QMSS – Data Analysis (5015) Midterm, Fall 2018Name _________________________________
Please answer these questions as completely as possible. Make sure you answer all parts of the question. The exam is open-note, open-book, open-Internet. If you do consult sources, do not directly quote them, but
paraphrase and synthesize (and provide citations). You do not need to provide citations for my slides or Wooldridge, but do not just copy-and-paste from us either. Do not discuss this exam with anyone until after the
due date. Thanks!
The exam is due on Courseworks under Assignments on Wednesday, Oct. 24th, 1pm.
Part 1-- 43 points
Using the 2016 GSS, a researcher wanted to answer the following question: What influences what people say is the the ideal number of children for a family to have? The outcome is Ideal Number of Children, which
represents the number of children in a family a respondent would ideally want, ranging from 0 to 7 children.
In Model 1, she predicts this outcome only using the number of actual children a respondent has, ranging from 0 to 8 children (Number of Children).
In Model 2, she further includes the number of siblings the respondent has, ranging from 0 to 8 siblings
(Number of Siblings), and the respondent’s age, which ranges from 18 to 89 (Age).
In Model 3, she further includes measures for the socioeconomic status (SES) of the respondent: a series of indicator variables for the highest degree the respondent achieved, with “Less than High School” as the
reference category, then ascending to High School Degree, then Some College, then a BA Degree and ultimately, Graduate Degree; and the natural log of total family income, ranging from 5.46 to 11.79
(Ln(Income)).
Lastly, in Model 4, she wants to consider if a person’s sense of their own past might affect their ideal number of children. So she includes a measure of the respondent’s sense of how much their standard of living has
improved now compared with their parents then: the answers range from much worse (1), somewhat worse (2), about the same (3), somewhat better (4), or much better (5) than their parents was (Improved Standard of
Living).
Table 1: Summary Statistics for Part I, GSS 2016
P. 1 of 6
Table 2: Models Predicting Ideal Number of Children, GSS 2016
1. (a) In Model 1, interpret the coefficient on Number of Children (8 points). (b) Is it statistically significant and how do you know? Please talk about t-statistics in your answer. (9 points) (c) In Model 1, does the
correlation (or R-squared) between Ideal Number of Children and Number of Children seem low or high to you? Based on your prior expectation, give one reason for why you think that relationship is either low or
high (4 points).
2. In Model 1, interpret the constant/intercept in this model-- i.e., to whom does it refer? Also note statistical significance. (10 points)
3. (a) In Model 2, interpret the coefficient on Age. Note statistical significance. (8 points). (b) In Model 2, Age
has a negative relationship with Ideal Number of Children. Does this relationship make sense to you? Give me one explanation for why you think Age would have a negative relationship with Ideal Number of
Children (4 points).
P. 2 of 6
P. 3 of 6
Part 2-- 57 points
Another researcher wanted to answer a different question using the 2016 GSS. Almost 80% of respondents to the GSS are now given a monetary incentive to take the survey; it is a variable called Fee in $, ranging from $5
to $75. The question is: What determines the amount of a fee someone will get, if they request money to complete the survey?
In Model 1, the researcher predicts the size of the incentive (Fee in $) as a function of the total family income of
the respondent, measured in 1000s of dollars, which ranges from 2.34 to 131.68 (Income in 1000s).
In Model 2, the researcher predicts the natural log of the incentive, ranging from 1.61 to 4.32 (Ln(Fee)), as a function of the natural log of the total family income (Ln(Income)), which ranges from 5.46 to 11.79.
In Model 3, the researcher includes indicator variables for the potential work statuses of the respondent, with
the reference group being “Full Time Work,” and the other options being: Working Part-Time, Not Working Currently, Laid Off/Unemployed, Retired, In School, "Keeping House," and Miscellaneous Work Status.
In Model 4, the researcher considers one more factor that might affect how much of an incentive the respondent
might get: how physically attractive the respondent was, as determined by the interviewer (How Attractive Respondent Is (Interviewer Assessment)), ranging from very unattractive (1), unattractive (2), about average (3),
attractive (4), very attractive (5).
Table 3: Summary Statistics for Part II
P. 4 of 6
Table 4: Models Predicting Amount of Incentive Respondents Got to Take the Survey, GSS 2016
6. Interpret the coefficient on Income (1000s) in Model 1. Note statistical significance. (6 points). Interpret the
coefficient on Ln(Income) in Model 2. Note statistical significance. (9 points). Why do economists often prefer the specification in Model 2 over the specification in Model 1? Give two reasons. (8 points).
7. Interpret the coefficient on “Miscellaneous Work Status” in Model 3. Note statistical significance. (8 points)
8. The researcher theorized that maybe one reason people would expect a larger incentive amount is if they
were not home as much (because they are working), they might feel their time was more valuable, even if we control for overall income levels of the family. He included a series of indicators for work status to test this
idea. Then he ran a partial F-test on the whole set of Work Status dummy variables (“Part-Time Work” through “Misc. Work Status”). The result was an F-statistic of 3.48 (p<.001). (a) Explain what a partial F-test tests for
(9 points), and (b) explain whether, from a statistical standpoint, this F-test indicates that the researcher made a good decision to include the Work Status dummy variables. (8 points)
P. 5 of 6
9. Interpret the coefficient on How Attractive Respondent Is (Interviewer Assessment) in Model 4. Note statistical significance. (5 points). What conclusion do you draw from the sign and statistical significance of
this variable and its relationship to the size of incentive given to the respondent? Provide one possible conclusion. (4 points)