EPISODIC PREDICTABILITY IN MODELS WITH PERSISTENT VARIABLES AND ENDOGENEITY: DETECTION AND ESTIMATION

Lead Research Organisation: University of Southampton
Department Name: School of Social Sciences

Abstract

Abstracts are not currently available in GtR for all funded research. This is normally because the abstract was not required at the time of proposal submission, but may be because it included sensitive information such as personal details.

Publications

10 25 50
publication icon
Pitarakis J (2014) A joint test for structural stability and a unit root in autoregressions in Computational Statistics & Data Analysis

publication icon
Pitarakis J (2017) A Simple Approach for Diagnosing Instabilities in Predictive Regressions in Oxford Bulletin of Economics and Statistics

publication icon
Banerjee A (2014) Functional cointegration: definition and nonparametric estimation in Studies in Nonlinear Dynamics & Econometrics

publication icon
Gonzalo J (2013) Handbook of Research Methods and Applications in Empirical Macroeconomics

publication icon
Gonzalo J (2017) Inferring the Predictability Induced by a Persistent Regressor in a Predictive Threshold Model in Journal of Business & Economic Statistics

publication icon
Pitarakis J (2012) Jointly testing linearity and nonstationarity within threshold autoregressions in Economics Letters

publication icon
Gonzalo J (2012) Regime-Specific Predictability in Predictive Regressions in Journal of Business & Economic Statistics

 
Description This grant allowed me to develop a series of statistical methodologies designed to empirically uncover behavioural interactions and relationships between economic and financial variables (e.g. stock returns and valuation ratios, economic growth and interest rates) when these interactions can occasionally become inactive during particular states of the economy (e.g. recessions versus expansions). My research has established that ignoring such regime specific behaviour can lead to very misleading conclusions on the effects of a particular economic variable (e.g. interest rates) on the future values of another variable of interest (e.g. economic growth, stock returns).

Through the numerous tests I developed I was able to establish for instance that valuation ratios such as dividend yields can be very powerful predictors of future stock returns but only during periods of recessions while during expansions or normal times the same predictors may lose their predictive powers. The search for useful predictors of future economic and financial activity clearly needs to take into account the possibility that predictability may be kicking in only occasionally.
Exploitation Route My research has started gathering citations (see URL) from other academics who picked up on my idea of regime specific predictability within a particular statistical environment commonly known as predictive regressions and that allows for a very rich variety of economic variables to be considered as predictors.

The research agenda I initiated through this grant nearly ten years ago has generated a vast literature with numerous papers published by econometricians making use and expanding the methods and modelling framework I introduced. This agenda was one of the first initiatives to introduce the notion of regime specific predictability in the context of predictive regression models. Considering this type of predictability has now become the norm in applied work with numerous initiatives with titles such as "pockets of predictability", "episodic predictability", "time varying predictability" having emerged in recent years.
Sectors Financial Services, and Management Consultancy,Government, Democracy and Justice
URL https://sites.google.com/site/jypjypsoton/episodic-predictability
 
Title Detecting the presence of regimes in time series data 
Description The agenda explored in this project led to numerous novel applications by practitioners seeking to investigate the presence of dynamic instabilities and regimes in economic and financial data. The title of this research agenda ("Episodic/regime specific predictability) has in fact become a commonly used terminology in the time series literature that this project initiated. Numerous "spinoffs" such as "pockets of predictability", "time varying predictability" etc., followed thereafter. The statistical tests developed in Gonzalo and Pitarakis (2012, 2017) for instance have been taken up and used in a variety of applications by other authors. As of today, the outputs from this agenda collectively generated about 100 citations. The methods mentioned above allow practitioners to test whether past values of one or more time series have predictive power for the future values of another series in a way that can accommodate the possibility that this predictive power may be episodically inactive (e.g., a variable may have predictive power for another one only during recessionary periods). Such episodes of inactivity considerably weaken the ability of traditional tests to detect predictability. The methods developed in this agenda offer a novel environment that can explicitly account for such periods of inactivity 
Type Of Material Improvements to research infrastructure 
Year Produced 2012 
Provided To Others? Yes  
Impact Spinoff research agendas on related themes. Numerous publications using or expanding the methods developed in this grant (e.g., Gonzalo and Pitarakis (2012) in particular). 
URL https://sites.google.com/site/jypjypsoton/episodic-predictability
 
Title Novel estimation method for uncovering predictability whose strength changes over time 
Description This research has developed a theoretical toolkit that allows researchers to explicitly test the predictive power of a variable when the strength of predictability may be changing over time. An accompanying and publicly available computer code is provided to researchers interested in using the methodology. 
Type Of Material Data analysis technique 
Provided To Others? No  
Impact The research developed in this project has started being regularly cited by other researchers who use the toolkit we have developed. 
URL https://sites.google.com/site/jypjypsoton/episodic-predictability
 
Description Collaboration with Universidad Carlos III de Madrid 
Organisation Charles III University of Madrid
Country Spain 
Sector Academic/University 
PI Contribution This ESRC funded research has led to joint research outputs with Professor Jesus Gonzalo based at the Universidad Carlos III de Madrid (UC3M). The funding allowed me to visit the Department of Economics at UC3M and to also host my Spanish collaborator in the UK. This joint research has led to the development of new and flexible statistical tools for detecting predictability in economic time series. The key novelty of our work comes from its robustness to situations where predictability may only be kicking in occasionally (e.g. during economic expansions or recessions).
Collaborator Contribution My partner has collaborated to the development of our statistical toolkit.
Impact Gonzalo, J. and Pitarakis, J. "Inferring the Predictability Induced by a Persistent Regressor in a Predictive Threshold Model," Journal of Business and Economic Statistics, Forthcoming (2016) Gonzalo, J. and Pitarakis, J. "Regime Specific Predictability in Predictive Regressions," Journal of Business and Economic Statistics, Vol 30, Issue 2, 2012 (pp. 229-241) Pitarakis, J. "Jointly Testing Linearity and Nonstationarity within Threshold Regressions", Economics Letters, Vol. 117, 2, 2012 (pp. 411-413) Pitarakis, J. "A Joint Test for Structural Stability and a Unit Root in Autoregressions," Computational Statistics and Data Analysis, Vol. 76, 2014 (pp. 577-587) Gonzalo, J. and Pitarakis, J. "Estimation and Inference in Threshold Type Regime Switching Models" in. Handbook of Research Methods and Applications in Empirical Macroeconomics, N. Hashimzadeh and M. A. Thornton, eds. Edward Elgar, 2013.
Start Year 2010