台灣工作風險與薪資關係之追蹤資料分析研究

Abstract

國內外政府及學者為衡量降低死亡風險之效益，通常使用特徵工資模型估計薪資的風險彈性，並且計算一個統計生命的價值（value of a statistical life，VSL），然而此專有名詞容易被錯誤解讀，因此本研究依照Cameron（2010）的方法改為計算風險變動百萬分之一的願付價值(the willingness to pay for a micro-risk reduction，)。\r\n過去補償性工資差異的研究常使用橫斷面資料，並假設不存在內生性問題，直接以普通最小平方法來分析，然而Garen（1988）認為風險是一個內生變數，因此他改用兩階段最小平方法處理風險之內生性問題。\r\n本研究使用的勞工特徵與風險特徵的資料，分別來自中央研究院人文社會科學研究中心華人家庭動態資料庫、勞工保險局勞工職業災害保險資料庫，華人家庭動態資料庫是1999年至2014年的追蹤資料。\r\n我們將全體就業者分成技藝有關人員、管理與專業人員等二類職業別，並用追蹤資料固定效果模型進行迴歸分析，可以處理部分內生性問題，其全體就業者、技藝有關人員、管理與專業人員薪資的風險彈性分別為0.0338、0.0567、0.0220，前兩者統計上顯著，代表其他情況不變下，行業風險越高工資也越高，因此存在補償性工資差異，然而管理與專業人員面對的工作風險較低，此職業別不會和雇主協議風險特徵的報酬，雇主在工作特徵中也不會放入風險特徵，所以管理與專業人員不存在補償性工資差異。\r\n最後，由加入環境特徵變數的迴歸估計結果，發現受僱者偏好較舒適的氣溫與較清淨的空氣，所以一月氣溫較高的地區有較多的勞動供給與較低的薪資，另一方面，七月氣溫越高的地區有較低的勞動供給與較高的工資，同樣的，懸浮微粒濃度較高的地區須用較高的薪水來吸引員工來此地就業。此結果符合Roback（1982）之理論預期與實證結果。

In general, the risk elasticity coefficient of wage and the value of a statistical life (VSL) are estimated by the hedonic wage model in order to evaluate the benefit of reducing mortality risk for governments and scholars. However, the term of VSL can easily be misinterpreted literally, thus this thesis instead utilized the approach proposed by Cameron (2010) to calculate the willingness to pay for a micro-risk reduction ().\r\nPast studies on compensating wage differentials usually use cross-sectional data, assume that there is no endogeneity problem, and directly conduct analysis with ordinary least squares regression. However, Garen (1988) considers that risk is an endogenous variable and adopts a two-stage least squares approach to deal with the endogeneity problem of risk.\r\nThe data of labor and risk characteristics used in this thesis are from the Panel Study of Family Dynamics during 1999 to 2014 published by the Research Center for Humanities and Social Sciences in Academia Sinica and from the Bureau of Labor Insurance, Ministry of Labor respectively.\r\nThe total employees are divided into two categories: technical workers and managers/professionals. In order to tackle the partial endogeneity problem, the panel data and fixed effects model are utilized to conduct a regression analysis. We found that the risk elasticity coefficients of total employees, technical workers, and managers/professionals are 0.0338, 0.0567, and 0.0220 respectively. The first two regressions are statistically significant, indicating that the higher the risk, the higher the wage if other things being equal, hence the results confirm the existence of compensating wage differentials. Nevertheless, managers/professionals face less risk at work so they won’t ask for any risk premium. Their employers also won’t add risk to the characteristics of the job; therefore, there is no compensating wage differentials for managers and professionals.\r\nIn addition to employer’s and employee’s characteristics, environmental variables are also important factors of the wage equation, and the result shows that employers have to pay higher wages to attract people to work in areas with less comfortable temperatures and dirtier air quality because employees prefer more comfortable temperatures and cleaner air. This finding is consistent with Roback’s (1982) theory and empirical results.