{"title":"通过虚拟变量测量结果差异的偏差:注","authors":"Shawn W. Ulrick","doi":"10.3233/JEM-140390","DOIUrl":null,"url":null,"abstract":"Disparity in an outcome between two groups is often measured via the coefficient of a dummy variable in a regression that pools both groups. The dummy is interpreted as the disparity. A casual search of the literature in economics and other social sciences reviews far too many examples of this method to catalog. Unfortunately, if the impact of one (or more) of the control variables differs between the two groups, the measured disparity (i.e., the coefficient on the group dummy) will be biased. We illustrate and derive this bias. Given the bias, we believe that one is better running separate regressions for each group and then implementing decomposition methods or predicting adjusted gaps in outcome (i.e., predicting the but-for world that would exist if the two groups had identical characteristics).","PeriodicalId":53705,"journal":{"name":"Journal of Economic and Social Measurement","volume":"39 1","pages":"153-161"},"PeriodicalIF":0.0000,"publicationDate":"2014-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.3233/JEM-140390","citationCount":"3","resultStr":"{\"title\":\"The bias in measuring disparity in outcomes via a dummy variable: A note\",\"authors\":\"Shawn W. Ulrick\",\"doi\":\"10.3233/JEM-140390\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Disparity in an outcome between two groups is often measured via the coefficient of a dummy variable in a regression that pools both groups. The dummy is interpreted as the disparity. A casual search of the literature in economics and other social sciences reviews far too many examples of this method to catalog. Unfortunately, if the impact of one (or more) of the control variables differs between the two groups, the measured disparity (i.e., the coefficient on the group dummy) will be biased. We illustrate and derive this bias. Given the bias, we believe that one is better running separate regressions for each group and then implementing decomposition methods or predicting adjusted gaps in outcome (i.e., predicting the but-for world that would exist if the two groups had identical characteristics).\",\"PeriodicalId\":53705,\"journal\":{\"name\":\"Journal of Economic and Social Measurement\",\"volume\":\"39 1\",\"pages\":\"153-161\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2014-01-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"https://sci-hub-pdf.com/10.3233/JEM-140390\",\"citationCount\":\"3\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Journal of Economic and Social Measurement\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.3233/JEM-140390\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"Social Sciences\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Economic and Social Measurement","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.3233/JEM-140390","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"Social Sciences","Score":null,"Total":0}
The bias in measuring disparity in outcomes via a dummy variable: A note
Disparity in an outcome between two groups is often measured via the coefficient of a dummy variable in a regression that pools both groups. The dummy is interpreted as the disparity. A casual search of the literature in economics and other social sciences reviews far too many examples of this method to catalog. Unfortunately, if the impact of one (or more) of the control variables differs between the two groups, the measured disparity (i.e., the coefficient on the group dummy) will be biased. We illustrate and derive this bias. Given the bias, we believe that one is better running separate regressions for each group and then implementing decomposition methods or predicting adjusted gaps in outcome (i.e., predicting the but-for world that would exist if the two groups had identical characteristics).
期刊介绍:
The Journal of Economic and Social Measurement (JESM) is a quarterly journal that is concerned with the investigation of all aspects of production, distribution and use of economic and other societal statistical data, and with the use of computers in that context. JESM publishes articles that consider the statistical methodology of economic and social science measurements. It is concerned with the methods and problems of data distribution, including the design and implementation of data base systems and, more generally, computer software and hardware for distributing and accessing statistical data files. Its focus on computer software also includes the valuation of algorithms and their implementation, assessing the degree to which particular algorithms may yield more or less accurate computed results. It addresses the technical and even legal problems of the collection and use of data, legislation and administrative actions affecting government produced or distributed data files, and similar topics. The journal serves as a forum for the exchange of information and views between data producers and users. In addition, it considers the various uses to which statistical data may be put, particularly to the degree that these uses illustrate or affect the properties of the data. The data considered in JESM are usually economic or social, as mentioned, but this is not a requirement; the editorial policies of JESM do not place a priori restrictions upon the data that might be considered within individual articles. Furthermore, there are no limitations concerning the source of the data.