In my dissertation, I devise novel methodological contributions that incorporate my expertise in qualitative and quantitative methods with an application to the study of comparative political economy or violent conflict.
In the first paper, Bayesian Integrative Meta-Analysis, I challenge the traditional quantitative framework of meta-analysis, which cannot meaningfully incorporate quantitative and qualitative scholarship into a single analysis. I articulate a novel method for incorporating both these types of studies into the same meta-analysis to account for the diverse evidence types that compose political science scholarship. I create a process called conversion elicitation— a structured judgment process in which coders evaluate the information obtained from a qualitative study. I demonstrate how the information gathered from this process can be input into a meta-analysis through both a simulated case study and an applied example concerning the effect of wage increases on bureaucratic corruption.
The second paper, Hybrid Sampling Strategies: A Design Framework for Merged Survey Designs, develops a logic for merged probability and non-probability survey samples. As survey costs grow and population information remains variable across contexts, probability-based survey techniques are increasingly unfeasible. This reality significantly impacts scholars studying limited information contexts in the Global South. With these intractable issues in mind, this project asserts a design framework for merging probability and non-probability samples using knowledge about the structure of the population of interest. I argue that context-driven decisions related to combined sampling designs can help to mitigate cost burden while still retaining means to estimate uncertainty.
The third paper, Measuring Costly Concepts: A Multi-Method Strategy for Measurement of Many-N Cases, provides a solution for what I call costly concepts, concepts for which a direct measure exists, but data collection is too expensive to engage in over many cases. Typically, scholars use proxy measures of these concepts. But when proxies are not a one-to-one match with the accurate concept measurement, there is non-random measurement error. I suggest that a mixed-method approach to measurement, in which true measurement of the concept is obtained over a set of cases, can be useful to define the contours of the bias between the true measure of concepts and a chosen proxy. I further demonstrate how case insights can inform the basis of improved measurement of the concept. I provide an illustration of this method to measure access to essential services, such as running water or sewage, at sub-municipal levels. In an additional paper, Ana Arjona and I extend this method to the measurement of municipal-level armed group presence in Colombia.