Strengthening Causal Reasoning in Research on Inequality, Poverty and Mobility: New Methods for Answering Old Questions?

New Publication in ASA Section Inequality, Poverty & Mobility Newsletter

My colleague Ulrike Schwabe (@SchwabeUlrike) and I were invited to contribute to the symposium in the current newsletter of the ASA Section on Inequality, Poverty & Mobility (@asa_ipm): "(New) Qualitative and Quantitative Methodological Approaches to Capturing Inequality, Mobility, or Poverty". After hosting our workshop on "Causality in the Social Sciences II", our contribution revolves around causality and the added value for research on inequality. Many thanks to Jessica Ordemann (@OrdemannJessica) for this opportunity!

The whole newsletter with all the contributions is avaiable for download here.

Strengthening Causal Reasoning in Research on Inequality, Poverty and Mobility: New Methods for Answering Old Questions?

Lang, Sebastian | Schwabe, Ulrike

The question of why has become increasingly important and popular across different fields of research (c.f. Pearl & Mackenzie 2019). However, this is not really surprising as many research questions – for example, in the area of inequality, poverty and mobility – are definitely of a causal nature. Nevertheless, it seems as if economists are more successful in providing causal inference.

In 2019, the Nobel Prize in Economics was awarded to the research group of Abhijit Banerjee, Esther Duflo and Michael Kremer for using randomized controlled trials (RCT) to analyze worldwide poverty (e.g., Duflo et al. 2011). As this example shows, random assignment to treatment and non-treatment conditions in experimental studies is definitely one way to detect underlying causal mechanisms. For that reason, experiments (either in laboratory or field) are regarded as the “gold standard” for causal explanations. Due to constraints like research ethics issues or limited financial resources (c.f. Carlson et al., 2019 as example of a very expensive field experiment), experimental designs, however, cannot always be implemented.

Although research design in observational studies is not ex ante perfect to handle (self-)selection into treatment, various ex-post estimation strategies are able to ensure causal interpretation of empirical findings ‑ at least to a certain extent. Depending on the kind of data available, there are numerous different methods that can be applied to assure causal inference: e.g. matching and instrumental variable approaches, models for selection correction, difference-in-differences estimation, regression discontinuity, or panel regression models. For us, the phrase “causal reasoning”, however, does imply much more than only applying these estimation tools: a comprehensive overall research design including analytical model building. For purposes of causal argumentation, graphical causal models, e.g. directed acyclical graphs (DAGs), are especially helpful (e.g. Elwert 2013). Accordingly, we argue that sociological inequality research benefits from more extensive use of causal-analytical reasoning (Gangl 2010, Morgan & Winship 2014).

Let us provide two examples from research on inequality to strengthen our position. First, the so-called Coleman report on “Equality of Educational Opportunity” was already published in the 1960s (Coleman et al. 1966). Since then, an almost unmanageable amount of empirical literature on the relevance of context for educational opportunities has been circulated worldwide. However, the core question on how and why class and neighborhood composition determines individual learning outcomes in school is not fully answered yet. Amongst others, one core challenge are processes of (self-)selection that account for different context conditions and thus learning environments. Besides experimental studies (Chetty et al. 2016), applying estimation strategies that ensure causal inference is one way to explicitly address the “selection-problem” for investigating this old question (Altonji & Mansfield 2018, Legewie 2012). At the core, social problems today are often of the same or similar nature as in the past, but the awareness for causal reasoning has risen considerably among inequality researchers.

Second, educational reforms provide excellent opportunities from a causal perspective, as they provide quasi-experimental or even indeed experimental settings. Furthermore, for evaluating intended and especially unintended consequences of single reforms, causal reasoning is required. Many of such political restructurings have been taken place in the German higher education system within the last two decades. The German Initiative of Excellence, for example, was introduced with the goal to foster universities’ research output. However, until now, we do not know whether increased vertical stratification also affects graduates’ labor market outcomes in the sense of a wage premium. And, if such an “excellence premium” exists, is it used differently by socially advantaged and disadvantaged groups? Hence, by applying causal reasoning more frequently, inequality researchers can make a substantial scientific as well as social contribution by giving evidence-based policy advice.

Closing with a personal note, we want to stress that both perspectives are necessary for a comprehensive understanding of social phenomena: (i) the more explorative, descriptive perspective, which uncovers dimensions of inequality in existing social structures; and (ii) the causal-analytical perspective, which studies the concrete mechanisms behind persistent patterns of inequality observed. Besides describing social action, providing causal explanations is a central concern in sociology, going back to Weber’s definition of the subject (Weber 2019). And, finally, for a broader understanding of causality, other disciplines like philosophy of sciences make essential contributions. Thus, causality itself emerges to be an important and interdisciplinary paradigm.

References:

Altonji, J. G., & Mansfield, R. K. (2018). Estimating Group Effects Using Averages of Observables to Control for Sorting on Unobservables: School and Neighborhood Effects. American Economic Review, 108, 2902-46.

Carlson, D. E., Elwert, F., Hillman, N., Schmidt, A., & Wolfe, B. L. (2019). The effects of financial aid grant offers on postsecondary educational outcomes: New experimental evidence from the Fund for Wisconsin Scholars (No. w26419). National Bureau of Economic Research.

Chetty, R., Hendren, N., & Katz, L. F. (2016). The Effects of Exposure to Better Neighborhoods on Children: New Evidence from the Moving to Opportunity Experiment. American Economic Review, 106, 855-902.

Coleman, J. S., Campbell, E. Q., Hobson, C. J., McPartland, J., Mood, A. M., Weinfeld, F. D.,& York, R. L. (1966). Equality of Educational Opportunity. Washington, D.C.: Government Printing Office.

Duflo, E., Dupas, P.,& Kremer, M. (2011). Peer Effects, Teacher Incentives, and the Impact of Tracking: Evidence from a Randomized Evaluation in Kenya. American Economic Review, 101, 1739–1774.

Elwert, F. (2013): Graphical causal models. In Morgan, S. L. (Ed.). Handbook of causal analysis for social research. Dodrecht: Springer, pp. 245-273.

Gangl, M. (2010): Causal Inference in Sociological Research. Annual Review of Sociology, 36, 21-47.

Legewie, J. (2012). Die Schätzung von kausalen Effekten: Überlegungen zu Methoden der Kausalanalyse anhand von Kontexteffekten in der Schule. Kölner Zeitschrift für Soziologie und Sozialpsychologie, 64, 123-153.

Morgan, S.L. & Winship, C. (2014). Counterfactuals and causal inference. Methods and principles for social research. Cambridge: Cambridge University Press.

Pearl, J. & Mackenzie, D. (2019): The book of why: the new science of cause and effect. Basic books.

Weber, M. (2019): Economy and Society: A New Translation. Harvard University Press.