Académique Documents
Professionnel Documents
Culture Documents
Spring 2012
Instructions
This is a closed book examination. No written materials are allowed. You can use a calculator.
YOU MUST ANSWER BOTH PARTS I AND II. YOU MUST OBTAIN AT LEAST 75% IN EACH PART TO PASS
THE PUBLIC FINANCE FIELD EXAM.
1.
(10 points) Review the methods that are used for estimating the economic value of
changes in life expectancy. In particular discuss the differences between methods that
estimate the value of a statistical life and calibration methods such as the one used in
Becker et al (2005).
2. (10 points) What does the existing literature say about the causes in the rise of health care
spending over time?
3. (10 points) Briefly review the basics of propensity score estimation. What are the
advantages of propensity score estimation compared to OLS?
4. (10 points) Review Manskis concepts on social effects. What are endogenous social
effects? Contextual effects? Correlated effects? Give an example of which and explain
why we are interested in differentiating between them.
5. (10 points) What is the existing evidence of the effect of insurance on health care
utilization? Explain how this relates to the issue of moral hazard in health care. Draw a
picture and explain what the deadweight loss associated with insurance is.
6. (10 points) What is the existing evidence of the extent to which there is crowding out in
the insurance market? Why is this question important?
7. (10 points) What are the main results of the Rothshild and Stiglitz (1976) seminal model
of insurance?
8. What is the empirical evidence for adverse selection? Does it support the predictions of
the Rothschild Stiglitz model?
9. (10 points) What economic rationale is there for national mandated insurance schemes?
What are the downsides?
10. (10 points) According to Summers (1989), what are the labor market effects of mandates
through firms? Why does he argue this provision mechanism dominates public provision?
1.
(10 points) Older studies investigating the effect of insurance on outcomes simply
regressed outcomes on an indicator for having insurance such as Medicare.
What assumptions do you need to make for the OLS estimates of the effect of
insurance to be unbiased?
Ex-ante do you expect OLS estimates to be upwards or downward bias?
Why are we interested in estimating the effects of insurance?
2. (10 points) Card et al use an RD design and estimate the effect of having Medicare
insurance on health care utilization and health outcomes. Card et al estimate the
following regression:
where the dependent variable is a utilization (or health) measure for individual i in group j at age
a, and C is dummy equal to one after age 65 (since individuals become eligible for Medicare at
that age).
-What is the key identifying assumption? How can you verify that it is likely to be met?
-Give an example of how RD might fail in this case.
-Does this strategy estimate the effect of being insured (relative to being uninsured)?
3.
(10 points) Some individuals are not eligible for Medicare (for example foreign nationals and
those that never worked). A researcher proposes to use a DD design and look at eligible and
ineligible populations above and below 65 to identify the effect of Medicare.
Write down the estimating equation
What assumptions do you need for this DD to be identified?
Is this design better or worse than the RD design? Explain
4. (10 points) A totally different approach consists in finding an instrument for having Medicare.
Currie and Gruber (1995) follow such an approach by using changes in Medicaid legislation at
the state level that affect the number of individuals that are eligible for public insurance.
Because Medicare is a federal program this approach is not feasible here, but we will imagine
that it is here and think about how that might help identification:
Write down the equation you would estimate using laws as instruments. Assume that
there are changes in legislation over time in many states and that the legislation
regulated who is eligible for insurance.
What assumptions are needed for the IV estimator to be consistent?
What parameter will we estimate using IV?
5. (10 points) The Card study uses contemporary cross sectional data to infer the effects of
Medicare. Instead, one could look at the historical evidence and see what happened to elderly
health after Medicare was passed in 1965, as a recent paper by Ken Chay does.
Putting aside identification concerns, do we expect the results from the historical study
to match the ones that Card et al get? Why or why not?
An older paper by Doyle (2005) uses car accidents as a random health shock and looks at
how the uninsured fare when they reach the hospital after an accident. He compares
how the uninsured are treated in the hospital after an accident relative to those that are
insured and are also in similar car accidents. What identification problems does this
approach solve? What are the potential issues?
An important limitation of almost all the studies that look at the question of the effect
of insurance on health is that they look at the immediate effect of becoming eligible on
outcomesthat is they measure outcomes within a few years. However insurance
increases access to preventive care, the benefits of which may not be observable for
years. Why is it difficult to estimate the long term effect of insurance?
Directions: Answer all parts and questions. If you need to make additional assumptions
beyond what is given in the statement of the problem to answer the questions, please make
your assumptions clear. Show all work. There are 160 points total.
Part A - Short Answer. In a few sentences, discuss the validity of each of the following
claims.
1. (10 points) The Tax Reform Act of 1986 lead to a substantial increase in incomes
reported at the top of the distribution. Therefore, the top marginal tax rate cut
allowed the government to collect more tax revenue from the rich.
2. (10 points) Womens labor supply is more elastic with respect to taxes than mens.
Therefore, there should be gender based taxation with lower tax rates for womens
earnings than for mens earnings.
3. (10 points) A simple OLS regression of the number of Workers Compensation claims
(divided by the total number of jobs) on the level of benets using several years of
data aggregated at the state level produces a large and positive coe cient. This is
evidence of a substantial individual behavioral response to the generosity of workers
compensation benets.
4. (10 points) Social insurance programs reduce total economic output while potentially
raising welfare by reducing agentsexposure to risk.
5. (10 points) Explain why there is so much attention in the public economics literature
devoted to estimating the elasticity of taxable income. Cite one study that estimates
this elasticity, what is the nding? What is one shortcoming of focusing on this
elasticity?
6. (10 points) The disability insurance program is one reason that unemployment rates
did not rise as much in the last recession (2009) as they did in previous recessions.