Ch3 - Empirical Methods in Corporate Governance Research

Question 1

Endogeneity

Answer 1

• Arises when explanatory variables are correlated with the error term in a regression model
• This correlation leads to biased and inconsistent parameter estimates

Question 2

Sources of Endogeneity

Answer 2

• Omitted Variables
• Reverse Causation
• Measurement Errors
• Signaling Effects

Question 3

Reverse Causation

Answer 3

• direction of causality between variables is unclear.
• change in one variable causes a change in another or if the relationship is bidirectional (-> Egg Chicken)

Question 4

Measurement Errors

Answer 4

• Discrepancies between the true values of variables and their measured ones
• Especially for proxy variables

Question 5

Signaling Effects

Answer 5

These occur when actions taken by firms or individuals convey information to stakeholders (bad signal when firing CG officer)

Question 6

Addressing Endogeneity

Answer 6

• Advanced Econometric Techniques (two-stage least squares (2SLS))
• Smart Empirical Design (randomized trials)
• Natural Experiments (exogenous shocks)
• Improved Data and Proxy Variables (better data sources)

Question 7

Randomized trials (RTs)

Answer 7

• gold standard
• treatment and control

Question 8

Randomized Experiments vs Endogeneity

Answer 8

– reverse causation and simultaneity (only the input variable is changed),
– omitted variables (all controls are orthogonal to the shock),
– measurement errors in the covariates (all controls are independent of the
shock),
– signaling effects (the shock is exogenous).

Question 9

Natural Experiments

Answer 9

exploit exogenous shocks as treatments to assess causal relationships (e.g. regulatory changes) for comparing outcomes before and after the shock

Question 10

Difference-in-differences (DID) designs

Answer 10

compares changes in the variable of interest between treatment and control groups over time (natural experiments)

Question 11

Why are Natural Experiments and DID difficult for CG-Research?

Answer 11

It’s not possible to have to variants of markets running at the same time

Question 12

Regression discontinuity designs (RDD)

Answer 12

Leverage a threshold that determines treatment assignment. Entities above the threshold receive treatment, while those below do not.
-> RDD regressions utilize a dummy variable indicating treatment status and interactions between the forcing variable and the treatment dummy

Question 13

Sharp RDD

Answer 13

Treatment assignment is deterministic based on the forcing variable. Knowing the forcing variable’s value definitively determines treatment status

Question 14

Fuzzy RDD

Answer 14

The probability of treatment assignment changes discontinuously at the threshold. While the forcing variable influences treatment assignment, there is not a perfect one-to-one relationship