Replication data for: Where's Waldo? Searching for the Hidden Variable of \"Real\" Corruption

Corruption is a commonly studied phenomenon, not only in political science, but across the social sciences. However, the difficulty of measuring corruption presents a significant barrier to this research program. In this paper, we use a well-specified model for explaining corruption presented by Campante, Chor, and Doh (2009) as a starting point for evaluating different measures of corruption. Testing aggregated corruption indices, corruption perception measures, and official records of corruption through non-parametric, parametric, and simulation methods, we find that different measures of corruption produce different results and lead to different substantive conclusions. Furthermore, by treating bias as measurement error, we propose a method to obtain a common denominator across different measures of corruption, which can be used as an instrumental variable to isolate the ``non-perception component\" of corruption measures from the ``perception component.\" By running a bivariate regression on these two components, we contribute to the theoretical debates of the field by shedding light on the relationship between corruption and common covariates such as stabilty and democracy.

腐败是政治学乃至整个社会科学领域均被广泛研究的现象。然而，腐败的测量难题却成为该研究议程推进的重大阻碍。本文以Campante、Chor与Doh（2009）提出的规范腐败解释模型为起点，对不同的腐败测度方式展开评估。通过非参数、参数及模拟方法，本文对综合腐败指数、腐败感知测度与腐败官方记录三类测度手段进行检验，结果显示不同的腐败测度会得出相异的结果，进而导向不同的实质性研究结论。此外，本文将偏差视作测量误差，提出了一种可获取不同腐败测度间共同基准的方法，该方法可作为工具变量，将腐败测度中的“非感知成分”与“感知成分”加以分离。通过对这两类成分开展二元回归分析，本文阐明了腐败与稳定性、民主等常见协变量之间的关联，为该领域的理论辩论贡献了新的研究视角。