Deck 8: Using Fixed Effects to Fight Endogeneity in Panel Data and Difference-In-Difference Models

Question 1

A pooled model will produce biased coefficient estimates when factors in the error term are not correlated with the independent variable.

Answer 1

False

Question 2

The fixed effect ( $\alpha$ ) is the part of the unobserved error and it is different for every observation for unit i.

Answer 2

False

Question 3

The value of the $\alpha$ _i parameter can potentially take on a different value for each unit i.

Answer 3

True

Question 4

When dealing with fixed effects, a de-meaned model approach is superior to the LSDV approach because the de-meaned model gives us more accurate coefficient estimates.

Question 5

Having a model that controls for fixed effects even when the fixed effects are 0 ( $\alpha$ _i=0 for all units) will not cause bias.

Question 6

Which of the following is true about fixed effects models?

A) We cannot use fixed effects models when we suspect there are fixed effects but no such effect exists ( $\alpha$ _i=0 for all units)
B) Fixed effects models cannot generate a coefficient on a variable that is constant within all units.
C) Fixed effects will generate a coefficient on a variable as long as that variable differs across units.
D) A fixed effects model is not possible when there are a large number of units in the model.

Question 7

Pooled OLS models are biased when:

A) When the fixed effects are correlated with the independent variable.
B) When the fixed effects are correlated with the dependent variable.
C) When the fixed effects model is used when in reality there are no fixed effects in the data.
D) When we don't know the direction of the fixed effects.

Question 8

In order to run a two-way fixed effect model, we can use:

A) A de-meaned approach
B) A LSDV approach
C) Both a and b
D) None of the above.

Question 9

Given Y_it = B₀ + B₁Treated_i + B2After_t + B3(Treated_i*After_t) + e_it, the expected value of the dependent variable for those who got the treatment before they received it is:

A) B0+B1+B2+B3
B) B0+B1+B3
C) B0+B1
D) B0+B1+B2

Question 10

Given Y_it = B₀ + B₁Treated_i + B2After_t + B3(Treated_i*After_t) + e_it, the expected value of the dependent variable for those who in the control group after the treatment was administered is:

A) B0+B1+B2+B3
B) B0+B1+B3
C) B0+B1
D) B0+B2

Question 11

Which of the following is a reason why fixed effect models can't estimate coefficients on variables that do not vary within unit?

A) Because no fixed effects exists in such a case.
B) The de-meaned variables will not vary.
C) We can estimate such a variable with fixed effects.
D) The errors will be homoscedastic.

Question 12

In order to employ the LSDV approach in a one-way fixed effects model, we have to:

A) Create a dummy variable for every unit and include it in the model.
B) Create a dummy variable for every unit and include all but one in the model.
C) Subtract the unit specific averages from both the dependent and independent variables.
D) Create dummy variables for every unit and subtract the unit specific average from the dependent variable.

Question 13

The benefit of the de-meaned approach over the LSDV approach is:

A) It gives more accurate coefficient estimates.
B) The coefficient estimates have a lower standard error.
C) Easier to implement if there are a lot of different units.
D) We will be able to estimate coefficients on variables that do not vary within unit.

Question 14

Which of the following factors could lead us to believe that a fixed effect model is necessary to control for endogeneity when examining the relationship between corruption (dependent) and regime type (independent) across countries over time?

A) Religion
B) Culture
C) External threat
D) All of the above.

Question 15

Which of the following is most likely not a factor that would cause endogeneity (fixed effect is zero) when examining the relationship of bilateral trade (dependent) and trade alliances (independent).

A) Shared language
B) Distance between trading partners
C) Regime type
D) Inflation rate

Question 16

List and provide a brief explanation of three potential fixed effects that could bias a model in the analysis of the relationship between attendance at a private school and test scores.

Question 17

Describe the issues involved with using a pooled model and given an example of why a pooled model could lead to biased results.

Question 18

Explain why a fixed effects model is not necessary in a case where fixed effects exist but the fixed effects are not correlated with the independent variable.

Question 19

Write down and explain the basic difference in difference equation. Point out which of the coefficients we are most interested in.

Question 20

Explain why coefficients on variables that do not vary within the unit cannot be estimated in a fixed effects model.