Essays on Regional Recessions, Spatial Interactions and Forecasting

Loading...
Thumbnail Image
Date
Authors
Shibaev, Sergei
Keyword
Econometrics , Bayesian Statistics , Business Cycles , Endogenous Clustering , Clustering , Bayesian , Regime-switching , Regional economic analysis , Spatial Econometrics , Spatial , Time Series Econometrics , Forecasting , Fractional cointegration , Fractional integration , Opinion poll data , Political Opinion , Vector Autoregressive Model , Cointegration , Spatial Interactions , Spatial Effects , Economic Policy , Elections , Markov-switching , Multivariate Econometrics , FCVAR , CVAR , ARIMA , ARFIMA , Spatial Autoregressive Models , Regime-switching models
Abstract
This thesis contains three essays spanning the fields of econometrics and empirical macroeconomics. The first essay develops an econometric procedure that enables applied researchers to quantify spatial interactions from panel data where variables exhibit recurrent abrupt shifts in behavior. In empirical macroeconomics, omitting spatial effects is restrictive in many contexts because the units of analysis are regions, the characteristics of which are rarely independent. The second essay employs this methodology to investigate how recessions propagated through small regional economies in the United States from 1990 to 2015. The empirical results identify regions that are potentially at risk of collective economic distress, which is useful for national and regional policy makers. The analysis shows the importance of the spatial (or geographical) dimension in explaining how regional shocks amplify in the economy. The third essay, co-authored with Morten Ø. Nielsen, investigates a unique data set of daily political opinion polls in the United Kingdom from 2010 to 2015. This work explores the forecasting capabilities of the recently developed fractionally cointegrated vector auto-regressive (FCVAR) model. The results show that the FCVAR model delivers superior forecast accuracy relative to a portfolio of existing alternatives. Furthermore, the forecasts generated by the FCVAR model leading into the UK 2015 general election provide a more informative assessment of the current state of public opinion than that suggested by opinion polls.
External DOI