economics assignment

1  

C171/151 – ECONOMIC DEVELOPMENT SPRING 2019

Problem Set 2

-- Due at the beginning of your assigned sections 15, 16 or 17 April –

(Please avoid hand-written answers if possible.) Question 1 This question is based on “Does Trade Cause Growth?” by Frankel and Romer (1999).

a) Explain in no more than a few sentences how the authors construct an instrumental variable to test for the causal effect of trade openness on economic development.

b) Interpret the regression coefficients of columns 1 and 2 in Table 3 of the paper, and comment on their statistical significance.

c) List the assumptions that need to hold for a causal interpretation of the point estimates on

trade openness in column 2 of Table 3.

d) List your two main concerns that you think could pose a challenge to the validity of these assumptions.

e) Suggest one way in which the authors could try to address each of your two concerns (two

suggestions). Think of this as two additional robustness checks that the authors could estimate and report in their analysis to check if your concerns are relevant.

f) How would you expect the point estimates of the effect of trade openness to differ between

columns 1 and 2 of Table 3? If this is not what the authors find, then what could be an explanation for the difference we observe in the table?

Question 2 In regression equation (3) of the “Worms” paper (Kremer and Miguel, 2004), the authors compare pupils in schools assigned to Group 1 relative to pupils in schools assigned to Group 2 at the end of the first year of the intervention. Schools in Group 1 had been treated by that time, and schools in Group 2 were about to receive the treatment, but had not yet been treated during the first year.

i = school j = student t = survey round (only one period here at the end of first year) d = distance range from school Y = student health outcome

2  

alpha = intercept/constant T1 = 1 if school assigned to Group 1 and 0 if school assigned to Group 2 D = 1 if student actually receives treatment as part of Group 1 or Group 2 X = student, school and time controls N = number of pupils within distance d N_T = number of pupils in treated schools within distance d u and e = error term/residual The authors then report estimation results from the following two regressions (displayed in the table below). Both regressions include only one cross-section of pupil data on health outcomes that were measured at the end of the first year (beginning of 1999). At that time, no schools in Group 2 had been treated yet, and the furture treatment status D of students going to schools in Group 2 was measured in the second round after this data on health outcomes had been collected.

a) Which one of these two columns corresponds to regression equation (3) above? Using the same variable notation as above, write down the regression equation of the other column.

b) Using the results in column 1, how would you compute the average effect of the intervention in treatment schools after the first year? Would this estimate be likely equal, larger or smaller than that of a simple difference in health outcomes between treatment and control groups at the end of the first year, and why?

3  

c) Interpret the final two regression coefficients of Column 1, and comment on their level of

statistical significance. Explain why the authors include these two variables in the regression.

d) Using the results in column 2, how would you compute the average effect of the intervention after one year for children in treatment schools who did not receive the de-worming medicine? How is this different from the effect for children in treatment schools who did receive the medicine?

Question 3 This questions asks you to estimate the causal effect of tourism on local household incomes in Mexico. To answer this question, we will use Stata and the dataset “Mexico_PS2.dta” that you can download from BCourses. This dataset contains 1153 Mexican municipalities that reported some amount of local tourism activity (measured by local hotel sales) in the year 2000. Write up the answers to a)-e) below in the same document you used for Questions 1 and 2 above. In addition, also attach the do file that you used to answer the following questions.

a) Open the dataset in Stata. Visualize and export a table that lists the number of observations, the mean, the standard deviation, the minimum value and the maximum value for each of the variables in the dataset (edit and include the table in your written up answer). Briefly describe what we learn from the table about the sample of Mexican municipalities.

b) Use the data to obtain an OLS point estimate of the effect of local tourism activity (measured by the logarithm of local hotel sales) on the logarithm of local average monthly household incomes. Export your result in a regression table (that you can edit and include in your written up answer), and comment on the interpretation and statistical significance of your result.

c) List three plausible arguments why the point estimate in b) could be biased upwards or downwards relative to the true causal effect of local tourism activity on monthly household incomes.

d) Now your GSI suggests that the kilometer distance between the center of the municipality

and the nearest segment of the US-Mexico border could be a valid instrumental variable for your measure of local tourism activity. List the assumptions that need to hold true for this to be correct.

e) Verify if the assumption of instrument relevance is satisfied, and export the results into the

same regression table that you used before. Comment on the interpretation and statistical significance of your result.

f) Now use Stata to estimate the 2nd stage IV point estimate as suggested by the GSI, and export

your result in the same regression table you used before. Comment on the interpretation and statistical significance of your result. In reference to your answer to c), is the difference between the OLS and IV point estimates as you expected or rather not?

g) Now one of your friends suggests that the distance to the US border is likely correlated with other local characteristics that affect local incomes, such as the logarithm of the average

4  

temperature, the logarithm of the average precipitation, the average years of education and the proportion of indigenous population. Propose a way to verify whether these concerns are relevant and export your regression results in the same table as before. Comment on the results and what they imply about the validity of the instrumental variable strategy.