Econ Papers Daily / 期刊论文
面试官作为工具：估计吸毒的工资惩罚
Interviewers as instruments: estimating the wage penalty of drug use
摘要
We exploit quasi-random interviewer assignment in the Australian HILDA Survey to instrument for self-reported drug use and estimate its effect on wages. Applying the optimal formula IV framework to six substance categories, we find that IV estimates largely mirror WLS for cannabis, ecstasy, and hallucinogens but diverge sharply for methamphetamine and inhalants — notably, the meth estimate reverses sign, indicating a wage penalty of 10% for frequent users, equivalent to the earnings loss from forgoing a high school diploma. Point estimates under IV suggest a larger premium for high cocaine use, but remain statistically similar to WLS. Romano-Wolf multiple-inference adjustment confirms substance-level significance for five of six drugs; the methamphetamine result, while just outside conventional thresholds after adjustment, is substantively large and clinically plausible. These findings demonstrate the empirical value of interviewer-driven variation for studying the labour-market effects of sensitive behaviours.
中文摘要
我们利用澳大利亚HILDA调查中访谈员分配的准随机性，作为自报药物使用的工具变量，并估计其对工资的影响。应用最优公式工具变量框架对六类物质进行分析，我们发现，对于大麻、摇头丸和致幻剂，工具变量估计与加权最小二乘估计基本一致，但对于甲基苯丙胺和吸入剂则出现显著分歧——值得注意的是，甲基苯丙胺的估计结果符号反转，表明频繁使用者面临10%的工资惩罚，相当于放弃高中文凭所带来的收入损失。工具变量下的点估计表明，高可卡因使用具有更大的溢价，但在统计上与加权最小二乘估计相似。Romano-Wolf多重推断调整确认了六种药物中五种在物质层面的显著性；甲基苯丙胺的结果在调整后虽略超出常规阈值，但实质上较大且在临床上是合理的。这些发现证明了访谈员驱动的变异对于研究敏感行为的劳动力市场效应具有实证价值。