Missing data often pose a problem for researchers. For political scientists who rely heavily on cross-national datasets, missing data may be the bane of their work. The central insight of this chapter is that missing data are not just a problem to overcome. Missing data stand as a phenomenon worthy of investigation and explanation. Patterns of missing data can serve as an approximation of the availability of credible information in a polity – a measure of transparency.

Broadly defined, transparency concerns the full flow of information within a polity. As the previous chapter explains, though, no single measure can capture all facets of transparency. In much of the theoretical literature on political accountability, policy outcomes drive the decisions of citizens to support or oppose incumbent governments. One might expect scholars therefore to employ a measure of data dissemination that captures the transparency of policy outcomes in their empirical research. So far, however, this has not been the case. While several measures of institutional transparency and media freedom are commonly used in empirical studies, few projects employ measures of data dissemination (for some notable exceptions, see Bueno de Mesquita et al. 2003; Islam 2006; Williams 2009; Hollyer, Rosendorff, and Vreeland 2011; Copelovitch, Gandrud, and Hallerberg 2018; Mahdavi 2017). Perhaps scholars have ignored this facet of transparency in their empirical work because a thorough and theoretically rigorous measure of data disclosure has not been available.

Here, we discuss an index of transparency that attempts to fill the demand for such a measure: the HRV index.We treat a state's tendency to disclose data as a latent (unobserved) term predicting the missingness/non-missingness of data on 240 variables from the World Development Indicators (WDI) data series. This latent term is extracted using a dynamic model grounded in item response theory (IRT), which ensures minimal loss of information from collapsing a 240 dimensional observation into a single-dimensional representation. The index has a consistent meaning over time and covers 125 countries from 1980 to 2010.We believe that this approach offers several advantages over earlier measures of data disclosure, as we elaborate below.