Research Method II Rules of Constructing Causal Theory : Causality vs. Correlation Jaechun Kim
CAUSALITY vs. Correlation Correlation is not causation! Some examples?? e.g., Incumbency advantage of Congressional election in… pp. 77-79 (yI,J – yN,J) realized causal effect of incumbency advantage… But we cannot rerun the history. We don’t have perfect experimental control… The fundamental problem of causal inference in social science…Holland (1986) p. 79
Assumptions required for estimating causal effects Unit Homogeneity rerun history (or experiment) in two different units that are homogenous! Weaker version of this is “Constant Effect Assumption!” p. 92-93 Constant Effect Assumption is the assumption underlying the method of comparative case studies p. 93 Conditional Independence – assumption that values of IV should be independent from values of DV… p. 94 Otherwise, Problem of Endogeneity… Examples?? p. 186, 187
Rules when you select your DV DV should be really dependent – cf. problem of endogeneity That is, your IV should be exogenous (independent), whereas your DV is endogenous (dependent)… Do not select observations based on DV! otherwise, “Selection Bias!” If you do, you research design is subject to selection bias… What is it?? p. 128 US investment Internal Violence If possible, vary your DV!
Additional advice of KKV Ceteris paribus, use concepts that can be operationalized ! Culture, intentions, motivations, norms are hard to be operationlized…
Rules for Constructing Causal Theories Construct Falsifiable Theories Build Theories that Are Internally Consistent (coherent definitions…) Select Dependent Variables Carefully (when selecting cases, avoid selection bias… make sure DV is dependent on your IV….) Maximize Concreteness (observable implications…) State Theories in as Encompassing Ways as Possible (generalizable theories…)