Map and document relationships between PUF variables and IRS spreadsheets

[donboyd5]

To set national targets based on IRS data, we need to know how PUF microdata variables correspond to totals published in IRS spreadsheets and reports -- totals that we have read into a "potential targets" data file.

Overview:

1. Determine appropriate mapping for 2015, the year of the PUF
2. Determine what to do if the IRS-published values for 2022 differ in concept from IRS-published values (and PUF concepts) for 2015.

A PR (to come) associated with this issue will construct the necessary mapping and document it.

[donboyd5]

Basic process for 2015

• think of a target concept that we want (e.g., total AGI for tax filers)
• read PUF documentation to figure out what variable is supposed to represent that concept (in 2015 it would be E00100)
• examine IRS spreadsheets and the main IRS document (Publication 1304) for 2015 to see what variables purport to represent the same concept (for example, in 2015, AGI would be found in column D of 15in11si.xls, as well as in other spreadsheets)
• calculate the weighted PUF sum for 2015 for the variable in question, which gives us the sample-based value
• verify, in PUF documentation, that the reported sample-based value is the same as the calculated value; it should be
• lookup, in PUF documentation, the universe value for the same concept (for most important variables this will be within a percentage point or so of the sample-based value, although for some variables the difference will be much greater)
• compare the PUF-documentation-reported-universe total for the PUF concept to the IRS-spreadsheet-based total for 2015; if you've got the right correspondence, the values should be the same or nearly so
• read the variable definitions in the PUF documentation and in Publication 1304 to ensure that they appear to be measuring the same thing
• examine tax forms for 2015, and the 2015 tax forms in the PUF documentation, and the discussion of tax forms in Publication 1304, to see if the PUF value generally comes from the same line and form as the IRS-published statistics
• if the values are the same or nearly so, and the PUF documentation and IRS documentation say nearly the same thing, and the location on tax forms appears to be the same according to the 2 documentation sources, then we probably have the right correspondence for 2015

Usually this works like a charm for 2015, although it is quite labor-intensive: there usually is a one to one correspondence between PUF and IRS-published statistics. Sometimes further detective work is required - multiple columns in IRS spreadsheets may need to be combined, or PUF variables may need to be decomposed into positive and negative values to find the right match between 2015 PUF variables and 2015 IRS-published statistics.

[donboyd5]

Process for 2022

The latest year for which we have detailed IRS data is 2022 so we want our 2022 PUF-based data to be as representative of 2022 as possible - we want to get aggregates right, as well as the distribution of income, and other important measures. Thus, we want to target 2022, not 2015.

Unfortunately, some IRS-published concepts have changed dramatically between 2015 and 2022. For example, the number of itemized returns published for 2022 will be much lower than the number for 2015 because the TCJA of 2017 increased the standard deduction dramatically and curtailed some deductions. Similarly and perhaps more importantly, for a specific variable of great policy interest, the deduction for state and local taxes, the IRS-published data for 2022 will only report SALT deductions for those who itemized. This is a small fraction of the number of people who might be able to claim the deduction if the SALT cap were increased or the standard deduction were reduced. In this case, we want a PUF (or, properly tax-microdata-benchmarking -- TMD) variable that represents potential SALT deductions, not actual SALT deductions claimed. In previous iterations of the TMD, we constructed this by "growing" the 2015 SALT deductions to plausible 2022 values and targeting those values. In other words, we constructed a target that, in essence, represented potential SALT deductions in 2022 for the kinds of people who filed tax returns in 2015. We found this to work quite well.

SALT is probably the most difficult challenge but some other variables are challenging, too.

Where 2022 concepts are different from 2015 concepts in ways that require custom solutions (as with SALT) we will document the custom solutions.

[martinholmer]

@donboyd5, when resolving enhancement issue #380, you should check that you have a correct understanding of the e03400 and e09900 PUF variables. See issue #405 for details.

[donboyd5]

Thank you, @martinholmer. I will make sure to fix it. This issue would not have been uncovered in the mapping because we don't attempt to target either of those variables but it may be a few days.

[donboyd5]

Completed in the course of preparing PR #424