FB, 03/03/25
ECON UH-4000 : Economic Policy
Assignment option: Simulating tax reforms in US, including basic income
Notes:
a) This assignment is an alternative option to the essays that will count for the final grade
Students choosing to complete this assignment do not have to write an essay, and vice-versa
b) The completion of this assignment requires a basic knowledge of Stata, but nothing
elaborate.
This assignment consists of simulating on a sample of around 60,000 real households the effect of
changing some features of the redistribution system in the US
Material made available
The material made available for the assignment comprises the following.
1) The original dataset "hbasicinc_base.dta"
It consists of a large subset of households drawn every year in March -so-called current population
survey - by the US Census Bureau to analyze the distribution of income. Only a few variables have
been kept. They are listed in Appendix A. Each record in the dataset corresponds to a single
household but note that individual characteristics like age, gender or work status are those of the
person who answered the questionnaire, that is the 'head' of the household or his/her
spouse/partner.
(A dictionary of the value labels of the key variables is available in
[Link]
extracts/2017/cps-asec-research-file/[Link] )
2) A Stata do file : "[Link]"
This do file generates the base redistribution data needed for the analysis. It includes a simplified
version of the calculation of the income tax as well as of the Earned Income Tax Credit (EITC),
according to the 2019 schedule, as presented in chapter 4's slides in the course. Three variables are
key:
- hprimary_val = primary income of the household (earnings, pensions, financial income and
unemployment compensation)
- disp_inc = disposable income (before EITC) resulting from adding transfers to primary income
and resting the income tax (for those who would like to disentangle the two effects, they can
use the variable htotval wich is the sum of primary income and transfer income only.)
- Incweitc = disposable income + Earned Income Tax Credit
Note that you'll have to edit this do file in order to simulate reforms in the redistribution system.
3) A Stata do file "[Link]"
1
This file evaluates the mean income and percentage of total income accruing to the percentiles of
the distribution of a variable named "y" with associated observational weights named "w". From the
dataset you are working with, it is just necessary to just create these y and w variables to get the
percentiles of any income variable.
This do file creates 4 variables for all observations: the percentile it belongs to ("rr"), that percentile's
mean y-value ("my"), the percentage of the total value of accruing to the percentile ("smy"), the
overall mean of y ("ybar"), and the Gini "gini") coefficient of the whole distribution. The variables rr,
my and smy are identical for all observations belonging to the same percentile, and the variable ybar
and gini are the same for all the observations. Yet, note that individual data are still available for each
observation. In other words, a record now includes individual household data and the mean and
income share of the percentile it belongs to.
Note that the do file "percentile" may need to be modified if you want to obtain other characteristics
of percentiles, like mean age or race composition.
To filter that dataset and obtain only the percentile information you can enter the following Stata
statements:
keep rr my smy ybar gini
sort rr
by rr: drop if _n>1
If you prefer to work with deciles you can easily transform this dataset with the few following
statements:
gen rrd = 1+floor(rr/10)
replace rrd=10 if _n==_N
sort rrd
by rrd: egen myd=mean(my)
by rrd: egen smyd=sum(smy)
by rrd: drop if _n>1
Then you may want to export the resulting files to an Excel folder which will allow you to make easer
comparison across various income concepts and redistribution systems.
Issues to be studied
A) The present system
Identify and discuss the nature of the changes in the distribution of income due to the
various redistribution instruments – transfers, taxes and EITC. What part of the distribution
do they affect? How big is the change they produce in the welfare of individual percentiles.
Note: Instead of working with 100 percentiles, you may choose to focus on some specific groups, for
instance, the first four quintiles, the 9th decile, the 19th vintile, and the 4 percentiles from 96 to 99,
and finally the top 1%.) You may also look at alternative inequality measures to the Gini, like the ratio
between the mean income of the top 1% or top 10% to bottom 50% and other similar inter-decile
ratios. These are operations you can easily do starting from the Excel files comprising the 100
percentiles.
2
Note that the welfare measure of people should be defined as the income of the household where
they live divided by the number of consumption units (hn_csunit) in that household. To get the
distribution of that measure among individuals, it is sufficient to weight each household observation
by the number of persons in it (on top of the sampling weights included under "hsup_wgt")
B) The 2018 Trump reform
The tax schedule implemented in "datagen" is the 2019 schedule, after the reform passed by the
Trump administration. That the income data refer to 2017 is of lesser importance because inflation
has been low during this 2-year interval. It is possible to get an idea about the impact of Trump's
reform by simply changing the tax schedule used to compute the income tax in datagen for what is
described next.
Note that, to simplify the calculation of the income tax, it is assumed in datagen that everything is as
if all households were filing a tax return separately, sharing the total income of the household and
the deduction equally among potential filers – 1 for a household with a single adult, 2 within
households with more than 2 adults. A more detailed description of the income tax calculation would
have taken too much of your time.
The 2017 tax schedule- prior to the Trump reform was the following.
By changing the various parameter of this system in datagen you'll be able to see how the disposable
income of households has been modified by the Trump reform and the difference it made in the
distribution of welfare and its degree of inequality by generating the same indicators as in the
previous question.
C) Budget neutral increase in progressivity
Going back to the 2019 tax and EITC schedule, change the last bracket to a marginal tax rate of 45 %
and allocate the proceeds to an increase in the EITC, for instance in raising the eligibility ceilings in
the same proportion for all schedules – and adjusting the phasing-in and phasing-out rates
appropriately. The trick is that you want this reform to be budget neutral, so that you don't know a
3
priori by how much the EITC ceilings must be adjusted. You must then proceed by tatonnement,
trying different values of the relevant parameters.
Once you have found the right adjustment, compare the distribution of living standards that you
obtain with the base results in question A.
D) Basic income
Now you want to simulate a major reform many politicians talk about, a basic-income/flat tax
system.
The new redistribution system is now given by:
A basic income B given to everybody
A flat tax at rate t on all primary incomes from the first dollar earned
A super-tax at rate T for the richest people (above some limit L $ per consumption unit) in order to
maintain some progressivity at the top of the distribution.
You can see that if YP is the primary income per consumption unit of a household, its disposable
income, yp, with this redistribution will be given by:
yp=B – t . YP−T . Max [(YP−L),0 ]
But you want this reform to be budget neutral. Thus the mean disposable income per consumption
unit obtained for the whole population with this reformed system, should be identical to the mean
obtained in the base run in question A).
1) To avoid your wasting too much time, start with the following parameters : t = 20%, T= 20%
and L =$ 200,000 and find out the value of B that ensures budget neutrality. After making the
necessary changes in "datagen", compare your results with the base case in A and most
importantly identify the winners and the gainers of the reform. Discuss.
2) Depending on the conclusion of the preceding discussion, what kind of change would you
recommend among the 3 parameters, t, T and L ?
3) Are there limits to these parameters? Why? What do you think should be the maximum
values for them, and what dose that imply for B?
Note: you could run additional simulations to answer the last question. At this stage,
however, you musty have understood what is at stake in a basic income reform and be able
to discuss it qualitatively rather than quantitatively.