Science, Environment and Development Group

Documents

Reverse Chronological | By Author

Citation:

Agrawal, Arun and Ashwini Chhatre. 2005 Interpreting long tails: Qualitative analysis of regression residuals. Political Methodology (Working Paper Series of the IPSA Committee on Concepts and Methods), No. 6.

Abstract:

Much recent writing in Political Science has elaborated the potential contributions of case studies, QCA (qualitative comparative analysis), and structured-focused comparisons. Proponents of quantitative techniques, on the other hand, maintain that testing and validation of theoretical propositions is best accomplished by statistical analysis using appropriate safeguards. This paper contributes to the fertile debates in Political Science on the uses and relative merits of qualitative and quantitative approaches for theory development and validation. Our approach hinges on careful analysis of residuals as a source of information. We suggest, in the spirit of a robust qualitative tradition, that focusing on residuals can illuminate aspects of theory that would be inaccessible without deep knowledge of the cases under scrutiny. Qualitative analysis of residuals can help check for possible omitted variables, measurement errors, non-linearities in posited relationships, and possible interaction effects. Qualitative examination of cases corresponding to extreme residuals can also assist in analyzing whether posited causal mechanisms apply to extreme values of independent variables, and thereby help specify the range within which theoretical predictions concerning specific variables are valid. Finally, an examination of residuals can help improve the understanding of conjunctural causation. Our paper outlines an integrated analytical approach in which the greatest strengths of qualitative and quantitative strategies can be combined to produce better causal accounts. We develop a systematic strategy to analyze residuals qualitatively that can be used by other scholars. Our paper illustrates the strategy by using a dataset on 205 cases of environmental governance from the Indian Himalayas.