Applied Panel Data Analysis

Lecturer: Oliver Lipps and Ursina Kuhn

Modality: Presence

Week 2: 21 - 25 August 2023

Workshop contents and objectives

Does a constant poverty rate of 10% mean no change? While cross-sectional data would suggest so, panel data might show that while half of the poor remain poor, the other half exits from poverty and half of the newly poor come from the former non-poor. In addition to such descriptive advantages, panel models need fewer assumptions for causal claims compared to cross-sectional models since – by relying on within-variation – they suffer less from selection effects. To understand the within perspective of panel data as opposed to the between perspective of cross-sectional data is the core of this workshop.

Panel data are repeated measures of the same individuals, the same companies or other units. This course introduces different strategies for panel analysis by providing the skills for choosing the appropriate models, data preparation and result interpretation. We focus on understanding the mechanism behind the different approaches. Participants learn to assess the advantages and limitations of different panel models, given their research questions and available data (sample size, number of waves, amount of within and between variance).

Panel regression models include pooled OLS, change score, fixed effects, random effects, difference-in-difference, and first difference models. In applications, difficulties for data analysis arise due to non-response and attrition or changed measurement tools (e.g., change of survey modes). We use real data and discuss the implications of these issues for data analysis. Examples are mainly from the Swiss Household Panel (SHP), which contains data over more than 20 years, covering topics from different social science disciplines and including different kinds of data files (individual, household, one-time collected information, events). We work in addition with two and three-wave panels, as well as small-N panel data.

In the morning sessions, we present the theory and applications using panel data. In the afternoon sessions, participants will work on practical exercises either with data prepared by the instructors from the SHP or with panel data of their own choice. Other types of panel data with different observational units (e.g., schools, companies) are welcome. As two are two instructors, there is time available to discuss and solve individual problems.

Prerequisites

We will use the Stata software and assume some knowledge of Stata with cross-sectional data. For students not sufficiently proficient in Stata we offer a two-day preparatory online course shortly before the workshop introducing Stata. We will also provide syntax for R users for those who wish to work with R, but we will only discuss syntax in Stata. 

Recommended Reading

Andress, H. J., Golsch, K., & Schmidt, A. W. (2013). Applied panel data analysis for economic and social surveys. Springer Science & Business Media.

Angrist, J. and J. Pischke (2009). Mostly harmless econometrics: An Empiricist's Companion. Princeton University Press.

Bell, A. and K. Jones (2015). Explaining Fixed Effects: Random Effects Modeling of Time-Series Cross-Sectional and Panel Data. Political Science Research and Methods 3(1): 133-153. doi:10.1017/psrm.2014.7

Kohler, U. and F. Kreuter (2012). Data Analysis Using Stata, 3rd ed. College Station, Texas: Stata Press.

Morgan, S. and C. Winship (2015). Counterfactuals and Causal Inference. Methods and Principles for Social Research. 2nd ed. Cambridge University Press.

Vaisey, S. and A. Miles (2017). What You Can—and Can’t—Do With Three-Wave Panel Data. Sociological Methods & Research 46(1), 44–67.