitivity by constructing 1440 plausibly defensible specifications to estimate the employment effect of California’s $20 fast-food minimum wage. While over 96 percent of the estimated employment effects are negative, the range is vast, illustrating that synthetic control methods are sensitive to specification choice. Estimates from simpler and far more transparent difference-in-differences methods lie near the centers of the comparable distributions of synthetic control estimates.
中文摘要
合成控制法有望通过紧密复现处理前的结果路径来识别可信的反事实。然而，其可靠性部分取决于它们如何有效地约束研究者那些不太可见的自由度。我们通过构建1440个看似合理的可辩护设定来估计加州20美元快餐最低工资的就业效应，以此探讨这些方法的敏感性。尽管超过96%的估计就业效应为负，但范围极大，表明合成控制法对设定选择敏感。来自更简单且透明得多的双重差分法的估计值，位于可比的合成控制估计分布的中心附近。