Replication Exercise #1
Incentive effects of pension plans on retention
1.
Read Cunha, Menichini, and Crockett (2015) “The retention effects of high years of service cliff-vesting pension plans,”
Economics Letters.
2.
Reproduce the 2 tables and the 3 survival curves presented in the paper using the Stata dataset “retirement_data.dta” You are also welcome to use R. I should be able to run your script without errors by only changing the parent directory.
3.
Additional questions to answer:
a.
Do you need to drop any observations? Why? Was there a different way to handle the missing data?
b.
In a few sentences, discuss the issues surrounding causal inference in this paper. Do you believe whether the removal of the 20-year cliff vest
causes
higher attrition? What are some threats to this causal interpretation of the empirical results?
c.
Extra Credit: See Footnote #2, and replicate the Kolmogorov-Smirnov tests of the equality of the distributions of YOS. What are the p-values of the 3 pair-wise tests?
4.
You are welcome to work with a partner – if you do, only one student should submit the answers with both students’ names on the documents.
5.
Submit (1) Stata do-file or R script; (2) Word or PDF file with 3 survival curve graphs copy/pasted from Stata/R. Include your answers to “Additional questions” in the Word file. Filenames should have your name, e.g. “Bacolod_replication.do”
Helpful hints:
·
The file “retirement_data.dta” is fully labeled with the following variables:
o
hiredate: This is the day/month/year in which the individual started service.
o
birthdate: Date of birth
o
sex: M==male, F==female
o
service: branch of service, Air Force, Army, or Navy
o
firstsepartiondate: Date of first separation from service (data on subsequent active duty and separations is not available)
o
superfund: The superannuation (retirement) fund the service member is subject to, either DFRDB or MSBS
o
officer: Officer==1, Enlisted==0
·
If you are not familiar working with dates, reading the Stata manual
https://www.stata.com/manuals/u25.pdf
may be a good first step. Keep in mind that you can calculate the difference between 2 dates by subtracting them.
·
Use the Stata stset function in order to perform the survival analysis. You went over this in MN4111.
·
R’s survfit does not quite replicate the plots. I think Stata filters out survivors, that is, plots the information for data with a separation date entry. It’s the only way I can replicate Stata results in R, at least.