在過往文獻中，使用母數模型研究BMI與薪資關係的文獻不勝枚舉，但這些母數模型只能捕捉到BMI對薪資的平均效果，因而忽略了BMI對於薪資較為細微的變化。另外，也較少有文獻使用相對BMI做與薪資相關的研究。因此，本文擬利用較敏感的半母數模型(Semi-parametric Model)，探討相對BMI對於薪資的影響，並與兩階段最小平方法(Two-stage Least Squares)的母數模型進行比較。研究資料主要為「華人家庭動態資料庫」中2006年調查資料，將其區分為30到60歲的中老年群體和17至25歲的年輕族群分別估計。另外，除了考慮薪資的樣本選擇性偏誤，也考慮了薪資與BMI的內生性，分別運用Inverse Mill’s Ratio和工具變數解決此問題，嘗試使用的工具變數包含區域肥胖盛行率、種族與基因變數。半母數模型的實證結果顯示，在男性方面，中老年群體的相對BMI和預期薪資大致上呈現倒U型關係，與二次函數模型估計相類似；而年輕男性相對BMI和預期薪資亦呈現倒U型關係，與二次函數模型估計結果差異較大。在女性方面，無論是在中老年齡層或是年輕族群，相對BMI對薪資大致上皆呈現負向影響，與二次函數估計較不符合，中老年齡層比較接近於線型，年輕族群呈傾斜起伏的倒U型，體型相對苗條的女性其預期薪資會高於體型相對肥胖者。總而言之，體型相對較為纖細的男性對薪資的負面衝擊較大。相反的，女性則是體型較為肥胖者對薪資的負面衝擊較大。 Many articles use parametric models to estimate the relationship between BMI and wage, but it only can capture the average effect of BMI on wage. It ignores the subtle changes of BMI on wage. Besides, relatively few articles estimate the relationship between relative BMI and wage. This thesis hence intends to use the more sensitive semi-parametric model to estimate the effect of relative BMI on wage, in comparison with the results of two-stage least squares. The data mainly come from the 2006 Panel Study of Family Dynamics of Taiwan. This thesis analyzes data divided into 30-60 years-old and 17-25 years-old cohorts. It uses Inverse Mill’s Ratio (IMR) to deal with sample selection bias and the approach of instrumental variables to deal with endogeneity problem. The instrumental variables include prevalence rate of obesity, race and a genetic variable. The empirical results of semi-parametrics show an inverted U-shaped between relative BMI and wage for the old males, and the results of two-stage least squares indicate a similar quadratic function. As for young men, the relationship between relative BMI and wage is like an inverted U-shape, which is quite different from the results of two-stage least squares. Young men with slim figure get lower wage than those young men who are overweight. For both old and young females, the effects of relative BMI on wages are negative. For old women, the relation is close to linear pattern but are not similar with the quadratic function from two-stage least squares. For young women, the relation appears to be a tilted inverted U-shape, which is also quite different from the results of two-stage least squares. Both old and young females with relatively slender figure get higher wages than those who are overweight and obese. In summary, relatively slende2slsr body size has a negative impact on wage for males. In contrast, relatively larger body size has a negative impact on wage for females.