Visiting Speaker: Julian Ludwig, Texas Tech University
Presentation Title: Disentangling Intrinsic and Informational Uncertainty
Abstract:
This paper distinguishes between two types of uncertainties in macroeconomic forecasting: intrinsic uncertainty and informational uncertainty. Intrinsic uncertainty arises from the variance of the residuals in the data-generating process, making it more difficult to predict future outcomes based on past information. Informational uncertainty occurs when increased noise in the data reduces the quality and precision of information, thereby increasing uncertainty due to external factors. This distinction has two important implications.
First, I demonstrate that when informational uncertainty changes over time, existing measures of uncertainty—such as volatility indices (which track fluctuations in financial data) and forecast dispersion (which reflects the level of disagreement among forecasters)—may become unreliable. This is because changes in uncertainty driven by informational factors can lead to a negative relationship between actual uncertainty and these proxies. Specifically, when individuals have less access to information, they place less weight on new data and rely more on outdated information, which reduces the variability and disagreement in their forecasts.
Second, to measure the causal effect of uncertainty on economic variables, I argue that it is more effective to use informational uncertainty as the treatment or instrumental variable rather than intrinsic uncertainty. Altering intrinsic uncertainty requires modifying the data-generating process itself, potentially confounding results since this process also generates the dependent variable. In contrast, informational uncertainty can increase exogenously without changing the underlying process, ensuring that any observed effects are due to changes in uncertainty alone.
To disentangle these two types of uncertainty, I leverage the fact that intrinsic uncertainty affects the proxies of uncertainty proportionally, while informational uncertainty has a non-linear impact. I identify the two types in two ways. First, I use existing measures of uncertainty and dispersion from the literature to extract them. Second, I analyze dispersion across different forecast horizons from the Philadelphia Fed's Survey of Professional Forecasters. Using these new measures, I show that informational uncertainty is a significant component of overall macroeconomic uncertainty. Furthermore, I find that the economic impact of uncertainty is reduced when shifting uncertainty through the more exogenous informational uncertainty measure.
Host: Christiane Baumeister