Conference Agenda

Monetary policy, as an important application of narrative economics, carries uncertainty in its decision-making process as a forward guidance tool, as well as being a strategic aspect of its communication policy. It has witnessed several attempts to construct uncertainty indices using uncertainty-related word counts, yielding questionable measurements that overlook key technical terms not included in word lists, thus rendering the computed indices biased and semantically agnostic regarding the actual jargon. This work proposes an in-depth assessment of uncertainty in a collection of international central bankers' speeches (1997-2022) and identifies its key drivers using semantic search models, namely Top2vec, to uncover nested semantic topic structures at the national and international levels as proxies for uncertainty sources. These were found to be robust in discriminating uncertainty features related to probability and risk, in line with the Keynes-Knight debate on uncertainty in macroeconomics. In addition, a robust uncertainty index could be constructed from the similarity between each document and the notion of uncertainty, either country-specific or international, which was found to follow the major global financial and banking events of the last two decades. The method of constructing an uncertainty indicator that measures uncertainty in central bank communications has been shown to have significant predictive power for GDP growth. This ability is comparable to other measures of uncertainty, such as the Economic Policy Uncertainty Index and the Monetary Policy Uncertainty Index. Unlike the other two, the proprietary World Uncertainty Indicator cannot predict a country's economic growth.

This paper provides a review and critical analysis of how economic uncertainty has been defined and measured across foundational and modern literature. One primary contribution of this paper is the critical assessment of whether empirical measures are in lines with their foundational theoretical definitions. By developing a conceptual framework derived from fundamental definitions, this paper evaluates the consistency of recent approaches and challenges the bias to equate uncertainty with observable fluctuations. It argues that although many approaches apply innovative measures, they are inconsistent with the core theoretical and interpretive frameworks of uncertainty. This paper emphasises the importance for clearer linkage between theory and measurement and calls for careful consideration in treating model driven volatility as representative of uncertainty.

This study introduces a novel monthly index of political instability, the WIPI, covering four dimensions for 182 countries from 1996 to 2023. Employing text mining techniques, we extract data from the Economist Intelligence Unit's country reports. We show that our monthly index carries the same information as other frequently used indices of political instability but with a much higher frequency. Using local projections, we identify that political instability reduces economic output significantly. Decomposing the index into its dimensions reveals that instability within the political regime and mass civil protests are the key drivers of the sizable decline in prosperity. We further show that estimates of models employing annual data suffer from a sizeable temporal aggregation bias.

We construct EU-MD, a novel database similar to FRED-MD featuring more than 100 monthly economic time series, to construct macroeconomic uncertainty indices for the set of EU countries and the euro area aggregate. For each cross-section, macroeconomic uncertainty relies on the common uncertainty estimated via stochastic volatility models in the forecast errors of factor augmented autoregressive models for all economic time series. To deal with the large impact of the COVID-19 pandemic, we use COVID-indicators to de-COVID the time series prior to factor estimation. The proposed econometric-based uncertainty indicator has several advantages over existing measures in the literature based on surveys and newspapers. In contrast to survey-based uncertainty measures, our index can be updated regularly at low costs. With regards to newspaper-based indices, our index reflects the uncertainty over a lot of domestic economic variables and is less vulnerable to being influenced by global events or ideological distortions from periodicals in different countries. We use the constructed macroeconomic uncertainty index to estimate a state-dependent Phillips curve with a monthly panel starting in 1999M1 of the euro area countries. The results imply different signs in the response of inflation to the unemployment gap for various threshold values used to distinguish across uncertainty states. In contrast, the persistence of inflation is relatively independent of the uncertainty level.