Choosing Spatial Units of Analysis (Module 19)

Monday, June 23

Hillel David Soifer (UC Berkeley)

Recent decades have seen major advances in the methodology of both qualitative and quantitative research, as scholarship has become much more purposive and precise both about the selection of cases and about the analysis of the data generated in the investigation of those cases. Yet the selection of cases and analysis of data is predicated on a prior, less studied, research design choice: the identification of the spatial unit of analysis. This issue, moreover, is no small detail: as geographers have shown, phenomena vary according to how spatial units are defined - this is what is known as the modifiable areal unit problem. As the size, shape, and location of the borders of a spatial unit change, the association we will find between social and political phenomena that we seek to study will change in fundamentally unpredictable ways. Indeed, even studies of individual-level attributes and behavior that seek to control for characteristics of context will be affected by how the researcher chooses to draw the boundaries within which context is measured. In short, the spatial units we choose affect the answers we get, and even the questions we ask. Yet even as the discipline of geography has been roiled by this issue for several decades, cognate social science disciplines have not grappled with it. The result is that we have little guidance for choosing an appropriate unit of analysis, and little sense of how choices about the units of analysis in existing studies might shape findings we take as robust instantiations of conventional wisdom. This module will explore these issues. We will first outline the depth of the problem in terms of the threats it poses to inference, and the range of scholarship to which it potentially applies. We will then turn to the role of theory in justifying the validity of the spatial units that we choose, and to ways to demonstrate the reliability of our findings through empirical analysis. Participants will have the opportunity to work through issues in their own research designs in addition to exploring the implications of this problem for their confidence in the findings of existing research in their areas of interest.

8:45am - 10:15am –The stakes of choosing spatial units

This session will explore the little-understood but fundamental ways in which the spatial units we choose affect the conclusions we draw from our analyses - this is the concern known to geographers as the ‘modifiable areal unit problem’. It will show that these effects are not only substantively important but deeply unpredictable. Since previous work on this question (in the discipline of geography) has focused on the quantitative setting, our discussion will begin there, but we will move to the ways in which these impacts unsettle our confidence in the conclusions of qualitative research as well.

Required readings:

• Openshaw, S. 1983. The Modifiable Areal Unit Problem’ Working Paper Series on Concepts and Techniques in Modern Geography. No.38: Institute of British Geographers.

• Tam Cho, Wendy K., and Neil Baer. 2011. “Environmental Determinants of Racial Attitudes Redux: The Critical Decisions Related to Operationalizing Context.” American Politics Research 39(2): 414–36.

Suggested readings:

• Colantone, Italo, and Piero Stanig. 2018. “Global Competition and Brexit.” American Political Science Review 112(2): 201–18. doi:10.1017/S0003055417000685. [We will use this paper as an exemplar across the first two modules, so it may be useful to be familiar with the argument and the basic structure of the analysis]

• Nemerever, Zoe, and Melissa Rogers. 2021. “Measuring the Rural Continuum in Political Science.” Political Analysis 29(3): 267–86.

• Arjona, Ana 2019. “Subnational Units, the Locus of Choice, and Concept Formation: Conceptualizing Civilian Behavior in Contexts of Civil War.” In Inside Countries: Subnational Research in Comparative Politics, Cambridge University Press, 214–42.

• Bisbee, James & Jan Zilinsky. 2022. “Geographic Boundaries and Local Economic Conditions Matter for Views of the Economy” Political Analysis 31(2): 288-294.

1:30pm - 3:00pm – a menu of responses to the MAUP

This module will explore a set of ways in which scholars can respond to the concerns raised by the modifiable areal unit problem. It will emphasize the primary role of conceptual and theoretical precision in attenuating the threats to inference that it poses, show some complementary empirical strategies that scholars can use, and highlight some intuitive-seeming but ineffective responses.

Required readings:

• Soifer, Hillel David. 2019. ‘Units of analysis in subnational research’ chapter 3 in Inside Countries: Subnational Research in Comparative Politics edited by Agustina Giraudy, Eduardo Moncada, and Richard Snyder (Cambridge University Press) pp.92-112

• Lee, Dong Wook, Melissa Z. Rogers, and Hillel David Soifer ‘‘The MAUP as a threat to inference in political science’ Political Analysis (online first)

Suggested readings:

• Cook, Scott J., and Nils B. Weidmann. 2022. “Race to the Bottom: Spatial Aggregation and Event Data.” International Interactions: 1–21.

• Wong, Cara, Jake Bowers, Tarah Williams, and Katherine Drake Simmons. 2012. “Bringing the Person Back In: Boundaries, Perceptions, and the Measurement of Racial Context.” The Journal of Politics 74(4): 1153–70.

• Wong, Cara et al. 2020. “Maps in People’s Heads: Assessing a New Measure of Context.” Political Science Research and Methods 8(1): 160–68.

• McCartan, Cory, Jacob R. Brown, and Kosuke Imai, 2024 ‘Measuring and modeling neighborhoods’ American Political Science Review 118(4): 1966-1985

• Córdova, Abby. 2022. “Living in Gang-Controlled Neighborhoods: Impacts on Electoral and Non-electoral Participation in El Salvador.” Latin American Research Review 54(1): 201–21.

3:30pm - 5:00pm – Collaborative research design

This module will apply the lessons from the previous two modules to an empirical setting - we will work together to design multi-method research projects intended to explain subnational variation in violence in contemporary Mexico. We will begin by working together to determine which research questions we seek to ask, and then consider how we should choose the appropriate spatial unit of analysis to operationalize our study. The instructor will provide a brief introduction to the case, and some data that can be used to design a research project, so there are no required readings for this exercise. After completing this exercise, we will turn to discussion of participants’ research projects and how they might approach them given the issues raised in preceding modules.