條件獨立下的隨機邊界處理效果模型

A Stochastic Frontier Model with Unconfounded Treatment Assignment

趙新曜(Sin-Yao Chao)

指導教授 ： 王泓仁

共同指導教授 ： 許育進(Yu-Chin Hsu)

國立臺灣大學/社會科學院/經濟學研究所/碩士(2020年)

https://doi.org/10.6342/NTU202001838

摘要

為了設計適當的政策進而提高生產率，部分決策者會將目標放在提升生產技術上；然而可能有些人會希望改善生產的效率。在本文中，我們提出了一個新的隨機邊界模型(stochastic frontier model)來衡量條件獨立(unconfounded)下的處理效果(treatment effect)，藉此進一步評估政策的成效。在條件獨立的假設下，我們透過機率倒數加權法(inverse probability weighting)得到對應參數的認定結果(identification)。因此我們能利用加權非線性最小平方法(weighted nonlinear least square)來估計政策對於生產前沿和效率水準的影響。我們分別以非參數化及參數化兩種方式來估計作為權重的傾向分數(propensity score)，也相對應地推導出了參數估計式的漸進常態分佈。我們進行蒙地卡羅模擬來評估模型的表現，結果顯示當樣本數足夠大時，會得到較小的估計誤差。

關鍵字

隨機邊界模型 ； 處理效果

並列摘要

To design an appropriate policy for productivity enhancement, some policy makers aim at production technological progress, while the others may focus on how to improve efficiency during the process. In this paper, we propose a new stochastic frontier model with unconfounded treatment to further evaluate the policy effect. Under unconfoundedness assumption of treatment assignment, we obtain identification of parameters by inverse probability weighting. Therefore, we could estimate average treatment effect on production frontier and inefficiency level through weighted nonlinear least square method. In addition to non-parametric propensity score, we also estimate propensity score parametrically for weighting. Accordingly, we derive asymptotic normality for the estimator of parameters respectively. To evaluate our model, we conduct Monte Carlo simulations and show that bias of estimation is small if sample size is large enough.

並列關鍵字

Stochastic frontier model ； Treatment effect

參考文獻

[1] Aigner, D., C. Lovell, and P. Schmidt (1977). “Formulation and Estimation of Stochastic Frontier Production Function Models,” Journal of Econometrics 6, p21–37.

[2] Ashenfelter, O. (1978). “Estimating the Effect of Training Programs on Earnings,” Review of Economics and Statistics 60, p47–57.

[3] Abadie, A., and Imbens, G. (2002). “Simple and Bias-Corrected Matching Estimators for Average Treatment Effects,” NBER working paper, No. 283.

[4] Abadie, A., and Imbens, G. (2011). “Bias-Corrected Matching Estimators for Average Treatment Effects,” Journal of Business and Economic Statistics 29, p1–11.

[5] Berger, A., and Humphrey, D. (1991). “The Dominance of Inefficiencies over Scale and Product Mix Economies in Banking,” Journal of Monetary Economics 28, p117–48.