Methods of Causal Inference with Panel Data

online course

1. Introduction to panel-data econometrics:
a. Ordinary least squares
b. Instrumental variables
c. Heteroskedasticity
d. Unit roots and stationarity
2. Static panels:
a. Clustering the data structure
b. Endogeneity, simultaneity, measurement error
c. Dealing with unobserved heterogeneity
d. Variance decompositions
e. Correlated random effects and slopes
f. Instrumental variables in panel data: testing and identification
g. Unbalanced and weakly balanced panels in the presence of sample selection bias
University of Ljubljana SCHOOL OF ECONOMICS AND BUSINESS
3. Dynamic panels:
a. Nickell’s bias
b. Overfitting versus weak instrumental variables
c. Arellano-Bond estimator and its extensions
d. First-differences and instrumental variables
e. Generalized method of moments estimator and instrumental variables
f. Principal components analysis and latent instruments
g. Multidimensional panels: estimating models with high-dimensional / multi-way
fixed effects
4. Heterogeneous and non-stationary panels:
a. Integration and cointegration
b. First- and second-generation unit root tests
c. Testing and correcting for cointegration through Pedroni and Westerlund tests
5. Difference-in-differences policy evaluation
a. Standard difference-in-differences model with q-parallel trends
b. Difference-in-differences estimator with staggered adoption
c. Difference-in-differences estimator with multiple-period heterogeneous
treatment
d. Fuzzy difference-in-differences and two-way fixed-effects extension
e. Event-style difference-in-differences
6. Synthetic control analysis:
a. Preliminaries
b. Model construction and estimation in the presence of parallel trend assumption
violation
c. Testing and specification issues
d. Estimating and fitting the synthetic control models
e. Sparsity and confidence intervals
f. Temporal and spatial placebo analyses
g. Prediction-based confidence intervals with synthetic controls
h. Non-parametric synthetic control analysis with kernel-based regularization of
bandwidth
i. Machine learning extensions of synthetic controls through ridge regression,
elastic nets and LASSO

Course leader

Rok SPRUK, University of Ljubljana, School of Economics and Business, Slovenia

Target group

Doctoral Winter School is an online programme with courses intended for PhD students, post-docs, academics and professionals from different areas and all around the world.

Course aim

In recent years, panel datasets dominate the empirical research paradigm with observations
at increasingly lower levels and research spanning many topics ranging from behavioural
economics to political economy. The aim of the course is to provide a theoretical,
methodological and applied overview of econometric models for panel data where
observations are available on at least two dimensions. The first part of the course relates to
static and dynamic panel data with emphasis on the endogeneity of covariates through the
correlation with individual heterogeneity and correlation with idiosyncratic shocks.
Instrumental variables estimators will be discussed in-depth through the generalized method
of moments. The second part of the course consists of problem evaluation techniques such
as difference-in-differences, synthetic controls and event studies to evaluate the impact of
policies on economic outcomes and behaviour. At the end of the course, students will be able
to critically evaluate the empirical literature based on panel data and policy evaluation
techniques, estimate their own issue of interest and write a scholarly paper on the topic of
interest.

Credits info

4 EC
Participants who need ECTS credits for their PhD studies, can obtain an official Transcript of records upon completion of all course obligations with passing final examination/assesment.

Fee info

EUR 600: For more information please see our website.