Time Series Analysis for the Social Sciences

Lecturer: Chendi Wang

Modality: In-presence

Week 1: 10-14 August 2026

 

Workshop Contents and Objectives

Time series data lie at the core of political, economic, and social processes: public opinion shifts, government approval, macroeconomic indicators, media sentiment, policy changes, crises, and shocks are all phenomena that unfold over time. Yet most standard statistical tools assume independent observations and no temporal structure, assumptions that time-series data violate fundamentally. As a result, traditional regression approaches often produce misleading inferences, spurious findings, and incorrect conclusions, especially when series contain persistence, trends, unit roots, structural breaks, or long memory.

This course offers a systematic introduction to time series analysis tailored to social scientists. Using real political and economic applications, participants learn how to diagnose temporal dependence, model dynamic relationships, test for stationarity, work with integrated processes, estimate autoregressive distributed lag (ADL) and error-correction models (ECMs), and understand concepts such as cointegration, equation balance, fractional integration, and bounds testing.

The course draws on examples from political science, economics, and public opinion research to highlight why theory must align with the data-generating process, and why many published time-series analyses fail due to mis-specification. The course is hands-on and emphasizes both conceptual understanding and applied modeling skills. R and Stata (minimally) will be used throughout, with exercises designed to build reproducible workflows applicable to participants’ own research.

By the end of this course, participants will be able to:

  1. Understand the properties of time series data and the implications for inference
  2. Estimate and interpret ARIMA, ADL, ECM, and cointegration models
  3. Diagnose stationarity, unit roots, and fractional integration
  4. Apply equation balance tests and bounds approaches for inference
  5. Design and implement complete empirical strategies for time-series-based research questions
  6. Build reproducible analysis pipelines

Evaluate and compare dynamic models in applied social science research

 

Workshop design

Each day on the course will consist of two parts:

  • The morning lectures introducing key concepts, illustrated with political science and social-science applications, with live coding demonstrations.
  • The afternoon session will feature applied work involving guided exercises, replication tasks, and supervised project work, where participants can bring their own datasets and receive feedback on how to apply the methods.

During the course, participants are also welcome to work on their own projects. The instructor will happily assist with any individual project that requires skills and knowledge covered during the course.

All lecture slides, sample datasets, code scripts, and exercise solutions will be provided through an online repository. The course encourages constant interaction, group discussion, and hands-on engagement with data.

 

Detailed lecture plan (daily schedule)

Day 1 – Introduction and Foundations of Time Series Analysis

Morning:

  • Time series notation and terminology
  • Autocorrelation, autoregression, serial dependence
  • Stationarity and weak dependence
  • Trends, cycles, structural breaks
  • Integration and instability

Afternoon:

  • Hands-on exploration of time series data in R
  • Descriptive analysis using ACF/PACF
  • Checking for stationarity and structural breaks
  • Exercise: students identify a series from their own research and run diagnostic plots
Day 2 – Dynamic Models, Equation Balance, ADL & ECM

Morning:

  • Time-series regression assumptions
  • Dynamic regression models and lag structure
  • The ADL model
  • The relationship between ADL model and ECM
  • Equation balance
  • General-to-specific model building

Afternoon:

  • Estimating ADL models, lag selection, diagnostics
  • Computing LRMs and interpreting dynamic effects
  • Exercise: students estimate an ADL model using their own series, check equation balance, and produce preliminary diagnostics
Day 3 – Cointegration & Error Correction

Morning:

  • Cointegration   and equilibrium relationships
  • The Engle-Granger two-step approach
  • One-Step, GECM
  • Inference and Interpretation from GECM

Afternoon:

  • Step-by-step Engle–Granger implementation in software
  • Determining cointegrating vectors
  • Testing for cointegration in sample datasets
  • Exercise: students identify whether pairs of their variables cointegrate; interpret cointegrating equilibrium
Day 4 – Fractional Integration & Advanced ECM/GECM Inference

Morning:

  • Fractional integration and ARFIMA models
  • I(1) vs FI processes
  • Fractional cointegration
  • GECM inference pitfalls

Afternoon:

  • Estimating ARFIMA and interpreting memory
  • Applying FI ideas to pre-whitening
  • Fractional cointegration practical
  • Exercise: students compute ACFs, test for FI, and compare ARIMA vs ARFIMA fits with their own project data
Day 5 – Bounds Approaches & Full Time-Series Research Designs

Morning:

  • Pesaran–Shin–Smith bounds testing (ARDL bounds)
  • Long-run multiplier bounds
  • Interpreting bounds results for theory-driven social-science questions
  • Building full empirical pipelines: diagnostics, specification and interpretation

Afternoon:

  • Guided estimation of bounds models in R

Student mini-presentations: My time series design in 3 slides

Course feedback and Q&A.

 

Class materials

All materials will be provided online.

Prerequisites

Students who intend to enroll in this course are required to have basic knowledge in statistical analysis including linear regression models and hypothesis testing. Participants are expected to have basic computer and statistical analysis skills. Basic familiarity with R is necessary to participate in practical exercises and activities. 

Recommended readings or preliminary material

  • De Boef, S. and Keele, L., 2008. Taking time seriously. American Journal of Political Science, 52(1), pp.184-200.
  • Pickup, Mark. 2015. Introduction to Time Series Analysis. Thousand Oaks, California: Sage Publications, Inc.
  • Suzanna Linn, Lebo, Matthew, and Clayton Webb. 2026. A Practical Guide to Time Series Analysis. Cambridge University Press

Chendi Wang

Vrije Universiteit, Amsterdam

Chendi Wang is an Assistant Professor of Political Science at the Department of Political Science and Public Administration, Vrije Universiteit Amsterdam. He earned his Ph.D. in Political Science from the European University Institute in 2021. His research spans comparative politics, political behaviour, political economy, and political methodology, with current work on European politics, comparative political economy of crisis and macro-policy, party and electoral politics, and political mobilisation. Methodologically, his agenda emphasises Bayesian and non-parametric statistics, time-series analysis, measurement, machine learning and AI, and the integration of causal-inference and computational techniques. His work has appeared in outlets such as the British Journal of Political ScienceComparative European Politics, European Political Science ReviewEuropean Union PoliticsJournal of European Public Policy, and West European Politics, as well as in books published by Cambridge University Press. He is currently the PI of the funded research project NEST – Navigating the Storm: European Political Contestation in Geopolitical Transformation and co-PI of PoliBias: Cross-National Analysis of Political Bias in Language Models.

In his teaching, Chendi has developed and delivered undergraduate and postgraduate courses in comparative politics and research methods at VU, and has taught advanced methods workshops at the IPSA Summer School and the ECPR Methods School, where he received the Cora Maas Award for best course.

Read