Hi, you are logged in as , if you are not , please click here

Time Series Analysis for Political and Social Data

More Info

Course Information

Time Series Analysis for Political and Social Data

This course provides an introduction to time series methods and their application to social science research. Many of our theoretical questions in the social sciences are implicitly temporal in nature – such as whether a change in public policy leads to a change in social behaviour, whether that relates to recycling, voting or offending, and the widespread availability of social and political longitudinal data make it possible to address them. There are specific issues associated with time series data – due to their temporal structure and dependence – that requires careful attention. This course will introduce participants to the time serial structure of social and political data, fundamental concepts in time series analysis, diagnostic tests for different time series processes (i.e. stationarity and unit root), and static and dynamic regression models (including “ARIMA”, autoregressive distributed lag and error-correction models) for social and political variables.


The course covers:

  • The structure of political and social time series.
  • Fundamental concepts in time series analysis.
  • Diagnostic tests for autocorrelation, moving average and stationary/integrated processes.
  • Univariate and static/dynamic regression models.


By the end of the course participants will:

  • Have developed an understanding of the theoretical structure of time series data, and be able to organise their own data in this format.
  • Be able to apply diagnostic tests for time series processes to their own data.
  • Be able to select the appropriate model for univariate and multivariate specifications, and estimate and interpret the short- and long-run effects of variables, lag distributions and rates of ‘error-correction’.


The target audience for this event is academics or government researchers from the social sciences (not including economics) with some background in quantitative methods in general, but no experience of time series analysis specifically. This may range from PhD students to more advanced researchers looking for an introduction to a new method.

If possible, please bring one or more social/political time series that are relevant to your research (e.g. crime rates, survey data on vote intentions or support for government cuts, social trust indicators).




Some knowledge of linear regression models is assumed but prior training in time series analysis is not required or expected. The course will make use of basic algebra. The computer workshop will use Stata. Some familiarity with Stata would be helpful but for those without preparatory materials will be provided ahead of the course and the teaching materials will provide a crash course during the session.


Preparatory Reading


Paul Kellstedt and Guy Whitten. (2013). The Fundamentals of Political Science Research. Cambridge University Press. Chapter 11.


Mark Pickup. (2014). Introduction to Time Series Analysis. Sage.


Janet M. Box-Steffensmeier, John R. Freeman, Jon C. W. Pevehouse and Matthew Perry Hitt. (2014). Time Series Analysis for the Social Sciences. Cambridge University Press.


Course Code


Course Dates

20th June 2017 – 21st June 2017

Course Leader

Professor Will Jennings
Course Description

Day 1:


Morning session:


-       What is a time series?

-       A practical introduction to the organisation of time series (and cross-sectional time series) data.

-       A crash course in using Stata.

-       The importance of preliminary descriptive analysis of time series data.

-       An introduction to the vocabulary (and algebra) of time series: white noise, autoregressive, integrated and moving average processes.


Afternoon session:


-       How to conduct diagnostic tests for stationary/integrated processes (Augmented Dickey-Fuller and KPSS tests), autocorrelation and partial autocorrelation.

-       A practical introduction to estimation of univariate ARIMA models using social and political data (based on the diagnostic tests above).

-       An extension of univariate models: intervention analysis as a method for estimating the effect of social and political events.

-       Reflecting on the material so far: how would you go about devising a research design using time series analysis in relation to a specific theoretical question?


Day 2:


Morning session:


-       An introduction to static regression models.

-       Using dynamic regression models to distinguish between short- and long-run effects: finite distributed lag (FDL) models and autoregressive distributed lag (ADL) models.

-       How and why to be aware of violations of time series regression assumptions (exogeneity).

-       How to test for serially correlated errors.


Afternoon session:


-       An introduction to generalized error correction models (ECMs).

-       Based on the material covered, plan and execute time series analysis on social or political data that you have chosen – going through the diagnostic, estimation and post-estimation process.

-       Feedback.

How would you rate your experience today?

How can we contact you?

What could we do better?

   Change Code