Keep Factually independent
Whether you agree or disagree with our analysis, these conversations matter for democracy. We don't take money from political groups - even a $5 donation helps us keep it that way.
Which data sources show historical top 1% federal tax shares (IRS SOI, Treasury, Piketty Saez)?
Executive summary
Three major types of data series are commonly used to show historical top 1% federal (or national) tax/income shares: the IRS Statistics of Income (SOI) administrative tax tables and time‑series (which provide raw tax-return based shares and tables) [1] [2], the academic Piketty–Saez (and Piketty–Saez–Zucman, “PSZ”) series built from tax returns plus adjustments and distributional national accounts [3] [4], and Treasury / Office of Tax Analysis critiques and alternative estimates that rework tax data or reconcile tax and national accounts (Auten & Splinter reply; Treasury analyses cited in reviews) [5] [6]. SOI provides the underlying IRS micro and tabulated series; Piketty–Saez/Zucman publish adjusted top‑share series and methods; Treasury/other academic work produces alternative series and points of methodological disagreement [1] [4] [5].
1. The IRS SOI: the raw administrative record journalists point to
The Statistics of Income (SOI) program is the official IRS source that compiles samples and tabulations of individual income tax returns, publishes historical tables and “individual time series” files, and releases topical SOI Bulletins and data books that researchers use to compute top‑share measures [7] [1] [8]. SOI’s time‑series tables explicitly provide income and tax items by AGI class, high‑income tables (including “Top 400” and other high‑income returns), and downloadable historical Excel tables suitable for constructing a top‑1% share directly from tax returns [9] [1] [2].
2. Piketty–Saez (and Piketty–Saez–Zucman): adjusted “fiscal income” series
Thomas Piketty and Emmanuel Saez (and later with Gabriel Zucman) produced long‑run top‑share series that start in 1913 by building on tax return data and then adjusting them—adding back certain items and reconciling to national accounts—to create what they call “fiscal” or distributional national‑account measures capturing a very large share of national income [3] [4] [10]. Their papers and databases are the canonical academic series repeatedly cited for the sharp rise in the top 1% since the 1970s and for including capital gains in some versions of the series [3] [4] [11].
3. How the series differ in practice: concept, coverage, and adjustments
Key differences trace to definition and adjustment choices. SOI tabulations are direct aggregates from returns (AGI, taxes, Top‑400 lists) and are indispensable for replication, but PS and PSZ re‑weight, add non‑tax items, and reconcile to national accounts to estimate “pre‑tax national income” shares and to include capital gains differently across series [9] [4] [11]. Independent reviewers note that fiscal‑income concepts used by Piketty–Saez cover roughly two‑thirds of national income unless supplemented, and that unit of observation choices (tax unit vs. adult) also shift levels and trends [12] [11].
4. Alternate and government re‑analyses: debate and critique
Government analysts and other academic teams have produced alternative top‑share estimates and methodological replies. For example, Treasury/OTA and independent economists (Auten & Splinter) have published replies arguing PSZ can be an outlier on some measures and stressing different treatments of missing income, non‑filers, and capital gains [5] [6]. Review pieces and policy guides (e.g., JEC guide, CEPR commentary) show multiple series (Piketty–Saez, PSZ, Auten & Splinter, CPS‑based series) that broadly agree on rising concentration but differ on magnitude and timing because of methodological choices [6] [12].
5. Practical guidance for researchers and journalists
To reproduce or compare top‑1% federal tax shares: start with SOI individual time‑series and historical tables (public Excel downloads) for the raw tax‑return aggregates [1] [2]; then document whether you (a) include realized capital gains, (b) add non‑tax income or reconcile to national accounts (the PSZ approach), and (c) choose tax units, families, or adults as your unit [3] [4]. Cite both the SOI source for raw numbers and the Piketty/Saez/PSZ methodology papers when using adjusted distributional accounts [1] [3].
6. What sources say about uncertainty and “outliers”
Researchers explicitly flag uncertainty: Piketty–Saez series rely on internal SOI micro‑files for some periods and on published tabulations for earlier years, and critics point to specific methodological choices that can make PSZ appear as an outlier on some pre‑tax measures [4] [5]. The JEC and CEPR discussions stress that while different series show the same qualitative rise in top shares, quantitative differences reflect definitional and reconciliation choices—so no single series is uncontroversial [6] [12].
7. Bottom line for readers
If you want a direct, reproducible federal tax‑return picture: use IRS SOI tables (individual time series, Top‑400 reports) [1] [9]. If you want a long‑run, economy‑wide “pre‑tax” or distributional national‑account estimate that attempts to capture all income—including capital gains and off‑return items—use Piketty–Saez or Piketty–Saez–Zucman series but note their documented adjustments and the existence of government/academic critiques [3] [4] [5]. Available sources do not mention a single universally accepted “best” series—choices depend on concept and audience (not found in current reporting).