Labour Economics Problem Set 2: Human Capital and Job Search
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
Labour Economics
Problem Set 2: Human Capital and Job Search
Human Capital
Returns to Education: Long-run Accumulation/Loss The following questions refer to the paper Winter-Ebmer/Ichino (2004): The Long-Run Educational Costs of World War II, Journal of Labor Economics (uploaded on Blackboard) . Section IV and the Appendix are not relevant to answering the questions .
1. Recall from the lecture: which biases arise when we estimate a Mincer equation to measure the e↵ect of an additional year of education on earnings? How does this motivate the paper by Ichino and Winter-Ebmer?
2. How does Table 1 support the authors’ idea of using World War II as a quasi-experiment to estimate the returns to education?
Describe the regressions run by the authors to produce Table 1 [hint: look at the table notes] . Why do the authors estimate the e↵ects of cohort dummies on the residual variation in education (conditional on age), instead of estimating the e↵ects directly on the number of years of education?
The authors write: “Looking at the first column of table 1, the average educational loss of the Austrian cohort born in the thirties amounts to approximately 16% of a year of schooling with respect to the previous cohort. ” How are the 16% calculated?
3. The authors argue that the decrease in education experienced by the cohort born in 1930-39 in Austria and Germany is due to the war . What is the assumption behind this claim? What are potential threats to it?
4. Why does the coefficient of “Born 1930-39” on education become insignificant from column 1 to 6 in Table 2? How does this support the idea that in Germany, the decrease in education experienced by the 1930-39 cohort was due to the war?
5. In the first column of Table 4, the authors report results from the following specification: yi = ↵ + βedui + 6femalei + ui (1)
where yi are the earnings of individual i and her/his years of education . femalei is a gender dummy and ui the residual . In column 2, edui is instrumented with a dummy for being born in 1930-39 . Denote this dummy as Bi .
What is the formal condition on the relation between ui and Bi for Bi to be a valid instru- ment? Intuitively, what do we need to assume on the earnings di↵erence between Austrian individuals with Bi = 1 and the other Austrians in the sample? Which factors could threaten this assumption?
6. In the last column of Table 4, the data include all four countries . Bi is now not the IV, but a control variable in the regression . What is now the IV?
What do the authors achieve by introducing dummies for each country in the regression (think of the intuition behind fixed e↵ects)? How do you interpret the fact that the coeffi- cients of these dummies are insignificant?
The IV approach of the last column resembles another identification technique seen in the (exercise) lecture . Which one is it?
7. How do you interpret the coefficient on education in the last column of Table 4? Can you use it to derive general conclusions on the returns to education? Why (not)?
Externalities of Education: Firm-level Evidence The following questions are based on Moretti (2004): Workers’ Education, Spillovers, and Productivity: Evidence from Plant-Level ProductionFunctions, American Economic Review (uploaded on Blackboard) . You do not need to read the full paper; it is sufficient to consider the elements mentioned below .
Moretti outlines a small theoretical model which considers the firms of two cities, A and B . There are two types of labour, high skilled (H) and low skilled (L), paying city-specific wages wH and wL , respectively. Each city is a competitive economy (implying zero-profit condition) that produces y using a Cobb-Douglas technology: y = AH ↵ H L↵ L K β . The (composite) good y is sold nationally, i .e . at the same price across the two cities; K denotes capital at price r . Human capital spillovers are modeled by allowing the productivity of plants in a city to depend on the aggregate level of human capital in the city: A = f() . Variation in the cost of living depends only on variation in cost of land, p (”rent”), which is the same for all workers in a city.
1. In this model, the equilibrium is obtained when the utilities of workers in both cities are equal and firms in the di↵erent cities have equal unit costs . Note that VH and VL are the indirect utilities of high- and low skilled level, which reach equilibrium values kh and kL , respectively. The firm’s constant unit costs equal at 1 . Moreover, note that B > A . The following two graphs illustrate the equilibria in the two cities . Explain and interpret the graphs . What do the points 1 and 3 represent in the two graphs? Explain their location and interpretation using the outlined model .
2. Point 2 represents the equilibrium in city B without externalities . How can thus the magni- tude of the spillover in city B be measured in the graphs? Explain .
3. In equilibrium, firms in city B are more productive than firms in city A . Since firms are free to relocate from A to B, why is productivity not driven to equality?
4. Derive the firm’s relative first order condition (ratio of marginal products equals relative wages) for high- vs . low-skilled labour . On which components does it depend? Does the spillover/externality have an impact or not? Why (not)?
5. Looking at the graphs, it seems that the relative di↵erences between wages, i .e . vs . , are not the same across the two cities . The wages seem to di↵er more in city A than in B, i .e . > . Discuss how A = f() has to be specified distinguishing A vs B and high vs low skilled workers such that the before-mentioned di↵erence in the wage ratio can be obtained (ceteribus paribus, i .e . conditional on unchanged relative labour supply) .
6. Now consider the results Table 2 of the paper . Is there evidence for spillover e↵ects or not? How are they measured? Comment as well on the quantity (taking into account that the dependent variable is output ln y) .
7. The results in Table 2 are based on cross-sectional data . Discuss some possible (empiri- cal/econometric) problems that could bias the causal estimate of the spillover e↵ect . Which issues are taken into account by the reported control variables?
8. Finally, consider the estimation results in Table 3 . Now, the author uses longitudinal data to estimate the same spillover e↵ects . These panel data allow for more controls and thus for a more credible identification of the e↵ect . By going through the (Cobb-Douglas) specifications (columns 1–5, focus on row 1), explain which additional potential identification problems the author tries to tackle by use of further controls (hint: use the intuition of fixed e↵ects models) . Are the results sensitive to the di↵erent specifications?
Job Search
Theory: Job Search Model with E↵ort In the following, we consider a job search model in which the job seeker chooses her optimal amount of search e↵ort e . The model di↵ers from the one in the lecture through two main simplifying assumptions: (1) Everyone receives benefits of amount b, unconditionally and until having found a job . (2) Once a job seeker has found a job, she keeps it forever .
maxrVu = b − c(e)+ λ(e) (Ve (w) − Vu )dF(w) (2)
where r is the discount rate, b is the amount of unemployment benefits, c(e) is the e↵ort cost function and jobs arrive at rate λ(e) .
rVe (w) = w (3)
where w is the wage of the job match .
1. The job seeker’s reservation wage is denoted x . What is the intuition behind the property that rVu = x? Use it to obtain an expression of the reservation wage that depends on b , c(e), λ(e), r, and (w − x) . Give an interpretation of each term in the expression .
2. Suppose c(e) = c0 e2 and λ(e) = λe . What is the job seeker’s optimal search e↵ort e ⇤ ?
3. Now, compare two individuals who have di↵erent levels of human capital . The agent with higher human capital will getter better job o↵ers from the market . Show how this is can be represented in the above-derived equation: which component is a↵ected by this di↵erence in human capital? How do the two individuals di↵er in their optimal search e↵ort?
4. Next, compare two individuals who di↵er in their fixed cost of search c0 . How do they di↵er in their reservation wage, given the optimal search e↵ort decision? Hint: combine the equations derived in 1 . and 2 .
5. Suppose the government decides to decrease benefits b. What is the e↵ect on x and e* ? Why are these e↵ects expected to induce an increase in the transition rate out of unemployment? Explain (using the derived equations and the knowledge from the lecture) .
6. Consider an individual who is not eligible for UI benefit compensation . How is she a↵ected in her job search by the mentioned benefit decrease? Discuss the impacts of the disincentive e↵ect and the eligibility e↵ect .
Empirics: Reduced-Form Estimation of Hazards Using Unemployment Duration
1. Recall from the lecture: what does the hazard rate (in continuous time) measure?
2. You want to model a proportional hazard function with the following specification: ✓(tu | x) = λ(tu )exp(x\ β)
with x being the explanatory variables in the regression and
10
λ(tu ) = exp(X λk Ik (4k < tu < (4k + 4))
k=0
where k=( 0, . . . , 10) are the number of weeks in unemployment . Why do we call this a “piecewise-constant” duration dependence function? If you run a regression, what do the dummies contained in λ(tu ) capture?
3. You want to estimate the e↵ects of a decrease in benefits b on the hazard . To this purpose, you exploit a reform which decreased b for individuals aged less than 30 years starting in the year 2000 .
A friend suggests to estimate the reform e↵ects on the log proportional hazard using a di↵erences-in-di↵erences approach . He proposes the following specification:
log ✓(tu | Post,A30) = log λ(tu )(β1 · Post + β2 · A30)
where Post is a dummy that turns one starting in the year 2000 and A30 is a dummy that turns one if the individual is aged above 30 .
Is this a di↵erence-in-di↵erences specification? If yes, which coefficient gives you the re- form e↵ect? If not, which term needs to be added to estimate the di↵erence-in-di↵erences coefficient?
4. Give at least one identifying assumption that is required to hold if you interpret the di↵erence- in-di↵erences coefficient as the causal e↵ect of the reform . Can you think of a way to test its validity?
5. Suggest another quasi-experimental approach which could be used to evaluate the above- mentioned reform .
6. Suppose you would like to know how the causal e↵ects of the reform di↵er, on one hand, by gender and, on the other hand, by education (e .g . , four education groups) . Show how you would adapt the empirical specification of the estimation model to do this .
2023-04-17