This exam has two sections: true/false questions and one longer question. You should write your answers in this document. The exam will go live on Classes at 12:30, and you should upload a  PDF before 1:45 (give or take a few minutes).

There are 67 points on the exam, so you should aim to spend about a minute per point. I am not  trying to trick you – do not spend 20 minutes crafting a perfect answer to a 5 point question. If something seems unclear, please email me and Lena to clarify.

This is a closed-book & no-talking-with-friends exam (like it would be in the classroom). The    midterms are a useful way for you to figure out if you understand the material (and for me to      figure out how to improve for next year. NYU forces me to curve your final grades. If you cheat, you aren’t cheating me so much as you are cheating your honest classmates. Please follow the    NYU honor code.

Section I: True/False (21 total points)

Consider the following statements.  Write 3-5 sentences (no more) on whether you think these    statements are: (a) true; (b) false; or (c) depend on certain conditions.  If your answer is “true” or “false,” explain why. If your answer is that “it depends,” make sure to say what it depends on.    Your grade will come from the quality ofyour answer – if you only circle the “right” answer,     you won’t get any credit.

1.   Across the world, there is a lot of variation in the extent to which extended families share one household. In the theory, the Agricultural Household model tells us that the total number of hours spent working on a family’s farm (by everybody: family members and employees) should be unrelated to the household’s size. (7 points)

True / False/ Depends

True. From separability of production and consumption and maximization of profits, we know that labor on the farm is determined by the equilibrium condition (marginal product of labor = local wage). This means that the optimal labor (L) is unrelated to household’s size.

2.   Three of the Kaldor facts are: (a) output per worker has been increasing at a steady rate, (b) the ratio of capital to output has been stable, and (c) capital per worker has been increasing at a steady rate (This sentence is true, the question is the next sentence). The first two facts imply the third. (7 points)1

True / False/ Depends (pick one)

True.

( L ) = ( Y ) ∗ (L ) = ( Y ) + (L )

Since the first component is stable (according to b) and the second component is increasing (according to a), capital per worker would be increasing at a steady rate.

3.   If an individual is stuck in a nutrition-based poverty trap, then it must be the case that the increase in energy from an extra bit of nutrition is low. (7 points)

True / False/ Depends (pick one and use a graph to illustrate. Make sure to label your axes!)2

False. It’s possible to have high increase in energy for an extra bit of nutrition for an individual   stuck in a nutrition-based poverty trap. Consider the example where the increase in income is      low from an extra bit of energy: for an individual who receives an extra bit of nutrition and gains lot of energy at time t, their energy does not translate to income and in the next period t+1, with  their low income, they are unable to afford significantly more nutrition.

What matters is the composition of 2 functions 1) income as a function of energy (health) 2) energy (health) as a function of income.

We plot the composition of these functions below. On both axis we have energy (health), with energy at t on the x-axis and energy at t+1 on the y-axis.

People to the left of h* are stuck in a poverty trap.

Part II: Longer answer. 42 total points.

The picture below plots the relationship between female life expectancy and female literacy rate around the world (each country is one point, the data is from 2011). You are probably not          stunned to see that they are positively correlated. This question is about interpreting this picture.

Before getting to the data, it’s always worthwhile to have a theory. Recall the set-up for the Mincer Equation (although here I’ll use simpler assumptions):

Utility is log future earnings minus the current costs.

Log earnings as an adult (ln(y)) is equal to ability (A) plus years of schooling (s) times the return to schooling (x):

ln(y) = A + XS

The cost of schooling is convex

ℎ(S) = CS + 1/2 S 2

Given utility equal to log future income minus the cost, the first order condition gives us the

Mincer Equation: the optimal amount of schooling is

s ⋆  =

i)          Use the Mincer equation to describe the optimal amount of education if people work for α periods instead of one. That is to say, if future earnings are ln(y) = a(A + XS).  (10    points)

If people can work for multiple periods, they can earn aln(y) relative to only for one period, which in return increases the return to schooling.

So optimal schooling is

s*  =

ii)        Given the picture, describe one reason why observed relationship might be too steep       relative to the underlying causal effect (that life expectancy has on education). (5 points)

Give any reason that biases the results upwards

For example: country with better institutions may have better health care and education systems, leading to both higher female literacy rate and life expectancy

iii)       Given the picture, describe one reason why observed relationship might be too shallow  relative to the underlying causal effect (that life expectancy has on education). (5 points)

Give any reason that biases the results downwards

For example, gender norms in a country may require women to work a lot. This could drive up female literacy rate (need to be literate to work) and lower female life expectancy (work related stress).

Between 1946 and 1953 there were enormous investments into maternal health in Sri Lanka,       increasing female life expectancy by around 1.5 years (relative to men). The number of hospitals, clinics, and health centers rose considerably, and many of these facilities were specifically for     maternal and child care. The number of trained birth attendants also increased. Importantly, most of the services were provided for free. Transportation to health facilities was also improved: a      system of free ambulances was developed. Sri Lanka also adopted new technologies from the      West, such as penicillin.

The graph below shows the overall decline in maternal mortality in Sri Lanka

There was a lot of spatial variation in this decline, shown in the graph below

iv)       As you can see, the declines are largest in the places with the highest initial levels. This can be shown more clearly in this graph, where the y axis is the change in the female-    male life expectancy gap (so larger numbers mean women live relatively longer) and the x-axis is the maternal mortality ratio in 1946.

Consistent with your answer to (i), not only do places with a higher maternal mortality  ratio see a larger decline, but they also see increases in female literacy relative to males, as shown in the graph below.

Explain how the intuition of difference-in-differences means that we can use these Figures 2iv (a) and (b) to uncover the causal effect of increased life expectancy on educational attainment. (10 points)

Difference-in-difference allow us to deal with both time and individual fixed effect, as we compare the treatment group and control groups in terms of their difference in outcome    before and after treatment.

We can think of 2iv(a) as the treatment that was applied: places with high initial maternal mortality ratio in 1946 received a bigger treatment in the form of higher change in life     expectancy.

2iv (b)  shows the change in the outcome variable female literacy rate. Calculating the difference of two points on this graph gets us the difference between control and         treatment in change in literacy rate.

By dividing the slope in these graphs, b/a, we get the causal effect of increased life expectancy on female literacy rate.

The slope of Figures 2iv (a) implies that increasing the initial maternal mortality ratio by 10% increases the change in life expectancy for women by 5%. The slope of Figures 2iv (b) implies that increasing the initial maternal mortality ratio by 10% increases educational attainment of women by 10% (This is a simplification: the graph is on literacy rates, not the s* of the Mincer equation, but it will be easier to assume that they are the same thing. Similarly, these are not the actual numbers for the graphs, but these make the math easier).