기업규모별 직업훈련 격차 연구

Abstract

본 연구의 목적은 인적자본기업패널조사(Human Capital Corporate Panel, 이하 HCCP)의 기업조사 균형패널 데이터(Balanced Panel Data)를 사용하여 중소기업과 대기업간 훈련격차의 원인을 실증분석하는 것이다. 패널로짓모형(Panel Logit Model), 패널토빗모형(Panel Tobit Model)을 적용하고 기업규모별 훈련격차 요인분해를 시도한 결과, 근로자 1인당 훈련비용은 중소기업이 대기업에 비해 유의하게 낮았다. 중소기업과 대기업 간 근로자 1인당 훈련비용 격차 요인분해 결과, 훈련제공에 영향을 미칠 수 있는 부존자원에서는 기업규모별로 유의한 차이가 없으나 중소기업에 비해 대기업의 훈련투자 성향이 유의하게 높은 특징을 보였다. 기업규모별 훈련격차 극복을 위해서는 중소기업의 인적자본투자에 대한 수익성 확보가 중요하며, 노사간의 신뢰회복, 대기업과 중소기업 간의 공정하고 수평적인 거래관계, 산업별 인적자원개발 거버넌스의 확립 등이 추진될 필요가 있다.

The provision of vocational training to their employees of small & medium enterprizes is below the level of training provided by the large firms. Certain characteristics by firm size are considered to be the main source of training provision gap to their employees by firm size. This study aims to estimating the training provision gap by firm size and figure out the main cause of the gap based on Oaxaca decomposition method using the Balanced Panel Data of Human Capital Corporate Panel(HCCP) from 3rd wave(2009) to 7th wave(2017). Based on the panel logit, panel Tobit estimation method and Oaxaca decomposition method, I have found several important results. Firstly, there exists significant training expenses gap per employee by firm size but no significant training opportunity gap by firm size measured by wether firms are giving training opportunity to their employees or not. Secondly, various firm characteristics have different effects on firms training expenses per employee by firm size. Labor union has significant positive effect on SME’s training expenses per employees but no significant effect for large firms. Lastly, the difference in training expenses per employee between SMEs and large fims are decomposed to find out the cause of training expenses gap. Difference in training expenses per employee due to the level difference in firm characteristics by firm size does not have significant effects but difference in estimated coefficients for firm characteristics by firm size has an important effects.