This thesis presents four papers on the role institutions (defined as Douglass North defined institutions) play in generating uncertainty. Specifically, we focus on how uncertainty changes as institutions change. The thesis is empirically focused. In answering our research questions, we employ, often in combination, tools from the ‘traditional’ econometric literature—in particular, structural vector autoregressions— and tools from the natural language processing literature. Two of the papers in this thesis focus on New Zealand’s wide-ranging institutional reforms, which took place from the mid-1980s to the mid-1990. Drawing on the purpose Douglass North stated for institutions: ‘to reduce uncertainty in exchange’ (North, 1991a, p. 97), we argue that institutional reform is only successful if it lowers uncertainty. We find evidence that New Zealand’s reform lowered uncertainty in general (at least, from firms’ point of view), as well as uncertainty about tax and transfers. We also argue in this thesis that institutional change might affect not just the level of uncertainty but also its composition (in terms of the types of uncertainty experienced). Motivated by the ‘Mundellian trilemma’, we argue that if every monetary regime/ institutional setup is imperfect in its own way, then the types of uncertainty associated with the regime will be idiosyncratic. Based on an analysis of a large corpus of historical newspaper articles, using natural language methods, we find the different monetary regimes adopted by the United States and the United Kingdom since 1890 are generally idiosyncratic in terms of the types of uncertainty they generate. In this thesis we also exploit the institutions—uncertainty link for another purpose: the identification of uncertainty shocks. A problem in research on the business cycle impact of uncertainty is finding movements in uncertainty that are plausibly exogenous to the business cycle (‘uncertainty shocks’). We narratively identify (using the New Zealand parliamentary record) a set of quarters where heightened uncertainty was the result of institutional or policy change, rather than the state of the business cycle. We use these quarters as instruments in a proxy structural vector autoregression to quantify the effect of uncertainty on output (amongst other variables). Comparing our results with the conventional recursive method for identifying uncertainty shocks, we find uncertainty shocks have a much larger impact on output. This suggests the estimates from the conventional recursive method suffer from attenuation bias.