Empirical political scientists agree that good science should generate testable predictions. However, in practice there is some confusion over what the term ‘prediction’ means. One school, which we term ‘pragmatic prediction’, argues that prediction should refer to unconditional probabilistic forecasts of future events. Recent data driven advances in political forecasting accuracy have encouraged some scholars to argue that success at pragmatic prediction ought to be a new gold standard for political science work. In this paper, however, while we recognize the value of pragmatic prediction, we advance the rival concept of scientific prediction, which is conditional and explanatory and refers to the logical implications of a theoretical model. We argue for the value of political science work which is scientifically even if not pragmatically predictive. We contend that success at pragmatic prediction is neither necessary nor sufficient for good political science work.

Keith Dowding arrived at the ANU as Professor of Political Science in July 2007 from the London School of Economics. Prior to the LSE, he held teaching appointments at Brunel University in West London and at the University of Oxford. Professor Dowding is the Chair of a network of scholars called 'Selection and De-selection of Political Elites' (SEDEPE) who are interested in the career paths of political elites. He edited the Journal of Theoretical Politics from 1996 to 2014, and since 2014 has been Associate editor of Research and Politics.

Charles Miller received his PhD in political science from Duke University in 2013. His work centres on military organizations and public opinion and foreign policy. His work has been published in the Journal of Peace Research, Rationality and Society, the International Journal of Public Opinion Research and International Relations of the Asia-Pacific.