7. Over And Under Specification

Question 1

Over specification

Answer 1

Irrelevant regression included in the model

Question 2

What happens to the OLS estimators if we over specify?

Answer 2

They will be unbiased but inefficient

Question 3

When will over specification lead to inefficient estimators?

Answer 3

When there is correlation between relevant and irrelevant variables

Question 4

What happens if there is over specification and the relevant and irrelevant variables are correlated?

Answer 4

• standard error increases- inefficient
• t ratios are smaller
• power is reduced so it’s more likely we will make type 2 errors

Question 5

Under specification

Answer 5

The omission of important regressors

Question 6

When are estimators biased?

Answer 6

During under specification if the included and omitted variables are correlated

Question 7

When is the OLS estimator positively biased

Answer 7

When the omitted variable is positive and the correlation between variables is positive.
Or
When the omitted variable is negative and the correlation between variables is negative

Question 8

When are OLS estimators negatively biased?

Answer 8

When the omitted variable is negative and the correlation between variables is positive.
Or
When the omitted variable is positive and the correlation between variables is negative.

Question 9

How does under specification effect standard errors

Answer 9

It leads to bias in them but we often don’t know if it will make the standard error larger or smaller

Question 10

Is it better to over specify or under specify?

Answer 10

Over specify