The Methodology of Randomized Experiments in Economics

經濟學中隨機控制實驗的方法論分析

鄭仲棠(Chung-Tang Cheng)

指導教授 ： 趙相科

國立清華大學/科技管理學院/經濟學系/碩士(2013年)

https://doi.org/10.6843/NTHU.2013.00647

摘要

摘要

二十世紀六零年代，隨機控制實驗 (Randomized Controlled Trial，簡稱RCT) 於醫藥實驗取得了成功，成為了科學方法上的黃金標準 (gold standard)。自八零年代以來，RCT亦逐步地被應用在經濟學的田野實驗 (field experiment) 中，並一定程度地影響了實證經濟學的研究走向。奠基於醫藥實驗的成功上，一些實證經濟學家認為：RCT作為一種黃金標準，也能夠對社會政策的制定提供客觀證據。本論文從思想史及科學方法論的角度，試圖探討：（一）為何RCT在實證經濟學家之間逐漸蔚為主流？（二）RCT真的能夠確證因果關係嗎？其中的理論限制為何？（三）有沒有任何實作 (practice) 能夠映證第二個問題的解答？

茲陳述本論文的研究結果如下。首先，在RCT的架構下實證經濟學家解決了傳統計量經濟學「被動觀察」 (passive observation) 的問題，並且為政策工具找到了客觀依據。其次，在「理想RCT→真實RCT→真實世界」的方法論圖像中，我們認為，實證經濟學家或其他科學實作者於田野RCT將面臨理想化缺口(idealization gap) 和一般化缺口 (generalization gap) 此兩類方法上的限制，各包含了實作時可能造成之偏誤 (bias) ，進而分別威脅其內部有效性 (internal va-lidity) 和外部有效性 (external validity)。最後在案例研究中，我們了回顧田納西州的小班實驗與加州的小班政策。結果顯示，此案例存在偌大的一般化缺口，至於理想化缺口是否出現則仍有待驗證。

關鍵字

隨機控制實驗 ； 因果關係 ； 田野實驗 ； 內部有效性 ； 外部有效性 ； 小班教學實驗

並列摘要

Abstract

In the 1960s, Randomized Controlled Trial (RCT) was established as a ‘gold standard’ in clinical trial. Since the 1980s, RCT-method has gradually adopted in field experi-ments in economics and greatly influenced empirical economics. Based on the success of clinical trial, some empirical economists feel that RCT is also a ‘gold standard’ that provides objective evidence in policy-making procedures. This thesis, will consider (a) why the RCT-method becomes popular among empirical economists? (2) Can RCTs in economic research truly confirm causal hypotheses? Are there any theoretical limi-tations in the RCT methodology? (3) Does any real practice correspond to these limi-tations?
Three brief conclusions are, first, RCT solves the problem of passive observation, which bothered traditional econometrics, and RCT as a policy tool is proved to be ‘objective’. Second, in the methodological figure we proposed, practicing economists and scientists may confront ‘idealization gaps’ and ‘generalization gaps’ that threaten the internal validity and external validity of field RCTs, respectively. Lastly, we take two class-size deduction cases: the Project STAR in Tennessee and the CSR Program in California as examples. The results indicate that the ‘generalization gaps’ are huge, and the ‘idealization gaps’ in these are still to be verified.

並列關鍵字

randomized controlled trial ； RCT ； causal relation ； field experiment ； internal validity ； external validity ； Project STAR

參考文獻

Abadie, A. 2005. “Semiparametric Difference-in-Differences Estimators.” Review of Economic Studies, 72(1): 1-19.

Angrist, J.D. and Lavy, V. 1999. “Using Maimonides’ Rule to Estimate the Effect of Class Size on Scholastic Achievement.” Quarterly Journal of Economics, 106(4):979-1014.

Ackermann, R. 1989. “The New Experimentalism.” British Journal for the Philoso-phy of Science 40: 185-90.

Averett, S.L. and McLennan, M.C. 2004. “Exploring the Effect of Class Size on Student Achievement: What Have We Learned Over the Past Two Decades?” in International Handbook on the Economics of Education, Geraint Johnes and Jill Johnes, eds. Edward Elgar Publishers.

Angrist, J.D. and Krueger, A.B. 2001. “Instrumental Variables and the Search for Identification: From Supply and Demand to Natural Experiments.” The Journal of Economic Perspectives, 15(4): 69-85.