INTRODUCTION TO SPATIAL PANEL DATA MODELS USING STATA

Many phenomena in the economics, medical and social fields, such as unemployment, crime rates or infectious diseases, tend to be spatially correlated. Spatial econometrics, in contrast to standard econometric modelling, exploits geo-referenced cross-sectional and/or panel data for dealing with spatial dependence and spatial heterogeneity. More specifically, spatial panel data sets contain repeated observations over time for a set of geo-referenced statistical units.

Our “Introduction to Spatial Panel Data analysis using Stata” course offers participants the opportunity to acquire the necessary theoretical and empirical toolset for modelling data which are correlated in time and space using both official and community written Stata spatial estimation commands. The opening session reviews Stata’s inbuilt sp command suite and illustrates how one prepares data for a spatial longitudinal analysis, before moving on to discuss different estimation techniques for both spatial fixed- and random-effects “static” models and for dynamic models with additive and/or interactive fixed-effects.

Throughout the course a series of empirical applications are used in order to highlight and discuss important issues such as model selection, average direct and indirect marginal effects, multiple spatial interactions and/or endogenous covariates, global stationarity, short-versus long-run marginal effects, and strong versus weak crosssectional dependence. Moreover, in common with TStat’s training philosophy, each individual session is composed of both a theoretical component (in which the techniques and underlying principles behind them are explained), and an applied segment, during which participants have the opportunity to implement the techniques using real data under the watchful eye of the course tutor. Throughout the course, theoretical sessions are reinforced by case study examples, in which the course tutor discusses and highlights potential pitfalls and the advantages of individual techniques. Particular attention is also given to both the interpretation and presentation of empirical results.

Upon completion of the course, it is expected that participants are able to identify and evaluate which specific spatial econometric methodology is more appropriate to both their dataset and the analysis in hand and subsequently apply the selected estimation techniques to their own data customizing the Stata do-file routines specifically developed for the course.

Ph.D. Students, researchers and professionals working in public and private institutions interested in acquiring the latest empirical techniques to be able to independently implement Spatial panel data estimation techniques in Stata.

Knowledge of the arguments covered in our “Introduction to Spatial Analysis using Stata”, “Linear Panel Data Models in Stata” and “Dynamic Panel Data Analysis” training courses is strongly suggested. Experience with Stata’s do-file programming is required.

SESSIONE I

Introduction

Spatial data analysis using Stata: overview of the sp suite
Space, spatial objects and spatial data

Preparing data for the spatial longitudinal analysis

Spatial and panel data declarations
Data with shapefile: Creating and merging a Stata-format shapefiles
Data without shapefile

SESSIONE II

Panel data models: first generation

The W (eights) matrix: types and normalization
Fixed- vs random- effects (static) models
Quasi Maximum Likelihood estimation
Hypothesis testing and model selection

SESSIONE III

First generation: further topics

Partial effects: direct, indirect and total effects
Fixed-effects Instrumental Variables estimation

(Selection) Internal instruments
Multiple spatial interactions and/or endogenous covariates

SESSIONE IV

Panel data models: second generation

Dynamic models
Estimation and testing

Global stationarity
Short- vs long-run marginal effects
Cross-sectional dependence (CD) and exponent of CD tests for Residuals

SESSIONE V

Panel data models: third generation

Dynamic models with weak and strong CD (Halleck Vega and Elhorst, 2016)

Quasi Maximum Likelihood estimation (Shi and Lee, 2017;Bai and Li, 2021)

Heterogeneous coefficients (Aquaro, Bailey and Pesaran, 2020)

We are currently putting the finishing touches to our 2024 training calendar. We therefore ask that you re-visit our website periodically or contact us at training@tstat.it should the dates for the course which you are interested in following not yet be published. You will then be contacted via email as soon as the dates are available.