Stage III Documentation

February 2017

Stage I and Stage II of OSPC’s extrapolation procedure blow up and re-weight each variable in the PUF-CPS matched data file (PUF) in order to match macroeconomic projections of growth rates and aggregate totals. Stage I determines per capita grow (or “blowup”) factors using both macro targets and population growth rates. In Stage II, a linear programming algorithm is applied in order to adjust the weight given to each record so that all targeted variables will sum up to their respective targets, while non-targeted variables also stay in reasonable ranges.

While Stage II does target wage distribution, only the aggregate totals for all other income variables are targeted and the resulting distributions of those variables can be inconsistent with what publicly available tax data shows. Stage III fixes this error by applying an adjustment factor to each record in the PUF based on the level of AGI in the PUF. The factor is calculated so that the aggregate value of the variable targeted does not change, but the distribution more accurately reflects the data.

Procedure

For the targeted variable, Stage III uses information from SOI Tax Stats tables to determine what percent of its total is in each AGI bin for 2009-2014 (see appendix for the bin breakdown). The distribution is assumed to hold at 2014 levels for the years 2015-2026. These are the goal distributions.

For each year adjustment factors are needed, the targeted variable is extrapolated to that year using the same routine as in Tax-Calculator. The goal bin amounts are then found using the goal distributions and aggregate total found in the PUF.

The goal bin amounts are divided by the actual bin amounts, which are calculated using the AGI variable found in the PUF before final processing, in order to find a set of adjustment factors that can be multiplied by each record in each AGI bin so that bin totals reach their targeted levels.

While this process does benefit from its simplicity, there are some trade-offs:

• Because AGI bin is determined using 2009 AGI levels, it does not account for any change in AGI overtime that could result in a tax unit moving into a different AGI bin. Thus, the final distribution will not be identical to what is found in IRS tax data.

• Because the factor is only being applied to one element of income, any possible relationship between two types of income will be lost. However, in the case of interest income, there do not appear to be any strong correlations with other income items (see appendix).

Results

Despite relying on AGI level found in the original PUF rather than recalculating AGI each year to account for growth as Tax-Calc does, the final distribution when including the adjustment is much more representative of the distribution found in SOI data, as seen in figure A.

fig. A

Tax-Calculator Implementation

Running stage3.py in the stage3 directory will output a file called puf_ratios.csv. This file is placed in the taxcalc directory within the larger Tax-Calculator package. To apply the proper adjustment factor, Tax-Calc uses the variable agi_bin in puf.csv. This variable is added during the final preparation of puf.csv and is between 0 and 18 for each tax unit dependent on which AGI bin they fall in. Tax-Calculator reads in puf_ratios.csv in the Records class and applies the adjustment factors based on the value of agi_bin for each tax unit.

Appendix

AGI Level Bins:

AGI Level
Less than zero
\(1-\)5,000
\(5,000 -\)10,000
\(10,000-\)15,000
\(15,000-\)20,000
\(20,000-\)25,000
\(25,000-\)30,000
\(30,000-\)40,000
\(40,000-\)50,000
\(50,000-\)75,000
\(75,000-\)100,000
\(100,000-\)200,000
\(200,000-\)500,000
\(500,000-\)1,000,000
\(1,000,000-\)1,500,000
\(1,500,000-\)2,000,000
\(2,000,000-\)5,000,000
\(5,000,000-\)10,000,000
$10,000,000 and over

Correlations:

Income Item	No Adjustment	Adjustment
Wages and Salaries	0.091	0.089
Ordinary Dividends	0.217	0.206
Qualified Dividends	0.162	0.154
Business Income	-0.005	0.002
Capital Gains	0.171	0.170