The Unassuming Economist

From a paper by Thomas Chalaux, Dave Turner and Steven Cassimon:

“Recent research by authors from the IMF, ECB and the Bank of England has identified Random Forests as the most effective method for predicting crisis episodes and superior to the more traditional approach of probit/logit modelling. We challenge that finding when predicting recessions for 20 OECD countries. A customised algorithm that selects probit models, matches the performance of Random Forests in out-of-sample quarterly predictions using real-time data over a two-year horizon, including the period of the Global Financial Crisis. This enhanced performance is attributed to a “wisdom of crowds” feature whereby predictions are averaged from many well-fitting probit equations, comparable to Random Forests averaging many trees. Both country-specific and pooled Random Forests are estimated, and, although the latter has a superior out-of-sample performance, a disadvantage of the pooled approach is that recession risks are highly correlated across countries and rarely very elevated, so that it is unlikely that a recession is ever ‘more likely than not’. The estimation framework is also separately applied to 8 consecutive quarterly horizons and demonstrates that different explanatory variables matter at different horizons; activity variables are more important for immediate quarters, but financial cycle variables (especially credit and house prices) dominate at horizons beyond that.”

From a paper by Tim Köhler:

“This study presents an examination of the predictive power of narrative reports from German economic institutes beyond traditional quantitative forecasts in anticipating economic recessions and directional changes in the business cycle. I transform qualitative narratives into quantitative sentiment scores using four different dictionaries and methods and use fixed-effect logistic regression to analyse their impact. To evaluate model performance, I use the Area under the Receiver Operating Characteristic Curve (AUROC) to compare models with versus without sentiment scores. Additionally, I employ DeLong’s test and bootstrapping to test the significance of AUROC improvements. Furthermore, I explore the potential of combining multiple sentiment scores to enhance forecasting accuracy. The results show that sentiment scores significantly enhance forecasting accuracy. This suggests that narrative information provides valuable insights beyond quantitative forecasts alone.”

From a paper by Nigel Meade and Ciaran Driver:

“Policy makers are concerned with the accuracy of GDP forecasts and want to understand the reasons for the revision of forecasts. We study these issues by examining forecasts of annual UK GDP growth by a panel of agents, published monthly by HM Treasury. We focus on two main issues: the developing accuracy of the group-mean forecast as horizons shorten and the identification of information categories underlying agents’ forecast revisions. The accuracy of the group-mean forecast is poor; there is evidence of information rigidity in forecasts within the target year, and accuracy only improves in May of the target year when contemporary information flows lead to increased accuracy. We find a pessimism bias; the median errors of group-mean forecasts are increasingly positive for horizons shorter than 17months. We seek to explain revisions to both long- and short-horizon group-mean forecasts and individual agent forecasts. Modeling individual agents’ forecast revisions using a moving window, we note a consistent tendency by agents to revise their forecast towards the group-mean. Although their importance varied over time, the main information categories explaining revisions were, over longer horizons, the cost of finance, production, and a business confidence indicator. FX rates and inflation were influential over shorter horizons.”

From a paper by Tim Köhler:

“This study presents an examination of the predictive power of narrative reports from German economic institutes beyond traditional quantitative forecasts in anticipating economic recessions and directional changes in the business cycle. I transform qualitative narratives into quantitative sentiment scores using four different dictionaries and methods and use fixed-effect logistic regression to analyse their impact. To evaluate model performance, I use the Area under the Receiver Operating Characteristic Curve (AUROC) to compare models with versus without sentiment scores. Additionally, I employ DeLong’s test and bootstrapping to test the significance of AUROC improvements. Furthermore, I explore the potential of combining multiple sentiment scores to enhance forecasting accuracy. The results show that sentiment scores significantly enhance forecasting accuracy. This suggests that narrative information provides valuable insights beyond quantitative forecasts alone.”

From a paper by Lin Li, and Guoping Li:

“Using data on analysts’ individual forecasts from all listed companies in China’s A-share market for the period 2007–2022, we document two stylized facts. First, the average forecast error can be predicted by forecast revision. Second, individual forecasts appear to over-react to own revisions and salient public signals, we show that the first fact is inconsistent with standard models of full information rational expectations. The second fact suggests that individuals may be irrational with respect to their use of information. Expectation formation theory suitable for China’s capital market may need to combine information friction and irrational belief.”