1. Visit the BHPS website (https://www.iser.essex.ac.uk/bhps) and familiarize yourself
with the basic structure and contents of the BHPS data. What features make it a
suitable data set for the estimation of the BM model?
2. Open the file BM_data.dta and create a do-file in which you will code answers to
the following questions.
(a) What is the sample size? What is the sex ratio in the sample?
(b) What is the sample unemployment rate? What is the sample unemployment rate
2
of men? Of women? Or workers in each education category?
(c) What proportion of initial spells are right-censored (end of the spell not observed
in the data)? Answer the same question for each type of first spell (employment or
unemployment spell).
3. Still in the same do-file, construct the initial (spell-1) CDF of paid log wages logw1
(call it G) and its density. Produce plots of those two objects.
4. Create a variable categorizing logw1 into 25 percentile bins, and a variable contain-
ing the mean spell-1 duration (spelldur1) within each of these bins. Plot those mean
durations against the wage percentiles. Is this consistent with the BM model?
5. Derive the formulas for per-period average job-to-job (J2J), job-to-unemployment
(J2U ), and unemployment-to-job (U2J) transition rates in the BM model. Show that
those are “distribution-free”, i.e. do not depend on the wage offer distribution F or the
wage paid distribution G.
6. Estimate the monthly average J2J, J2U, and U2J rates from the sample of 2,263
workers.Show that x, λe and λu are identied from those three moments alone.
For students of EF5411, you can instead choose to write a referee report for one of
four papers: Pissarides (1985), Shimer (2005), Lucas and Prescott (1974) or Alvarez
and Shimer (2011); those who do the data analysis will be rewarded with better grade
though.
See https://scholar.harvard.edu/files/apassalacqua/files/refguidelines_tepe.pdf for guide-
lines to write a referee report.
For students of EF8078, also answer the following question.
7. Explain how one can obtain a non-parametric estimate of the wage sampling distri-
bution F from the data in BM_data.dta. Construct this non-parametric estimate, and
plot it on the same graph as G. Is this consistent with the theory? What else can you
say about this estimate of F?
