General
Finding a small value of the p-value (e.g., less than .05) indicates evidence _____ the null hypothesis.
against
The two conditions for a valid instrument.
Instrument relevance and instrument exogeneity.
When would you use logit or probit?
When the dependent variable is binary (pass/fail, win/lose, alive/dead, healthy/sick).
What are the two things panel data can control for?
Time-fixed effects and entity fixed effects.
What is partial compliance?
The failure of individuals to follow completely the randomized treatment protocol.
Define heteroscedasticity & homoscedasticity.
You should know this.
Instrument relevance is ...
Correlation of X and Z is not 0.
You get a predicted value of 1.96 after running the logit model. What do you conclude?
That you did something wrong. The logit and probit both have a domain between 0 and 1.
When is it useful to use panel data?
As long as the answer makes sense
Name two threats to external validity.
Nonrepresentative program or policy, Nonrepresentative sample, and General equilibrium effects
In an OLS regression, if we change the dependent variable from GDP to log(GDP), the R- squared does not change.
False. R-squared is ESS/TSS, or the fraction of the total variation in GDP explained by the X variables. Taking the log of GDP changes TSS, since there is a different amount of variation in log(GDP) than there is in GDP. As a result, R-squared will change when the dependent variable is put into log form.
Instrument exogeneity is ...
Z and U (error term) are not correlated.
Using a logit model is a useful way to correct for omitted variable bias (True/False).
False. A logit model is used in the case of a binary dependent variable, and is unrelated to omitted variable bias. Instrumental variables strategies, or sometimes panel data, are used to correct for omitted variable bias.
Why do we use it?
Give three examples of threats to internal validity.
failure to randomize, failure to follow the treatment protocol, attrition, experimental effects, and small sample sizes.
Define omitted variable bias. How can you reduce it?
When factors that are not included in the regression affect the regression. You can add controls / more variables.
Define weak instruments. Why are they a problem?
Instruments are weak when the instruments in linear IV regression are weakly correlated with the included endogenous variables. It is an issue because the TSLS estimator may not be normally distributed, even in large samples
How do you interpret the coefficients in logit and profit models?
You don't.
What is the difference between a balanced panel and an unbalanced panel?
Balanced panel: Variables are observed for each entity and each time period. Unbalanced panel: Some missing data for at least one time
period.
What is the differencing estimator and how do you compute it?
The differences estimator is the difference in the sample averages for the treatment and control groups.
Yi =b0 +b1Xi +ui, i=1, ... ,n.
If you have n binary entities, and include n variables in your regression, what will this result in?
Your regression will not run; you have multi-collinearity.
The rule-of-thumb for checking for weak instruments is as follows: for the case of a single endogenous regressor, the first-stage F-statistic must be statistically significant to indicate a strong instrument.
False. The rule-of-thumb for checking for weak instruments is that the first-stage F-statistic must be greater than 16. This indicates a strong correlation between the instrument(s) and the endogenous regressor.
In the expression Pr(deny = 1| P/I Ratio, black) = Ф (-2.26 + 2.74P/I ratio + 0.71black), the effect of increasing the P/I ratio from 0.3 to 0.4 for a white person results in...
4.8 percentage points
If you have n binary entities, how many variables should you have in your regression?
n-1.
Suppose I want to study the return to a Brandeis education. You regress an OLS model that looks like this: earnings ~ brandeis.
Name three issues with this study.