Any researcher on corruption has faced at some point the dataset dilemma. How can one assess the incidence of a phenomenon on corruption levels if we cannot determine how much corruption is there in the first place? The problem compounds when the research has a transnational or comparative element. How can one assess how different corruption levels are in different jurisdictions if we cannot be sure if the measurements are comparable? It becomes critical when the research has a diachronic component, and tries to incorporate changes over time, as the stability and consistency of datasets become essential. This article reviews the literature on the topic from the last fifteen years and evaluates all the main options available today for researchers and policy designers in terms of validity and reliability, explaining first the particularities of these two concepts in the context of measurements of the prevalence of corruption. It pays particular attention to the limitations of the different datasets and determines the validity and reliability of the oft-used Corruption Perceptions Index post-2012 (CPI) and the Control of Corruption (CoC) indicator for the whole data series. This conclusion partially vindicates those researchers and policy designers who have used these datasets in the past, especially when compared to all the other options, while firmly warning users about the kind of conclusions that can be extracted from them.

任何腐败研究人员都曾在某个时刻面临数据集困境。如果我们首先不能确定多少腐败，那么如何评估腐败现象的发生率呢？当研究具有跨国性或比较性因素时，问题就会加剧。如果我们不能确定这些衡量标准是否可比，那么如何评估不同司法管辖区的腐败程度如何？当研究具有历时性并试图随着时间的推移纳入变化时，这一点就变得至关重要，因为数据集的稳定性和一致性变得至关重要。本文回顾了过去15年中有关该主题的文献，并从有效性和可靠性方面评估了当今可供研究人员和政策设计者使用的所有主要选项，首先在衡量腐败发生率的背景下解释这两个概念的特殊性。它特别注意不同数据集的局限性，并确定整个数据系列常用的2012年后腐败感知指数（CPI）和腐败控制（CoC）指标的有效性和可靠性。该结论部分证明了过去使用过这些数据集的研究人员和政策设计者，尤其是与所有其他选项进行比较时，同时还向用户发出可以从中提取结论的坚定警告。它特别注意不同数据集的局限性，并确定整个数据系列常用的2012年后腐败感知指数（CPI）和腐败控制（CoC）指标的有效性和可靠性。该结论部分证明了过去使用过这些数据集的研究人员和政策设计者，尤其是与所有其他选项进行比较时，同时还向用户发出可以从中提取结论的坚定警告。它特别注意不同数据集的局限性，并确定整个数据系列常用的2012年后腐败感知指数（CPI）和腐败控制（CoC）指标的有效性和可靠性。该结论部分证明了过去使用过这些数据集的研究人员和政策设计者，尤其是与所有其他选项进行比较时，同时还向用户发出可以从中提取结论的坚定警告。