You are on page 1of 2

EC310K

Wilfrid Laurier University


Winter 2021
Individual STATA Assignment # 1

You have been provided with a simplified dataset from a paper by Nathan Nunn and Nancy Qian (AER,
2014). In the paper, the authors explore the effect of US food aid on conflict in recipient countries.

If you run a simple OLS regression where conflict is the dependent variable, and US food aid is the
dependent variable, it is likely that the regression coefficient on the food aid variable will be biased. In
order to solve this identification problem, the authors employ US wheat production as an instrumental
variable for US food aid. The strategy is based on the idea that US wheat production in a given year will
be correlated with the amount of food aid disbursed by US in that year (relevancy condition). However,
wheat production in the US in any given year should not have any direct effect on the incidence of civil
conflicts in the recipient countries (exclusion condition).

The dataset contains the following variables:

1) recipient_country - a variable that identifies the country that the observation is related to
2) year - a variable that identifies the year that the observation is related to
3) intra_state - a variable that captures the incidence of intra-state (civil war) in recipient countries
in that year
4) wheat_aid - a variable that captures the amount of U.S. wheat aid (in 100,000 of tonnes) received
by the recipient country in that year
5) US_wheat_production - a variable that captures total U.S. wheat production (in 100,000 of
tonnes) in that year

By the assignment deadline, you are to upload a do file, and a document that provides answers to the
questions below. The do file should perfectly replicate all aspects of your analysis, including loading the
dataset. Upload your files to the designated dropbox folder on MyLS by 11:59 pm on Tuesday, February
23rd.

Download the dataset Conflict_foodaid.dta from MyLS under Content->STATA Assignment # 1

1. Estimate an OLS regression where intra-state conflict is the dependent variable, and wheat aid is
the independent variable. Report the regression coefficient and standard error. [1 mark]
2. Discuss at least two reasons/examples why the coefficient on wheat aid might be biased. For each
source of bias you identify, determine the direction of bias. You may find the discussion in page 2
of the introduction of the paper "US Food Aid and Civil Conflict" by Nathan Nunn and Nancy Qian
(AER, 2014) helpful. You will find the paper in MyLS -> Content -> STATA Assignment # 1. [3 marks]
3. Estimate the first-stage regression. Report the regression coefficient and standard error. Does
this mean that the relevancy condition holds? (In the context of IV estimation, the first-stage
regression refers to the regression where the endogenous explanatory variable of interest is the
Y variable, and the proposed instrumental variable is the X variable.) [2 marks]
4. Estimate the "reduced-form" regression. Report the regression coefficient and standard error. (In
the context of IV estimation, "reduced-form" regression refers to the regression where the
original dependent variable of interest is the Y variable, and the proposed instrumental variable
is the X variable.) [1 mark]
5. Now, estimate the effect of food aid on civil conflicts by using the instrumental variables
approach. Report the coefficient and standard error. [2 marks]
6. Compare the two coefficients from (5) and (1): what does this tell you about the direction of bias
in (1). [1 mark]

You might also like