Part 1: Hillary Clinton's Favorability in the 2016 Election
When Clinton first ran for the presidency in 2008 she distanced herself from her association to Bill Clinton. Many thought this was a mistake, since he was a popular president and left office when most experienced economic prosperity. In 2016, Hillary's campaign instead embraced the connection to her husband. Bill Clinton vigorously campaigned for Hillary and she sometimes touted her husband's accomplishments in the ‘90s.
In this part, we will explore the impact of individuals' views of Bill Clinton on favorability for Hillary in 2016. We use the American National Election Studies 2016 study. This is a large survey of 4,271 respondents on voting behavior and U.S. presidential elections. You will use dataset anes2016.dta.
We test the hypothesis that favorability for Hillary Clinton is strongest among those that hold positive views for Bill Clinton. We also test the hypothesis that favorable views of Bill Clinton depend on age. The idea is that individuals that were children or had not yet come of age in the ‘90s are not as likely to view Bill Clinton as favorably as those older people.
The dependent variable is Hillary, a feeling thermometer on a scale of 0 to 100, where higher values indicate stronger favorability. A score of 70 for example, suggests the respondent has a strongly favorable view of Hillary, whereas a score of 20 indicates low favorability. The other variable definitions are on the table below.
Variable Descriptions:
1. Run three regression models using the anes16.dta dataset and report the results in Table 1.
- You will need to create an interaction term for Model 3.
- Each cell should contain a coefficient and the robust standard error for that coefficient report below in the parentheses.
- State whether the coefficients are significant using asterisks.
- Note: Gray cells and dashes indicate the variable is not included in the model.
Table 1: The Effect of Support for Bill Clinton on Favorability for Hillary in 2016
2. Interpret the coefficient on "Bill" in Model 1. Are results consistent with expectations?
3. Interpret the adjusted R2 in Model 1. Does the adjusted R2 tell us if the coefficient on "bill" is biased? Why or why not?
4. Using Model 2, can we expect an increase in favorability for Hillary among individuals older than 30? Explain.
5. Does the constant have a meaningful interpretation in Model 2?
6. Using Model 3, calculate the effect of "Bill" on favorability for Hillary among individuals 30 and older.
7. Using Model 3, calculate the effect of "bill" on favorability for Hillary among individuals younger than 30.
8. Does this regression model indicate that there is an interactive effect between "Bill" and "age30"? Explain.
9. Is "Democrat" an omitted variable in Models 1 and 2? If so, what is the sign of the bias? Explain.
10. Comparingall three models, what do you conclude? Can we confidently establish that favorability for Hillary is explained by all or any of these models? Discuss.
Part 2: Who is More Likely to Support the Repeal of the Affordable Care Act (ACA)?
To study this question, we will look at the ANES 2016. The survey asked respondents whether they favored or opposed the Affordable Care Act. The news media are reporting that the American public is for the most part opposed to the repeal. But opposition varies among individuals. The issue is highly partisan, with most Republicans more inclined to oppose the ACA. Here, we predict the probability of someone supporting the ACA based on their party id, whether or not the individual has insurance, and race.
Using the variables described below, we run a linear probability model, and probit and logit models.
With the Stata output below, answer the questions for this part.
Description of Variables:
11. Are the findings from the Stata Output #1 consistent with expectations?
12. Interpret the coefficient on nonwhite.
13. According to the calculations conducted in Stata (below), what is the probability of supporting the ACA for someone who is a Democrat, does not have insurance and is not white. We obtained the probability below.
scalar z = _b[_cons] + _b[democrat]*1 + _b[insurance]*0 + _b[nonwhite]*1
display z
.63108475
14. Next, we examined this question using a probit model. What is the direct interpretation of the coefficient on nonwhite?
15. According to the probit model, what is the probability of someone supporting the ACA who is Democrat, does not have insurance and is not white?
16. According to the logit model in Stata output #3, what is the probability of someone supporting the ACA who is a Democrat, does not have health insurance and is not white?
17. Table 1 shows the predicted probabilities for all three models. The first row shows the predicted probabilities of supporting the ACA for someone that is a Democrat, has no insurance and is not white. The second row shows the probability of supporting the ACA for someone who is not a Democrat, has insurance and is white.
What do you conclude about the predicted probabilities for all three models for both individuals?
18. Which model is best and why?
Attachment:- Problem Set.rar