Total Points: 10
1) You are interested in the causal effect of a country’s GDP on its birth rate. Because you have panel data for every country over the last 20 years (2004-2024), you can add fixed effects for both country and year. Your friends have some concerns about the validity of your findings. The following statements people raise as concerns that are not accounted for in the model with country and year FE. Pick true or false for each statement.
a) The overturning of Roe v. Wade in 2022 is a policy that can affect birth rates in the U.S. and not be captured by either country or year fixed.
b) Assuming the effects of the Great Recession of 2008 on GDP are country-invariant (i.e. countries had a similar downturn in GDP in 2008), we would think that country fixed-effects dummies would be netting things out.
c) What about changes in government? For example, there was a coup in Gabon (an African country) just this past year. This varies both within our unit and across time, so our fixed effects are not capturing its impact on birth rates.
2) In the context of the question above, here is another statement “COVID-19 certainly reduced GDP, but it also made people feel less sure about the future; this could reduce the birth rate on its own, and thus, we would be conflating the impact of GDP and COVID-19 on birth rates.” Pick the assumptions you would have to make to argue that these are captured by either country or year fixed-effects.
a) COVID-19 affected countries differently and it’s effect vary over time.
b) COVID-19 started in March of 2020.
c) COVID-19 does not affect GDP.
d) COVID-19 had a similar impact on GDP and Birth rates across countries.
3) Below is a screenshot of some panel data. What are the two dimensions of the data?
a) Latitude and temperature
b) Year and Latitude
c) Temperature and Month
d) Latitude and Day
4) We would now like to examine predictors of GDP. You want to know if there is a relationship between educational attainment and GDP in the United States:
where is the GDP of state s in year t, is measured in thousands of dollars and is the percent of individuals in a given state (s) and year t, who have at least a Bachelor’s degree. Which of the following would be appropriate fixed effects to include in your regression?
a) Year-fixed effects, because post-secondary attainment and GDP may reflect economic changes or shocks that happen in given years.
b) State-fixed effects, because educational investment may vary with state characteristics in ways that are related to their economic productivity.
c) We do not need fixed effects because we are interested in this relationship at the national level.
5) You run the following regression using the sample between years 2013-2019:
Where captures year fixed effects and captured state fixed effects. Which of the following variables can we add controls for?
a) The amount of federal funds (in dollars) allocated to states for post-secondary education.
b) National economic shocks or recessions that decreased the economic output, that roughly affected states the same.
c) Some colleges are near the border of other states, which could affect PctBA and GDP, so we should account for the number of states that a state borders.
d) Ideology measured as the percentage of voters for the democratic candidate in the presidential election during this sample period.
6) (True/False) If we use the Least Squares Dummy Approach (LSDA) to implement fixed-effects in our regression of GDP on the percentage of people with BAs, we are creating one dummy for every state in our dataset.
a) True
b) False, we use LSDA so that we do not have to control for individual states.
c) False, it omits one state as our baseline group.
d) False, LSDA creates a “Treatment” dummy based on if states have greater or less than 50% BA attainment.