Projecting into the Future as Methodology

• In Event: Conflict Modeling, Prediction and Forecasting

Abstract

Political methodology has focused on descriptive inference that accurately portrays the world as it is (King, Keohane, Verba 1994), on interpretive approaches that explore the nature of meaning and our understandings of the world in which we live (Bevir and Blakely 2018), and on causal inference that gets at the causes of social outcomes (Brady and Collier 2004). These are powerful tools for understanding the facts of the world, the social meanings holding the world together, and the counterfactual possibilities that might be created.
Political scientists have often ignored projective inference (Goodman 1954) – imaginative projections into the future of new possibilities that employ the opportunities for change identified by causal inference and that start from current circumstances captured by descriptive and interpretive inferences. Through projection, humans contrive projects that put together diverse elements to solve a problem – a Constitution for a fledgling democratic republic in 1789; a Harlem Children’s zone combining education, health, and neighborhood services; or a Global Warming Solutions Act (AB 32) in California employing market mechanisms and regulatory standards to reduce Green House Gases. The problem of projective inference is evaluating the likely impact of these projects.
Not surprisingly, while social science methodology mostly ignores projection, public policy analysis gives it pride of place. After telling the analyst to “define the problem” and to “assemble evidence,” Gene Bardach’s canonical “Eightfold Path to More Effective Problem Solving” -- now in its sixth edition (Bardach and Patashnik 2019) – calls on the analyst in the crucial third through sixth steps to “construct the alternatives,” “select the criteria,” “project the outcomes,” and “confront the trade-offs.” Only then can the analyst “decide” and “tell the story.” It might be possible for those intent on just describing and understanding the world through descriptive, interpretive, and causal inference to ignore projective inference, but for those who want to change the world, it is an essential activity.
Forecasting and simulation methods provide ways to develop scenarios for the future based upon realistic assumptions that allow researchers to evaluate the likely results. For climate-change mitigation strategies, researchers can use climate simulation models to project into the future using varying assumptions about efforts to reduce CO2 emissions. These simulation methods, however, are often given short shrift in discussions of methodology, and they typically provide only limited assessments of the impacts of the project. Even more neglected are the ways that the possibilities for the social construction of reality (Berger and Luckman 1966) should enter into our projections of the future. How should we think about, for example, new institutions and understandings – carbon taxes, Kyoto Protocols, environmental justice movements -- that change the way we think about the environment and climate?
This paper defines the problem of projective inference by considering how it differs from simple extrapolation or from identifying a policy lever and assuming its impact in an experiment will be replicated in the larger society. It will discuss how projective inference must take into account the complexities of a modern society in which there will be unanticipated consequences, general equilibrium effects, multiple causes, learning effects, strategic adjustments, rational expectations, and many other possibilities that make the inference problem especially difficult.
Forecasting and simulation techniques used by quantitative methodologists to project into the future are the starting point and baseline because so much is known about their strengths and weaknesses (e.g., Saltelli, Tarantola, and Campolongo 2004; Oreskes, Stainforth, Smith 2010; Lucas 1976; Darnell and Evans 1990; Freedman 2009). These techniques demonstrate that evaluations must take into account uncertainties in data (as in standard statistical analysis), in causal inferences (as in meta-analyses), and in modeling (due to the approximations contained in the model of reality). The paper does not confine itself to quantitative methods, and it explores the utility of other projection methods used to understand the future. These include analogies (Hesse 1966, Rentetzi 2005), process tracing (George and Bennett 2004), history (1986), and other methods (Jaccard and Jacoby 2010). The paper considers the methodological problems associated with evaluating these qualitative methods.

Author

• Henry E. Brady, University of California, Berkeley