Difference in differences in reverse

Kimin Kim, Myoung jae Lee

    Research output: Contribution to journalArticlepeer-review

    13 Citations (Scopus)

    Abstract

    In the usual difference in differences (DD), there is a control group that is never treated and a treatment group that is treated at some time point. However, there are DD cases where the control group is always treated (instead of always untreated), which we call ‘DD in reverse (DDR)’. This paper examines how the usual DD identification and estimation procedures change for DDR. As it turns out, DDR estimation can be performed in the same way as DD estimation. In contrast, the identification procedure is quite different, because DDR essentially identifies pre-treatment-period effects, whereas DD identifies post-treatment-period effects. An empirical illustration of the effects of a work-hour limit law on actual work hours and wages is provided, where the law is applied to large firms first and then small firms 1 year later in South Korea so that in the second year, the large firms constitute the always-treated control group and the small firms constitute the treatment group. We find that the law raised South Korean workers’ well-being, as their work hours decreased while their real weekly wage increased.

    Original languageEnglish
    Pages (from-to)705-725
    Number of pages21
    JournalEmpirical Economics
    Volume57
    Issue number3
    DOIs
    Publication statusPublished - 2019 Sept 15

    Bibliographical note

    Publisher Copyright:
    © 2018, Springer-Verlag GmbH Germany, part of Springer Nature.

    Keywords

    • Difference in differences
    • Difference in differences in reverse
    • Repeated cross sections
    • Work-hour limit

    ASJC Scopus subject areas

    • Statistics and Probability
    • Mathematics (miscellaneous)
    • Social Sciences (miscellaneous)
    • Economics and Econometrics

    Fingerprint

    Dive into the research topics of 'Difference in differences in reverse'. Together they form a unique fingerprint.

    Cite this