Econ B2000, MA Econometrics

Kevin R Foster, the Colin Powell School at the City College of New York, CUNY

Fall 2020

Not all of these questions are strictly relevant; some might require a bit of knowledge that we haven’t covered this year, but they’re a generally good guide.

Of course you should be able to do all of the Practice Problems for Exam 1.

2019

(20 points) Please answer the following; you might find it useful to make a sketch.

For a Normal Distribution that has mean 12 and standard deviation 1.3, what is the area to the left of 9.92?
For a Normal Distribution that has mean 1 and standard deviation 3.6, what is the area to the right of -4.76?
For a Normal Distribution that has mean -8 and standard deviation 5.2, what is the area to the left of -2.8?
For a Normal Distribution that has mean -15 and standard deviation 8.3, what is the area to the right of -3.38?
For a Normal Distribution that has mean -5 and standard deviation 9.9, what is the area in both tails farther from the mean than -10.94?
For a Normal Distribution that has mean 15 and standard deviation 2.9, what is the area in both tails farther from the mean than 8.04?
For a Normal Distribution that has mean 8 and standard deviation 5.7, what is the area in both tails farther from the mean than 19.4?
For a Normal Distribution that has mean -8 and standard deviation 2.3 what values leave probability 0.396 in both tails?
For a Normal Distribution that has mean -14 and standard deviation 1.9 what values leave probability 0.673 in both tails?
A regression coefficient is estimated to be equal to -4.9 with standard error 4.9; there are 5 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?
A regression coefficient is estimated to be equal to -3.6 with standard error 7.2; there are 11 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?
A regression coefficient is estimated to be equal to -14.72 with standard error 6.4; there are 33 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?

For the next few questions we’ll consider the (seasonally appropriate!) question of who gets flu shots, using BRFSS data.

(20 points) Using a subset of the BRFSS data (so you can do this question without R), the following coefficients are estimated in a logit model where FluShot depends on $Age$, $Age^2$, $Female$, and interaction of $Female$ with the age terms.

Variable	Coeff	Std Error
$Age$	-0.030	0.004
$Age^2$	0.001	0.00003
$Female$	0.115	0.125
$Age*Female$	0.015	0.005
$Age^2*Female$	-0.0002	0.00004
$Constant$	-0.436	0.094

What is the predicted probability that a 30-year-old person (not a female) will get a flu shot?
What is the predicted probability that a 30-year-old female person will get a flu shot?
Describe and explain the impact of the squared age terms in the model, for male and female.

For next questions, load BRFSS data in R. I have prepared a dataset for the exam that you can download along with a file, “Notes for using BRFSS for Exam 2.R” that tells a bit about the variabe names. It’s big! 278,021 observations so you might want to take another subset to manage some of your estimations.

(60 points) Now using R, consider estimating a model for the probability that someone will get a flu shot.

Explain what subset of people you will use. Provide summary statistics of the important variables.
Estimate a simple OLS (LPM) model of whether a person gets a flu shot.

i. Explain what variables you choose to use as predictors. Do they seem exogenous? Do the estimates seem plausible?  
ii. Are the estimates each statistically significant?
iii. Construct a joint test of whether a reasonable set of coefficients are all zero.
iv. What are some predicted probabilities for particular people?
v. How many Type I and Type II errors are made by the model?

Next estimate a logit model of whether a person gets a flu shot. Explain same as in previous.

(20 points) Estimate one or more additional models with other methods (not LPM or logit) to predict if a person gets a flu shot. Explain as in previous.

2018

(20 points) Please answer the following; you might find it useful to make a sketch.
1. For a Normal Distribution that has mean -10 and standard deviation 6.6, what is the area to the right of -19.9 ?
2. For a Normal Distribution that has mean 5 and standard deviation 3, what is the area to the right of 10.4 ?
3. For a Normal Distribution that has mean 3 and standard deviation 6.1, what is the area to the right of -7.98 ?
4. For a Normal Distribution that has mean 13 and standard deviation 4, what is the area to the left of 16.2 ?
5. For a Normal Distribution that has mean 2 and standard deviation 8.9, what is the area to the left of 10.01 ?
6. For a Normal Distribution that has mean -15 and standard deviation 4.6, what is the area to the left of -22.82 ?
7. For a Normal Distribution that has mean -11 and standard deviation 2.8, what is the area in both tails farther from the mean than -5.96 ?
8. For a Normal Distribution that has mean 8 and standard deviation 9.8, what is the area in both tails farther from the mean than 23.68 ?
9. For a Normal Distribution that has mean 13 and standard deviation 4.7 what values leave probability 0.804 in both tails?
10. For a Normal Distribution that has mean 8 and standard deviation 5.2 what values leave probability 0.315 in both tails?
11. A regression coefficient is estimated to be equal to -0.9 with standard error 0.6; there are 4 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?
12. A regression coefficient is estimated to be equal to 1.08 with standard error 3.6; there are 23 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?

I have created a dataset for you based on the Fed’s Survey of Consumer Finances, which you can download from Blackboard as the file fed_scf_exam.RData. For this exam we will examine the amount of credit card debt that people have (CreditCardBalance is the name of the variable).

For each question below, carefully specify the statistical tests that you perform, including the null hypothesis and test statistics such as t-stat and p-value.

(10 points) Find some basic summary statistics about the credit card balances. How common is a big balance (define “big”)? Are you measuring absolute dollar value or fraction of income? Can you find some interesting correlates – is education a factor? Race/ethnicity? Test if the average balance for people over 35 is significantly different from the average balance for people under 35. Explain what you initially expected to find (before you ran the numbers) and how/if the results shift that view.
(15 points) Estimate an OLS model of the credit card balance. What are some important variables to include in the regression? Discuss and explain the results; which coefficients are significant? Are education variables jointly significant? Discuss issues of endogeneity.
(20 points) Add your choice of polynomial Age terms along with interactions with dummy variables to the OLS model. Explain these interaction effects. Explain the variation with age. Are the interaction coefficients jointly significant - does adding the interaction significantly help in prediction? Do heteroskedastic-consistent errors change the interpretation?
(10 points) For a particular subset (don’t try to replicate it), I estimate these coefficients for a logit estimation of whether the credit card balance is greater than zero.

Varb	Coeff	Std Err
AGE	-0.011390	0.001173
ed_hs	0.372705	0.064133
ed_somecoll	0.629340	0.068319
ed_colldegree	0.229324	0.062082
inLF	0.404572	0.043195
(Intercept)	-0.664143	0.091779

Explain what these coefficients mean and whether each is statisically significant. Calculate the estimated probability of a nonzero balance, for a 35 year old person with a high school diploma who is in the labor force. What is the change in estimated probability if the person has some college? What is the change in estimated probability if the person has a college degree?

(20 points) Next estimate a logit or probit model, where the dependent variable is now whether the balance is “big” (explain your choice of “big” and why). Explain what variables ought to be included or excluded. Discuss the results of the model and hypothesis tests. Calculate some predicted probabilities for representative people.
(25 points) Next estimate another model or several (of your choice) to the credit card balance and/or whether it’s big. Explain and discuss - impress me with your econometric virtuosity. Can you improve some of the predictions of previous models? Discuss.

2017

(25 points) Please answer the following; you might find it useful to make a sketch.

For a Normal Distribution that has mean -7 and standard deviation 7.1, what is the area to the right of 6.49?
For a Normal Distribution that has mean 11 and standard deviation 4.1, what is the area to the right of 6.08?
For a Normal Distribution that has mean 5 and standard deviation 3, what is the area to the left of 3.5?
For a Normal Distribution that has mean -7 and standard deviation 3.8, what is the area to the left of 1.74?
For a Normal Distribution that has mean -10 and standard deviation 5.1, what is the area to the left of -18.67?
For a Normal Distribution that has mean -10 and standard deviation 3.4, what is the area in both tails farther from the mean than -12.04?
For a Normal Distribution that has mean 8 and standard deviation 8.6, what is the area in both tails farther from the mean than -5.76?
For a Normal Distribution that has mean 12 and standard deviation 2.2 , what is the area in both tails farther from the mean than 10.02 ?
For a Normal Distribution that has mean -5 and standard deviation 1.3 what values leave probability 0.15 in both tails?
For a Normal Distribution that has mean 11 and standard deviation 7.6 what values leave probability 0.782 in both tails?
For a Normal Distribution that has mean 9 and standard deviation 3.1 what values leave probability 0.077 in both tails?
A regression coefficient is estimated to be equal to 6.09 with standard error 8.7; there are 40 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?
A regression coefficient is estimated to be equal to -20.16 with standard error 8.4; there are 34 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?
A regression coefficient is estimated to be equal to 8.8 with standard error 4.4; there are 5 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?
A regression coefficient is estimated to be equal to -17.64 with standard error 9.8; there are 11 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?

Since there has been a great deal of recent news about age differences in couples, I used the ACS data to look at ages of spouses.

The data are arranged in a particular way: each household answers the survey and picks a person to be “head” then the other (if applicable) as “spouse”. There is no presumption of gender for either role. In the way I’ve arranged the data, the line for the person who is labeled “spouse” has some demographics about the household head (all that data is prefixed with h_) so for example here are the first few lines of data:

AGE	h_age	age_diff	SEX	h_sex	RELATE
61	56	5	Female	Male	Spouse
53	53	0	Female	Male	Spouse
61	63	-2	Female	Male	Spouse
37	35	2	Male	Female	Spouse
51	57	-6	Male	Female	Spouse
34	32	2	Male	Female	Spouse

In the first case the male took role of “head” and female took role of “spouse” so her line of data includes her own age (61) and her husband’s age (56) so the “age_diff” is +5. In later lines the female took role of “head” and male took role of “spouse”.

You might consider statistics like summary(age_diff[(SEX == “Female”)&(h_sex == “Male”)]), summary(age_diff[(SEX == “Male”)&(h_sex == “Female”)]), summary(age_diff[(SEX == “Male”)&(h_sex == “Male”)]), and summary(age_diff[(SEX == “Female”)&(h_sex == “Female”)]).

For each question below, carefully specify the statistical tests that you perform, including the null hypothesis and test statistics such as t-stat and p-value.

(10 points) Find some basic summary statistics about the age gap for married people. How common is a big age gap (you can define “big”)? Can you find some interesting correlates – is geography a factor? Education? Working status? Test if the average gap for people over 50 is significantly different from the average gap for people under 50. Explain what you initially expected to find (before you ran the numbers) and how/if the results shift that view.
(15 points) Estimate an OLS model of the age difference. What are some important variables to include in the regression? Discuss and explain the results; which coefficients are significant? Discuss issues of endogeneity.
(20 points) Add your choice of polynomial Age terms along with interactions with dummy variables to the OLS model. Explain these interaction effects. Explain the variation with age. Are the interaction coefficients jointly significant - does adding the interaction significantly help in prediction?
(10 points) For a particular subset, I estimate these coefficients for a logit estimation of whether the age difference is especially large.

Varb	Coeff	Std Err
educ_hs	-0.327	0.010
educ_smcoll	-0.403	0.011
educ_coll	-0.615	0.011
educ_adv	-0.611	0.012
SEXFemale	-0.086	0.032
h_sexFemale	-0.096	0.032
Constant	-0.369	0.033

Explain what these coefficients mean. Should we include education levels of the head as well as the spouse? Calculate the estimated probability that a spouse with some college, who is male and the head of household is female, has a large age difference. What is the change in estimated probability if the spouse is female and head of household is male? What is the change in estimated probability if, instead, the spouse gets an advanced degree?

(20 points) Next estimate a logit or probit model, where the dependent variable is now whether the age difference is more than a few years (explain your choice of “a few” and why). Explain what variables ought to be included or excluded. Discuss the results of the model and hypothesis tests. Calculate some predicted probabilities for representative people.
(20 points) Next estimate another model or several (of your choice) to the age difference. Explain and discuss - impress me with your econometric virtuosity. Can you improve some of the predictions of previous models?

2016 and earlier

(20 points) Please answer the following; you might find it useful to make a sketch.

For a Normal Distribution that has mean -11 and standard deviation 6.6, what is the area to the right of -15.62?
For a Normal Distribution that has mean -5 and standard deviation 2.8, what is the area to the right of -9.48?
For a Normal Distribution that has mean 12 and standard deviation 4.7, what is the area to the left of 6.83?
For a Normal Distribution that has mean -1 and standard deviation 6.5, what is the area to the left of 4.85?
For a Normal Distribution that has mean 3 and standard deviation 8.6, what is the area in both tails farther from the mean than -9.9?
For a Normal Distribution that has mean 7 and standard deviation 0.6, what is the area in both tails farther from the mean than 6.04?
For a Normal Distribution that has mean -5 and standard deviation 9.8, what is the area in both tails farther from the mean than -5?
For a Normal Distribution that has mean 13 and standard deviation 9 what values leave probability 0.16 in both tails?
For a Normal Distribution that has mean -13 and standard deviation 7 what values leave probability 0.486 in both tails?
A regression coefficient is estimated to be equal to 2.6 with standard error 5.2; there are 33 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?
A regression coefficient is estimated to be equal to 20.47 with standard error 8.9; there are 9 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?

The next questions use the dataset provided, ATUS_for_exam2.RData, (details in file ATUS_for_exam2.R) of just working people (who are coded as “Employed - at work”). You should consider the relevant factors for how much time people spend on social fun activities, ACT_SOCIAL. This gives minutes of a typical day spent socializing, relaxing, and doing other leisure activities (games, TV, hobbies). For each question below, carefully specify the statistical tests that you perform, including the null hypothesis and test statistics such as t-stat and p-value. To get from the ATUS data that I had given in class to this exam data, I ran these simple lines,

load("ATUS_2003_2013_a.RData")
use_varb <- (dat2$EMPSTAT == 'Employed - at work')
dat_use <- subset(dat2,use_varb)
is.na(dat_use$EARNWEEK) <- which((dat_use$EARNWEEK == 9999999))
dat_use$EARNWEEK <- dat_use$EARNWEEK/100
dat_use$lots_social <- (dat_use$ACT_SOCIAL > 180)

(10 points) Start with some basic statistics: how does the time spent on social activities vary among educational groups? Are these differences statistically significant? Discuss. (Perhaps relate to opportunity cost.) What other explanatory variables might be relevant?
(10 points) There are data items that give the amount of time spent on other activities (time spent working, sleeping, etc) - might there be worries about endogeneity? Explain. Are there particular correlations that are significant?
(15 points) Next estimate a linear OLS model for the dependent variable, time spent on social activities. Explain what variables you believe are important to include. Discuss the results.
(15 points) The ATUS includes a great number of dummy variables including the state the person lives in (STATEFIP), the metro area (if they are in a metro area, METAREA), their occupation (OCC) and industry (IND). Add one or more of these factors to your OLS model and discuss how this changes the other estimated coefficients. Are the dummy variable coefficients jointly significant - does adding the factor significantly help in prediction?
(15 points) Next estimate a logit or probit model, where the dependent variable is now whether the person spends a lot of time (more than 180, so more than 3 hours per day) on social activities. Explain what variables ought to be included or excluded. Discuss the results of the model and hypothesis tests.
(35 points) Next estimate another model or several (of your choice) to predict time spent on social activities. Explain and discuss - impress me with your econometric virtuosity. Can you improve some of the predictions of previous models?
(20 points) Consider the following simple distribution questions. Your answer can be as detailed as necessary. You might sketch each case.

For a Normal Distribution with mean -1 and standard deviation 5.2, what is area to the right of -9.32?
For a Normal Distribution with mean -6 and standard deviation 7.2, what is area to the right of 2.64?
For a Normal Distribution with mean 15 and standard deviation 8, what is area to the left of -3.4?
For a Normal Distribution with mean -3 and standard deviation 0.5, what is area to the left of -2.1?
For a Normal Distribution with mean 11 and standard deviation 5.8, what is area in both tails farther from the mean than 19.12?
For a Normal Distribution with mean -9 and standard deviation 5.1, what is area in both tails farther from the mean than -1.86?
For a Normal Distribution with mean 11 and standard deviation 6.1 what values leave probability 0.357 in both tails?
For a Normal Distribution with mean 13 and standard deviation 4.3, what values leave probability 0.035 in both tails?
A regression coefficient is estimated to be equal to -5.25 with standard error 2.1; there are 12 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?
A regression coefficient is estimated to be equal to 5.88 with standard error 8.4; there are 13 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?
A regression coefficient is estimated to be equal to -4.5 with standard error 9; there are 3 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?

(15 points) Using a subset of BRFSS data (so don’t bother trying to replicate), I estimated a regression for BMI with the following quadratic terms on age (including an interaction with gender). [BMI is a person’s weight in kg divided by their squared height in m, so a number over 25 is interpreted as overweight.]

Variable	Coeff	Std Error
Constant	18.968	1.023
Age	0.523	0.022
Age2	-0.006	0.0003
Female	1.723	0.605
Age*Female	-0.122	0.030
Age2*Female	0.001	0.0004

Are these coefficient estimates each statistically significant? Calculate t-statistics and p-values for each (there are 105409 df).
What is the predicted BMI for a 35-year-old male? For a female of the same age? At what age are men predicted to have BMI over 30? At what age are women predicted to have BMI over 30?
At what age does male BMI peak? Female? At what levels for each?
Does this model seem reasonable? Discuss.

(15 points) Continuing with the BMI data from BRFSS, I estimate a logit model where the dependent variable is whether the person’s BMI would classify them as “overweight” or “obese” but again estimate a quadratic in age with gender interaction. These results are:

Variable	Coeff	Std Error
Constant	-2.355	0.430
Age	0.197	0.008
Age2	-0.0020	0.0001
Female	1.360	0.206
Age*Female	-0.100	0.010
Age2*Female	0.0011	0.0001

Are these coefficient estimates each statistically significant? Calculate t-statistics and p-values for each (there are again 105409 df).
What is the predicted probability of being overweight for a 35-year-old male? For a female of the same age?
At what age does male probability of being overweight peak? Female? At what levels for each?
Does this model seem reasonable? Discuss.

The next questions use the dataset provided, BRFSS2013_forexam.RData, (details in file brfss_for_exam2.R). You should make a further examination of factors important in determining whether a person is overweight or obese. You can decide the best measurement: the data includes a continuous variable (BMI_measure), a 0/1 dummy for overweight (d_overweight), and/or a 4-category classification (X_BMI5CAT). For each case, carefully specify the statistical tests that you perform, including the null hypothesis and test statistics such as t-stat and p-value.

(10 points) Start with some basic statistics: how does the tendency to overweight vary among educational groups? Are these differences statistically significant? Discuss.
(10 points) Next estimate a linear model, something like the one in Q 2 above (more extensive of course!). Explain what variables are important to include? Discuss the results.
(15 points) Next estimate a logit or probit model, something like Q 3 above (more extensive of course!). Explain what variables ought to be included or excluded. Discuss the results of the model.
(35 points) Next estimate another model or several (of your choice) to predict BMI. Explain and discuss – impress me with your econometric virtuosity.
Consider the following simple distribution questions. Your answer can be as detailed as necessary. You might sketch each case.

For a Normal Distribution with mean 4 and standard deviation 3.6, what is area to the right of 1.48?
For a Normal Distribution with mean -1 and standard deviation 3.2, what is area to the left of 4.12?
For a Normal Distribution with mean -14 and standard deviation 6.2, what is area in both tails farther from the mean than -4.08?
For a Normal Distribution with mean 10 and standard deviation 0.6, what is area in both tails farther from the mean than 10.18?
For a Normal Distribution with mean -12 and standard deviation 3.3, what is area in both tails farther from the mean than -5.73?
For a Normal Distribution with mean -1 and standard deviation 6.3, what values leave probability 0.058 in both tails?
For a Normal Distribution with mean -1 and standard deviation 5.3, what values leave probability 0.225 in both tails?

Consider the following table of regression coefficients explaining wage and salary as a function of the given variables with heteroskedasticity-consistent Eicker-Huber-White standard errors. Fill in the blanks. There are 47,528 degrees of freedom with this data from CPS. (I used a random subset so don’t try to re-estimate.)

Variable	Coeff	Std Error	t-stat	p-val
(Intercept)	-60367	_____	-6.45	<0.01
female	27009	11333	_____	_____
Age	_____	503	7.99	<0.01
Age^2	-38	6	_____	_____
female*Age	-1949	_____	-3.19	_____
female*Age2	_____	8	2.41	_____
AfAm	-8953	676	_____	_____
Asian	-1740	_____	-1.26	0.21
Hispanic	-7042	_____	-7.01	<0.01
married	5322	635	8.39	<0.01
divwidsep	150	889	_____	_____
union_m	-2052	1701	_____	_____
veteran	-1063	1325	_____	_____
educ_hs	9408	855	11.01	<0.01
educ_smcoll	18135	979	18.52	<0.01
educ_as	19349	996	19.43	<0.01
educ_bach	39289	1013	38.77	<0.01
educ_adv	66908	1575	42.49	<0.01

Consider a model, based on CPS data using only fulltime workers with four-year degrees, estimating coefficients on Age and Age-squared, interacted with gender (there are other explanatory variables as well, don’t worry about those for now).

Variable | Coeff | Std Error ————|-_——-|———- (Intercept) | -62851 | 18405 Age | 5809.56 | 994.97 Age^2 | -54.46 | 12.79 female | 37523 | 22090 femaleAge | -2216.21| 1192.02 femaleAge2 | 16.98 | 15.39

What is the predicted wage for a 40-year-old female? For a 40-year-old male? What is the estimated peak age for earnings for each gender?
Which coefficient estimates are statistically significantly different from zero? Discuss.
Can you estimate other models (e.g. quantile, loess, nonparametric) that would cast more light on the question of relative earnings by gender, age, and/or other factors Discuss.

I created a dataset based on the National Health Interview Survey (NHIS) from 2013, on Blackboard as nhis_2013.RData. We will analyze the important determinants of whether a person is not covered by health insurance (the variable, NOTCOV, where 1 means they are not covered and 0 means they are covered by some type of health insurance). The data frame has the following variables: data_use1 <- data.frame(NOTCOV, educ_nohs, educ_hs, educ_smcoll, educ_as, educ_bach, educ_adv, AGE_P, female, AfAm, Asian, RaceOther,Hispanic, Hispan_PR, Hispan_Mex, Hispan_DR, married, widowed, divorc_sep, REGION, borninUSA, region_born, veteran_stat, inworkforce, ERNYR_P, disabl_limit, person_healthstatus, MEDICARE, MEDICAID, private_ins, RRP, HHX, FMX, FPX, SCHIP, sptn_medical)

Most of those variables should be familiar: a series of education dummies, the person’s age, gender, race/ethnicity, marital status, what REGION (Census breaks the US into 4 broad regions), where the person was born (in US or in various regions of the globe), whether the person is a veteran, if they’re in the workforce. Note that there is a factor for their earnings (ERNYR_P is broken into broad amounts, it is not a continuous variable). Each person reports if they have any limitations on normal activity; this is disabl_limit. Each has a health status from 1-5. They might be covered by various types of insurance: Medicare, Medicaid, or private insurance (there are others, which I omitted). RRP is relationship – there is the reference person in each household plus others possibly including spouse, children, other family. SCHIP is a government program supporting children. There is a factor for broad amount spent on medical care, sptn_medical. Then HHX, FMX, FPX are identifiers.

Create a basic classification table for the person’s health status and their type of insurance (Medicaid, Medicare, private insurance, other insurance or not covered). You might choose a certain subgroup of the data – explain.

Conditional on a person being in “Fair” or “Poor” health, what is the likelihood that they have government insurance (Medicaid or Medicare)? Of being uninsured? Carefully explain a hypothesis test for whether these are equal.
Are people with government insurance in worse health? Discuss possible interpretations of causality.
Create another classification table of person’s type of insurance and a different explanatory variable (choose one). Discuss and explain.

Estimate simple logit and/or probit models of whether a person is likely to be covered by health insurance. You might choose a certain subgroup of the data (by age or other factor). Consider which variables belong and are not endogenous.

Explain what (if any) subgroup of the data you’re using. Explain the rationale for included variables.
Discuss the signs of the coefficient estimates. Which ones are statistically significantly different from zero? Are there differences between probit/logit?
Compare the predicted probabilities for a few different people. What are some of the estimates for the change in likelihood of being covered, for a person getting an advanced degree?
Create and discuss the classification table (those who are predicted 0/1 versus actual 0/1).
Discuss pro/con of including the amount spent on medical care as an explanatory variable in the regression.

Estimate some additional classification models of whether a person is likely to be covered by health insurance. Discuss the results (such as classification table) for models such as k-nn, tree, random forest, etc. Compare with the previous estimation. Discuss.
You might find it useful to sketch the distributions.

If a variable has a Normal Distribution with mean 9 and standard deviation 9, what is area to the right of -8.3?
For a Normal Distribution with mean 5 and standard deviation 0.4, what is area to the left of 4.7?
For a Normal Distribution with mean 6 and standard deviation 0.3, what is area in both tails farther from the mean than 5.7?
For a Normal Distribution with mean -2 and standard deviation 3.8, what is area in both tails farther from the mean than 2.9?
For a Normal Distribution with mean 6 and standard deviation 7.5, what is area in both tails farther from the mean than -2.3?
For a Normal Distribution with mean 14 and standard deviation 3.4, what values leave probability 0.292 in both tails?
For a Normal Distribution with mean 8 and standard deviation 2.6, what values leave probability 0.253 in both tails?
For a Normal Distribution with mean -11 and standard deviation 2.6, what values leave probability 0.420 in both tails?
For a Normal Distribution with mean 2 and standard deviation 4.7, what values leave probability 0.007 in both tails?
For a Normal Distribution with mean -10 and standard deviation 7.9, what values leave probability 0.156 in both tails?

A regression coefficient is estimated to be equal to 1.902 with standard error 1.5; there are 26 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?
A regression coefficient is estimated to be equal to 12.942 with standard error 9.6; there are 8 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?
A regression coefficient is estimated to be equal to 3.647 with standard error 2.6; there are 15 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?
A regression coefficient is estimated to be equal to -5.130 with standard error 3.5; there are 17 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero?
A regression coefficient has standard error 2.40; there are 14 degrees of freedom. The t-statistic is 2.5994. What is the coefficient?
A regression coefficient has standard error 3.40; there are 28 degrees of freedom. The t-statistic is -1.4877. What is the coefficient?
A regression coefficient has standard error 2.30; there are 12 degrees of freedom. The t-statistic is -1.0175. What is the coefficient?
A regression coefficient is estimated to be equal to 11.219; there are 7 degrees of freedom. The t-statistic is 1.6259. What is the standard error?

A recent research paper, looking at how much attractiveness and personal grooming affects wages, used data from The National Longitudinal Study of Adolescent Health in 2001-2.

Are there gender differences? Among the 6074 people (48.4% female), 38.8% of the males were rated as being well groomed or very well groomed; 50.6% of the females were rated that way. Is this a statistically significant difference?
The study considers interrelations between physical attractiveness and grooming. People were ranked on a 4-point scale (where 1 is below average, 2 is average, 3 is above average, and 4 is very much above average) for each attribute. The full details are: Physically 4 Very Attractive 3 Attractive 2 Average 1 Less Attractive 4 Very well groomed 297 199 57 30 3 Well groomed 290 1169 607 54 2 Average grooming 75 788 2013 167 1 Less than average grooming 1 25 164 138 Conditional on a person being ranked physically 3 or 4 in attractiveness (above average), what is the chance that they are above average (3 or 4) in grooming as well. Conditional on being above average physically, what is the chance that they are average or below average (1 or 2) in grooming? Are these statistically significantly different?
The study also considers the attractiveness of someone’s personality (charisma), with the same 4-point scale. These data are: Personality 4 Very Attractive 3 Attractive 2 Average 1 Less Attractive 4 Very well groomed 326 171 60 26 3 Well groomed 416 1186 467 51 2 Average grooming 212 966 1729 136 1 Less than average grooming 11 49 184 84 Conditional on having an above-average personality, what is the chance that someone has above-average grooming? Conditional on having an above-average personality, what is the chance that their grooming is at or below average? Is there a statistically significant difference?
Comment on the study. If overall attractiveness is a combination of these 3 factors, is there evidence that they are gross substitutes or complements in production? PK Robins, JF Homer, MT French (2011). “Beauty and the Labor Market: Accounting for the Additional Effects of Personality and Grooming,” Labour, 25(2), pp 228-251.

The next two questions ask you to use the dataset, scf2010 data exam2. This is the Federal Reserve’s Survey of Consumer Finances (SCF), which is oddly weighted (so don’t worry that the averages seem a bit off! They oversample rich people since they have interesting financial portfolios) but has interesting data on the financial situation of households. 10. Consider a regression to explain household liquidity (the variable LIQ, which adds up checking account, saving account, and other liquid assets – very much what we might consider M1 in macro). Carefully designate the range of data that you will explore, then show and describe simple statistics, and only then create one (or more) interesting regressions. (Interesting regressions might include polynomial terms, interactions, quantiles, nonparametric, etc.) Explain why each variable is in the model and consider whether it is endogenous. Describe what you learn from the regression model. 11. With the same dataset, look at the ratio of monthly house debt payments to income (PIRMORT) and the ratio of consumer debt payments to income (PIRCONS). For housing debt, many financial planner suggest a ratio of higher than 33% is worrisome. Create a dummy 0/1 variable for whether a household has mortgage payments more than 33% of income. Show descriptive statistics. Then estimate interesting probit and/or logit regressions to explain this variation. (“Interesting” as defined in previous question.) Again explain your variable choices. Explain what you learn from these models.

I’ve combined the dataset on beer consumption and taxes with data* from Prof Nagler on car accidents, fatalities, and social capital (thanks to him for the generosity!). Now we have data on the fraction of population over 65, with high school or bachelor’s degree, fraction African American, rates of divorce and suicide, traffic fatalities (number and rate) overall and in summer, miles of road and fraction unpaved, population density and per road mile, gas price and gas stations per population, attendance at church and if the fraction who pray daily, and many responses to ‘values’ questions: is it OK to cheat on taxes, is life dull, are most people trusted or fair, etc. Use instrumental variables techniques to try some of these new variables as instruments and/or controls in interesting regression specification(s) about economic growth rates across US states. Again ensure that you designate the range of data and show simple statistics. Explain about the rationale for the instrument(s).
You might find it useful to make a sketch.

For a Normal Distribution with mean 1 and standard deviation 1.7, what is area in both tails farther from the mean than 1.9? A. 0.4013 B. 0.6915 C. 0.2630 D. 0.6171
For a Normal Distribution with mean 3 and standard deviation 4.7, what is area in both tails farther from the mean than -0.3? A. 0.7384 B. 0.9679 C. 0.2616 D. 0.4839
For a Normal Distribution with mean -7 and standard deviation 8.1, what is area in both tails farther from the mean than -17.5? A. 0. 2551 B. 0.1936 C. 0.7422 D. 0.3872
For a Normal Distribution with mean -4 and standard deviation 5.7, what is area in both tails farther from the mean than -16.0? A. 0.9821 B. 0.1587 C. 0.7586 D. 0.0357
For a Normal Distribution with mean -11 and standard deviation 5.7, what is area in both tails farther from the mean than -12.1? A. 0.5398 B. 0.8415 C. 0.5793 D. 0.9732
For a Normal Distribution with mean -10 and standard deviation 8.3, what is area in both tails farther from the mean than -8.3? A. 0.1141 B. 0.5793 C. 0.8859 D. 0.8415
For a Normal Distribution with mean 11 and standard deviation 0.8, what is area in both tails farther from the mean than 10.6? A. 0.4314 B. 0.1587 C. 0.6171 D. 0.5987
For a Normal Distribution with mean -3 and standard deviation 0.8, what is area in both tails farther from the mean than -4.0? A. 0.7422 B. 0.9032 C. 0.0001 D. 0.1936
For a Normal Distribution with mean -3 and standard deviation 1.0, what is area in both tails farther from the mean than -4.8? A. 0.9641 B. 0.1437 C. 0.8159 D. 0.0719
For a Normal Distribution with mean -15 and standard deviation 0.4, what is area in both tails farther from the mean than -15.7? A. 0.7317 B. 0.1783 C. 0.0891 D. 0.8023
For a Normal Distribution with mean 8 and standard deviation 0.8, what is area in both tails farther from the mean than 9.7? A. 0.7757 B. 0.1587 C. 0.1469 D. 0.0357
For a Normal Distribution with mean 1 and standard deviation 3.5, what values leave probability 0.469 in both tails? A. (-1.5344 3.5344) B. (0.7278 1.2722) C. (6.3837 -4.3837) D. (-0.4384 1.0098)
For a Normal Distribution with mean -8 and standard deviation 1.7, what values leave probability 0.114 in both tails? A. (-8.0494 -5.9506) B. (-10.6868 -5.3132) C. (-9.2673 -6.7327) D. (-6.2863 -3.1254)
For a Normal Distribution with mean -9 and standard deviation 9.0, what values leave probability 0.489 in both tails? A. (-9.2482 -8.7518) B. (-15.2271 -2.7729) C. (-0.6919 0.6919) D. (9.1268 -27.1268)
For a Normal Distribution with mean 3 and standard deviation 2.6, what values leave probability 0.301 in both tails? A. (1.6554 4.3446) B. (0.3108 5.6892) C. (3.6722 2.3278) D. (-1.0343 1.0343)
For a Normal Distribution with mean -5 and standard deviation 4.2, what values leave probability 0.254 in both tails? A. (-9.7909 -0.2091) B. (-1.1407 1.1407) C. (-4.9158 -5.0842) D. (-7.3954 -2.6046)
For a Normal Distribution with mean 13 and standard deviation 7.3, what values leave probability 0.244 in both tails? A. (-2.0048 15.0048) B. (12.7804 13.2196) C. (7.9375 18.0625) D. (4.4952 21.5048)
For a Normal Distribution with mean 1 and standard deviation 5.2, what values leave probability 0.116 in both tails? A. (-5.2152 7.2152) B. (-7.1733 9.1733) C. (-7.6733 8.6733) D. (-3.0866 5.0866)
For a Normal Distribution with mean -4 and standard deviation 9.9, what values leave probability 0.020 in both tails? A. (-2.3263 2.3263) B. (-21.3318 13.3318) C. (-27.0308 19.0308) D. (-24.3321 16.3321)
For a Normal Distribution with mean -11 and standard deviation 7.2, what values leave probability 0.054 in both tails? A. (-19.3732 8.3732) B. (-22.5722 0.5722) C. (-24.8732 2.8732) D. (-17.9366 -4.0634)
For a Normal Distribution with mean 12 and standard deviation 8.9, what values leave probability 0.167 in both tails? A. (5.8505 18.1495) B. (8.1828 15.8172) C. (-0.2990 24.2990) D. (-1.3819 1.3819)
For a Normal Distribution with mean 12 and standard deviation 0.3, what values leave probability 0.245 in both tails? A. (11.9925 12.0075) B. (38.8374 41.1626) C. (11.6512 12.3488) D. (14.8256 18.1744)

You might find it useful to make a sketch.

A regression coefficient is estimated to be equal to -5.968 with standard error 4.7; there are 4 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 1.7958 B. 0.5540 C. 0.2730 D. 0.7270
A regression coefficient is estimated to be equal to -9.880 with standard error 9.7; there are 21 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.6830 B. 0.3200 C. 0.8694 D. 0.6500
A regression coefficient is estimated to be equal to -10.781 with standard error 6.6; there are 20 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.1180 B. 0.0003 C. 0.2390 D. 0.9700
A regression coefficient is estimated to be equal to 4.456 with standard error 4.9; there are 12 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.0060 B. 0.2346 C. 0.6190 D. 0.3810
A regression coefficient is estimated to be equal to 6.696 with standard error 4.2; there are 31 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.1210 B. 1.8891 C. 0.8790 D. 0.6609
A regression coefficient is estimated to be equal to 32.530 with standard error 10.0; there are 7 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.9860 B. 0.0140 C. 0.0017 D. 0.6991
A regression coefficient is estimated to be equal to -6.463 with standard error 4.8; there are 5 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.0089 B. 1.8218 C. 0.2360 D. 0.8581
A regression coefficient is estimated to be equal to 7.335 with standard error 3.4; there are 19 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.5060 B. 0.0440 C. 0.9560 D. 0.1750
A regression coefficient is estimated to be equal to -6.564 with standard error 3.1; there are 7 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.0720 B. 0.9239 C. 0.9280 D. 0.1386
A regression coefficient is estimated to be equal to -8.619 with standard error 6.1; there are 25 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.1493 B. 0.1700 C. 0.9500 D. 0.8300
A regression coefficient is estimated to be equal to 7.966 with standard error 8.8; there are 17 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.3780 B. 0.6347 C. 0.8510 D. 0.2575
A regression coefficient is estimated to be equal to 10.254 with standard error 6.7; there are 26 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.1380 B. 0.8620 C. 0.0965 D. 0.9352
A regression coefficient has standard error 4.60; there are 14 degrees of freedom. The t-statistic is 1.3942. What is the coefficient? A. 0.1376 B. 0.3357 C. 6.4132 D. 0.4581
A regression coefficient has standard error 4.70; there are 6 degrees of freedom. The t-statistic is 1.7159. What is the coefficient? A. 0.3849 B. 6.8630 C. 8.0647 D. 1.9138
A regression coefficient has standard error 4.30; there are 21 degrees of freedom. The t-statistic is 1.1602. What is the coefficient? A. 0.3331 B. 4.9887 C. 8.7410 D. 0.1866
A regression coefficient has standard error 6.60; there are 24 degrees of freedom. The t-statistic is 1.5223. What is the coefficient? A. 1.8721 B. 0.4186 C. 10.0472 D. 7.8590
A regression coefficient has standard error 5.40; there are 27 degrees of freedom. The t-statistic is -1.6674. What is the coefficient? A. -9.0038 B. 0.0954 C. 0.8930 D. 1.0433
A regression coefficient has standard error 1.10; there are 5 degrees of freedom. The t-statistic is -3.1285. What is the coefficient? A. -3.4414 B. -2.0018 C. 0.9740 D. 0.6120
A regression coefficient is estimated to be equal to 16.483; there are 10 degrees of freedom. The t-statistic is 2.7937. What is the standard error? A. 0.0102 B. 5.9000 C. 46.0716 D. 1.6483
A regression coefficient is estimated to be equal to -2.806; there are 8 degrees of freedom. The t-statistic is -1.0394. What is the standard error? A. 0.2986 B. 0.7436 C. -0.3508 D. 2.7000
A regression coefficient is estimated to be equal to -11.059; there are 11 degrees of freedom. The t-statistic is -1.1641. What is the standard error? A. 12.9570 B. -1.0054 C. 9.5000 D. 0.8512
A regression coefficient is estimated to be equal to 13.066; there are 25 degrees of freedom. The t-statistic is 1.5743. What is the standard error? A. 8.3000 B. 20.5795 C. 1.8846 D. 0.0831
A regression coefficient is estimated to be equal to 15.167; there are 6 degrees of freedom. The t-statistic is 3.0333. What is the standard error? A. 5.0000 B. 0.9770 C. 0.0076 D. 1.9976
A regression coefficient is estimated to be equal to 6.136; there are 28 degrees of freedom. The t-statistic is 1.2271. What is the standard error? A. 0.2191 B. 0.7700 C. 5.0000 D. 7.6024

Subsequent questions will use the BRFSS dataset (I know, all the variable labels are ALL CAPS like some elderly aol user typed them, sorry), the Behavioral Risk Factor Surveillance Study. There are many observations on a wide variety of risky behaviors: smoking, drinking, poor eating, flu shots, whether household has a 3-day supply of food and water… I put a dataset for the exam on Blackboard.

For now concentrate on BMI, which is Body Mass Index, a measure of whether a person is overweight (2500-2999) or obese (3000 or above). Note there are 2 implied decimals so that 2490 is 24.90, the upper limit for ‘normal weight’. And note that values of 9999 are missing so make sure to omit those.

What are descriptive statistics for BMI? Does this suggest some way you ought to limit the sample? Construct a hypothesis test for whether there is no statistically significant difference between men and women (you can limit to specific ages or by other variables if you wish). What is the standard error of the difference? What is the test statistic? What is the p-value? What is the chance of selecting a female who is underweight (BMI<1850)? What is the chance of selecting a male who is underweight? What is the chance of selecting a person who is underweight, of the females (i.e. conditional on being female)? What is the chance of selecting a person who is underweight, of the males?
What explanatory variables could be in the model? Which are available in the BRFSS data? Construct at least one simple regression model; discuss the estimates (including statistical significance but also relevance and whether the estimates accord with theory). Should income be in the model? Explain whether income and BMI might be endogenous.
Construct another reqression model with more complicated interactions. How could you improve the model? Consider nonlinear age terms, gender-age interactions, race-age interactions, state dummies and more. Note that INCOME2 is not a continuous variable but you would need to create dummy variables for the relevant income levels.
You might find it useful to sketch the distributions.

For a Normal Distribution with mean 1 and standard deviation 9.6, what is area in both tails farther from the mean than 23.1? A. 0.1251 B. 0.0214 C. 0.4585 D. 0.9893
For a Normal Distribution with mean 5 and standard deviation 7.6, what is area in both tails farther from the mean than 14.1? A. 0.2743 B. 0.1587 C. 0.2301 D. 0.4603
For a Normal Distribution with mean -2 and standard deviation 3.8, what is area in both tails farther from the mean than 2.9? A. 0.7007 B. 0.1936 C. 0.3872 D. 0.2578
For a Normal Distribution with mean -7 and standard deviation 5.1, what is area in both tails farther from the mean than -1.9? A. 0.3173 B. 0.0849 C. 0.6346 D. 0.9151
For a Normal Distribution with mean 13 and standard deviation 3.5, what is area in both tails farther from the mean than 7.8? A. 0.2672 B. 0.1336 C. 0.1587 D. 0.7734
For a Normal Distribution with mean -12 and standard deviation 9.6, what values leave probability 0.003 in both tails? A. (-2.9677, 2.9677) B. (-38.3787, 14.3787) C. (-36.1166, 12.1166) D. (-40.4903, 16.4903)
For a Normal Distribution with mean -2 and standard deviation 9.1, what values leave probability 0.092 in both tails? A. (-1.6849, 1.6849) B. (-17.3330, 13.3330) C. (-1.9047, 1.4652) D. (-16.3950, 14.3950)
For a Normal Distribution with mean 0 and standard deviation 4.0, what values leave probability 0.039 in both tails? A. (-5.6746, 5.6746) B. (-7.0496, 7.0496) C. (-2.0642, 2.0642) D. (-8.2567, 8.2567)
A regression coefficient is estimated to be equal to -12.684 with standard error 9.4; there are 16 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.1960 B. 0.4810 C. 1.8228 D. 0.9323
A regression coefficient is estimated to be equal to 10.030 with standard error 4.0; there are 5 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.2030 B. 0.3354 C. 0.0300 D. 0.0540
A regression coefficient is estimated to be equal to 0.559 with standard error 0.2; there are 3 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.6797 B. 0.9320 C. 0.0680 D. 0.7121
A regression coefficient is estimated to be equal to -3.564 with standard error 1.9; there are 22 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.2100 B. 0.0740 C. 0.8950 D. 0.9393

A Japanese study looked at the effect of “kawaii” (cute) on test performance. (I am inferring numbers from the graphs shown so these might not exactly match the study!) There were 24 subjects; half were shown pictures of cute baby animals and half were shown pictures of full-grown animals (not cute). The ability to complete tasks, for those shown cute pictures, changed by 4, with a standard deviation of 3.5. The ability to complete tasks, for the control group, changed by 0.5 with a standard deviation of 2.9. What is the difference in means? What is the standard error of the difference? What is the normalized value for the difference? What are the degrees of freedom? What is the p-value? Is this difference statistically significant? Comment on the study. Should you immediately Google images of cute baby animals to help the rest of your exam performance? Nittono H, Fukushima M, Yano A, Moriya H (2012) The Power of Kawaii: Viewing Cute Images Promotes a Careful Behavior and Narrows Attentional Focus. PLoS ONE 7(9): e46362.
Consider the BRFSS dataset, the Behavioral Risk Factor Surveillance Study. There are many observations on a wide variety of risky behaviors: smoking, drinking, poor eating, flu shots, whether household has a 3-day supply of food and water… For now concentrate on BMI, which is Body Mass Index, a measure of whether a person is obese. Construct at least one good model to explain BMI. What are descriptive statistics for BMI? Does this suggest some way you ought to limit the sample? What explanatory variables could be in the model? Which are available in the BRFSS data? Construct at least one good regression model; discuss the estimates (including statistical significance but also relevance and whether the estimates accord with theory). How could you improve the model? Consider nonlinear age terms, gender-age interactions, race-age interactions, state dummies and more. Note that INCOME2 is not a continuous variable but you would need to create dummy variables for the relevant income levels.
The BRFSS data includes information on household disaster preparedness. Two measures are whether the household has a 3-day supply of food and water. We might believe these to be similar so want to examine the marginal probabilities. Count Water– 3 day supply = “Yes” = “No”
Food – 3 day supply = “Yes” 5528 3190
= “No” 484 1130

What is the probability that a household has sufficient food and water for 3 days? Given that a household has sufficient food, what is the probability that it has sufficient water as well? If a household does not have sufficient food, what is the probability that it has sufficient water? Are the latter two proportions statistically significantly different?

Continuing with the BRFSS, examine the 0/1 dependent variable of whether the person ever smoked seriously (which they define as at least 100 cigarettes); the variable is SMOKE100. We are interested in de-tangling the effects of both income and education. (Recall note from previous question about INCOM2 variable.) Estimate both probit and logit models; explain different predictions from each model. Explore various specifications possibly including interactions, dummies, etc. Carefully explain the results that you find.
You might find it useful to sketch these.

For a Normal Distribution with mean 1 and standard deviation 9.6, what is area to the right of 23.1? A. 0.1251 B. 0.0107 C. 0.4585 D. 0.9893
For a Normal Distribution with mean 8 and standard deviation 4.9, what is area to the left of 6.5? A. 0.5596 B. 0.7642 C. 0.3821 D. 0.1587
For a Normal Distribution with mean 4 and standard deviation 7.1, what is area in both tails farther from the mean than 13.2? A. 0.1936 B. 0.3872 C. 0.2866 D. 0.1587
For a Normal Distribution with mean -11 and standard deviation 5.0, what is area in both tails farther from the mean than 0.5? A. 0.1251 B. 0.1587 C. 0.0429 D. 0.0214
For a Normal Distribution with mean 13 and standard deviation 3.5, what value leaves probability 0.197 in the left tail? A. 12.0588 B. 10.0166 C. 0.8030 D. 15.9834
For a Normal Distribution with mean -10 and standard deviation 2.6, what values leave probability 0.146 in both tails? A. (-1.4538, 1.4538) B. (-5.3000, -2.3923) C. (-13.7799, -6.2201) D. (-8.7799, -1.2201)
For a Normal Distribution with mean 12 and standard deviation 9.8, what values leave probability 0.220 in both tails? A. (-1.2265, 1.2265) B. (10.5205, 13.4795) C. (-0.0200, 24.0200) D. (4.4325, 19.5675)
A regression coefficient is estimated to be equal to -5.941 with standard error 3.9; there are 9 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.1620 B. 0.8491 C. 0.1080 D. 0.3380
A regression coefficient is estimated to be equal to -10.249 with standard error 3.5; there are 26 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.0966 B. 0.9930 C. 0.0070 D. 0.9999
A regression coefficient is estimated to be equal to 5.563 with standard error 3.0; there are 24 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.9240 B. 0.0363 C. 0.1710 D. 0.0760

Using data from the NHANES study, we find the following numbers of people classified as whether they report themselves to be overweight (person_overweight) and if they have ever tried marijuana (tried_pot). person is not overweight person is overweight has not tried marijuana 613 837 has tried marijuana 848 956

What fraction of people who are overweight have tried marijuana? What fraction of people who are not overweight have tried marijuana?
Are these statistically significantly different? What is the p-value of a hypothesis test for a difference in the means?
Does this data provide evidence that smoking marijuana helps people not be overweight? Discuss.

A study of Quantitative Easing in Japan (Kobayashi, Spiegel, and Yamori 2006) looked at the stock prices of particularly indebted firms to see if they were disproportionately impacted by the Bank of Japan’s easy money policy. They report the following (49 Observations and R2=0.08): Estimate T-Statistic P-Value
Capital -0.03 -1.879
Liquidity -0.51 -0.929
Bad Loan -0.00 -0.066
Where the “Capital” is the capital-asset ratio of the firm’s main bank; “Liquidity” is its main bank’s ratio of cash, reserves, and loan balances to assets; and “Bad Loan” is the ratio of nonperforming loans to total assets.

What are the p-values for each T-statistic?
Which regressors are statistically significant?
What does this imply?

Using the PUMS data for people in NYC, people are classified if they report that they speak English well or very well versus those who do not speak it well or at all. Among the 38,740 households who speak English well, 18.29% have children under 6 years old; for the 14,688 that do not speak English well, 14.46% have children under 6. Also, 38.72% of those who speak English well have children under 17; 35.27% of those who do not speak English well have children under 17.

Are these differences statistically significant?
What is the p-value for each difference in means?
Why do you think we might see this difference?

Use the PUMS data for people in NYC (download from Blackboard or InYourClass; pums_NYC_2.zip) examine people’s choice of rent or own, as well as how much to pay (the variable “own_rent_frac” gives the fraction of household income that goes to costs of either owning or renting).

What fraction of households own their apartment/house/dwelling? What fraction rent? What are some of the important factors that explain this difference?
Estimate a limited-dependent variable model to explain the choice to rent or own. What variables should be in this regression? Why might we believe that the “own_rent_frac” variable would be endogenous with the own/rent choice? What variables are statistically significant in this choice? Have you omitted any important variables? What are the predicted probabilities for different representative people? Discuss.
Estimate a linear regression to explain the fraction of income going to ownership or rental costs. What variables should be in this regression? What variables are statistically significant? Have you omitted any important variables? Discuss.

Answer each question; you might find it useful to make a sketch.

A regression coefficient is estimated to be equal to -1.417 with standard error 1.6; there are 30 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.9249 B. 0.7998 C. 0.6240 D. 0.3830
A regression coefficient is estimated to be equal to -15.901 with standard error 7.1; there are 3 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.1110 B. 0.9749 C. 0.0001 D. 0.9065
A regression coefficient is estimated to be equal to -16.558 with standard error 7.1; there are 8 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.9520 B. 0.1280 C. 0.0002 D. 0.0480
A regression coefficient is estimated to be equal to -0.322 with standard error 0.3; there are 29 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 1.0000 B. 0.8378 C. 0.2920 D. 0.7080
A regression coefficient has standard error 7.50; there are 9 degrees of freedom. The t-statistic is 1.3730. What is the coefficient? A. 1.8302 B. 10.2972 C. 1.1441 D. 0.8282
A regression coefficient has standard error 2.00; there are 9 degrees of freedom. The t-statistic is -1.2381. What is the coefficient? A. -2.4762 B. 0.7321 C. 0.7530 D. -0.5836
A regression coefficient has standard error 3.40; there are 22 degrees of freedom. The t-statistic is 2.0265. What is the coefficient? A. 1.9573 B. 0.3132 C. 6.8903 D. 0.6075
A regression coefficient is estimated to be equal to -14.943; there are 4 degrees of freedom. The t-statistic is -1.7176. What is the standard error? A. 0.0411 B. 0.8390 C. 8.7000 D. 0.9358
A regression coefficient is estimated to be equal to -8.636; there are 4 degrees of freedom. The t-statistic is -2.1590. What is the standard error? A. 0.9473 B. 18.6638 C. 0.0150 D. 4.0000
A regression coefficient is estimated to be equal to 7.693; there are 16 degrees of freedom. The t-statistic is 1.5699. What is the standard error? A. 1.8836 B. 4.9000 C. 0.1057 D. 0.4808

Suppose a student is answering 50 multiple-choice questions on an exam where each question has 4 choices.

If the student guesses randomly, what is the expected number of correct answers? If the questions are worth 2 points each, what is the expected score for a student who is completely ignorant?
If a student guesses randomly, what is the standard error of the fraction guessed correctly?
What is a 95% confidence interval for scores of students who guess randomly?
If a student scores 27 points, what is the probability that the student was guessing randomly?

Peter Gordon, in his talk at CCNY, presented results from linear regressions to explain the growth of metropolitan areas. He begins with a simple model to explain population growth from 1990-2000: Log Population Growth 1990-2000 Coefficient t-stat p-value Constant term -0.0229 -0.12
Population in 1990 (log) 0.0192 1.33
Pop. Density in 1990 -0.0504 -1.65
% in manufacturing -0.0028 -1.63
R2 0.57
Where he also includes dummy variables for Census Regions (New England, Mid Atlantic, etc.). There are 79 observations and 67 degrees of freedom.

What are the p-values for the 3 coefficients? Are they significant? The averages and standard deviations are: Average Standard deviation Population in 1990 (log) 14.52 14.89 Pop. Density in 1990 1.80 1.02 % in manufacturing 18.69 7.75
What is the predicted population growth for a metropolitan area that is exactly average?
What is the predicted population growth for a metro area that is one standard deviation above average in 1990 population? For a metro area one standard deviation above average in density? In manufacturing concentration?
Give a careful explanation for why we would observe coefficients of these signs.

Since several groups decided that they wanted to look at changing patterns of interracial marriage, I created a dataset from the CPS 2009, that has information for each spouse (with A_FAMREL=2; note the spouse might be male or female) as well as the race, ethnicity, and citizenship of their partner. This data, cps_2009_spouseinfo_famrel2, is on Blackboard. Estimate a linear regression to explain the person’s wage as a function of age. Carefully explain which variables you choose to include in the estimation and why they might be important. Show the regression results and note which coefficient estimates are statistically significant. What results are surprising to you? Which results are not surprising? You might estimate and compare several different models.
Please answer the following questions on Blackboard. It might help to make sketches.

For a Normal Distribution with mean 9 and standard deviation 9.1, what is area in both tails farther from the mean than -8.3? A. 0.8387 B. 0.0574 C. 0.1587 D. 0.9713
For a Normal Distribution with mean 1 and standard deviation 9.6, what is area in both tails farther from the mean than 23.1? A. 0.1251 B. 0.0214 C. 0.4585 D. 0.9893
For a Normal Distribution with mean 5 and standard deviation 7.6, what is area in both tails farther from the mean than 14.1? A. 0.2743 B. 0.1587 C. 0.2301 D. 0.4603
For a Normal Distribution with mean -14 and standard deviation 2.8, what is area in both tails farther from the mean than -20.4? A. 0.0214 B. 0.8235 C. 0.0429 D. 0.0971
For a Normal Distribution with mean -2 and standard deviation 3.8, what values leave probability 0.382 in both tails? A. (-3.6610, -0.3390) B. (-5.3220, 1.3220) C. (0.7331, -4.7331) D. (-3.1409, -0.8591)
For a Normal Distribution with mean 13 and standard deviation 3.5, what values leave probability 0.099 in both tails? A. (10.0292, 15.9708) B. (7.2260, 18.7740)C. (0.5785, 12.2990) D. (8.4946, 17.5054)

Please answer the following questions on Blackboard.

A regression coefficient is estimated to be equal to 10.527 with standard error 8.3; there are 19 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.2200 B. 1.7953 C. 0.1323 D. 0.7800
A regression coefficient is estimated to be equal to 0.521 with standard error 0.5; there are 17 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.6380 B. 1.7026 C. 0.3120 D. 0.6988
A regression coefficient is estimated to be equal to 2.885 with standard error 1.7; there are 19 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.2423 B. 0.2810 C. 0.9980 D. 0.1060
A regression coefficient is estimated to be equal to 1.902 with standard error 1.5; there are 26 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.4410 B. 0.3363 C. 0.9714 D. 0.2160
A regression coefficient is estimated to be equal to -7.981 with standard error 4.3; there are 28 degrees of freedom. What is the p-value (from the t-statistic) against the null hypothesis of zero? A. 0.2131 B. 0.1630 C. 0.9547 D. 0.9260
A regression coefficient has standard error 5.10; there are 14 degrees of freedom. The t-statistic is -2.3978. What is the coefficient? A. -0.8735 B. 0.9730 C. -0.4571 D. -12.2287
A regression coefficient has standard error 9.40; there are 16 degrees of freedom. The t-statistic is -1.3494. What is the coefficient? A. -12.6842 B. -0.0543 C. 0.1772 D. 0.8861
A regression coefficient is estimated to be equal to 4.873; there are 12 degrees of freedom. The t-statistic is 0.9093. What is the standard error? A. 5.3588 B. 0.2279 C. 0.4061 D. 0.7702
A regression coefficient is estimated to be equal to -4.488; there are 7 degrees of freedom. The t-statistic is -1.8992. What is the standard error? A. 0.8650 B. -0.6411 C. 0.0232 D. 2.3631

An economic study modeled the log price of gasoline (each week) as a simple function of the log futures price of the week. In the futures market a firm can buy gasoline at a price set today, to be delivered next month, so we anticipate that this should have some predictive power. The model is , where Pt is this week’s price of gasoline (in natural logs) and Ft-1 is the price from the week before (also in natural logs) in the futures market. When I estimate this relationship (slightly different data so no use looking online!) I get the following relationship: , where the standard error of the intercept coefficient is 0.006, the standard error of the slope coefficient is 0.008, and the R2=0.96. (M. Chinn, “Gasoline Prices Implied by Futures,” 2012)

Is the slope coefficient statistically significant? What is its t-statistic? P-value? What is the null hypothesis and what can you conclude?
The most recent price of gas was $4.02. The futures price declined from$2.97 to $2.94 (this is the price at the harbor, some of the difference is transportation cost). What does this recent change imply for the price of gasoline next week? (Careful! Disentangle the logs!)
A regression of the log price of diesel fuel on the log of last week’s futures price of gasoline gives the result, where the standard error of the intercept coefficient is 0.013, the standard error of the slope coefficient is 0.017, and the R2=0.81. How would you evaluate the predictive power of gasoline futures for diesel prices (note that there are no futures traded for diesel)?

An experiment in a bar recorded the amount of alcohol ordered, depending whether the waitress put her hand on the shoulder of the first patron while asking if he wanted a drink. There were 24 pairs of men drinking together; 12 were touched and 12 were not. When the waitress touched one, the average amount drunk by the man touched was 36 oz (standard deviation is 6) and the other man in the group drank 29.6 oz (std. dev. 5). When the waitress did not touch either man, they drank 30.6 (std. dev. 5) and 26.6 oz (std. dev. 5). The units are beer-equivalent ounces. [The standard errors were not explicitly reported in the study, I’m making them up.] (Kaufman & Mahoney, J Social Psychology, 1999)

There are two slightly different effects: what is the effect of touching a patron on the shoulder, and what is the effect on the second man who sees his drinking buddy get a touch on the shoulder? What is the standard error of the difference in consumption of the first man in each group (who was actually touched)? What is the t-stat for the null hypothesis test that there was no difference in drinking quantity? P-value? Should the alternative hypothesis be one-sided or two-sided?
What is the standard error of the difference in consumption of the second man in each group? What are the t-stat and p-value? Should the alternative hypothesis be one-sided or two-sided? Explain the null hypothesis and your conclusions.

A study examined whether people were more likely to lie by email or in person, finding in a study of business students that 24 out of 26 lied when writing an email while 14 out of 22 lied when writing on paper. (The lie was in the context of a fictitious business transaction.) (Naquin, Kurtzberg, Belkin 2010 J Applied Psychology)

What is the standard error of the fraction of people lying by email? What is the standard error of the fraction of people lying on paper?
What are the t-stat and p-value for the null hypothesis that the medium of communication (email or paper) had no effect on the probability of lying?

We use the most recent data to assess the relation between changes in GDP and changes in the unemployment rate (so-called Okun’s Law), comparing the relation in the entire period since 1948 with the relation in the period since 1990. Data are from FRED Stats. A regression has the dependent variable as the quarterly change in the unemployment rate (denotedUR). The independent variable is the quarterly percent growth rate of nominal GDP (denoted %Y). The estimated regression is .

Using data for the entire period, 1948-2012, the estimated equation is UR = 0.37 – 0.22%Y, where the standard error of the intercept is 0.03, the standard error of the slope is 0.02, and the R2 is 0.39. Is the slope coefficient statistically significant? What is its t-statistic? P-value?
Using data for the period 1990-2012, the estimated equation is UR = 0.38 – 0.30%Y, where the standard error of the intercept is 0.05, the standard error of the slope is 0.04, and the R2 is 0.45. Is the slope coefficient statistically significant? What is its t-statistic? P-value?
Compare the two regressions. What are the arguments in favor of using the whole sample versus only more recent data?
For the most recent data (first quarter of 2012), GDP growth was 0.93 while UR was -0.4. What was the predicted value from each model for that time? How would you interpret this?

Use the data on Blackboard, CPS_finalexam. We want to compare the acquisition of wage income, as a function of education, between native-born and immigrant workers. Examine a regression with wage as the dependent variable, and with age, age-squared, educational attainment dummies (choose which are appropriate), and other reasonable independent variables (as many as you think are appropriate).

Estimate two such regressions: one for natives and one for immigrants (explain what subgroup you’re examining – all people, workers, fulltime workers?).
What are the effects of a college degree for the two groups? (Carefully explain the null hypothesis, t-stat, and p-values before explaining the results of the hypothesis test.)
Compare the age-wage profiles for natives and immigrants.
If you were to estimate two additional regressions but with the logarithm of wages as dependent variable, how would your conclusions change? Explain the hypothesis tests.
How would industry/occupation dummy variables change these results?
What other models can you estimate that could be informative – explain.

You might sketch a picture.

For a Normal Distribution with mean 4 and standard deviation of 1, what is the area to the left of 3.3? 0.484 0.758 0.242 0.363
For a Normal Distribution with mean -13 and standard deviation of 7, what is the area to the left of -3.2? 0.162 0.081 0.919 0.758
For a Normal Distribution with mean 1 and standard deviation of 4, what is the area to the right of -6.6? 0.829 0.029 0.971 0.057
For a Normal Distribution with mean -6 and standard deviation of 2, what is the area to the right of -9.8? 0.057 0.829 0.029 0.971
For a Normal Distribution with mean -3 and standard deviation of 5, what is the area to the right of -8? 0.691 0.317 0.841 0.159
For a Normal Distribution with mean -12 and standard deviation of 5, what is the area in both tails farther from the mean (in absolute value) than -21.5? 0.057 0.029 0.971 0.351
For a Normal Distribution with mean -9 and standard deviation of 5, what is the area in both tails farther from the mean (in absolute value) than -10? 0.579 0.421 0.841 0.087
For a Normal Distribution with mean -13 and standard deviation of 8 what value leaves 0.22 in the right tail? -3.188 -3.607 -8.303 -11.792
For a Normal Distribution with mean -7 and standard deviation of 5 what value leaves 0.24 in the right tail? -4.026 -6.749 -1.052 -1.125
For a Normal Distribution with mean 12 and standard deviation of 2 what value leaves 0.03 in the right tail? 15.110 16.340 13.024 14.048

You might sketch a picture.

For a t Distribution with sample average of 1.43, standard error of 1.22, and 11 observations, what is the area in both tails, for a null hypothesis of zero mean? 0.133 0.181 0.412 0.266
For a t Distribution with sample average of 2.9, standard error of 1.82, and 13 observations, what is the area in both tails, for a null hypothesis of zero mean? 0.068 0.541 0.012 0.135
For a t Distribution with sample average of 3.31, standard error of 2.16, and 9 observations, what is the area in both tails, for a null hypothesis of zero mean? 0.009 0.160 0.530 0.080
For a t Distribution with sample average of 1.47, standard error of 1.47, and 16 observations, what is the area in both tails, for a null hypothesis of zero mean? 0.332 0.166 0.332 0.161
For a t Distribution with 20 observations and standard error of 2.53, what sample mean leaves 0.08 in the two tails, when testing a null hypothesis of zero? 0.922 1.844 3.689 4.666
For a t Distribution with 5 observations and standard error of 2.78, what sample mean leaves 0.2 in the two tails, when testing a null hypothesis of zero? 0.738 1.476 4.103 2.952
For a t Distribution with 20 observations and standard error of 0.53, what sample mean leaves 0.24 in the two tails, when testing a null hypothesis of zero? 1.211 0.606 0.642 2.422
Sample A has mean 4.28, standard error of 0.21, and 4 observations. Sample B has mean 4.99, standard deviation of 0.33, and 23 observations. Test the null hypothesis of no difference. 0.005 0.002 0.906 0.517
Sample A has mean 1.6, standard error of 0.68, and 9 observations. Sample B has mean 4.83, standard deviation of 2.81, and 9 observations. Test the null hypothesis of no difference. 0.360 0.009 0.010 0.004

You are given the following data on the number of people in the PUMS sample who live in each of the five boroughs of NYC and who commute in each specified manner (where ‘other’ includes walking, working from home, taking a taxi or ferry or rail). Bronx Manhattan Staten Is Brooklyn Queens car 5788 2692 5526 10990 16905 bus 3132 2789 1871 4731 4636 subway 6481 13260 279 18951 14025 other 2748 10327 900 6587 4877

Find the Joint Probability for drawing, from this sample, a person from Queens who commutes by bus. Find the Joint Probability of a person from the Bronx who commutes by subway.
Find the Marginal Probability of drawing, from among the people who commute by subway, someone who lives in Brooklyn. Find the Marginal Probability, of people who commute by bus, someone who lives in the Bronx.
Find the Marginal Probability of drawing, from among the people who live in Staten Island, someone who drives a car to work. Find the Marginal Probability, of people in Brooklyn, who commute by subway.
Are these two choices (which borough to live in, how to commute) independent? Explain using the definition of statistical independence.

To investigate an hypothesis proposed by a student, I got data, for 102 of the world’s major countries, on the fraction of the population who are religious as well as the income per capita and the enrollment rate of boys and girls in primary school. The hypothesis to be investigated is whether more religious societies tend to hold back women. I ran two separate models: Model 1 uses girls enrollment rate as the dependent; Model 2 uses the ratio of girls to boys enrollment rates as the dependent. The results are below (standard errors in italics and parentheses below each coefficient): Model 1 Model 2 t-stat p-value Intercept 137 1.12
1. (0.09)
  Religiosity -0.585 -0.0018
  (0.189) (0.0009)
  GDP per capita 0.00056 0.0000016
  (0.00015) (0.0000007)

Which coefficient estimates are statistically significant? What are the t-statistics and p-values for each?
How would you interpret these results?
Critique the regression model. How would you improve it?

Download the data, “PUMA_nyc_for_exam” from Blackboard, which gives PUMA data on people living in the 5 boroughs. Run a regression that models the variable, “GRPIP,” “Gross Rent as Percent of Income,” which tells how burdensome are housing costs for different people.

What are the mean, median, 25th, and 75th percentiles for Rent as a fraction of income? Does this seem reasonable?
What is the fraction spent on rent by households in Brooklyn? In Queens? Is the difference statistically significant? Between Brooklyn and the Bronx?
What variables might be important in explaining this ratio? Find summary statistics for these variables.
Run a regression and interpret the output. Which variables are statistically significant? How do you interpret their coefficients? Are these reasonable?
What variables are omitted? How could the regression be improved (using actual real data)? Can you estimate a better model (with squared terms, interaction terms, etc)?

A random variable is distributed as a standard normal. (You are encouraged to sketch the PDF in each case.)

What is the probability that we could observe a value as far or farther than 1.3?
What is the probability that we could observe a value nearer than 1.8?
What value would leave 10% of the probability in the right-hand tail?
What value would leave 25% in both the tails (together)?

Using the CPS 2010 data (on Blackboard, although you don’t need to download it for this), restricting attention to only those reporting a non-zero wage and salary, the following regression output is obtained for a regression (including industry, occupation, and state fixed effects) with wage and salary as the dependent variable.

Fill in the missing values in the table.
The dummy variables for veterans have been split into various time periods to distinguish recent veterans from those who served decades ago. If you knew that the draft ended at about the same time as the Vietnam war, how would that affect your interpretation of the coefficient estimates?
Critique the regression: how would you improve the estimates (using the same dataset)? ANOVAb Model Sum of Squares df Mean Square F Sig. 1 Regression 8.201E+13 152 5.395E+11 324.098 .000a Residual 1.639E+14 98479 1.665E+09
Total 2.460E+14 98631

Coefficientsa

Model Unstandardized Coefficients Standardized Coefficients t Sig.
B Std. Error Beta
1 (Constant) 12970.923 2290.740 5.662 .000
Demographics, Age 2210.038 62.066 .605 ____ ____
Age squared -21.527 .693 -.504 ____ ____
Female -14892.950 ____ -.149 -47.872 .000
African American -3488.065 ____ -.022 -7.809 .000
Asian -2700.032 ____ -.012 -2.782 .005
Native American Indian or Alaskan or Hawaiian ____ 824.886 -.009 -3.442 .001
Hispanic ____ 483.313 -.024 -6.847 .000
Immigrant ____ 632.573 -.032 -6.728 .000
1 or more parents were immigrants 989.451 541.866 .008 ____ ____
immig_india -456.482 1675.840 -.001 ____ ____
immig_SEAsia 821.730 1252.853 .003 ____ ____
immig_MidE -599.852 2335.868 -.001 ____ ____
immig_China 3425.017 1821.204 .006 ____ ____
Education: High School Diploma 2786.569 492.533 .025 5.658 .000
Education: Some College but no degree 5243.544 528.563 .042 9.920 .000
Education: Associate in vocational 6530.542 762.525 .028 8.564 .000
Education: Associate in academic 7205.474 736.838 .032 9.779 .000
Education: 4-yr degree 17766.941 576.905 .143 30.797 .000
Education: Advanced Degree 36755.485 703.658 .227 52.235 .000
Married 4203.602 414.288 .042 10.147 .000
Divorced or Widowed or Separated 830.032 501.026 .006 1.657 .098
kids_under18 3562.643 327.103 .036 10.891 .000
kids_under6 -721.123 404.818 -.006 -1.781 .075
Union member 4868.240 976.338 .013 4.986 .000
Veteran since Sept 2001 2081.909 4336.647 .001 .480 .631
Veteran Aug 1990 - Aug 2001 -1200.688 1788.034 -.002 -.672 .502
Veteran May 1975-July 1990 -1078.953 1895.197 -.001 -.569 .569
Veteran August 1964-April 1975 -6377.461 3195.784 -.005 -1.996 .046
Veteran Feb 1955-July 1964 -7836.420 4904.511 -.004 -1.598 .110
Veteran July 1950-Jan 1955 -19976.382 10570.869 -.005 -1.890 .059
Veteran before 1950 -15822.026 12943.766 -.003 -1.222 .222

Using the NHANES 2007-09 data (on Blackboard, although you only need to download it for the very last part), reporting a variety of socioeconomic variables as well as behavior choices such as the number of sexual partners reported (number_partners), we want to see if richer people have more sex than poor people. The following table is constructed, showing three categories of family income and 5 categories of number of sex partners: number of sex partners
family income zero 1 2 - 5 6 - 25 >25 Marginal: < 20,000 11 63 236 255 92 ______ 20 - 45,000 7 117 323 308 117 ______ > 45,000 3 234 517 607 218 ______ Marginal: ______ ______ ______ ______ ______

Where is the median, for number of sex partners, for poorer people? For middle-income people? For richer people?
Conditional on a person being poorer, what is the likelihood that they report fewer than 6 partners? Conditional on being middle-income? Richer?
Conditional on reporting 2-5 sex partners, what is the likelihood that a person is poorer? Middle-income? Richer?
Explain why the average number of sex partners might not be as useful a measure as, for example, the data ranges above or the median or the 95%-trimmed mean.
(5 points) (You will need to download the data for this part) Could the difference be explained by schooling effects? How does college affect the number of sex partners?

I provide a dataset online (stock_indexes.sav on InYourClass) with the S&P 500 stock index and its daily returns as well as the NASDAQ index and its returns, from January 1, 1980 to December 9, 2010.

What is the mean and standard deviation?
If the stock index returns were distributed normally, what value of return is low enough, that 95% of the days are better?
What is the 5% value of the actual returns (the fifth percentile, use “AnalyzeStatistics” and check “Percentiles” in “Options”)? Is this different from your previous answer? What does that imply? Explain.

Using the CPS 2010 data online, examine whether children are covered by Medicaid or other insurance plan. Run a crosstab on “CH_HI” whether a child has health insurance, and “CH_MC” if a child is covered by Medicaid.

What fraction of children are covered by Medicaid? What fraction of children are not covered by any policy?
What is the average family income of children who are covered by Medicaid? Of children who are not? What is the t-statistic and p-value for a statistical test of whether the means are equal?

The oil and gas price dataset online, (oil_gas_prices.sav on InYourClass, although you only need to download it for the very last part), has data on prices of oil, gasoline, and heating oil (futures prices, in this case). Compare two regression specifications of the current price of gasoline. Specification A explains the current price with its price the day before. Specification B has the price of gas on the day before but also includes the prices of crude oil and heating oil on the day before. The estimates of the coefficient on gasoline are shown below: Coefficient estimate Standard error Specification A 0.021 0.028 Specification B 0.153 0.048

Calculate t-statistics and p-values for each specification of the regression.
Explain what you could learn from each of these regressions – specifically, would it be a good idea to invest in gasoline futures?
Explain why there is a difference in the estimated coefficients. Can you say that one is more correct?

A random variable is distributed as a standard normal. (You are encouraged to sketch the PDF in each case.)

What is the probability that we could observe a value as far or farther than -0.9?
What is the probability that we could observe a value nearer than 1.4?
What value would leave 5% of the probability in the right-hand tail?
What value would leave 5% in both the tails (together)?

[this question was given in advance for students to prepare with their group} Download (from Blackboard) and prepare the dataset on the 2004 Survey of Consumer Finances from the Federal Reserve. Estimate the probability that each head of household (restrict to only heads of household!) has at least one credit card. Write up a report that explains your results (you might compare different specifications, you might consider different sets of socioeconomic variables, different interactions, different polynomials, different sets of fixed effects, etc.).
Explain in greater detail your topic for the final project. Include details about the dataset which you will use and the regressions that you will estimate. Cite at least one previous study which has been done on that topic (published in a refereed journal).
You want to examine the impact of higher crude oil prices on American driving habits during the past oil price spike. A regression of US gasoline purchases on the price of crude oil as well as oil futures gives the coefficients below. Critique the regression and explain whether the necessary basic assumptions hold. Interpret each coefficient; explain its meaning and significance. Coefficients(a)

Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta
1 (Constant) .252 .167 1.507 .134 return on crude futures, 1 month ahead .961 .099 .961 9.706 .000 return on crude futures, 2 months ahead -.172 .369 -.159 -.466 .642 return on crude futures, 3 months ahead .578 .668 .509 .864 .389 return on crude futures, 4 months ahead -.397 .403 -.333 -.986 .326 US gasoline consumption -.178 .117 -.036 -1.515 .132 Spot Price Crude Oil Cushing, OK WTI FOB (Dollars per Barrel) 4.23E-005 .000 .042 1.771 .079 a Dependent Variable: return on crude spot price

You estimate the following coefficients for a regression explaining log individual incomes: Coefficients(a)

Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta B Std. Error 1 (Constant) 6.197 .026 239.273 .000 Demographics, Age .154 .001 1.769 114.120 .000 agesq -.002 .000 -1.594 -107.860 .000 female -.438 .017 -.184 -25.670 .000 afam -.006 .010 -.002 -.590 .555 asian -.011 .015 -.002 -.713 .476 Amindian -.063 .018 -.009 -3.573 .000 Hispanic .053 .010 .016 5.139 .000 ed_hs .597 .014 .226 43.251 .000 ed_smcol .710 .014 .272 50.150 .000 ed_coll 1.138 .015 .379 74.378 .000 ed_adv 1.388 .018 .355 78.917 .000 Married .222 .009 .092 25.579 .000 Divorced Widowed Separated .138 .011 .041 12.311 .000 union .189 .021 .022 8.951 .000 veteran .020 .012 .004 1.646 .100 immigrant -.055 .013 -.017 -4.116 .000 2nd Generation Immigrant .064 .012 .022 5.268 .000 femaleed_hs -.060 .020 -.017 -2.948 .003 femaleed_smcol -.005 .020 -.002 -.270 .787 femaleed_coll -.104 .022 -.026 -4.806 .000 femaleed_adv -.056 .025 -.010 -2.218 .027 a Dependent Variable: lnwage a. Explain your interpretation of the final four coefficients in the table.
b. How would you test their significance? If this test got “Sig. = 0.13” from SPSS, interpret the result.
c. What variables are missing? Explain how this might affect the analysis.

Fill in the blanks in the following table showing SPSS regression output. The model has the dependent variable as time spent working at main job. Coefficients(a)

Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta
1 (Constant) 198.987 7.556 26.336 .000 female -65.559 4.031 -.138 ? ? African-American -9.190 6.190 -.013 ? ? Hispanic 17.283 6.387 .024 ? ? Asian 1.157 12.137 .001 ? ? Native American/Alaskan Native -28.354 14.018 -.017 -2.023 .043 Education: High School Diploma ? 6.296 .140 11.706 .000 Education: Some College ? 6.308 .174 14.651 .000 Education: 4-year College Degree 110.064 ? .183 16.015 .000 Education: Advanced degree 126.543 ? .166 15.714 .000 Age -1.907 ? -.142 -16.428 .000 a Dependent Variable: Time Working at main job

Suppose I were to start a hedge fund, called KevinNeedsMoney Limited Ventures, and I want to present evidence about how my fund did in the past. I have data on my fund’s returns, Rett, at each time period t, and the returns on the market, Mktt. The graph below shows the relationship of these two variables:

I run a univariate OLS regression, . Approximately what value would be estimated for the intercept term, 0? For the slope term, 1?
How would you describe this fund’s performance, in non-technical language – for instance if you were advising a retail investor without much finance background?

Using the American Time Use Study (ATUS) we measure the amount of time that each person reported that they slept. We run a regression to attempt to determine the important factors, particularly to understand whether richer people sleep more (is sleep a normal or inferior good) and how sleep is affected by labor force participation. The SPSS output is below. Coefficients(a)
Model Unstandardized Coefficients Standardized Coefficients
B Std. Error Beta t Sig. 1 (Constant) -4.0717 4.6121 -0.883 0.377 female 23.6886 1.1551 0.18233 20.508 0.000 African-American -8.5701 1.7136 -0.04369 -5.001 0.000 Hispanic 10.1015 1.7763 0.05132 5.687 0.000 Asian -1.9768 3.3509 -0.00510 -0.590 0.555 Native American/Alaskan Native -3.5777 3.8695 -0.00792 -0.925 0.355 Education: High School Diploma 2.5587 1.8529 0.01768 1.381 0.167 Education: Some College -0.3234 1.8760 -0.00222 -0.172 0.863 Education: 4-year College Degree -1.3564 2.0997 -0.00821 -0.646 0.518 Education: Advanced degree -3.3303 2.4595 -0.01590 -1.354 0.176 Weekly Earnings 0.000003 0.000012 -0.00277 -0.246 0.806 Number of children under 18 2.0776 0.5317 0.03803 3.907 0.000 person is in the labor force -11.6706 1.7120 -0.08401 -6.817 0.000 has multiple jobs 0.4750 2.2325 0.00185 0.213 0.832 works part time 4.2267 1.8135 0.02244 2.331 0.020 in school -5.4641 2.2993 -0.02509 -2.376 0.017 Age 1.1549 0.1974 0.31468 5.850 0.000 Age-squared -0.0123 0.0020 -0.33073 -6.181 0.000

Which variables are statistically significant at the 5% level? At the 1% level?
How much more or less time (in minutes) would be spent sleeping by a male college graduate who is African-American and working full-time, bringing weekly earnings of $1000?
Are there other variables that you think are important and should be included in the regression? What are they, and why?

You are given the following output from a logit regression using ATUS data. The dependent variable is whether the person spent any time cleaning in the kitchen and the independent variables are the usual list of race/ethnicity (African-American, Asian, Native American, Hispanic), female, educational attainment (high school diploma, some college, a 4-year degree, or an advanced degree), weekly earnings, the number of kids in the household, dummies if the person is in the labor force, has multiple jobs, works part-time, or is in school now, as well as age and age-squared. We include a dummy if there is a spouse or partner present and then an interaction term for if the person is male AND there is a spouse in the household. There are only adults in the sample. Descriptive statistics show that approximately 5% of men clean in the kitchen while 20% of women do. The SPSS output for the logit regression is:

B S.E. Wald df Sig. Exp(B) female 0.9458 0.0860 120.945 1 0.000 2.5749 African-American -0.6113 0.0789 60.079 1 0.000 0.5427 Hispanic -0.2286 0.0765 8.926 1 0.003 0.7956 Asian 0.0053 0.1360 0.001 1 0.969 1.0053 Native American -0.0940 0.1618 0.338 1 0.561 0.9103 Education: high school 0.0082 0.0789 0.011 1 0.917 1.0082 Education: some college 0.0057 0.0813 0.005 1 0.944 1.0057 Education: college degree 0.0893 0.0887 1.013 1 0.314 1.0934 Education: advanced degree 0.0874 0.1009 0.751 1 0.386 1.0914 Weekly Earnings 0.0000007 0.0000005 1.943 1 0.163 1.0000 Num. Kids in Household 0.2586 0.0226 131.473 1 0.000 1.2952 person in the labor force -0.5194 0.0694 55.967 1 0.000 0.5949 works multiple jobs -0.2307 0.1009 5.223 1 0.022 0.7940 works part-time 0.1814 0.0733 6.130 1 0.013 1.1989 person is in school -0.1842 0.1130 2.658 1 0.103 0.8318 Age 0.0551 0.0088 38.893 1 0.000 1.0567 Age-squared -0.0004 0.0001 22.107 1 0.000 0.9996 spouse is present 0.5027 0.0569 78.074 1 0.000 1.6531 Male * spouse is present -0.6562 0.1087 36.462 1 0.000 0.5188 Constant -3.3772 0.2317 212.434 1 0.000 0.0341

Which variables from the logit are statistically significant at the 5% level? At the 1% level?
How would you interpret the coefficient on the Male * spouse-present interaction term? What is the age when a person hits the peak probability of cleaning?

Use the SPSS dataset, atus_tv from Blackboard, which is a subset of the American Time Use survey. This time we want to find out which factors are important in explaining whether people spend time watching TV. There are a wide number of possible factors that influence this choice.

What fraction of the sample spend any time watching TV? Can you find sub-groups that are significantly different?
Estimate a regression model that incorporates the important factors that influence TV viewing. Incorporate at least one non-linear or interaction term. Show the SPSS output. Explain which variables are significant (if any). Give a short explanation of the important results.

This question refers to your final project.

What data set will you use?
What regression (or regressions) will you run? Explain carefully whether the dependent variable is continuous or a dummy, and what this means for the regression specification. What independent variables will you include? Will you use nonlinear specifications of any of these? Would you expect heteroskedasticity?
What other variables are important, but are not measured and available in your data set? How do these affect your analysis?

Estimate the following regression:: S&P100 returns = 0 + 1(lag S&P100 returns) + 2(lag interest rates) + ε using the dataset, financials.sav. Explain which coefficients (if any) are significant and interpret them.
A study by Mehran and Tracy examined the relationship between stock option grants and measures of the company’s performance. They estimated the following specification: Options = 0+1(Return on Assets)+2(Employment)+3(Assets)+4(Loss)+u where the variable (Loss) is a dummy variable for whether the firm had negative profits. They estimated the following coefficients: Coefficient Standard Error Return on Assets -34.4 4.7 Employment 3.3 15.5 Assets 343.1 221.8 Loss Dummy 24.2 5.0 Which estimate has the highest t-statistic (in absolute value)? Which has the lowest p-value? Show your calculations. How would you explain the estimate on the “Loss” dummy variable?
A paper by Farber examined the choices of how many hours a taxidriver would work, depending on a number of variables. His output is:

“Driver Effects” are fixed effects for the 21 different drivers.
a. What is the estimated elasticity of hours with respect to the wage? b. Is there a significant change in hours on rainy days? On snowy days? 26. A paper by Gruber looks at the effects of divorce on children (once they become adults), including whether there was an increase or decrease in education and wages. Gruber uses data on state divorce laws: over time some states changed their laws to make divorce easier (no-fault or unilateral divorce). Why do you think that he used state-level laws rather than the individual information (which was in the dataset) about whether a person’s parents were divorced? Is it important that he documents that states with easier divorce laws had more divorces? If he ran a regression that explained an adult’s wage on the usual variables, plus a measure of whether that person’s parents had been divorced, what complications might arise? Explain. 27. Using the data on New Yorkers in 1910, we estimate a binary logistic (logit) model to explain labor force participation (whether each person was working for pay) as a function of gender (a dummy variable for female), race (a dummy for African-American), nativity (a dummy if the person is an immigrant and then another dummy if they are second-generation – their parents were immigrants), marital status (three dummies: one for married; one for Divorced/Separated; one for Widow(er)s), age, age-squared, and interaction effects. We allow interactions between Female and Married (fem_marr = Married * Female), and then between Age and Immigrant (age_immig = Age * Immigrant) and Age-Squared and Immigrant (agesq_immig = Age2 * Immigrant). Explain the following regression results: Variables in the Equation

B   S.E.    Wald    df  Sig.    Exp(B)

Step 1(a) female -1.890 .122 240.805 1 .000 .151 AfricanAmer 2.703 .235 132.625 1 .000 14.919 Married 1.144 .193 35.245 1 .000 3.141 fem_marr -4.946 .209 562.000 1 .000 .007 DivSep .251 .568 .195 1 .658 1.285 Widow -1.238 .131 89.790 1 .000 .290 immigrant 1.575 1.167 1.822 1 .177 4.831 immig2g .068 .117 .338 1 .561 1.070 Age .114 .047 5.858 1 .016 1.121 age_sqr -.00176 .001 7.137 1 .008 .998 age_immig -.035 .068 .263 1 .608 .966 agesq_immig 0.00027 .001 .080 1 .777 1.000 Constant 1.069 .795 1.809 1 .179 2.911 a Variable(s) entered on step 1: female, AfricanAmer, Married, fem_marr, DivSep, Widow, immigrant, immig2g, age, age_sqr, age_immig, agesq_immig.

At what age do natives peak in their labor force participation? Immigrants? Which is higher? The regression shows that women are less likely to be in the labor force, married people are more likely, African-Americans are more likely, and immigrants are more likely to be in the labor force. Interpret the coefficient on the female-married interaction.

Calculate the probability in the following areas under the Normal pdf with mean and standard deviation as given. You might usefully draw pictures as well as making the calculations. For the calculations you can use either a computer or a table.

What is the probability, if the true distribution has mean -15 and standard deviation of 9.7, of seeing a deviation as large (in absolute value) as -1?
What is the probability, if the true distribution has mean 0.35 and standard deviation of 0.16, of seeing a deviation as large (in absolute value) as 0.51?
What is the probability, if the true distribution has mean -0.1 and standard deviation of 0.04, of seeing a deviation as large (in absolute value) as -0.16?

Using data from the NHIS, we find the fraction of children who are female, who are Hispanic, and who are African-American, for two separate groups: those with and those without health insurance. Compute tests of whether the differences in the means are significant; explain what the tests tell us. (Note that the numbers in parentheses are the standard deviations.) with health insurance without health insurance female 0.4905 (0.49994) N=7865 0.4811 (0.49990) N=950 Hispanic 0.2587 (0.43797) N=7865 0.5411 (0.49857) N=950 African American 0.1785 (0.38297) N=7865 0.1516 (0.35880) N=950
Explain the topic of your final project. Carefully explain one regression that you are going to estimate (or have already estimated). Tell the dependent variable and list the independent variables. What hypothesis tests are you particularly interested in? What problems might arise in the estimation? Is there likely to be heteroskedasticity? Is it clear that the X-variables cause the Y-variable and not vice versa? Explain. [Note: these answers should be given in the form of well-written paragraphs not a series of bullet items answering my questions!]
In estimating how much choice of college major affects income, Hamermesh & Donald (2008) send out surveys to college alumni. They first estimate the probability that a person will answer the survey with a probit model. They use data on major (school of education is the omitted category), how long ago the person graduated, and some information from their college record. Their results are (assume that the 0 coefficient is 0.253): pr(respond to survey) t-statistic Major (Dummy variable) Architecture and Fine Arts -0.044 1.61 Business—general 0.046 1.72 Business—quantitative 0.038 1.45 Communications 0.023 1.00 Engineering 0.086 2.51 Humanities -0.013 0.54 “Honors” 0.087 2.08 Social Sciences 0.052 2.28 Natural Sciences, Pharmacology 0.04 1.52 Nursing, Social Work 0.061 1.57 dummy variables Class of 1980 0.025 1.61 Class of 1985 -0.009 0.61 Class of 1990 0.041 2.65 Class of 1995 0.033 2.20 GPA 0.027 2.59 Upper Div. Sci. & Math Credits 0.0001 0.21 Upper Div. Sci. & Math Grades 0.002 0.51 HS Area Income ($000) 0.001 1.92 Female 0.031 3.06 What is the probability of reply for a major in quantitative Business, from the Class of 1995, with a GPA of 3.1, with 31 upper-division Science & Math credits, with a 2.9 GPA within those upper-division Science & Math courses, from a high school with a 40 HS Area Income? How much more or less is the probability, if the respondent is female?
Consider the following regression output, from a regression of log-earnings on a variety of socioeconomic factors. Fill in the blanks in the “Coefficients” table. Then calculate the predicted change in the dependent variable when Age increases from 25 to 26; then when Age changes from 55 to 56 (note that Age_exp2 is Age2 and Age_exp3 is Age3).

Model Summary

Model R R Square Adjusted R Square Std. Error of the Estimate 1 .613 .376 .376 .94098

ANOVA

Model Sum of Squares df Mean Square F Sig. 1 Regression 53551.873 26 2059.687 2326.152 .000(a) Residual 88995.531 100509 .885
Total 142547.403 100535

Coefficients(a)

Model Unstandardized Coefficients Standardized Coefficients t Sig.
B Std. Error Beta
(Constant) 3.841 0.059 65.581 0.000
Education: High School Diploma 0.106 0.008 0.040305 ? ?  Education: AS vocational ? 0.015 0.051999 19.644 0.000  Education: AS academic 0.344 ? 0.062527 23.574 0.000  Education: 4 year College Degree 0.587 0.009 0.195326 65.257 0.000
Education: Advanced Degree 0.865 0.011 0.221309 77.658 0.000
geog2 0.070 0.013 0.017072 5.220 0.000
geog3 0.005 0.013 0.001232 ? ?  geog4 -0.050 0.013 -0.01345 ? ?  geog5 0.062 0.012 0.019974 ? ?  geog6 -0.061 0.017 -0.01039 ? ?  geog7 0.026 0.014 0.006106 ? ?  geog8 0.056 0.013 0.014445 4.303 0.000
geog9 0.102 0.012 0.030892 8.357 0.000
Married ? 0.009 0.062911 17.213 0.000  Widowed ? 0.025 -0.00191 -0.697 ?  Divorced or Separated ? 0.012 0.022796 7.042 0.000  female ? 0.006 -0.19408 -76.899 0.000  union 0.208 ? 0.024531 9.808 0.000  hispanic -0.106 ? -0.03211 -12.012 0.000  Af_Amer -0.038 ? -0.00995 -3.774 0.000  NativAm -0.100 ? -0.01342 -5.322 0.000  AsianAm -0.061 ? -0.01147 -4.420 0.000  MultRace 0.001 0.066 1.93E-05 0.008 ?  Demographics, Age 0.377 0.005 4.332516 83.265 0.000
Age_exp2 -0.00689 0.00011 -6.70717 -65.345 0.000
Age_exp3 0.0000384 0.0000008 2.65889 49.301 0.000

a Dependent Variable: ln_earn

Use the dataset brfss_exam2.sav. This has data from the Behavioral Risk Factors Survey, focused on people under 30 years old. Carefully estimate a model to explain the likelihood that a person has smoked (measured by variable “eversmok”). Note that I have created some basic dummy variables but you are encouraged to create more of your own, as appropriate. Explain the results of your model in detail. Are there surprising coefficient estimates? What variables have you left out (perhaps that aren’t in this dataset but could have been collected), that might be important? How is this omission likely to affect the estimated model? What is the change in probability of smoking, between a male and female (explain any other assumptions that you make, to calculate this)?
Using the CPS 2010 data (you don’t need to download it for this), restricting attention to only prime-age (25-55 year-old) males reporting a non-zero wage and salary, the following regression output is obtained for a regression (including industry, occupation, and state fixed effects) with log wage and salary as the dependent variable.

(17 points) Fill in the missing values in the table.
(3 points) Critique the regression: how would you improve the estimates (using the same dataset)? Model Sum of Squares df Mean Square F Sig. 1 Regression 11194.359 145 77.202 127.556 .000a Residual 21558.122 35619 .605
Total 32752.482 35764

Coefficientsa Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta
1 (Constant) 8.375 .112 74.714 .000 Demographics, Age .078 .005 .705
Age squared -.00085 .00006 -.617
African American -.184 .015 -.058
Asian .022 -.025 -4.620 .000 Native American Indian or Alaskan or Hawaiian .027 -.025 -5.674 .000 Hispanic -.051 -.020 -2.172 .030 Mexican -.021 -.007 -.868 .386 Puerto Rican .014 .002 .319 .750 Cuban .007 .059 .001
Immigrant -.094 .019 -.039
1 or more parents were immigrants .001 .018 .001
Education: High School Diploma .219 .105 13.582 .000 Education: Some College but no degree .333 .130 18.332 .000 Education: Associate in vocational .362 .081 14.919 .000 Education: Associate in academic .025 .080 14.642 .000 Education: 4-yr degree .019 .236 28.773 .000 Education: Advanced Degree .023 .253 33.757 .000 Married .011 .140 25.219 .000 Divorced or Widowed or Separated .016 .021 3.992 .000 Union member .030 .031 7.168 .000 Veteran since Sept 2001 -.047 .094 -.002
Veteran Aug 1990 - Aug 2001 -.053 .038 -.006
Veteran May 1975-July 1990 .035 .048 .003
Veteran August 1964-April 1975 .078 .129 .003
35. Using the BRFSS 2009 data, the following table compares the reported health status of the respondent with whether or not they smoked (defined as having at least 100 cigarettes) SMOKED AT LEAST 100 CIGARETTES
Yes No Marginal GENERAL HEALTH Excellent 27775 49199 ____ Very good 58629 77357 ____ Good 64237 67489 ____ Fair 31979 26069 ____ Poor 15680 9191 ____ Marginal ____ ____
a. What is the median health status for those who smoked? For non-smokers? b. Fill in the marginal probabilities – make sure they are probabilities. c. Explain what you might conclude from this data. 36. Using the CPS data, run at least 4 interesting regressions to model the wages earned. Carefully explain what we can learn from each regression: does it accord with theory; if not, what does this mean? Explain what statistical measures allow us to compare different specifications.

Variable	Coeff	Std Error
\(Age\)	-0.030	0.004
\(Age^2\)	0.001	0.00003
\(Female\)	0.115	0.125
\(Age*Female\)	0.015	0.005
\(Age^2*Female\)	-0.0002	0.00004
\(Constant\)	-0.436	0.094

Exam 2 Practice

2019

2018

2017

2016 and earlier