Threshold Regression Panel Models
From 27 octobre 2016 to 28 octobre 2016 | Orléans - Hôtel Dupanloup

Duration and teaching language

The training (in English) is delivered over two consecutive days.

Course Aim

The aim of this course is to introduce the panel threshold regression models. In econometrics, the threshold regression models specify that individual observations can be divided into classes based on the value of an observed variable. The slope parameters of the regression model are then supposed to be specific to each of these classes. In the panel data context, these models have particular appealing features. They allow to take into account the slope parameters heterogeneity, while explaining these heterogeneity by an economic threshold variable.  Hence, they can be viewed as a parametric alternative to the random coefficient models. Notice that all the models presented in the course are developed for non-dynamic panels with individual specific "fixed” effects. The training will also propose an application of these models based on Matlab codes.

Course leader

Christophe Hurlin, University of Orleans (personal website)

Christophe Hurlin is a specialist of financial econometrics and panel data econometrics. He is professor at the University of Orleans and vice-director of the research laboratory of economics (LEO, UMR CNRS 7322). He was previously associate professor at the University Paris Dauphine and he taught at HEC Geneva and HEC Lausanne. His research has been published in academic journals like the Journal of Financial Econometrics, the Review of Finance, the European Journal of Operational Research, the Journal of Banking and Finance or the Journal of Empirical Finance.


  • Introduction: panel data regression models and the slope parameters heterogeneity issue. We will present some heterogeneous panel regression models, including the random coefficient models (Hsiao, 2007)

  • Panel Threshold Regression (PTR) model (Hansen, 1999). In this section, we will present the specification and the computation of the marginal effects in this threshold regression model with two (at least) extreme regimes for the parameters. We will also discuss the estimation method (Non-Linear Least Squares) and the inference.

  • Panel Smooth Threshold Regression (PSTR) model. This model proposed González, Teräsvirta and van Dijk (2005) is based on a smooth transition function. This model can be viewed as model with an infinity of values (regimes) for the slopes parameters that depend on the value of a threshold variable. We will discuss the specification of this model, with one or more transition functions, and the computation of the marginal effects. We will also detail the estimation method (Non-Linear Least Squares) and the inference on the slope parameters.

  • Conclusion: we will briefly present various possible extensions of the PTR and the PSTR.

Statistical software

We will implement most of the methods discussed in lecture, using data sets covering a variety of areas in economics and finance. We will use Matlab as the main statistical software. The participants are invited to bring their personal laptop.

Lecture Notes  

Main lectures

González, A., Teräsvirta, T., and van Dijk, D. (2005), Panel smooth transition regression model, Working Paper Series in Economics and Finance, No. 604.

Hansen, B.E. (1999), Threshold effects in non-dynamic panels: estimation, testing, and inference, Journal of Econometrics, 93, pp 334–368.

Hsiao, C. (2003), Analysis of Panel Data, Cambridge University Press.

van Dijk, D., Teräsvirta, T., and Franses, P. H. (2002), Smooth Transition Autoregressive Models - a Survey of Recent Developments, Econometric Reviews, 21(1), pp 1-47.

Other lectures (empirical applications and surveys)

Candelon B., Colletaz G., and Hurlin C. (2013), "Network Effects and Infrastructure Productivity in Developing Countries", Oxford Bulletin of Economics and Statistics, 75(6), pp 887-913.

Franses, P. H. and D. van Dijk, (2000), Nonlinear Time Series Models in Empirical Finance. Cambridge University Press.

Fouquau, J., Hurlin, C. and Rabaud, I. (2008), ''The Feldstein-Horioka Puzzle: a Panel Smooth Transition Regression Approach'', Economic Modelling, 25(2), pp. 284-299.