Population Economics Field Exam September 2010 Instructions You have 4 hours to complete this exam. This is a closed book examination. No materials are allowed. The exam consists of two parts each worth 100 points. DO NOT spend too much time on any one question. Partial credit will be given whenever possible. You need to get 75 points in each section in order to pass the Population Exam. Please answer Parts A and B in separate booklets. Make sure you write your identification information in ALL booklets.
Part I: Population Economics (2 hours) 1- (25 points) General review questions (1/2 page each at most) 1. (5 points) What are the theories and the evidence on the causes of fertility decline in the last 2 centuries? 2. (5 points) Discuss the evidence of the effect of changes in life expectancy and health on economic growth. 3. (5 points) Briefly describe the quantity-quality model and how various empirical studies have attempted to test it. 4. (5 points) Review the 2 main models of intra-household allocation that exist in the literature, detailing on how they differ in their predictions. Why is it important or interesting to distinguish between these? 5. (5 points) Review Manski s 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. 2- (75 points) Empirical Application: Analyzing the effects of Family Medical Leave Act on child health In 1993, Congress passed the Family Medical Leave Act (FMLA). The FMLA requires all employers with 50 employees or more to provide up to 12 weeks of unpaid, jobprotected parental leave a year to eligible employees. The Act was passed with the expressed purpose of allowing leave to care for a newborn, newly-adopted or foster child, child, spouse, or parent with a serious health condition, or for the serious health condition of the employee (including maternity-related disability). In this way, the FMLA could be used for both maternity leave and parental leave following childbirth, for up to 12 weeks in total. Health insurance benefits must be continued during the period of leave. When the FMLA was passed in 1993, 34 states and the District of Columbia had already passed some form of maternity or parental leave law. Most policies were unpaid and the duration of these policies varied from state to state. Suppose you have data containing state-level averages of infant mortality, mean birth weight and the average number of illnesses within the first year of birth for 1990 and 1995 (that is before and after the FMLA was passed). 1. (5 points) Suppose a researcher estimates the following regression: (1) H i =β 0 +β 1 *I(mom took leave from work=1) i + X i B + e i Where H i is a health measure for infant i (birth weight, infant mortality, or incidence of disease), X i is a set of infant characteristics, and I(mom took leave from work=1) is an indicator variable equal to 1 if i s mom took leave at or around the time of i s birth. Under which assumptions does β 1 give us the causal effect of leave taking on health? 1
2. Simple DD estimates: Given the information about the FMLA above and the data that is available to you, a. (5 points) Write down the estimating equation to estimate the effects of the FMLA on infant mortality using a linear difference-in-differences (DD) strategy. b. (5 points) What are the identifying assumptions we need to make to obtain causal effects of the FMLA? Provide examples where assumptions would fail. c. (5 points) Discuss the extent of bias in standard errors in the current setting. What could be wrong and how you would correct for it? d. (5 points) Suppose that the FMLA had been randomly assigned to states. Would you still estimate a DD model? Would you want to include other state or individual characteristics as covariates? e. (5 points) We have 2 observations per state, one before and one after. Would you want higher frequency data? More years of data? How would this help? f. (5 points) Would it be better to estimate a non-linear model? Why does this matter? 3. Since the legislation affected only individuals in large firms, one could consider estimating a Difference-in-Difference-in-Difference (DDD) model. Assuming you had data by state, year and firm size (large versus small firms), a. (5 points) How would you estimate the effect of the FMLA model using a DDD approach? Write down the equation. b. (5 points) Is this DDD a better strategy than the DD strategy? Why? 4. Suppose that you had access to firm level data including information on the exact number of employees at the firm that could be matched to individual data that includes health outcomes of the babies of the firm s employees. a. (5 points) How would you use these data to estimate the effect of the FMLA using a regression discontinuity approach (RD)? Write down the equation. b. (5 points) What assumptions would we need to make to get unbiased estimates using the RD approach and would you test that these assumptions are likely to be met? c. (5 points) Is the parameter estimated using this RD approach the same as the one we estimate using DD or DDD approaches? Is it the parameter of interest? d. (5 points) Would this approach be better than the DD or DDD approaches above? 5. Imagine you have individual data that would allow you to estimate equation 1 and that can be linked to state level laws. You can now consider estimating the effect of leave-taking using state laws as instruments in a linear framework. a. (5 points) Write down the assumptions we need to make for the instrumental variables strategy to provide unbiased estimates of the effects of leavetaking on infant health. How likely are these assumptions to hold? 2
b. (5 points) Do you expect the OLS estimates of (1) to be bigger or smaller then the instrumental variable estimates? Why can OLS differ from instrumental variables estimates? 3
Part II Development (2 hours). Total points: 100. 1 Questions (30 points total) Answer each question below. Cite speci c papers when you refer to empirical evidence. 1. (10 points) Very few sharecropping contracts are such that the return to the tenant (the share of harvest kept by the tenant) depends upon the yields of other farms in the area, or the weather. Is this surprising to you? Why? 2. (20 points) In 1964, T. W Schultz published Transforming Traditional Agriculture, a book in which he refuted the notion that farmers in developing countries are poor because of cultural characteristics such as lack of a work ethic, lack of an understanding of the idea of savings, or general ignorance of how to make best use of their resources. Instead, in his famous e cient but poor hypothesis, Schultz argued that low income levels in agriculture in developing countries are a result of the low productivity of the available factors of production, not of ine ciencies in their allocation. Since then, it has been argued that farmers in developing countries can be too poor to be e cient in other words, that richer farmers with access to the same land would be able to achieve a higher production level than poor farmers. What are the bases for this too poor to be e cient argument? Discuss the various circumstances under which the yield for a given piece of land might depend on the background wealth of the one cultivating it. 2 Credit and Insurance (20 points) Consider a cluster of small garment producing rms in the outskirts of Calcutta. Each household in the cluster has its own small rm, employing both family and non-family labor. Only adults and and teenagers work as garment producers. Firms produce stitched garments for the domestic market. The only xed capital consists of sewing machines. Working capital (to buy cloth needed 4
to stitch the garment) is crucial. Working capital credit is supplied by cloth merchants from another community, operating from Calcutta. 1. (a) Given the context and what you know about credit markets in developing countries, would you expect working capital credit to be freely available? (b) Suppose you want to collect data in this garment producing community to nd out whether rms are credit constrained or not. If they are not credit constrained, and all other markets are well functioning, what would be the e ect of a negative shock in the household (for example: the sickness of a child, necessitating a visit to the hospital) on labor supply, purchase of cloth, production and pro t for the household s rm? (c) If they are credit constrained, what would be the e ect of such a shock: i. In the presence of some formal or informal mechanism for health insurance (still considering the sickness example)? ii. In the absence of such a mechanism? (d) Assume you want to implement a test based on this idea. Explain: i. The data you would collect (Panel? Cross-section? What variables? ) ii. The main speci cations would you would run, and what the potential problems are with these speci cations. 3 Adverse Selection (20 points) Imagine that there are two types of projects t 2 f1; 2g that people want to nance. Type 1 projects have relatively low returns (R 1 ), but are also low risk (the probability of success of a type 1 project is denoted 1 ). Type 2 projects have higher returns (R 2 > R 1 ), but are also riskier than type 1 projects (the probability of success for a type 2 project is p 2 > p 1 ): Assume that p 1 R 1 = p 2 R 2 : Investment in both types of projects costs 1 unit. The borrowers do not have one unit on their own, that s why they need to borrow it. There is a problem of limited liability, however borrowers cannot repay their loan if their project fail, because they have no wealth at all. 5
If the project succeeds, the borrowers have to pay back the principal plus a share i of the principal charged as interest. The opportunity cost of capital for the lender is : There is perfect competition among lenders. The outside option for a borrower that doesn t get a loan is to work on a labor market and get U: 1. First, assume that the lenders can perfectly observe which type of project a given credit applicant is going to invest in. What credit contracts will the lender o er the applicant? Will the two types of projects be funded with the same loan or will one type get a lower interest rate? If so which type? 2. Now consider imperfect information: the lender cannot observe which credit applicant will invest in which project. The only thing that the lender knows is that the proportion of potential borrowers who want to invest in a type 1 project is : What are the possible loan contracts that the lender can o er to credit applicants? What types of projects will be undertaken under each contract? Will anyone be credit-constrained? 4 Health (30 points) 1. The capacity curve relates the work capacity to the employment income: (a) Explain the shape of the capacity curve in the gure 1 below. (b) Two wage curves have been drawn (v1 and v3). Reproduce this gure on your exam book, then plot the minimum piece rate at which the worker has some non-trivial capacity to work. Let s call this minimum rate v0. (c) Plot the individual labor supply as a function of the piece rate. Discuss what happens to the individual labor supply at v0. (d) Imagine an economy with a population of identical people whose capacity curve is that of gure 1. Describe the equilibrium if the labor demand intersect the labor supply at v < v0, v = v0 and v > v0. Identify in which case we will have involuntary unemployment. 6
(e) Assume now that we have landed and landless individuals whose capacity curves are represented in gure 2. De ne v0 for the landless and v0 L for the landed. Which one is larger? Why? Is the labor market increasing or decreasing the existing inequality in assets? How? (f) Show that if there is a mix of people with nonlabor income and some without nonlabor income, the individuals with nonlabor income will always do better than the individuals with no such income. 2. Can long-term contracts (such as indentured servitude) allow a worker to o er his labor power at a lower wage? Why would this strategy work? Why might it not be feasible? 7
8