● f_other: dummy variable, takes the value of 1 if this country is a former colony but neither a former British colony nor a French colony.

Abr.u uhe attigomeou:

In this assignment, we will investigate whether better property rights protection can lead to better eco- nomic development among ex-colonies. Understanding how property rights can affect economic devel- opment is a fundamental area of research in Public Economics, Development Economics, and Political

Economics.

The data are drawn from:

Acemoglu, D., Johnson, S. and Robinson, J.A., 2001. The colonial origins of comparative development: An empirical investigation. American Economic Review, 91(5), pp. 1369- 1401.

Throughout this assignment, unless otherwise mentioned, please use significance level a = 5% as the default option.

Please do ALL the relevant data exercises in Stata. Almost all commands needed for completing this assignment have been discussed in your tutorial sessions. It would be useful to review your tutorial notes. Also, Google is always your best friend when learning a new software language.

1. We will start by looking at the dataset AJR_base. Report summary statistics (number of observa- tions, mean, standard deviation, min, max) for avexpr and logpgp95. Comment of the min and max values, highlighting the extremes each variable takes on. [2 marks]

2. Compute 95% confidence interval for the respective means of avexpr and logpgp95. Use two dif- ferent ways to interpret the confidence intervals. [4 marks]

3. Create a new variable good_property_right which equals to 1 if avexpr above median and 0 oth- erwise. Plot 2 separate densities within the same graph for logpgp95 when good_property_right = 1 and for logpgp95 when good_property_right = 0. Interpret the differences in the conditional densities and provide a potential explanation for their differences in means. [Stata hint: use kden- sity to draw density and use twoway to draw two density on the same graph] [3 marks]

4. Conduct hypothesis testing for difference in mean:

H0: mean(logpgp95 if good_property_right=1)=mean(logpgp95 if good_property_right=0) H1: mean(logpgp95 if good_property_right=1)mean(logpgp95 if good_property_right=0)

Report the difference means, 95% confidence interval for the difference in means, the p-value for the test, and whether the test implies a statistically significant result at 5% level. Provide a brief in- terpretation of your findings by computing the per cent change in the conditional mean of logpgp95 when going from good_property_right=0 to good_property_right=1. [5 marks]

5. Draw a scatter plot in Stata with avexpr on the horizontal axis and logpgp95 on the vertical axis, with a fitted line to visualize the relationship. Also, report the correlation coefficient (corr(avexpr, logpgp95)). Does the pattern in the plot align with your findings from questions 3 and 4? [3 marks]

6. Run the regression using avexpr as dependent variable and logpgp95 as an independent variable, assuming homoskedasticity.

lОgogo95i = 80 + 81 auezopi + ri                                                           (1)

Report the coefficient estimates, the standard error, and interpret the magnitude of the predicted change in the dependent variable corresponding to a one-standard-deviation increase in the inde- pendent variable. Test wehther the predicted change is statistically significantly different from 0 at 5% level, and report the 95% confidence interval for the predicted change. [4 marks]

7. Now we consider the error term might be heteroskedastic instead. Rerun the above regression us- ing heteroskedasticity-robust standard error; report the coefficient estimates and the standard error. Comment on any differences and similarity you observed. [2 marks]

8. Now, suppose you think that colonial experience might matter to both property rights protection in a country and economic development of that country. You want to add a variable colony into the regression. Run this regression and comment on the coefficients estimates of both avexpr and colony assuming homoskedasticity. [1 marks]

9.  Suppose you think the identity of the colonizer (i.e., British colonies vs French colonies) matters to both property rights protection in a country and economic development of that country.

● merge your original dataset AJR_base with colonyID which contains information on the iden- tity of the colonizer, using countries’ 3-letter shortnam as the identifier. [Stata Hint: you will need to use merge 1:1. If you need any help, type help merge in the command line]

● Now run the regression that uses logpgp95 as dependent variable, and avexpr, f_french f_brit, f_other as independent variables (right-hand side variables). Again, assume homoskedastic

error term.  (1) comment on each coefficient estimated from this model.  (2) interpret coef- ficient on f_french f_brit, f_other, and (3) discuss whether the identity of the colonizer (i.e., British colonies vs French colonies) matters to log GDP per capita. [4 marks]

10. By running equation (1), we implicitly assume E(ri1auezopi)  =  0.   Explain in words what E(ri1auezopi)  = 0 means in this context.  Propose two possible scenarios in which the above assumption would be violated and, in each case, explain the expected direction of bias. [6 marks]

11. Please attach all your codes in your .do file at the end. [2 marks]

● 2/2 if your assignment presentation is neat and clean, and your Stata code is organized and for each step you do, there is a comment explaining your code. (i.e., think about you will need to make your code understandable when you read your code in 2 years time.)