Time Series Approaches to International Affairs

by

Jon C. W. Pevehouse, Sujeong Shim

• LAST REVIEWED: 24 July 2018
• LAST MODIFIED: 24 July 2018
• DOI: 10.1093/obo/9780199743292-0237

Introduction

Time series analysis is used to analyze data that are organized chronologically. Much of our data in international relations have a time series element: events data of international interactions, data on economic flows between states, historical data on international conflict and cooperation between states, as well as newer forms of social media data. As practitioners of quantitative methods are well aware: to ignore the time-series element in data is to invite bad statistical inferences. Yet, treating time series properties as merely a nuisance misses an important modeling opportunity, namely to substantively interpret the nature of the time series element and use that to assist in the diagnosis of the data-generating process. This article samples important works using different techniques in the family of time series methods. We begin with univariate and diagnostic approaches commonly referred to as Box-Jenkins modeling. We continue with univariate intervention models before moving to multivariate vector autoregressive and error-correction models. While the later types of models remain common in international relations, the former are now less commonly used. The penultimate section deals with a host of time-series regression models. Many of these choose to treat the time series element of the data as a nuisance to be controlled for rather than explicitly modeled. A large increase in the types of approaches to time series regression (especially combined with cross-sectional data) has been registered, and here we provide a number of international relations examples. Our final section examines newer developments in time series and some of their applications to international relations questions. A final note: while most of the entries are applications of time series techniques applied to questions of international relations, a few important papers containing key methodological innovations are included as well.

General Overviews

Overviews of time series modeling in the social sciences are numerous. Box-Steffensmeier, et al. 2014 is a recent example heavy on political science and international relations examples. Pickup 2014 is a time series guide for a non-econometrics audience. Enders 2014 is a mid-level economics treatment of the approach. Hamilton 1994 is an advanced economics treatment. McCleary, et al. 2017 is an updated version of the classic McCleary and Hay 1980 that constitutes an early touchstone text for univariate models in international relations. Brandt and Williams 2007 is a shorter treatment focusing on multiple equation models with many political science and international relations examples.

• Box, George E. P., Gwilym M. Jenkins, Gregory C. Reinsel, and Ljung M. Greta. Time Series Analysis: Forecasting and Control. 5th ed. Hoboken, NJ: John Wiley, 2015.

  Book discussing univariate and multivariate time series models written by the originators of several of the key models in the field. Many examples center on engineering control applications rather than the social sciences. Discussion of autoregressive integrated moving-average models, tests of non-stationarity, vector autoregression, and forecasting.

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

  DOI: 10.1017/CBO9781139025287

  General introduction to time series methods in social sciences; includes many international relations and other political science examples. Covers difference equations, autoregressive integrated moving-average models, tests of non-stationarity, vector autoregression, time series regression, error-correction models, and other recent advances in time series methodology. Online materials allow users to work through examples.

• Brandt, Patrick T., and John T. Williams. Multiple Time Series Models. Thousand Oaks, CA: SAGE, 2007.

  DOI: 10.4135/9781412985215

  An introduction to multivariate time series models including simultaneous equations, autoregressive integrated moving-average models, vector autoregression, and error-correction models. Examples related to international relations and political economy.

• Enders, Walter. Applied Econometric Time Series. 4th ed. Hoboken, NJ: Wiley, 2014.

  A recent econometric book on time series analysis. Excellent discussions of difference equations as well as multivariate time series models. Includes many examples from the field of economics.

• Hamilton, James D. Time Series Analysis. Princeton, NJ: Princeton University Press, 1994.

  The canonical econometric book on time series analysis. Detailed discussions of many major time series estimation procedures, including mathematical derivations.

• McCleary, Richard, and Richard A. Hay Jr. Applied Time Series Analysis for the Social Sciences. Beverly Hills, CA: SAGE, 1980.

  An early work on the diagnosis of temporal dynamics in social science time series. Extensive discussion of autoregressive integrated moving-average models and intervention models. Many general social science examples, including some from political science and public policy.

• McCleary, Richard, David McDowall, and Bradley Bartos. Design and Analysis of Time Series Experiments. New York: Oxford University Press, 2017.

  DOI: 10.1093/oso/9780190661557.001.0001

  A significant recent update of McCleary and Hay 1980, focusing on model-building as well as threats to statistical inferences. Examples are taken from a wide variety of social science disciplines.

• Pickup, Mark. Introduction to Time-Series Analysis. Beverly Hills, CA: SAGE, 2014.

  Basic and practical introduction to time series models. Demonstrates the use and the assumptions of commonly used models, including distributed lag, autoregressive distributed lag, moving average, differenced data, autoregressive conditional heteroskedastic, autoregressive integrated moving-average, and error-correction models.