经济与社会科学2016journals.plos.org
美国警察枪击事件中种族偏见的贝叶斯分析
Bayesian Analysis of Racial Bias in Police Shootings in the United States
专业 · 约 55 分钟读完前置知识：贝叶斯统计、多层回归模型、美国警务数据
导读
这篇论文用贝叶斯方法分析美国警察枪击数据（作者承认数据来自众包且不完整），估计手无寸铁的黑人被警察枪杀的概率约为白人的3.49倍。县级层面的分析还显示，这种种族偏见更可能出现在大都市、收入中位数低、黑人居民比例较高且财政不平等严重的县，而与犯罪率没有关系。
HN 关注点：HN 读者的讨论焦点是：3.49 倍差距是否因为没有控制警民接触率、众包数据是否完整、县级聚合是否落入辛普森悖论，以及数据之外的少数族裔社区信任与亲身叙述应占多大分量。
后来怎么样
该论文的“3.49倍”估计成为美国警务种族偏见公共讨论中的标志性数字，并促使后续研究更重视警民接触率与众包数据偏差。
HN 评论精选
这几条评论里：质疑 3 补充 1
质疑批评研究未控制警民接触率，可能把“接触更多”误读为“接触更致命”。
他们真正需要做的是把不同种族的警民接触概率纳入分析。也就是说，我们想知道：在所有与警察的接触中，如果受害者是黑人，是否更可能致命？如果不考虑警察接触率，结论可能只是说明黑人更可能遇到警察；这本身是个问题，但性质略有不同，更指向社会经济因素决定社区警务密度，而非警察种族主义。
akud查看英文原文
> the probability of being {black, unarmed, and shot by police} is about 3.49 times the probability of being {white, unarmed, and shot by police} on average What they really need to do is factor in the probability of encounters with police gI've race. That is, we want to know, out of all encounters with police, are they more likely to be fatal if the victim is black? Without factoring in the rate of police encounters, the conclusion could just be indicating that black people are more likely to encounter the police, which is a problem in itself but is slightly different. That would point more to socioeconomic factors determining the rate of policing in neighborhoods, rather than police racism.
质疑质疑众包数据完整性，并反驳作者“随机子集不影响均值”的说法。
这项分析基于众包的枪击数据。我不是说它因此无效，但在信任结果前需要更多了解数据集；这种数据很难完整，FBI 官方数据也有同样问题。我们需要要求地方警察部门以标准化方式报告所有枪击。补充：作者在结论部分承认数据不完整，但奇怪地认为这不太会影响均值，理由是所用样本是“待完成数据集的一个大而随机的子集”。我不太理解：对不完整数据做随机抽样怎么能改善结果？
jim-greer查看英文原文
This analysis is based on crowd-sourced data on shootings. I'm not saying that makes it invalid, but I'd need to know more about the dataset before trusting the results. It seems hard to have it be complete. The official FBI data has the same problem. We need to require local departments to report all shootings in a standardized way. Edit: the author acknowledges the incompleteness of the data in the conclusions section. Oddly, he doesn't think that's likely to affect the mean "since the sample used herein is a large and random subset of the to-be-completed data set". That doesn't really make sense to me. How would random sampling of incomplete data improve the results? http://regressing.deadspin.com/deadspin-police-shooting-database-update-were-still-go-1627414202
质疑批评县级聚合可能产生辛普森悖论，掩盖社区层面的真实机制。
结论还说，县级预测变量分析表明，警察枪击中的种族偏见最可能出现在大都市、收入中位数低、黑人居民比例可观且财政不平等高的县，并且与犯罪率（甚至按种族划分的犯罪率）无关。免责声明：结论可能仍正确，但我认为在县级做分析完全荒谬，因为你在社区层面很容易遭遇辛普森悖论。假设多数警察枪击发生在贫困社区，而贫困社区中黑人/白人比例很高；他们的分析会暗示存在种族偏见，但实际上这种偏见可能完全由人口构成解释，也可能不是，但在县级聚合会洗掉所有有用信号。
chris_va查看英文原文
Finally, analysis of police shooting data as a function of county-level predictors suggests that racial bias in police shootings is most likely to emerge in police departments in larger metropolitan counties with low median incomes and a sizable portion of black residents, especially when there is high financial inequality in that county. There is no relationship between county-level racial bias in police shootings and crime rates (even race-specific crime rates), meaning that the racial bias observed in police shootings in this data set is not explainable as a response to local-level crime rates. With the disclaimer that the conclusion might still be correct, I think looking at the county level is completely absurd. You leave yourself open to the Simpson's Paradox at a neighborhood level. For argument's sake, let's say that the majority of police shootings happen in poor neighborhoods. Let's also assume, sadly, that the ratio of black/white people in poor neighborhoods is high. Their analysis would imply that their is a racial bias to the shooting, when in fact, the racial bias could be entirely explained by the demographics. Or it might not, but doing it at a county level completely washes out all useful signal.
补充主张数据研究应放进更广泛的证据与社区信任语境中，不能只凭数据否定少数族裔的亲身叙述。
这里有些评论指出数据质量等问题，这些确实是真实担忧，更好的数据当然能改进分析。但不要孤立看这项研究：它是大量观察和证据中新添的一部分，支持“个人种族偏见影响美国警务”的理论。证据包括其他研究、刑事调查、刑事案件、联邦调查、与克利夫兰和西雅图等警局的改革协议、暴力执法视频和照片，以及少数族裔社区数十年来关于警察如何对待他们的叙述。最后一点很重要，因为它涉及信任，而信任是问题核心。许多少数族裔社区不信任警察会像保护更白、更富社区那样保护他们，并且有故事解释为何不信任。如果你只依赖数据研究来判断警务中的种族偏见，你就是在隐式地不信任或拒绝少数族裔个人和社区的说法。为什么会这样？值得想一想。这也回到数据：为什么数据如此匮乏？回答绕回偏见，因为直到最近，是警察部门自己在提供……
snowwrestler查看英文原文
Some of the comments here are pointing out the questions about data quality etc. that affect a study like this. And those are real concerns, and obviously analytical studies could only be improved if they had better data to work with. But let's not make the mistake of looking at this study only in isolation. It is a recent addition to a large collection of observations and evidence that support a theory that personal racial bias affects American policing. The evidence includes other studies, criminal investigations, criminal cases, federal investigations and reform agreements with police departments like Cleveland and Seattle, videos and photos of violent police encounters, and of course decades of stories and statements from minority communities about how the police treats them. The last one is important because it gets at trust, which is the heart of the issue. Minority communities, many of them, do not trust the police to protect them in the same way they protect whiter/richer communities, and they have stories that explain why not. If you are depending solely on data-driven studies to inform your opinion on racial bias in policing, then you're implicitly saying that you distrust or reject what minority people and communities say. Why is that? It's worth thinking about IMO. Which brings us back to the data. Why is it so lacking? You can't answer that question without coming back around to bias, because until recently, it was the police forces themselves who supplied the data, or not, or only part of the data. So discounting the bias reported in this study because of data problems is getting toward begging the question, logically speaking. The essential question, when it comes to whether you agree that racial bias affects policing, is: what level of evidence will convince you?
相关资源
从这篇论文所有 HN 讨论帖的全部评论里整理出来的链接。
历次讨论
- 2016Hacker News 讨论帖256 分 · 348 评
同领域还值得读
一切的回报率，1870–2015
The Rate of Return on Everything, 1870–2015
利用1870–2015年多国的长期数据，估算股票、住房等各类资本的回报率，并将其与经济增长率进行比较。
经济与社会科学2019 · 487 分 · 260 评进阶 · 约 100 分钟编辑精选一切资产的回报率，1870–2015
The Rate of Return on Everything, 1870–2015
构建了 1870–2015 年 16 个国家的长周期资产回报数据集，覆盖股票、住房、政府债券和短期票据，系统比较各类资产的真实回报与波动性。
经济与社会科学2018 · 215 分 · 151 评进阶 · 约 120 分钟编辑精选一切资产的回报率，1870–2015（2019）
The Rate of Return on Everything, 1870–2015 (2019)
这篇发表于《经济学季刊》的论文构建了 1870–2015 年间 16 个主要发达国家的股票、债券、房地产等资产收益率的长周期数据库。
经济与社会科学2024 · 190 分 · 148 评进阶 · 约 80 分钟编辑精选想法是否越来越难找？
Are ideas getting harder to find?
探讨“想法是否越来越难找到”：通过衡量多个领域的“研究生产率”，发现维持相同的增长需要投入指数级增长的研究资源，即研究效率在持续下降。
经济与社会科学2016 · 182 分 · 106 评入门 · 约 60 分钟编辑精选比特币与区块链的经济极限
The Economic Limits of Bitcoin and the Blockchain
芝加哥大学 Eric Budish 的论文从经济学角度分析比特币与区块链的根本限制：核心论点是只要攻击网络（双花或砸盘做空）的收益超过挖矿成本，51% 算力攻击就有利可图，而挖矿…
经济与社会科学2018 · 176 分 · 37 评进阶 · 约 90 分钟编辑精选