Trade unions and workers’ welfare in Vietnam

ABSTRACTThe role of trade unions in the labour market is currently a hot topic in Vietnam as there is a plan to transition from a single state-led union system to a new system that allows the participation of some types of independent unions. This study provides the first evidence on the links between union coverage and workers’ welfare utilising a Vietnamese national representative dataset. An instrument variable technique is applied for estimation. Using cross-sectional data, the study finds that union coverage negatively affects both sexes in terms of earnings and labour supply while positively affecting their participation in social insurance. However, the effects disappear for women in terms of earnings and for both sexes in terms of labour supply when investigating the non-state sector. The study also shows some important results using panel data.KEYWORDS: Trade unionworkerwelfareincomework hourssocial insurance Disclosure statementNo potential conflict of interest was reported by the authors.Notes1. In this study, we use the commands IVREG2 and XTIVREG2 of STATA for IV estimations and IV-FE estimations, respectively. For binary outcome (e.g. social insurance participation), we also use these commands because the IVPROBIT command works with continuous endogenous covariates only and a command like IVLOGIT does not exist.2. For each wave, VLFS is designed on a 2-2-2 rotation basis. A household is surveyed for two consecutive quarters. It is then temporarily excluded from the survey for another two consecutive quarters. It is surveyed again in the next two quarters. Therefore, we can construct a short-panel data sample with two periods from the 2018 wave of VLFS. Each worker in the sample was observed twice. The first observation of the worker conducted in the first half of that year is considered to be made in the first period. The second observation of the same worker conducted in the second half of that year is considered to be made in the second period (i.e. a part of the sample were observed twice, the first observed in Quarter II and the second observed in Quarter III of that year. The rest of the sample were also observed twice, the first observed in Quarter I and the second observed in Quarter IV of that year).3. For quantile fixed effect estimation, we cannot use XTQREG command of STATA because it works only with a large number of periods while we has only a short panel of data with two periods. Therefore, we use QREGPD for the estimations.