A multinominal LOGIT analysis of various employment and nonemployment choices in Taiwan: with emphasis on selectivity bias in wage equations of employment sector

The purpose of this paper is two/fold:first, apply the multi/nominal logit model to estimate the attachment to various employment and non/employment sectors; second, adopt the generalized two/stage estimation approach to correct for selectivity bias in the wage equation of each employment sector. The employment section is classified as four mutually exclusive sectors of atypical/part/time, atypical/full/time, typical/part/time, and typical/full/time employment, while$ the non/employment section including both sectors of unemployment and non/labor force.By fitting the related estimation models to household data from the Manpower Utilization Survey of Taiwan conducted in each May of 2011/2016, the results can be briefly$summarized as follows: (1) Those$with higher education level, married males, and single fathers, as well as generations of X Y,and Z,have a higher probability of attaining the typical/full/time employment than their counterparts.(2) The positive selection is found to be important for the four employment sectors. Especially, the two part/time samples are found to have greater positive selection bias than the two full/time samples mainly because the population wage dispersions of the former samples are greater than those of the latter ones. (3) Of those of the four employment sectors, only is the life/cycle wage profile of the typical/full/time employment sector compatible with what is predicted in human capital theory. (4) Holding other things equal, the magnitudes of wage profiles by gender and marital status are found to be in the order of married male (single father) > single male> single female (single mom) > married female. (5) The county/level aggregate unemployment rate imposes a non/negligible negative impact on the wage rate of each employment sector.