Scribd
Upload
9 views

DE Assignment

The assignment for Development Economics involves analyzing the impact of a 1996 policy by the Election Commission of India aimed at reducing frivolous candidates in elections by increasing deposit amounts. Students are required to conduct a difference-in-difference analysis using election data from 1978-2007, focusing on independent and party candidates in open versus reserved constituencies. The assignment includes data preparation, analysis through regression models, and interpretation of results regarding the policy's effectiveness.

Uploaded by

shravani.22-25
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
9 views

DE Assignment

The assignment for Development Economics involves analyzing the impact of a 1996 policy by the Election Commission of India aimed at reducing frivolous candidates in elections by increasing deposit amounts. Students are required to conduct a difference-in-difference analysis using election data from 1978-2007, focusing on independent and party candidates in open versus reserved constituencies. The assignment includes data preparation, analysis through regression models, and interpretation of results regarding the policy's effectiveness.

Uploaded by

shravani.22-25
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Development Economics, Winter 2024, GIPE

Assignment
Posted on February 20, 2024
Due on March 22, 2024 by 11:59 PM

Question 1: Difference-in-difference Estimation


Context:
Here we will analyze the effects of a policy announced by the Election Commission
of India in the year 1996 to reduce the number of “frivolous” candidates in elections.
In Indian national or state elections, if any candidate failed to get one sixth of the
vote in an election in a constituency, then the candidate would have to forfeit his
or her money deposit. Prior to 1996, however, this deposit was 500 rupees for non-
SC/ST candidates and 250 rupees for SC/ST candidates. In 1996, the ECI increased
these deposits to 10,000 rupees and 5,000 rupees, respectively, for the two groups
of candidates. The purpose of this policy was to increase the cost of “frivolous”
candidates to run in elections. The hypothesis is that it would significantly affect
the independent candidates as most of them fail to get more then one sixth of the
vote in most elections. Additionally, because of the large difference in deposit fees
between SC/ST and non-SC/ST candidates, we conjecture that the effect of the
policy will be larger in constituencies that are open (where mostly non-SC/ST can-
didates run) compared to those that are reserved for SCs or STs (where candidates
belonging to only those groups can run). We therefore use a difference-in-difference
analysis to estimate the effect of the policy on number of independent as well as
party candidates. We use open constituencies as our treatment group and reserved
constituencies as our control group; elections that happened after 1996 is the post
period and elections prior to 1996 is the pre period.

Data Description:
The data uploaded on Google Classroom contain election results for state assembly
elections during the period 1978-2007. Each row gives information about one can-
didate running in a constituency in a state in a given election year. state_name is
the name of the state, year is the election year, constituency_no is an id for the
constituency, candidate contains the name of the candidate, party gives his/her
party affiliation, and position gives the ranking of the candidate in the election
based on votes. Therefore, the winner in the election has position = 1, and so
on. constituency_type tells us whether the constituency is open (“GEN”) or is

1
reserved for some group.1

Data Preparation:
Step 1 Create a dummy variable open that takes value one if a constituency in an
election is open and zero if it is reserved for any population group (SC, ST
etc.).

Step 2 Create a dummy variable Post that takes value one if the year is 1996 or later
and zero otherwise.

Step 3 Create a dummy variable indcand that takes value one if the candidate in an
election is an independent candidate. The variable party takes value “IND”
for independent candidates.

Step 4 For each constituency in each election, calculate the total number of indepen-
dent candidates. Store this information in a variable n_indcand. (This is the
only hard step in this exercise.

Step 5 For each constituency in each election, calculate the total number of candi-
dates. Store this information in a variable n_cand.

Step 6 Keep only the winners in the data.

Step 7 Calculate the total number of party candidates in an election in a constituency,


n_partycand = n_cand - n_indcand.

Data Analysis:
1.1 Plot in a graph the average number of total candidates (n_cand) and indepen-
dent candidates (n_indcand) in a constituency over the years. In the graph
the x-axis will have years and the y-axis will have number of candidates. There
will be two plots, one each for the two variables.

1.2 Compute the average number of independent candidates in the reservation and
open constituencies in the pre period. Compute the same in the post period.
You will have 4 averages to report.

1.3 What is the difference-in-difference estimate that you get using these averages?
Show the work using a table as described in the lecture slides.

1.4 Using the same approach, compute the difference-in-difference estimate for the
number of party candidates (i.e., for n_partycand).
1
Type of constituency is a feature of the constituency, and therefore, its value is repeated for
rows (candidates) that correspond to a particular constituency.

2
1.5 Run the following regression and report the output:
n_indcandct = α + β1 openc + β2 postt + β3 openc ∗ postt + ϵct
where n_indcandct is the number of independent candidates in constituency c
and election year t.
1.6 Why does openc not have the t subscript? Why does postt not have the c
subscript?
1.7 What is the estimate of β3 ? What is its interpretation? How does it line up
with your answer in [1.3]?
1.8 Run the following regression and report the output:
n_partycandct = α + β1 openc + β2 postt + β3 openc ∗ postt + ϵct

1.9 What is the estimate of β3 ? What is its interpretation?


1.10 How does the estimate of β3 in [1.5] compare with the one in [1.8]? Why is it
like this?
Maybe the analysis is not satisfactory because it is comparing any open con-
stituency with any reserved constituency in the entire country to estimate the ef-
fect. However, constituencies across the country are very different in terms of their
political climate. Moreover, different states go for election in different years. Hence
comparing constituencies across states may not be ideal. Therefore, let’s refine our
regression by putting state fixed effects, i.e., dummy variables for each state. It will
make comparisons of constituencies across states more meaningful, by averaging out
differences in outcome variables across states.
1.11 First encode the variable state_name to create ids for each state. Then run
the following regression and report the output:
n_indcandct = αs + β1 openc + β2 postt + β3 openc ∗ postt + ϵct
where αs is the state fixed effect, i.e., it is the intercept for state s (corre-
sponding to the constituency c).
1.12 What is the estimate of β3 now? How does it compare with your answer in
[1.7]?
1.13 Run the following regression and report the output:
n_partycandct = αs + β1 openc + β2 postt + β3 openc ∗ postt + ϵct

1.14 What is the estimate of β3 now? How does it compare with your answer in
[1.9]?
1.15 Does the exercise tell you anything about the effectiveness of the policy in
removing “frivolous” candidates?

3
Output:
Upload a single PDF file containing the code, the output of the regressions and an-
swers to the questions asked. Please write your name and division in the document.

You might also like