Panel Data Econometrics: theory and applications

Bertinoro, 25 - 30 July 2022



Maria Elena Bontempi  
University of Bologna, Department of Economics, Piazza Scaravilli 2, 40126 Bologna  



Maria Elena Bontempi, University of Bologna
Roberto Golinelli, University of Bologna
Irene Mammi, University of Venezia Ca' Foscari


The maximum number of allowed participants in presence is 50.


Basic knowledge of econometrics to a level comparable with Introductory Econometrics SIDE course.
Examples of reference are represented by:
** Verbeek M. (2017), A guide to Modern Econometrics, 5th ed, Wiley, Ch. 1, 2, 3, 4, 5, 6  or
** Wooldridge J. M. (2020), Introductory Econometrics, a Modern Approach, 7th ed., Cengage, Ch. 1-6, 8-9, 15


Course outline, objectives and learning outcomes
Nowadays panel datasets permeate the empirical research on many topics, going from micro- and macro-economic towards behavioural and political economy. Panel datasets include alternative types of observations available at least in two dimensions: longitudinal, cross-sectional-time-series (CSTS), and multilevel where observations are nested at higher- and lower-levels. The aim of the course is to provide an advanced overview, both methodological and applied, of econometric models for panel data. During the course, to ease the comprehension and to introduce important topics, N will indicate cross-sections/higher-level units and T time-series/lower-level measurements. The first part of the course starts with N larger than T. After introducing the fixed and random effects models, with emphasis on their pros and cons, we will discuss about endogeneity of explanatory variables, intended as correlation with either individual heterogeneity (the heterogeneity bias) or idiosyncratic shocks (due to simultaneity, measurement errors, and dynamics), or both. The instrumental variables estimator, derived from the Generalized Method of Moments (GMM), is at the core of the lectures. The second part of the course relates to T larger than N. The main issues will be non-stationarity and cointegration, analysed and discussed in the light of parameters’ heterogeneity and cross correlated effects. At each step of the course, the methodologies will be accompanied by hands-on empirical applications with an econometric software. At the end of the course, students will get an overview of the literature on panel data and its recent developments. Participants will also be able to critically evaluate the empirical papers based on panel data, and to model and estimate their own issue of interest, according to the problems at hand: static versus dynamic approaches, heterogeneity in the parameters, exogeneity versus endogeneity of covariates, GMM, unit roots and long/short run relationships.


Reading list of the course:

The textbooks will be integrated with handouts and a list of papers provided during the lectures.


Schedule of the course:

0) Crash course: OLS, GLS, IV, heteroskedasticity, the introduction of dynamics and the role of unit roots into the empirical models’ specification.

(1) Static panels. Understanding the data structure and the role of unobserved heterogeneity. The decomposition of the total variability at two or more levels, and the heterogeneity bias. Methods: fixed, random, and correlated effects. Guided hands-on session on the topics.

(2) Dynamic panels. Understanding endogeneity. Instrumental variables (IV) in panel data models, and the problem of overfitting and weak instruments. Methods: alternative data transformations; first-differences and IV, GMM-DIF, -LEV, -SYS estimators, the principal components analysis applied to the set of instruments. Guided hands-on session on the topics.

(3) Heterogeneous panels. ARDL specification and Pesaran’s poolability. Methods: Mean-Group (MG) and Pooled-MG estimators; demeaning and cross-correlated effects. Guided hands-on session on the topics.

(4) Non-stationary panels. Integration and cointegration. Methods: first- and second-generation unit roots tests; Pedroni and Westerlund cointegration tests; PANIC and PANICCA. Guided hands-on session on the topics.


Hands-on Sessions

Theoretical lectures are associated with working sessions; during them you will receive the suggestions needed to use an econometric software and to learn how running your own empirical analysis. The empirical analyses will be presented through datasets and script files programmed to perform applied investigations on panel data. The material will be provided during the lectures.



Participants will use their laptops with the software STATA already installed on them. Any STATA version from 12 on is ok! Important remark: the course is not about STATA, but on Panel Data Econometrics. STATA is just a tool, like any other econometric package able to manage panel data. Moreover, in Bertinoro we do not have, and we cannot provide, STATA installation files and licenses. To actively interact during the lectures and the hands-on sessions, we strongly suggest that you have STATA pre-installed on your laptop before coming to Bertinoro.



The certificate of attendance  will be based on one final “take-home in Bertinoro” exam. The assignment will contain an empirical exercise, covering the issues faced during the lectures, to be performed by using the econometric software and your understanding of the taught theoretical issues. Working in groups is strongly encouraged to improve exchanges of views, but each of you must submit an individual take-home exam.


Venue and timetables

The Module will last one week and will be held in the University Residential Centre, Via Frangipane 6, 47032 Bertinoro (FC). Participants will be hosted in the Centre guest quarters, (as an exception, in case of reduced availability of rooms in the Centre, they will be accommodated in local hotels).

Lectures and tutorials will be in English, with the following schedule:

Monday to Friday: lectures 9:00-13:00, 15:00-17:00 individual hands-on sessions: 17:00-19:00.
Saturday: lectures 9:00-13:00.