Does modelling corporate governance variables in bank intermediation function in a data envelope analysis (DEA) framework resolve the endogeneity to some extent?
Regardless of the context, endogeneity is solved by identifying the source. This means: identify omitted variables, better measuring to avoid measurement error, identify simultanity effects, and so on.There is no shortcut to ensure endogeneity does not happen in your model. Otherwise, it is not one of the biggest hurdle in statistical regression analysis.
Christopher is right. Solving endogeneity starts from its sources. There are three sources of endogeneity and the reason why they cause endogeneity:
Omitted variables.: it happens when we do not have sufficient control for factors that might have impact in the relationship. So that the error term might include those factors, which in turn might correlate to one of the independent variables.
Reverse causalilty (simultanity effect): in a research model y = ax + e, is it x explains y, or y can explain x? If y actually explains x, then e and x might also have some kind of relation. In other words, the mechanism of the relationship might work the other way around, other than we thought it would be.
Measurement error: A and B have a statistically signification relation. You use X to measure A and Y to measure B. However, X is an estimate with error which is noisy enough to mislead the relationship and render statistics biased. This means X cannot represent A in its relationship with B (which in turn proxied by Y). The Y = aX + e regression generates an error term that might be correlate to the proxy variable (X) itself.
So, regardless of research topics, it is crucial to understand the relationship you are studying, understand your proxy variables, and make a reasonable literature reviewing. Hope that helps.
- Badges
- Science topic
- Similar topics
- Engineering
- Modeling