Science makes progress with rational debate, disagreement, and discussion. In that vein, I am grateful to the authors of the comment, Steven Durlauf and James Heckman, for engaging in one of the most important social issues of our time. Their comment can be synthesized into three points. First, Durlauf and Heckman (2020) believe that we make claims about discrimination, or the absence thereof, when no such conclusions using a direct regression approach are justified. Second, they describe issues with the endogeneity of police-civilian contacts and suggest modeling those interactions. And, third, they express skepticism of any analysis that relies on police narratives regarding interactions with civilians. Below, I respond to each of these in turn. Here is a brief summary. I disagree with their first point—we went through painstaking effort to belabor what our analysis can and cannot teach us about race and policing. Weuse the term “racial differences”114 times in lieuof themoreprescriptive wording—“racial discrimination.”Weuse the phrase “conditional on an interaction” 20 times throughout the paper and explicitly note at the outset—in the fifth paragraph—that if we do not use this phrase, it is implied. I am not sure howmanymore ways we would have needed to caveat our results to satisfy the authors. I agree with their second point, theoretically—that is why the section immediately following the data description is titled “A Note on Potential Selection into Police Data Sets”—but I am perplexed as to why they claim that our sample is “based on ‘stops.’” The truth is that the vast majority of the data they are referring to are gleaned from 911 calls for service in which a civilian requests police presence, not from police lurking to harass civilians. I agree that police shootings that begin with police stopping a civilian and that interaction escalating and becoming deadly are deeply

中文翻译：

---

对 Steven Durlauf 和 James Heckman 的回应

科学在理性辩论、分歧和讨论中取得进步。在这方面，我感谢评论的作者 Steven Durlauf 和 James Heckman，他们参与了我们这个时代最重要的社会问题之一。他们的评论可以综合为三点。首先，Durlauf 和 Heckman（2020 年）认为，当使用直接回归方法的此类结论不合理时，我们会提出歧视或不存在歧视的主张。其次，他们描述了警民接触的内生性问题，并建议对这些互动进行建模。第三，他们对任何依赖于警方关于与平民互动的叙述的分析表示怀疑。下面，我逐一一一回复。这里是一个简短的总结。我不同意他们的第一点——我们经过艰苦的努力，仔细研究了我们的分析可以和不能教给我们关于种族和警务的什么。我们使用“种族差异”一词 114 次代替更具说明性的措辞——“种族歧视”。我们在整篇论文中使用了“以互动为条件”一词 20 次，并在开头——第五段——明确指出，如果我们不用这个词，它是暗示的。我不确定我们需要多少种方式来警告我们的结果才能让作者满意。我同意他们的第二点，理论上——这就是为什么紧跟数据描述的部分标题为“关于潜在选择到警察数据集的说明”——但我很困惑他们为什么声称我们的样本是“基于”停止。’”事实是，他们所指的绝大多数数据都是从 911 服务电话中收集的，其中平民要求警察在场，而不是来自潜伏骚扰平民的警察。我同意以警察拦下平民开始的警察枪击事件以及不断升级和致命的互动是深刻的