Thelen and Mahoney 2015: Comparative-historical analysis in contemporary political science

1min assessment

 * Writing style: repetitive, rambling
 * Argument: Comparative-historical analysis (CHA) is useful because it can identify difficult causal relationships (interactions, specific scope conditions, macro/structural relationships), by way of in-depth case analysis, including process tracing. Therefore, CHA complements statistical large-N studies well. (Warning: I wrote this after reading 2 of 3 sections. I couldn't stand it any longer.)
 * No time? I wouldn't even read the conclusion (it's not clear, not summing up well). Read nothing or the summary below. :)

intro
CHA has "stood the test of time" because the advent of other methodologies (rational choice, quantitative methods, experimental, big data analysis) have not replaced it. So there must be sth valuable to it. Therefore, the article sets out to analytically define three aspects of CHA ("macroconfigurational", case-based, time-sensitive), explain the merits of each, and how CHA can be complemented by other methods.

(No definition or description given of what exactly CHA is.)

macro...
CHA focuses on and starts at macro level, but often also zooms to micro level, e.g. CHA insists on the importance of structural factors / the macro level in "shaping the interests of individual agents"
 * when deriving hypotheses for micro level (e.g. when process tracing)
 * locating causality at the micro level (e.g. theory of "critical junctures", which explains path-dependency of institutions by attributing institutional change to short periods during which individual agency (thus micro-level) has long-term impact, see Critical Junctures reading)

CHA denies that all processes have to be disaggregated into "individual-level choices and behaviors", claiming instead that this would "render much macro reserach infeasible or impossible".

.

...configurational
CHA assumes interaction effects to be "common", thus does not abstract causal effects from context

The paradigm of Causal Inference and with it the emphasis on experimental methods has led to a "dramatic narrowing of the type of studies that scholars are likely to undertake", because researchers are encouraged to leave aside those questions which are "empirically intractable". But many Polisci variables (e.g. power) are difficult to manipulate (while keeping external validity), so much research focuses on variables of limited explanatory value (e.g. information, which can easily be manipulated in the lab). Examples: .
 * The study of economic development is reduced to assessing the impact of randomly applied development programs.
 * Survey analysis has become very popular but "citizen preferences are not necessarily the main driver" of many important outcomes.

complements
Statistical studies, identifying broad patterns of individual variables, can be complemented by CHA studies looking at scope conditions and interactions.

Experimental research is not a good complement: It tries to isolate effects of individual variables, which cannot be summed up to explain the macro outcomes which CHA works on. Experimental research also misses reciprocal causation. (Example of welfare state research: Esping-Anderson found causal "syndromes" (feedback effects), causing coherence of welfare state types.)

Case-based research
...

...

Temporally oriented research
...

...

Conclusion

 * CHA methodologically "adaptable" and "pragmatic"
 * CHA's best complement are large-N studies
 * experimental methods unduly narrows down research program
 * "most productive research" when scholars bring "different methods" to bear on "substantively big and important programs"

Unclear to me
"In an era when cross-national regression research is often denigrated" ?? (p11)