DataScience@BI seminar with Jonas Vestby

Abstract:

"Integrating political factors in long-term forecasts of economic growth"
Jonas Vestby and Halvard Buhaug

While we know that countries' economic development is influenced by the political context, it has proven hard to find a statistical model that can utilize political co-variates and make better future expectations than models without such co-variates. One reason is that changes in the future political context are almost impossible to forecast. However, even when using the observed political context as co-variates in out-of-sample models of economic growth, it turns out it is hard to make good use of the co-variates. This article provides a comprehensive exploration of the role of two political co-variates in growth forecasting: electoral democracy and armed conflict. We find that a reason for why it is hard to utilize these co-variates is that their effects are affecting other distributional properties than the expectation. Another reason is that the effects are likely dynamic: the early effects of democratization are likely different than the effects of having stable democratic institutions, and while wars can induce dramatic costs to productivity, they also increase the chance of large bounce-backs at a later stage. A third reason is that it can be hard to separate causal and contextual factors, and both can vary over time. We explore different approaches to growth forecasting and compare flexible dynamic models to direct long-horizon forecasts and statistical time-series models. Our preferred model is a dynamic skewed Student-t distributed model that parameterize the variance using democracy and the kurtosis and degrees of freedom using armed conflict. Democracy reduces growth variance, while wars dramatically increase the risk of large negative growth. These findings are in line with existing research, but this is the first time both ideas are used in the same forecasting model. We argue for using such a model in long-term scenario forecasts where democracy and armed conflict projections are given exogenously.

• Research themes: Modeling social phenomena, causal inference, forecasting and data quality.
• Read about the speaker

The DataScience@BI seminar is organised by the Department of Data Science and Analytics, BI Norwegian Business School.

Visit our website to read more about the Department of Data Science and Analytics.