# T HE D ISTRIBUTION OF C APITAL G AINS IN

> Briefs, arguments, decisions, and more.

URL: https://www.frixlaw.com/law-library/documents/agency%3Airs%3A5e6da15f26e80b15

## Record

- **Collection:** Agency decision
- **Document type:** Agency decision

## Text

T HE D ISTRIBUTION OF C APITAL G AINS IN
THE U NITED S TATES ∗
Cole Campbell, Jacob A. Robbins†, Sam Wylde
December 23, 2024

Abstract
Booming stock, housing, and private business markets have driven
large capital gains in the United States, averaging 20% of national income
over the past two decades. Using internal IRS tax return data, this paper studies the distribution of these gains, and their contribution to income
inequality and tax progressivity. We fnd capital gains to be highly concentrated, with 75.7% fowing to the richest 10% and 45.3% to the top 1%.
Capital gains substantially increase inequality, raising the top 1% share of
income to 21.0%, compared to 18% in their absence. Due to low realization levels, effective tax rates on capital gains are only 5%. Accounting
for capital gains reduces the progressivity of the tax system, with fat rates
across the Haig-Simons distribution. We document evidence of substantial heterogeneity in returns and cap rates across income groups. Richer
individuals have higher owner and tenant occupied housing returns, own
businesses that sell for higher multiples, and lower property tax rates.

Keywords: capital gains, inequality, Haig-Simons income, wealth inequality,
tax progressivity, heterogeneous returns
JEL Classifcation: E12, E21, E52
∗
The views expressed in this paper are those of the authors and do not necessarily refect the
position of the IRS. Many thanks to the Statistics on Income division for their support of this
project, in particular Mike Weber, Mike Strudler, Justin Bryan, and Kelly Dauberman.
†
University of Illinois at Chicago, e-mail: jake.a.robbins@gmail.com

1

Introduction

The year 2021 yielded $16.2 trillion in aggregate capital gains on assets held by
households in the United States — 94% of net national income, and an amount
larger than wages, dividends, or interest income.1 While 2021 returns were unusually massive, economically signifcant gains are not uncommon: over the
past two decades, real asset appreciation averaged 20% of national income. This
paper studies the distribution of capital gains, and how they contribute to broad
measures of income inequality and tax progressivity. Most measures of income
inequality, including the Distributional National Accounts (DINAs) of Piketty,
Saez and Zucman (2018) (henceforth PSZ), do not include price appreciation as
part of their income measure.
We do so using internal tax data from 2002-2021. The advantage of tax data
is its comprehensiveness, spanning the entire income and wealth distribution.
The disadvantage is that only realized sales of assets directly show up on tax
forms, which are only a small fraction of total appreciation (the sum of unrealized and realized).

(a)

(b)

Figure 1: (a) Measures of capital gains, 1954-2021 (b) measures of capital gains,
2021. ‘Nominal KGs’ are total realized and unrealized aggregate capital gains.
‘Taxable’ gains exclude pension and nonproft gains. ‘KG in agi + Excl’ are
capital gains reported on tax returns plus an estimate of capital gains on certain
categories that are excluded by law (described in section 4). ‘Nominal KGs’
estimated from Financial Accounts data. ‘KG in AGI’ estimated from individual
tax fles.
Estimating total capital gains is necessary because most have never been realized on tax returns. Figure 1 presents data on aggregate nominal capital gains,
compared to realized. From 1954-2021, $116 T in total capital gains were ac1

Real capital gains were $5.97 trillion, or 39.2% of national income.

1

crued, but less than 20% of that was reported on tax returns.2 Total capital gains
are poorly proxied by tax realizations for three reasons: (i) a growing share of
realized capital appreciation is not subject to tax (ii) individuals can delay selling
assets, sometimes indefnitely3 (iii) capital gains reported to tax authorities are
nominal, and combine income over many periods rather than annual amounts.
We overcome these limitations to estimate total capital gains by following a
three step procedure: (i) link individuals to their specifc portfolio holdings (ii)
capitalize income (using heterogeneous capitalization factors when available) to
estimate wealth (iii) estimate capital appreciation by multiplying wealth by asset
class specifc returns (using heterogeneous returns when available).
For step (i), we develop new methods to capitalize private business wealth,
tenant occupied housing, and owner occupied housing, exploiting tax form data
that has become available through electronic fling. Matching individuals to their
specifc portfolio of assets is important because it allows for accurate valuation.
Knowing a property’s location and type or a business’s industry and size is crucial for estimating its market value. Our work linking individuals to their assets
allows the use of heterogeneous capitalization factors that vary based on the specifc characteristics of the asset. For private business wealth, we estimate the
businesses’ market values using price-to-earnings and price-to-sales multiples
that vary by industry, frm size, and legal form of organization. For rental and
owner occupied real estate, we capitalize property tax payments using type of
property and location specifc tax rates.
Business and real estate wealth is shrouded through opaque ownership structures, with individuals holding stakes in partnerships and trusts that themselves
own partnerships, tangled together in an immense ownership web.4 We develop
new methods to cut through the chain of ownership and assign wealth to the
underlying controlling stakes, using techniques from the network economics literature. We successfully trace 90% of total partnership assets to their ultimate
holders.
We document new evidence of substantial differences in capitalization rates
across the income distribution. Richer individuals own businesses that sell for
higher multiples: the average S-corp Enterprise Value (EV) to EBITDA ratio
for the top 1% is 9.0 , compared to 5.8 for the middle 40; for partnerships, the
top 1% EV/EBITDA is 8.6 compared to 5.9 for the middle 40. The rich also
live in areas with lower property tax rates, leading to higher wealth estimates
for this group. The average owner-occupied property tax rate is 1.02% for the
2

This fnding is in line with Bailey (1969), who estimate more than two thirds of accrued
gains on corporate securities were never realized during their holders’ lifetimes. Poterba and
Weisbenner (2001) fnd that for the richest estates in 1998, over half of the value is due to
unrealized capital gains.
3
Three common strategies for avoiding realizations are (1) borrowing again assets rather than
selling them (Ensign and Rubin (2021)) (2) passing on assets at death, which avoids taxation
through a loophole, the ‘step-up basis at death’ (Kopczuk (2016)) (3) funneling assets into Roth
IRAs which will not be subject to capital gains tax (Hemel and Rosenthal (2021)).
4
See Hess et al. (2024) for details on these partnership networks.

2

middle 40, compared with 0.90% for the top 1%. This also holds for rental real
estate: within property classes, richer individuals have lower property tax rates
for single family, multifamily, and commercial properties.
Higher cap rates for the rich lead to markedly larger estimates of wealth
concentration, relative to a baseline assumption of homogeneity. The top 1%
of the income distribution owns 73% of S-corp wealth with heterogeneous cap
rates compared to 64% with homogeneous; for partnerships, 81.3% compared
with 64.1%. We estimate the richest 1% own 8.93% of owner-occupied housing
compared with 8.30% if we assume homogeneous returns. For tenant occupied housing, even though we fnd evidence of heterogeneous cap rates within
property classes, there is heterogeneity in portfolio shares across classes which
largely cancels out the overall effect on wealth inequality.
We estimate capital gains allowing for heterogeneous returns for owner and
tenant occupied real estate, while imposing homogeneous returns for public equity, private equity, and pension wealth. We fnd that capital gains are distributed
highly unequally, and are the most concentrated form of income: the top 1%
(10%) received 45.3% (75.7%) of total revaluations over our sample period.
Capital gains are so concentrated that including them in a comprehensive
income measure substantially increases the level of inequality. Using a HaigSimons income defnition, which includes asset appreciation, the top 1% (10%)
of individuals received 21.0% (47.9%) of the income pie; their share without
capital gains is 18% (45%). Capital gains on public and private equity are the
main driver of the increased concentration, while housing and pension capital
gains lead to a more equal distribution.
We document new evidence of heterogeneous returns across income groups.
There is a positive gradient, with returns increasing in income rank. The average
capital gain yield for the top 1% for owner-occupied real estate is 2.00% compared with 1.64% for the middle 40. For tenant occupied housing, the average
total return for the top 1% is 9.06% compared with 7.55% for the middle 40.
Accounting for capital gains lowers the progressivity of the tax system. Because most appreciation is either unrealized or exempt from taxation, the effective tax rate on macro capital gains is 3.0% for nominal gains and 5.2% for real
gains, substantially below their statutory rates.5 Since this income fows mostly
to wealthier groups, measured tax rates decrease for the rich, causing a less progressive tax system. Overall, we fnd Haig-Simons tax rates that are largely
fat across groups: the middle 40% pays average rates of 27.3%, the 90th-99th
27.0%, and the top 1% 26.8%.
The results of this paper are directly applicable to current policy questions related to capital gains. President elect Donald Trump’s Project 2025 would lower
capital gains tax rates and index them to infation, whereas President Biden’s
5

Under current law, for married couples fling jointly, long-term capital gain rates are 0%
for income less than $89,250, 15% for income between $89,250 and $555,850, and 20% for
income above this level. In addition, for flers with over $250,000 in income, there is a 3.8%
net investment income tax, which applies to the lesser of net investment income (which includes
capital gains) and the amount by which income exceeds $250,000.

3

2025 budget, endorsed by Vice President and Democratic nominee Kamala Harris, would introduce a tax on unrealized capital gains,6 as well as increase top
tax rates on gains to 28%.7 Our empirical fndings highlight (i) the large magnitude of the tax base (ii) the distributive impacts, which almost entirely fall on the
top 10% (iii) the current disparate treatment of capital gain income, which has a
low effective tax rate. This potential revenue source motivates policies that can
overcome the substantial challenge of collection and enforcement.8

1.1 Prior literature
This paper builds on Piketty, Saez and Zucman (2018), who study the distribution of national income, which excludes most capital gains.9 Aside from
adding capital gains to their income series, we materially modify their estimates
of wealth by including heterogeneous cap rates for business, owner-occupied,
and tenant occupied housing assets, which lead to higher estimates of wealth
concentration.
Our work is also closely related to Smith, Zidar and Zwick (2023) (henceforth SZZ) and Smith et al. (2019), who pioneered linking individuals to private
businesses using tax data. We follow their methodology for private business
valuation, with three major exceptions. First, we value businesses using valuation multiples constructed from two database of private business, while SZZ use
multiples from public frms, with additional adjustments for liquidity. Second,
we develop a new method to link partnership networks to their ultimate owners.
Finally, SZZ ultimately scale their estimates of private business wealth to totals
from the Financial Accounts, while we leave ours unadjusted. As a result, our
estimates of business wealth are appreciably higher, and lead to greater estimated
wealth concentration.
Two categories of papers have previously studied the distribution of capital
gain income. Most common are studies of realized capital gains. Piketty and
Saez (2003) study the distribution of taxable income, and in some specifcations
include capital gains. Feenberg and Poterba (2000) also include realized capital
gains in their study of top income inequality. The second category studies the
distribution of capital gain income by imputing returns based on asset holdings,
with early contributions from Goldsmith et al. (1954), Bhatia (1974), and McElroy (1971). Two papers closest in scope to this study are Armour, Burkhauser
and Larrimore (2013), who use data from the Survey of Consumer Finances
(SCF), and Larrimore et al. (2021), who also use tax data to estimate the distribution of Haig-Simons income. Our main differences with Larrimore et al.
(2021) are: (i) the inclusion of a wider variety of assets, including pension and
tenant occupied real estate (ii) our focus on real rather than nominal gains (iii)
our inclusion of heterogeneous cap rates and returns. These make a considerable
6

See https://taxfoundation.org/blog/harris-unrealized-capital-gains-tax/
See https://taxfoundation.org/blog/harris-capital-gains-tax-rate-historical/.
8
See discussions in Sarin et al. (2022) and Slemrod and Chen (2023).
9
But see discussion in section 2.1 on retained earnings.
7

4

difference in the inequality series: while their top incomes shares are generally
below that of the DINAs, our concentration measures are higher. Our estimation
of capital gain and Haig-Simons tax rates is related to Yagan (2023) and Saez
and Zucman (2019), who estimate Haig-Simons tax rates for the Forbes 400 and
likewise fnd very low tax rates for upper income groups.
Our work on heterogeneous returns is related to the burgeoning literature
that documents persistent disparities across the income distribution, as in Bach,
Calvet and Sodini (2020) and Fagereng et al. (2020). SZZ show that the homogeneous returns assumption does not hold for fxed income, and that in fact richer
individuals have higher returns, which leads to lower estimates of wealth concentration. Our paper provides new evidence on within-asset class heterogeneity
for real estate and private business wealth, showing that richer individuals have
(i) lower property tax rates, and (ii) own businesses that sell for higher multiples;
this heterogeneity thus leads to higher estimates of wealth inequality.
Finally, we contribute to a series of papers that attempt to disentangle the
ownership networks of pass-through business entities. Building on work by May
(2012), Cooper et al. (2016) match partnership income to their ultimate owners
using a recursive matching algorithm. They successfully allocate 77% of total
income, but encounter circular ownership structures and missing data that prevent them from allocating the remainder. Love (2021), using additional data,
succesfully tracks 99% of partnership income. Our paper tackles a separate but
related question. Rather than trying to account for the proportion of partnership income fowing to different types of partners (which is mainly a question of
identifying what entity the partners are), our problem is to track specifc assets
through the partnership networks and to their ultimate owners.

2

Data and methodology

2.1 Income and wealth concepts
National income, as defned by the Bureau of Economic Analysis, consists of
the total labor and capital income earned in production.10 This defnition explicitly excludes capital gains (which are not earned by production).11 Indirectly,
however, capital gains are present through the retained earnings of corporations.
Under a standard Miller and Modigliani (1961) assumption, a frm that retains a
dollar of earnings increases its market value by $1. Indirectly, then, at least some
capital gain income is accounted for in the national accounts income measure.
To isolate capital gains from other income fows, we defne ordinary factor
income as national income minus the retained earnings of corporations. Capital gains for year t are the real increase in the market value of an individual’s
10

In theory, this is equal to net national product, however statistical discrepancies mean in
practice they are not always equal.
11
BEA Handbook, Chapter 2, the “NIPA measures of income and savings exclude ... capital
gains and losses.” Fox and McCully (2009)

5

fnancial and nonfnancial assets over the period. Pure capital gains are capital
gains that do not include any national income, and equal total capital gains minus
retained earnings.
Figure 2 compares capital gains to other capital income sources in the postwar era. Pure capital gains are large, especially in the post-1980 period, where
they average 10% of national income. Retained earnings, the component present
in national income, averaged 4% of national income for the same period. The
large magnitude and persistence of pure capital gains suggests that analyses that
do not include this income source (such as the DINAs) provide an incomplete
picture of inequality in recent times.

(a)

Figure 2: Comparison of capital income. ‘Pure capital gains’ are real capital
gains of households after subtracting retained earnings, 5 year moving average,
estimated from the Financial Accounts. Other components of capital income
adopted from Piketty, Saez and Zucman (2018).
Haig-Simons income12 equals ordinary factor income plus capital gains. As
described by Hicks, Haig-Simons income “is what [one] can consume during
the week and still expect to be as well off at the end of the week as at the beginning.” In practical terms, it is measured as consumption plus change in wealth,
or equivalently (as in this paper) factor income plus capital gains.13
12
13

See Haig (1921), Simons (1938), and Hicks (1946).
Haig (1921) wrote that income is “the money value of the net accretion to one’s economic

6

In tax and inequality studies, Haig-Simons is often described as the “gold
standard” measure (see JCT (2012) or Armour, Burkhauser and Larrimore (2013)),
however some recent theoretical work raises doubts whether capital gains should
be included as part of income, especially if they are driven purely by changes in
interest rates. A rentier owning a consul that yields a real income stream of $100
per annum, who does not plan to sell, is neither better nor worse off if a decline
in interest rates increases the security’s market value.14 The income effect of a
change in interest rates may be completely independent of the change in asset
value for the holders. Auclert (2019) shows that for a temporary change in interest rates, the welfare effect is proportional to a household’s unhedged interest
rate exposure. Fagereng et al. (2024) show for the general case that the change
in welfare from an interest rate change is proportional to the present value of an
individual’s net asset sales. If all changes in asset prices were from changes in
expected returns, it would not be appropriate to include capital gains as part of
the income measure.
There are, however, three drivers of capital gains unrelated to changes in interest rates, each with broad support in the literature and strongly backed by empirical evidence. The frst is the rise of markups and profts in the United States,
documented in De Loecker and Eeckhout (2017) and Barkai (2016), which has
lead to capital gains in the stock market. Eggertsson, Robbins and Wold (2018)
and Caballero, Farhi and Gourinchas (2017) show how these changes in markups
can explain a large fraction of revaluations over the period. In a recent paper,
Eeckhout (2024) fnds that changes in dividends can explain 80% of the rise
in capital gains, while discount change explains 20%. A second force is unmeasured intangible investment or sweat equity (see Bhandari and McGrattan
(2018), McGrattan and Prescott (2010), and Hall (2001)), which shows up in
measured capital gains on private business assets. Intangible investment leads
to capital gains on business wealth, since the value of businesses increase beyond any measured investment. Finally, hetergeneous returns on real estate by
their very nature cannot be due to aggregate discount rate changes.15 These well
supported forces justify the study of the distribution of the capital gains of asset
holders. To partially address remaining issues, for our main results we do not
include any capital gains on directly held fxed income assets, and strip out any
capital gain on indirectly held fxed income assets from ETFs and mutual funds.
Household wealth is the market value of fnancial and nonfnancial assets
held by households, minus the market value of their liabilities. We largely follow the Financial Accounts in determining the asset and liability composition
of the balance sheet, with the exception of two asset classes where we provide
power between two points of time”, and Simons (1938) wrote that income is “the algebraic sum
of (1) the market value of the rights exercised in consumption and (2) the change in the value of
the store of property rights between the beginning and end of the period in question.”
14
See Krugman (2021) or Cochrane (2020)
15
Demers and Eisfeldt (2022) and Kahn (2024) fnd substantial heterogeneity in real estate
capital gains across cities. In section 5 we provide new evidence of heterogeneous returns across
income groups.

7

bespoke estimates: private business wealth and tenant occupied real estate. Our
primary measure (excepting private business and tenant real estate) is equal to
the Financial Accounts net worth for households and nonprofts, minus the net
worth of nonproft organizations, minus consumer durables.16
Following PSZ, our basic unit of analysis is the individual, with tax-return
income split equally across spouses. To allow for nonflers, we include a sample
from the Current Population Survey.
Our measure of income inequality is the top income share. Individuals are
sorted into percentiles based on yearly income, and the fraction of total income
the group receives is calculated. Our quantiles exhaust the income distribution:
bottom 50%, middle 40% (50th to 90th percentile), 90th-99th%, top 1%. Within
the top 1%, we further break down income to the 99th-99.9th, 99.9-99.99th, and
top .01.
For Haig-Simons inequality, it is useful to estimate reranked income shares.
These are estimated by frst ranking individuals into percentiles using ordinary
factor income, and then calculating each group’s share of Haig-Simons income.
We do so because Haig-Simons income is volatile, and in years of large losses the
bottom of the distribution will contain wealthy individuals with capital losses.
Reranking by ordinary factor income in this case cuts down on dramatic reranking of individuals across years, and eases interpretation over time.
Our measure of wealth inequality is the top wealth share, and we likewise
construct reranked wealth shares by frst ranking individuals on factor income,
then calculating the share of wealth held by that group.
We estimate confdence intervals for top income and wealth shares through
a survey bootstrap procedure, which takes into account the sampling error of
income ranks as well as the income shares of top quantiles.

2.2 Data
Our goal is to construct a dataset of individual level Haig-Simons income and
wealth. The basic building blocks for this are the Distributional National Accounts of PSZ, which we replicate with internal tax records. The DINAs contain
data on ordinary factor income at the individual level. We add to this our own
estimates of wealth and individual level capital gains.
Our primary data source is internal individual and business tax records from
2002-2021, provided by the Internal Revenue Service through the Joint Statistical Research Program. We use a wide variety of forms and subforms in our
analysis: the primary forms and line items used are summarized in fgure 3.
Table 1 presents summary statistics for the sample and key income variables.
We use Form 1040, the individual income tax return, to measure capital income that fows to individuals. We use a .2% (approximately 300,000 observations per year) weighted sample constructed by the IRS Statistics of Income for
all inequality analysis. Schedule A of form 1040 lists the itemized deductions,
16

Wealth portfolio components are described in section A.1.

8

which we use to estimate owner-occupied real estate wealth. We use Schedule
C to estimate sole proprietorship wealth.
Table 1: Summary Statistics
2002

2007

2012

2017

2021

338,342
172,157,842
230,631,608

350,624
182,560,333
241,963,680

438,281
191,450,996
249,433,156

45,834
66,701
11,310
3,584
78,010
371,380

59,634
75,235
24,437
8,448
99,672
516,356

6,669
776

7,388
1,624

Population and sample
Tax Filer Sample
Filer Population (weighted)
Adults (weighted)

175,315
151,024,628
206,200,680

AGI
Ordinary factor income
Capital gains (5yr)
Realized capital gains
Haig-Simons (5yr) income
Wealth

29,418
44,025
4,473
1,146
48,498
204,974

335,920
160,706,016
218,456,658

Income and wealth averages (current $)
39,831
54,638
-7,044
4,175
47,593
298,932

39,727
57,980
13,229
2,660
71,208
283,310

Tax averages (current $)
Federal income tax
Federal capital gains tax

4,037
225

5,360
795

5,057
487

For tenant occupied (t.o.) real estate that is directly owned, we use Schedule
E. For tenant occupied real estate indirectly owned through partnerships and
S-corps, we use Form 8825. For estimating private business wealth, we use
Form 1120-S for S-corps and Form 1065 for partnerships. To identify who
owns the businesses, we use Schedule K-1 of Forms 1120-S and 1065. For
the employment of frms, we use Form 941 and Form W2. When analyzing
aggregate t.o. and business wealth, we use 100% samples. For the inequality
analysis, we use the .2% (matched) sample.

9

Figure 3: Primary tax forms used in estimating wealth and capital gains.
10

We combine the tax fles with a variety of secondary data sources. Data on
aggregate wealth and capital gains is from the Financial Accounts of the Federal
Reserve. Data on national income components is from the National Income and
Product Accounts (NIPAs) and Fixed Asset tables of the Burea of Economic
Analysis (BEA). We impute a number of wealth variables using the Survey of
Consumer Finance (SCF) of the Federal Reserve.

2.3 Estimating wealth
Our estimation of capital gains follows a three step procedure: (i) link individuals to their specifc portfolio holdings (ii) estimate the market value of said portfolio through capitalization (iii) multiply wealth by price appreciation to derive
capital gains.
Capitalization is the process wherein measured income fows from the asset
are used to estimate the market value by multiplying the fow by a capitalization
factor. For a general asset class j, individual i, and year t, the value of the asset
is given by
Valuejit = Capital fowitj · Cap factoritj .
(1)
For example, we trace the ownership of a partnership business in the warehousing industry thorugh a holding company to its ultimate owner, then estimate the
market value by multiplying its (fow) EBITDA by a cap factor of 8.5. In this
case, the cap factor was estimated from a database of sales of partnership businesses in the same year that were in the same industry and a similar size.
The specifc asset identifcation and capitalization method will differ depending on the asset class, which we now describe.

2.3.1 Owner-occupied housing
For tax units that are itemizers, we identify homeowners through property tax
deductions listed on Schedule A, and the specifc location of the home through
the zip code address listed on the tax return. To estimate the value of the property,
we capitalize an individual’s property taxes, with a capitalization factor equal to
the inverse average property tax rate for the county, estimated from the American
Community Survey and Decennial Census for the year. For an individual i living
in county c in year t, the value of their house is estimated as
O.O. House Valueict =

Property tax paymentit
.
Property tax ratect

(2)

Appendix fgure A.1 shows that this procedure captures about 80% of the aggregate value of owner-occupied housing from the Financial Accounts. Following
PSZ, we scale the value of itemizers to exactly equal 80% of FA values, and
allocate the remaining 20% to nonitemizers and non-flers using averages from
the SCF.

11

While our approach to the valuation of owner-occupied homes is sensitive
to geographic variation in property tax rates, it does not account for withinjurisdiction variation in assessment ratios. To the extent that richer individuals
face lower assessed value-to-market value ratios within taxing jurisidicitions, we
will underestimate owner-occupied housing wealth, and therefore capital gains,
´ and Howard (2022) provide evidence that withininequality. Avenancio-Leon
jurisdiction assessment gradients exist along racial lines such that minorities face
systematically higher assessed value-to-market value ratios, although they do
not explicitly investigate the assessment gradient according to income. Others,
such as McMillen and Singh (2020), who observe market values and assessed
values directly fnd that homes with higher market values face lower assessment
ratios, with some evidence that the same regressivity exists when homeowners
are ranked by income.
The Tax Cut and Jobs Act (TCJA) of 2018 increased the standard deduction from $6,500 to $12,500 for single flers (and likewise close to double for
other fling statuses as well), which correspondingly reduced the percentage of
itemized tax returns from 30% to 11%, hindering our ability to capture real estate wealth. In 2017, capitalized itemizers’ property tax payments accounted for
77% of Financial Accounts housing wealth, dropping to 40% in 2018. To accurately estimate housing wealth from 2018 onwards for non-itemizers, we use
the following procedure. First, we take the address of the tax fler post 2018,
and see if that same address fled a tax return in 2017. If they did, and itemized a property tax deduction, we can estimate the 2017 value of the house, then
update the value to the current year using the FHFA house price indices for the
area. This procedure increases the percentage of Financial Accounts housing
wealth accounted for to 75%. The residual wealth we impute using the same
SCF methodology as above.

2.3.2 Tenant occupied housing
We identify the tenant occupied properties owned directly by individuals via information on Schedule E, and those owned indirectly through partnerships and
S-corps from Form 8825. These data contain property level tax payments, the
location of the real estate, and the type of property, which we separate into three
broad categories: single family, multifamily, and commercial. We estimate property values using equation 2, where property tax rates are at the county-year-type
level. Data on effective tax rates is taken from the Lincoln Institute of Land Policy and the Minnesota Center for Fiscal Excellence.17
Our procedure is able to capture the majority of aggregate tenant occupied
real estate value, measured in either the Financial Accounts or SCF. Figure A.6
shows that in 2021, aggregate TO wealth is $16.3 T in the Financial Accounts,
$14.8 T by our reckoning, and $12.7 T in the SCF.
Appendix fgure A.5 provides details on the composition and ownership
17

For small cities and rural areas the effective tax rate is estimated; see appendix A.8.1.

12

structure of rental housing. About half is owned through partnerships, slightly
less than half owned directly, and the remainder through S-corps. Commercial
real estate is the largest component, followed by multifamily, then single family.

2.3.3 Private business wealth
We estimate the value of private businesses by capitalizing EBITDA and sales
from operating businesses, using a methodology developed in Campbell and
Robbins (2023). The enterprise value (EV) for company j is given by mutliplying j’s EBITDA by an EV to EBITDA multiple, M ultj ,
EBITDA valuationjt = EBITDAjt · Multjt .

(3)

To estimate appropriate cap rates, we follow the methodology of business appraisers, who form valuations from a comparison to similar businesses that have
previously sold.18 We estimate multiples at fnely grained levels: by legal form
of organization, industry, size, and year cells. Data on private business sales is
from two separate private transactions databases from Business Valuation Resources, which collects data on private business sales. Data on business level
sales and EBITDA comes from tax returns: Form 1120-S for S-corps, 1065 for
partnerships, Schedule C for sole proprietorships. We estimate a separate sales
valuation analogous to equation 3 by multiplying sales by an EV to Sales multiple, then take an average of the EBITDA and sales valuations to form our fnal
measure.
We estimate large totals for aggregate private business wealth (fgure A.17
and table A.1), comensurate with their economic importance in terms of employment and sales. Aggregate value is $17.4 T in 2017, much higher than the
Financial Accounts value of $8.6 T. The higher totals stems primarily from the
fact the Financial Accounts uses book values of partnerships for their valuations,
which are small in comparison to profts and sales. In 2017, for example, the
FA estimated partnership values to be only $2 T from book value, a year when
partnership businesses made $5.5 T in sales and $810 B in net income.19

2.3.4 Other wealth elements
To estimate fxed income wealth, we capitalize interest fows from fxed income
assets. The work of SZZ shows that wealthier individuals hold risky long duration assets such as corporate and government bonds, with high interest rates,
compared to individuals lower in the distribution who hold shorter duration safe
assets with low interest rates. To account for these heterogeneous returns, we
18

See Pratt (2006), chapters 11 and 12, or Goedhart, Koller and Wessels (2015), chapter 16.
The American Society of Appraisers Business Valuation Standards recognizes the “market approach” as one of the three pillars of business valuation. This is also recognized in the Institute
of Business Appraisers ‘Business Appraisal Standards’. For a recent paper using this approach,
see Smith, Zidar and Zwick (2023).
19
See section A.5.1 for a description of Financial Accounts business estimates.

13

use data from SZZ on the average fxed income yield by quantile in the wealth
distribution (by year) to capitalize taxable interest fows: 0-99th percentile, 99th99.9th, 99.9-99.99, and top .01%. Following SZZ and PSZ, we then gross up our
aggregates to match Financial Accounts totals for directly held bonds.
For indirectly held bonds held through mutual funds, money-market funds,
ETFs, and closed-end funds, we capitalize non-qualifed dividends under an
equal returns assumption using Financial Accounts totals. For munis, we capitalize tax-exempt interest. For currency, we follow PSZ and allocate Financial
Account wealth by income rank using SCF data.
For public equities directly held and held through mutual funds and ETFs, we
use the methodology of SZZ and capitalize a mixture of 90% qualifed dividends,
10% realized capital gains.

2.3.5 Allocating indirect business and real estate wealth
After estimating the value of private businesses and tenant occupied properties,
we trace the wealth to its ultimate owners. Using Form 1120-S K1, we measure
an individual’s S-corp business and real estate wealth as the fraction of the business they own multiplied by the estimated business value. Figure A.23 shows
the pass through of S-corp business values to their ultimate owners. In 2017, $7
T in S-corp business wealth is passed down to the K-1 level, where we capture
$6 T in value. The loss of $1 T is due to (i) some companies not fling K1s (ii)
insuffcient data on business ownership on the K1s. A total of $5.5 T in value is
accounted for at the individual level. This may be due to (i) some S-Corps being owned by nonprofts / estates / trusts (ii) foreign shareholders (iii) sampling
error.
The process of tracing wealth through partnerships is more diffcult due the
presence of partnership networks: the fact that a partnership can be owned by
another partnership. To cut through this chain of ownership, we use results from
the network economics literature20 to allocate value to its ultimate owner. The
key to this allocation is capturing the information contained in the network of
partnerships owning other partnerships.
A network of partnership ownership is denoted by matrix C, where Cij is the
fraction of partnership j that partnership i owns. The value of partnership i is
the value of its fundamental
assets, ai , plus the value of the shares of other frms
P
it owns: Vi = ai + j6=i Cij Vj . The vector describing the market value of all
partnerships is then given by V = a + CV . Solving for the total value of the
frm, V = (I − C)−1 a. To estimate V for either businesses or real estate, we
start with the estimate of the underlying assets a, then estimate the partnership
ownership matrix C using Form 1065-K1.
20

See, in particular, Elliott, Golub and Jackson (2014) and Galeotti and Ghiglino (2021).

14

Direct ownership value of partnership real estate assets
by partner type
Billions of Dollars
1,000 1,500 2,000 2,500

Full pass-through value of partnership real estate assets
by partner type

22%

42%

8%

500

0%
4%

Partnerships

(a)

d
te

ie

bu

tif

is

tri

en
ot
N

m

id

en

gn

tF

en

un

d

d

y
tit

of
it
rie

N

on

-p
r

or

ps

es
&

SC

at
Es
t

-C
C

Direct ownership

U
nr
ed

Not identified

ire

Foriegn entity

Retirement Fund

et

Non-profit

R

Trusts & Estates

S-Corps

Fo

C-Corps

Tr

In

di

vi

9%

Individuals

us
ts

11%

du

al

or
ps

s

0

1%
2%

Indirect ownership

(b)

Figure 4: (a) Direct ownership of partnership real estate (b) Indirect ownership
of partnership real estate
Figure 4 provides breakdown of partnership tenant occupied real estate ownership for the year 2017. A total of 42% of partnership t.o. wealth is owned directly by individuals, 22% are owned by other partnerships, and 34% are owned
by other entities (C-corps, trusts, and unidentifed entities being the largest shares).
After inverting the partnership ownership matrix, we are able to pass through the
full value of real estate to their ultimate owners. Figure 4 (b) shows that a full
51% of partnership real estate is ultimately held by individuals, 12% by C-corps,
16% by estates/trusts, and 21% by others. We allocate the 16% held by trusts to
individuals in proportion to their estate and trust income. For those owned by Ccorps, we do not attempt to allocate to individuals since this would be potentially
double counting the value that is already allocated through public equity wealth,
which includes REITs. For the fnal 23% we have little indication of ultimate
ownership status, and we leave unallocated.21
For partnership operating businesses, the aggregate enterprise business value
was $7 T in 2017. Of this, $6 T can be traced back to some K1 owner, of which
ultimately $2.7 T fows to individual ownership. Figure A.22 provides the full
breakdown of partnership business ownership.

2.4 Capital gains
We estimate capital gains by multiplying wealth by price appreciation. For individual i holding asset class j in year t, the capital gain is the market value of the
asset multiplied by the real price growth of the asset:
KGjt = MV Assetjit · % real price yieldjit .
21

Recent work by Love (2021) suggests these are likely foreign owned entities.

15

(4)

We estimate capital gains for the following asset classes: public equities, private equities, owner-occupied housing, tenant occupied housing, and pension
assets.22 We exclude capital gains on fxed income and debt.23
For owner-occupied housing, real housing price data is from FHFA24 price
indices at the 5-digit zip code level based on the location of the taxpayer. For
tenant occupied housing, price data on multifamily and commercial housing is
from Freddie Mac and the CoStar Group, based on the location of the property.
For other asset classes, capital gains are estimated with a homogeneous returns assumption: that individuals across the income distribution have the same
expected capital gain return. Real price yields are calculated at the macroeconomic level. For an asset class j, we compute yields by dividing the fow of
aggregate gains during year t by the total value of the corresponding asset at the
beginning of the year:
Yieldjt = KGjt /Wtj
(5)
Nominal yields are converted to real by adjusting for net national product infation. Data on aggregate nominal capital gains on assets held by US nationals, by
asset class, is taken from the Integrated Macroeconomic Accounts of the Financial Accounts.
The use of homogeneous returns is valid as long as individuals across the
income distribution have the same expected return on assets. We must use this
method because we lack data on heterogeneous returns. This is a serious limitation: for the asset classes we do have data for, we fnd strong evidence of heterogeneous returns and capitalization factors.25 This is line with prior work such as
SZZ who fnd heterogeneous returns on fxed income, and Fagereng et al. (2020)
who likewise fnd evidence of persistent heterogeneity in returns. To the extent
that the equal returns assumption is false, and richer individuals have higher returns, we will tend to understate the amount of capital gains inequality. To the
extent that richer individuals have lower returns, we will tend to overstate the
amount of capital gains inequality. Estimating heterogeneous returns on equity
wealth is an important question for future research.
Capital gains are volatile, and embodiment of the stock and housing markets
which drive them. This volatility poses a challenge for measuring and interpreting trends in Haig-Simons income inequality. In years when the stock and
housing markets boom, top-income shares increase, as capital gains are very concentrated. In turn, during stock market crashes, top-income shares drop. For this
reason, we will include smoothed measures of gains. Five year moving average
capital gains are calculated using equation 4, with yields for year t taken as a
fve-year geometric average of returns centered at year t.
22

For pension, we include only the capital gains that come from the equity component, and do
not include fxed income capital gains or other components.
23
We do so because they mainly consist of large yearly losses due to infation.
24
See section A.7.2 for details.
25
See section 5 below.

16

3

The distribution of capital gains in the United
States

(a)

(b)

Figure 5: (a) Average real capital gains per person, 2002-2021 (b) Average HaigSimons income per person, by income group. Haig-Simons income equals factor
labor plus (ordinary) factor capital plus capital gains. Capital gains totals are not
smoothed.
Figure 5 (a) presents average per person capital gains. Capital gains were large,
both in absolute terms and in comparison to other income sources. Across our
time frame, the average gain was $11,275 per person per year, about two-thirds
the magnitude of ordinary capital income ($18,719 per person), and about a ffth
of the size of ordinary factor income ($71,101 per person). Figure 6 (a) displays
capital gains by asset class. Although no one asset dominates, the largest component is public equities with an average of $5,548 per person, followed by
owner-occupied housing ($2,219 ), pension ($1,633 ), tenant occupied housing
($1,113 ), and private business ($762 ).
Figure 5 (a) shows the high degree of cyclicality of capital gains, with extreme lows in the Great Recession (a nadir of $-60,000 per person in 2008) and
dot-com bust, and large gains during the long post-2010 expansion. As shown
in Figure 5 (b), the cyclicality carries over into Haig-Simons income, which is
negative in 2008. The volatility of capital gains and Haig-Simons income using
1-year returns, and the fact that it is negative in 2008, pose a challenge for measuring and interpreting trends in income shares. For this reason, we next focus
on income series using 5-year moving average capital gain returns.

17

(a)

(b)

Figure 6: (a) Average real yearly capital gain income per person by asset class.
Every year the average capital gain per person is calculated, and bars display the
average of these across all years of the sample. (b) Average real yearly capital
gain income per person by factor income rank (log scale). Bars display averages
across all years of sample.
Figure 6 (b) shows that the upper parts of the distribution receive substantial
fows of capital gain income. Ranking individuals by factor income (i.e. without
capital gains), the average gain of the top 1% is $433,755 per person per year,
about 40 times greater than the average, declining to $30,535 for the 90th-99th
percentiles, $7,626 for the middle 40, and $2,278 for the bottom 50%. These
fows lead in turn to high concentration of capital gain income. Figure 7 shows
the share of overall capital gains for groups ranked by factor income. Over the
full sample period the top 1% grossed 37.7% of aggregate capital gains. The
share of gains by the top 1% is more than double their 18% share of ordinary
factor income, and similar to their 38% share of ordinary capital income. Their
share was the largest of any other income group; the next largest was the middle
40% (27.4%), 90-99th percent (24.7%), then bottom 50% (10.3%).

18

(a)

(b)

Figure 7: Share of capital gains by income group. Individuals are ranked by
factor income, and the share of total capital gains for that group calculated. For
graphing display purposes, extreme values above 150% have been trimmed.
The addition of capital gain income necessitates a reranking of individuals,
which in turn leads to a higher concentration of capital gain income. Figure
8 shows the share of capital income from individuals ranked on Haig-Simons
income. The top 1% received 45.3% of capital gains by this measure. As shown
in fgure 9, this high degree of concentration is driven by gains on public equity,
private business, and tenant occupied housing assets.

(a)

(b)

Figure 8: Share of real capital gains by income group. Individuals are ranked
by Haig-Simons (5 yr) income, and the share of total capital gains for that group
calculated. Bars display weighted averages across all years of sample, where
weights are the level of capital gains in 2021 dollars.

19

(a)

(b)

Figure 9: Share of real capital gain income by income group and asset class.
Individuals are ranked by Haig-Simons (5 yr) income, and the share of capital
gains for each asset class is calculated. Bars display weighted averages across
all years of sample, where weights are the level of capital gains of the given asset
class in 2021 dollars.
The large magnitude of capital gains combined with their concentration leads
to high levels of Haig-Simons inequality. As a baseline, we frst describe the
distribution of ordinary factor income, shown in fgure 10, red line. The top 1%
received 15% of factor income in 2002, steadily growing over time to 20% by
2021. The increase is relatively smooth, and there is little cyclicality in income
shares. The green series shows the distribution of Haig-Simons (5yr) income,
ranked on factor income. This series, which adds capital gains to factor income (while preserving the ranking of individuals) increases the top 1% share
to 20.5%, exceeding the factor income share of 18%. There is also more volatility and cyclicality, even in this smoothed capital gain series, with shares rising
as high as 28%. Finally, the gold series shows the distribution of Haig-Simons
(5yr) income, ranked on Haig-Simons income. The reranking has a moderate
effect on income shares, increasing the top 1% share to 21.0%.

20

(a)

(b)

Figure 10: Comparison of income shares: factor, Haig-Simons ranked on factor
(r.f.), and Haig-Simons, ranked on Haig-Simons. Haig-Simons calculated using
using 5 year average of capital gain returns.
Figure 11 shows the Haig-Simons (5yr) shares for the rest of the distribution.
The top 1% is the only group that shows an increase over the time period; this is
made up by declines for the bottom 50% and middle 40%, while the 90th-99th
is relatively fat. Even within the top 1%, it is the top part of this group, the
99.9-99.99 and top .01 group, that shows the largest gains.

(a)

(b)

Figure 11: Average 5-year Haig-Simons income share by income group. Individuals are ranked on Haig-Simons (5 yr) income.
Figure 12 summarizes the changes in income share across the distribution by
income measure. Income concentration increases moving from fscal to factor
income, factor to Haig-Simons (rank factor), and fnally to Haig-Simons (rank
Haig-Simons).
21

(a)

(b)

Figure 12: Income share comparisons. ‘Fiscal’ income is income reported on
tax returns. Haig-Simons income uses 5 year moving average for capital gain
income.
To better understand which specifc capital gains increase inequality, fgure
13 provides a decomposition. For each percentile ranking we start with their
share of factor income, and then add asset-specifc capital gains one at a time,
calculating changes in income shares after adding in the gain. Since we do not
rerank individuals, this process provides a lower bound on the concentration
effects of capital gain income. The two largest drivers of increased income concentration are public and private equities; this is not surprising, given that these
are the most concentrated forms of income. On the other hand, pension and
owner-occupied housing gains tend to decrease concentration— this decrease
comes about since the underlying assets are relatively evenly dispersed.

22

(a)

(b)

Figure 13: Contribution of capital gains to changes in income share. Bars represent the marginal effect of capital gain income source on the share of income for
the 1% (or top 10%) income group. Individuals ranked on factor income.
One additional implication of including capital gains as part of the income
measure is that the capital share of income increases. Figure 14 shows that for
the top 1%, the capital share increases from 55% without capital gains to almost
70% with capital gains. For the 90th-99th percentile, the capital share increases
from 25% to 35%.

(a)

(b)

Figure 14: Capital share of income, by measure. ‘Ordinary’ capital share equals
ordinary capital income divided by factor income. ‘Ord. + KGs’ equals ordinary
capital income plus capital gains, divided by Haig-Simons income.

23

4

Taxes and tax rates

4.1 Capital gain tax rates

(a)

(b)

Figure 15: (a) Nominal capital gain tax rates (b) real capital gain tax rates. Tax
rates calculated as total capital gains tax revenue, divided by alternate measures
of aggregate capital gain income. ‘Macro’ capital gains are total realized and
unrealized capital gains by households and nonprofts. ‘Macro hh’ are total realized and unrealized gains by households. ‘HH txable’ are total realized and
unrealized gains by households in taxable categories, i.e. excluding pension and
nonproft. ‘KG in agi + Excl’ are capital gains reported on tax returns plus an estimate of capital gains on certain categories that are excluded by law (described
in text). See section A.6.
The macroeconomic tax rate for capital gain income is defned as:
Aggregate household capital gain taxes paidt
Aggregate realized and unrealized gains households and nonproftst
(6)
Figure 15 shows that the macro tax rate for capital gains is quite low, averaging
3.0% for nominal gains and 5.2% for real. In 2021, total tax revenue from capital
gain income was $379 B, while aggregate nominal capital gains was $20.5 T, for
KGT RtMacro = 1.8%. For real gains, the total macroeconomic rate was 3.2%.
The realized tax rate is the tax rate on gains realized and reported on tax
returns:

KGT RtMacro =

KGT RtRealized =

Aggregate household capital gain taxes paidt
.
Total capital gains reported in AGIt

Figure 15 presents data on the realized rate, which greatly exceeds the macro
rate— in 2021, it was 18%.
24

The macro tax rate is much lower than the statutory rate because only a
fraction of macro capital gains are reported on tax returns— of the $20.5 T in
2021 macro gains, $2.1 T was reported. There are three reasons why aggregates
dwarf tax-return gains: (i) a signifcant portion of capital gains are not taxable
by law, including those in pension funds and held by nonprofts (ii) some part of
gains are not subject to tax due to exclusions, such as those for selling a primary
residence, 1031 exchanges when selling tenant occupied real estate, and small
business sale exemptions26 (iii) gains are only taxed when realized.
Figure 1 (b) provides a breakdown of the difference between aggregate and
tax return capital gains in 2021. Of the $20.5 T in gains, $3.8 T are held in
pensions or non-profts, and thus are nontaxable. $2.1 T were reported on tax
returns as part of AGI, while $.3 T in gains were excluded from returns. The total
that can be accounted for is thus $6.2 T, leaving $14.3 T which were unrealized
(or unmeasured).27 For the entire 1954-2021 period we have data on, there have
been $116 T in gains: $18 T nontaxable, $20 T realized and reported on tax
returns, $6 T excluded, and $72 T unrealized.
26

We estimate the total amount of sales that are excluded from tax returns using the following
method. First, the size of the ‘tax expenditure’ for each category (e.g., exemption for the sale of
primary residences) of excluded gains is taken from the Joint Committee on Taxation’s yearly
estimate (see JCT (2008)). This yields the estimated tax revenue that capital gain category would
have yielded in the absence of the exemption. We then use the average capital gain tax rate, in
combination with the tax expenditure, to back out the size of the capital gains not reported on
tax returns.
27
Prior research consistently shows that capital gains reported on tax returns captures only
a small fraction of total capital gains. Bourne et al. (2018) link federal estate tax returns from
decedents in 2007 to panel data on income tax returns prior from 2002-2006. Although this was a
period of very high returns in the stock and housing markets, the majority of wealthy individuals
reported nominal returns on capital to the IRS of less than 2%. Steuerle (1985) and Steuerle
(1982) also provide evidence that realized capital gains bear little relation to actual returns.

25

(a)

(b)

Figure 16: Average macroeconomic capital gains tax rate, by income group. Calculated as in equation 6. Individuals ranked by Haig-Simons (5yr) income. Bars
are weighted averages across years, with weights equal to the level of capital
gains in 2021 dollars.
How does the KGT RtMacro vary across the income distribution? Figure 16
shows that the top 1% macro tax rates exceed those of the lower percentiles.
This is a function of (i) greater percentage of gains that taxable by law (fgure
A.26, i.e., non-pension and housing gains) (ii) steeper tax rates on realized gains
due to being in a higher bracket (fgure A.27) (iii) a higher percentage of gains
that are realized (fgure A.25).

4.2 Haig-Simons tax rates
The Haig-Simons tax rate is defned as the total amount of direct and indirect
taxes paid, divided by Haig-Simons income:
HST Rt =

Total direct and indirect taxest
.
Haig-Simons incomet

(7)

For estimation of the taxes paid and their incidence, we follow PSZ’s construction of tax variables and tax incidence assumptions. Taxes are comprehensive
across individual, corporate, payroll, property, and sales. They include federal,
state, and local taxation. As a comparison to our Haig-Simons series, we also
estimate tax rates on ordinary factor income (i.e., excluding capital gains):
F AT Rt =

Total direct and indirect taxest
.
Ordinary factor incomet

26

(8)

(a)

Figure 17: Average Haig-Simons (5yr) tax rate vs factor income tax rate. HaigSimons tax rate defned in equation 7, factor income tax rate defned in equation
8.
Figure 17 compares Haig-Simons and factor tax rates. The addition of capital gains to the income defnition leads to a lower rate for Haig-Simons than
factor income. The Haig-Simons tax rate averaged 25% over the sample period,
compared to 30% for factor income.
Figures 18-19 compare Haig-Simons tax rates across the income distribution.
Because the top 10% receive the bulk of capital gains, in periods of high returns
tax rates drop, sometimes below the rates of the lower 90%. In the years of
capital losses, such as the Great Recession, the pattern is reversed. Over our
entire sample, however, the gains outpaced the losses, which means that top
groups saw the largest decrease in rates.
Figure 19 shows that Haig-Simons tax rates change our understanding of tax
progressivity: tax rates moderately decline across the income distribution. The
middle 40% pays average rates of 27.3%, the 90th-99th 27.0%, and the top 1%
26.8%. The decline of the top 1% is driven by the top .01% of the distribution,
which has the lowest tax rate of any group.

27

(a)

(b)

Figure 18: Haig-Simons (5yr) tax rates, by income group. Haig-Simons tax rate
defned in equation 7.

(a)

(b)

Figure 19: Haig-Simons (5 yr) tax rates, by income group. Haig-Simons tax rate
defned in equation 7. Bars are weighted averages across years, with weights
equal to the level of Haig-Simons income in 2021 dollars.
Figure 20 compares the progressivity of the different tax rate measures. While
factor income tax rates are progressive, with an increasing rate for higher percentiles, Haig-Simons tax rates are relatively fat, and decrease with higher levels
of income.

28

(a)

(b)

Figure 20: Comparison of tax rates. Haig-Simons tax rate defned in equation 7,
factor income tax rate defned in equation 8. PSZ tax rates taken from Distributional National Accounts.

5

Differential cap rates and returns

We fnd evidence of substantial heterogeneity in returns across the income distribution. Two separate forms of differential returns infuence our estimates: (i)
differential cap rates that affect how income fows are capitalized into wealth (ii)
differential returns on wealth that affect measured capital income. In addition,
there is the interaction between the two, since higher measured wealth in turn
can lead to higher measured income. The focus of our analysis will be documenting the differential returns, and studying how they affect our measures of
income and wealth inequality.
Our analysis of heterogeneous returns has several limitations. First, the returns are not adjusted for risk, either in absolute terms or through a CAPM or
other asset pricing model. Second, we have limited indications of the forces
driving the return differences.28 Finally, we cannot measure differential returns
for public equities and pension wealth.
28

Heterogeneity may be caused by higher skill or risk tolerance at all wealth levels, known as
type dependence, or from scale dependence, which posits that access to higher wealth changes
returns through differential information, opportunity, or decreasing relative risk aversion. See
Bach, Calvet and Sodini (2020) and Fagereng et al. (2020).

29

5.1 Owner-occupied housing

(a)

(b)

(c)

(d)

Figure 21: Owner-occupied housing.(a) - (b) Average property tax rates (c)-(d)
Average capital gain returns. Individuals ranked by factor income.
For individuals that are homeowners, fgure 21 presents average property tax
rates by factor income rank, showing declining rates as income increases. The
middle 40% pays average rates of 1.02%, declining across the distribution to
0.90% for the top 1%. Tax rates continue to decrease to the tails of the distribution, with the top .01% paying the lowest rates of all.
The source of our variation for fgure 21 is purely geographic; richer individuals have lower tax rates because they live in counties with lower rates.
Prior research suggests that the disparities would only increase if we took into
´ and Howard (2022) fnds
acount within geographic variation. Avenancio-Leon
that within neighborhoods, black and hispanic residents face 10-13% higher burdens, while McMillen and Singh (2020) also fnds regressivity in tax rates across
incomes.
30

Lower tax rates lead to higher estimated housing wealth and wealth concentration. Figure 22 (a) shows that the top 1%’s average housing wealth increases
from $780,000 under equal rates to $810,000 under heterogeneous taxes. The
share of housing wealth held by the top 1% increases from 8.30% to 8.93%.

(a)

(b)

(c)

(d)

Figure 22: Owner-occupied housing wealth and capital gains, comparison of
heterogeneous vs homogeneous returns. (a)-(b) Average owner-occupied housing wealth (c)-(d) Average housing capital gains. Individuals ranked by factor
income.
Figure 21 (c) shows average real house price growth across the factor income
distribution, displaying a strong relationship between income and returns. The
top 1% has an average return on owner-occupied real estate of 2.00% compared
with 1.64% for the middle 40.
Again, our source of variation is geographic, suggesting that higher income
individuals reside in counties that have seen larger growth in real estate prices.
This result is consistent with Demers and Eisfeldt (2022), who fnd that high
price-tier cities accrue more capital gains.
31

Heterogeneous returns make a substantial difference for capital gain inequality. Average owner-occupied housing capital gains for the top 1% are $ 17,500
under heterogeneous returns, compared to $ 7,500 under homogeneous returns.
Figures A.10 and 22 show that both heterogeneous cap rates and heterogeneous
returns contribute to an increase in income inequality. Moving from homogeneous cap rates and returns to heterogeneous ones, the top 1% share of housing
capital gains increases from 8.83% to 15.23%.

5.2 Directly owned tenant occupied housing

(a) All

(b) Homestead

(c) Multifamily

(d) Commercial

Figure 23: Average property tax rates, Schedule E real estate, by income group.
For individuals that directly own tenant occupied properties, fgure 23 calculates average property tax rates by income group. Results are broken out by
three primary property classes: homestead (single family and vacation homes),
multi-family, and commercial (which includes land and other types of proper32

ties). Within property classes, we fnd similar results to owner-occupied housing: there is a negative relationship between factor income and property tax rates,
suggesting that richer individuals invest in real estate located in areas with lower
property tax rates.
Looking at tenant occupied housing as a whole, however, there is little variation in rates (panel a). This result is due to a composition effect: higher income
individuals own more multifamily and commercial buildings, which have higher
average tax rates, canceling out the within-property class effects. Heterogeneous
property tax rates thus do not meaningfully change the concentration of property
wealth (see fgure A.13).

(a) All

(b) Homestead

(c) Multifamily

(d) Commercial

Figure 24: Average Schedule E real estate total return. Total return equals rental
return plus real capital gain yield. Individuals ranked by factor income.
Total tenant occupied housing returns equals the income return plus the cap-

33

ital gain return,29
Rettot
t =

Net rental incomet + Capital gainst
.
Market valuet

Figure 23 displays average total returns across the income distribution. There is
a positive relationship between income and total returns, with richer individuals
receiving higher average yields. As shown in appendix fgures A.12 and A.11,
this pattern is driven mainly by differences in the income return, with little variation in capital gain yields. Heterogeneity thus affects the overall concentration
of Haig-Simons income, but not capital gain income.
29

Net rental income equals rent roll minus maitenance, management, utilities, and other expenses from Schedule E.

34

5.3 Private business wealth

(a) S-corp

(b) Partnership

(c) Sole proprietorship

Figure 25: Private business, average EV/EBITDA valuation ratios, by income
group. Individuals ranked by factor income.
For owners of private business, we combine together all ownerships stakes and
calculate average statistics across investments. We then calculate average cap
rate by income group. Figures 25 and A.18 show that richer individuals own private businesses that sell for higher multiples. The average S-corp EV/EBITDA
ratio for the top 1% is 9.0, compared to 5.8 for the middle 40. The average
partnership EV to EBITDA for the top 1% is 8.6 compared to 5.9 for the middle 40. There is less heterogeneity in sole proprietorship valuation to earnings
ratios, ranging from 2.2 for middle 40 to 2.3 for the top 1%.
Richer owned businesses are worth more for two reasons: (i) the businesses
are larger (ii) they are in industries that sell for a premium. Figure A.19 shows
differences in business size across the distribution. Top 1% owned S-corps are
worth about $4.5 million, compared with $500,000 for the 90th-99th percentile,
35

and $100,000 for the middle 40. Top 1% partnerships are worth $1,000,000 on
average, compared with $50,000 for the middle 40. Figures A.20 and A.21 show
the industry composition: rich owned S-corps have more manufacturing, retail,
and wholesale than average, which tend to sell for higher multiples, and fewer
professional businesses, which have lower mutliples.

(a)

(b)

(c)

Figure 26: Average private business wealth, by measurement method. “FA”
columns present homogeneous capitalization using Financial Accounts totals.
“Hom cap” present totals assuming homogeneous capitalization rates across the
income distribution. “Het cap” represent our baseline results, which use heterogeneous capitalization rates.

6

Conclusion

This paper develops new methods to study the distribution of capital gains, and
fnds that:

36

1. Capital gains are large and highly concentrated. They average about 20%
of ordinary factor income, and are comparable in magnitude to ordinary
capital income. Capital gains in public equities, private equities, and tenant occupied housing are particularly concentrated, with the top 1% receiving a majority of gains. Capital gains on housing and pension assets
are more widely dispersed. Overall, the top 1% of the distribution receives
45.3% of capital gains.
2. Capital gains contribute substantially to income inequality. The top 1%
share increases from 18% without capital gains to 21.0% with gains included.
3. The U.S. tax system is less progressive when capital gains are taken into
account. Overall, only a small proportion of gains are taxed, leading to
an overall macro tax rate on nominal gains of 3%, signifcantly below the
statutory rate. Because gains are concentrated in the top 10%, the average tax rate on Haig-Simons income is lower for higher income groups,
leading to a tax rate that is fat across the income distribution.
4. Cap rates and returns exhibit marked differences across income levels.
Richer individuals have higher cap rates for real estate and private business
wealth, and higher returns for owner and tenant occupied housing. The
heterogeneity makes a material difference in measures of overall income
and wealth inequality.
These empirical fndings have direct relevance to capital gain tax policy.
Capital gain tax reform is a perennial issue, with the standard debate weighing
revenue against concerns over the effects on entrepreneurship as well as large
estimated elasticities of tax responses.30 While traditional scorekeepers such as
the Joint Committee on Taxation use revenue elasticities as high as -.7,31 recent
fndings suggest that -.3 to -.5 may be more reasonable for long-run responses.32
Our fndings suggest that wealthy taxpayers have historically been able to largely
shield their capital gains from taxation. In order to raise substantial revenue from
the tax it will be necessary to close one or more of the existing loopholes: the
step-up basis at death, charitable giving of appreciated property, or taxation at
realization.
There are several limitations of this study which we leave for future work.
More analysis needs to be done to estimate heterogeneous cap rates and returns
on public equities and pension wealth, which we could not do due to data limitations. In addition, it will be important to distinguish between capital gains
that refect pure changes in discount rates, which do not necessarily correspond
to welfare changes. Finally, there is still a portion of indirectly held partnership
and real estate wealth which cannot be traced back to individuals, which could
also affect measured income and wealth equity.
30

See Burman (2010) and Slemrod and Chen (2023).
See JCT (1990), JCT (2021),Dowd, McClelland and Muthitacharoen (2015).
32
Sarin et al. (2022) Agersnap and Zidar (2021).
31

37

References
Agersnap, Ole, and Owen Zidar. 2021. “The tax elasticity of capital gains and
revenue-maximizing rates.” American Economic Review: Insights, 3(4): 399–
416.
Armour, Philip, Richard V Burkhauser, and Jeff Larrimore. 2013. “Levels
and Trends in United States Income and Its Distribution A Crosswalk from
Market Income Towards a Comprehensive Haig-Simons Income Approach.”
National Bureau of Economic Research.
Auclert, Adrien. 2019. “Monetary policy and the redistribution channel.” American Economic Review, 109(6): 2333–2367.
Avenancio-León, Carlos F, and Troup Howard. 2022. “The assessment gap:
Racial inequalities in property taxation.” The Quarterly Journal of Economics,
137(3): 1383–1434.
Bach, Laurent, Laurent E Calvet, and Paolo Sodini. 2020. “Rich pickings?
Risk, return, and skill in household wealth.” American Economic Review,
110(9): 2703–2747.
Bailey, Martha J. 1969. Capital gains and income taxation. Brookings.
Barkai, Simcha. 2016. “Declining Labor and Capital Shares.” Mimeo. University of Chicago.
Bhandari, Anmol, and Ellen R McGrattan. 2018. “Sweat Equity in US Private
Business.” National Bureau of Economic Research.
Bhatia, Kul B. 1974. “Capital Gains, The Distribution of Income, and Taxation.”
National Tax Journal, 319–334.
Bourne, Jenny, Eugene Steuerle, Brian Raub, Joseph Newcomb, and Ellen
Steele. 2018. “More Than They Realize: The Income of the Wealthy.” National Tax Journal, 71(2): 335–356.
Burman, Leonard E. 2010. The labyrinth of capital gains tax policy: A guide
for the perplexed. Brookings Institution Press.
Caballero, Ricardo J, Emmanuel Farhi, and Pierre-Olivier Gourinchas.
2017. “Rents, Technical Change, and Risk Premia: Accounting for Secular Trends in Interest Rates, Returns on Capital, Earning Yields, and Factor
Shares.” National Bureau of Economic Research.
Campbell, Cole, and Jacob Robbins. 2023. “The Value of Private Business in
the United States.” Available at SSRN 4635369.
Cochrane, John. 2020. “Wealth and Taxes, part II.” Blog post.
38

Cooper, Michael, John McClelland, James Pearce, Richard Prisinzano,
Joseph Sullivan, Danny Yagan, Owen Zidar, and Eric Zwick. 2016. “Business in the United States: Who owns it, and how much tax do they pay?” Tax
Policy and the Economy, 30(1): 91–128.
De Loecker, Jan, and Jan Eeckhout. 2017. “The Rise of Market Power and the
Macroeconomic Implications.” National Bureau of Economic Research.
Demers, Andrew, and Andrea L Eisfeldt. 2022. “Total returns to single-family
rentals.” Real Estate Economics, 50(1): 7–32.
Dowd, Tim, Robert McClelland, and Athiphat Muthitacharoen. 2015.
“New evidence on the tax elasticity of capital gains.” National Tax Journal,
68(3): 511–544.
Eeckhout, Jan. 2024. “The Value and Profts of Firms.” Working paper.
Eggertsson, Gauti B, Jacob A Robbins, and Ella Getz Wold. 2018. “Kaldor
and Pikettys Facts: The Rise of Monopoly Power in the United States.” National Bureau of Economic Research.
Elliott, Matthew, Benjamin Golub, and Matthew O Jackson. 2014. “Financial networks and contagion.” American Economic Review, 104(10): 3115–
3153.
Ensign, R, and Richard Rubin. 2021. “Buy, borrow, die: How rich Americans
live off their paper wealth.” Wall Street Journal. July, 13.
Fagereng, Andreas, Luigi Guiso, Davide Malacrino, and Luigi Pistaferri.
2020. “Heterogeneity and persistence in returns to wealth.” Econometrica,
88(1): 115–170.
Fagereng, Andreas, Matthieu Gomez, Emilien Gouin-Bonenfant, Martin
Holm, Benjamin Moll, and Gisle Natvik. 2024. “Asset-price redistribution.”
Feenberg, Daniel R, and James M Poterba. 2000. “The income and tax share
of very high-income households, 1960-1995.” American Economic Review,
90(2): 264–270.
Fox, Douglas R, and Clinton P McCully. 2009. “Concepts and methods of the
us national income and product accounts.” NIPA Handbook.
Galeotti, Andrea, and Christian Ghiglino. 2021. “Cross-ownership and portfolio choice.” Journal of Economic Theory, 192: 105194.
Goedhart, Marc, Tim Koller, and David Wessels. 2015. “The real business of
business.” McKinsey on Finance, 53.

39

Goldsmith, Selma, George Jaszi, Hyman Kaitz, and Maurice Liebenberg.
1954. “Size distribution of income since the mid-thirties.” The Review of Economics and Statistics, 1–32.
Haig, Robert M. 1921. “The concept of income-economic and legal aspects.”
The federal income tax, 1(7).
Hall, Robert E. 2001. “The stock market and capital accumulation.” The American Economic Review, 91(5): 1185–1185.
Hemel, Daniel J, and Steve Rosenthal. 2021. “Mega-IRAs, Mega-401 (k) s,
and Other Mega-Retirement Accounts: Statement for the Record.” University
of Chicago Coase-Sandor Institute for Law & Economics Research Paper.
Hess, Ryan, Emily Black, Zaynah Javed, Jonathan Hennessy, Rebecca
Lester, Jacob Goldin, Daniel E Ho, and Annette Portz. 2024. “The Spiderweb of Partnership Tax Structures.” Available at SSRN.
Hicks, John. 1946. Value and Capital. Oxford: Clarendon Press.
JCT. 1990. “Explanation of Methodology Used to Estimate Proposals Affecting
the Taxation of Income from Capital Gains.”
JCT. 2008. “Estimates of Federal Tax Expenditures for Fiscal Years 2008–
2012.”
JCT. 2012. “Overview of the defnition of income used in distributional analyses.” JCT.
JCT. 2021. “Estimated Budgetary Effects Of An Amendment in the Nature of a
Substitute to the Revenue Provisions of Subtitles F, G, H, I, and J.”
Kahn, Matthew E. 2024. “Racial and ethnic differences in the fnancial returns
to home purchases.” Real Estate Economics, 52(3): 908–927.
Kopczuk, Wojciech. 2016. “US capital gains and estate taxation: a status report
and directions for a reform.”
Krugman, Paul. 2021. “Pride and Prejudice and Asset Prices.” International
New York Times.
Larrimore, Jeff, Richard V Burkhauser, Gerald Auten, and Philip Armour.
2021. “Recent trends in US income distributions in tax record data using more
comprehensive measures of income including real accrued capital gains.”
Journal of Political Economy, 129(5): 1319–1360.
Love, Michael. 2021. “Where in the world does partnership income go? Evidence of a growing use of tax havens.” Evidence of a Growing Use of Tax
Havens (December 14, 2021).
40

May, Larry. 2012. “Using link analysis to identify indirect and multi-tiered
ownership structures.” IRS Offce of Research.
McElroy, Michael Bancroft. 1971. “Capital gains and the theory and measurement of income.”
McGrattan, Ellen R., and Edward C. Prescott. 2010. “Unmeasured Investment and the Puzzling US Boom in the 1990s.” American Economic Journal:
Macroeconomics, 2(4): 88–123.
McMillen, Daniel, and Ruchi Singh. 2020. “Assessment regressivity and property taxation.” The Journal of Real Estate Finance and Economics, 60: 155–
169.
Miller, Merton H, and Franco Modigliani. 1961. “Dividend policy, growth,
and the valuation of shares.” the Journal of Business, 34(4): 411–433.
Piketty, Thomas, and Emmanuel Saez. 2003. “Income inequality in the United
States, 1913–1998.” The Quarterly journal of economics, 118(1): 1–41.
Piketty, Thomas, Emmanuel Saez, and Gabriel Zucman. 2018. “Distributional national accounts: methods and estimates for the United States.” The
Quarterly Journal of Economics, 133(2): 553–609.
Poterba, James M, and Scott J Weisbenner. 2001. “Capital gains tax rules,
tax-loss trading, and turn-of-the-year returns.” The Journal of Finance,
56(1): 353–368.
Pratt, Shannon P. 2006. The market approach to valuing businesses. John Wiley
& Sons.
Saez, Emmanuel, and Gabriel Zucman. 2019. “Progressive wealth taxation.”
Brookings Papers on Economic Activity, 2019(2): 437–533.
Sarin, Natasha, Lawrence Summers, Owen Zidar, and Eric Zwick. 2022.
“Rethinking how we score capital gains tax reform.” Tax Policy and the Economy, 36(1): 1–33.
Simons, Henry C. 1938. “Personal income taxation: The defnition of income
as a problem of fscal policy.”
Slemrod, Joel, and Xinyu Chen. 2023. “Are capital gains the Achilles heel of
taxing the rich?” Oxford Review of Economic Policy, 39(3): 592–603.
Smith, Matthew, Danny Yagan, Owen Zidar, and Eric Zwick. 2019. “Capitalists in the twenty-frst century.” The Quarterly Journal of Economics,
134(4): 1675–1745.

41

Smith, Matthew, Owen Zidar, and Eric Zwick. 2023. “Top wealth in america: New estimates under heterogeneous returns.” The Quarterly Journal of
Economics, 138(1): 515–573.
Steuerle, C Eugene. 1982. The Relationship Between Realized Income and
Wealth: Report from a Select Sample of Estates Containing Farms Or Businesses. Offce of Tax Analysis, US Treasury Department.
Steuerle, Eugene. 1985. “Wealth, realized income, and the measure of wellbeing.” In Horizontal equity, uncertainty, and economic well-being. 91–124.
University of Chicago Press.
Yagan, Danny. 2023. “What is the average federal individual income tax rate on
the wealthiest Americans?” Oxford Review of Economic Policy, 39(3): 438–
450.

42

Online Appendix for
The Distribution of Capital Gains
Cole Campbell, Jacob A. Robbins, Samuel Wylde

A

Data construction

A.1 Final wealth categories
Total household net worth equals the sum of fxed income, public equities, private equities, pension, owner occupied housing, and tenant occupied housing,
minus debt. Aggregate wealth for each category will equal Financial Accounts
totals, with the exception of bespoke estimates for private business wealth and
tenant occupied housing. For all Financial Accounts variables, we use the March
24 2024 data release. Wealth variables are mid-year totals. We estimate wealth
at the individual level through a combination of our own methods and those of
PSZ and SZZ. We use the latest PSZ (2020) updated methods, described here.
SZZ methods are described in their online appendix.
1. Fixed income, 2021 total = $20.77 T.
• Currency, 2021 total = $3.37 T. Follows PSZ methodology and
imputes based on tabulations from the SCF.
• Bond mutual fund / ETF / closed end funds, 2021 total = $4.38 T.
Estimated by capitalizing nonqualifed dividends from 2003 onward,
following SZZ. Prior to 2003, capitalized using dividends (following
PSZ).
• Munis, 2021 total = $2.32 T. Follows PSZ and capitalizes tax exempt interest.
• Taxable fxed income wealth, 2021 total = $10.70 T. Capitalizes
taxable interest, with heterogeneous cap rates following SZZ; see
section A.4.
2. Public corporate equities, 2021 total = $16.35 T. Uses SZZ methodology, capitalizing mix of 90% qualifed dividends, 10% capital gains.
3. Private business wealth, 2021 total = $25.78 T. Household total $18.66
T. See section 2.3.3.
• S-corporation, 2021 total = $8.47 T. Household total $6.70 T.
• Partnership, 2021 total = $11.62 T. Household total $6.28 T.
• Private C, 2021 total = $4.53 T.
A.1

• Sole proprietorship, 2021 total = $1.16 T.
4. Pension wealth, 2021 total = $47.59 T. Following PSZ, capitalizes a mixture of 60% taxable pensions, 30% wages, and 10% tax-exempt pensions.
• IRA, 2021 total = $13.56 T.
• Defned contribution, 2021 total = $10.15 T.
• Defned beneft, 2021 total = $16.57 T. Note that here we use the
full value of DB pensions, and not only the funded portion, as in
PSZ.
• Life insurance, annuity, other, 2021 total = $7.31 T.
5. Owner occupied housing, 2021 total = $35.94 T. See description in section A.8.
6. Tenant occupied housing, 2021 total = $11.06 T. See description in section A.8.
7. Household debt, 2021 total = $-21.58 T.
• Owner occupied mortgages, 2021 total = $-11.37 T. Allocated proportionally to o.o. housing wealth.
• Tenant occupied mortgages, 2021 total = $-4.73 T. Allocated proportionally to t.o. wealth.
• Other debt, 2021 total = $-5.47 T. Following PSZ, imputed using
tabulations from SCF.

A.2 Capital gain variables
Capital gains estimated (i) using totals from the Financial Accounts, using a homogeneous return assumption (ii) for owner and tenant occupied housing, using
real estate price indexes multiplied by the value of the real estate asset. We strip
out all fxed income capital gains from the totals, and do not include capital gains
or losses on debt. The breakdown by asset class is as follows:
1. Fixed income, capital gains not included.
2. Public corporate equities, 2021 total = $3,899 B. Totals from Financial
Accounts. Estimated using equal returns.
3. Private business wealth, 2021 total = $1,753 B. Household total = $1,243
B. See description in text. Estimated using equal returns.
• S-corporation, 2021 total = $437 B. Household total = $345 B.
• Partnership, 2021 total = $741 B. Household total = $323 .
• Private C, 2021 total = $352 B.
A.2

• Sole proprietorship, 2021 total = $223 B.
4. Pension wealth, 2021 total = $1,292 B. Only includes capital gains of
corporate equities; fxed income capital gains are stripped from Financial
Accounts totals. Totals from Financial Accounts, distributed using equal
returns.
• IRA, 2021 total = $772 B.
• Defned contribution, 2021 total = $560 B.
• Defned beneft, capital gains not included. Under our assumptions
individuals do not receive DB kgs, only sponsors do.
• Life insurance, annuity, other, capital gains not included.
5. Owner occupied housing 2021 total = $2,019 B. Totals estimated using
FHFA price index multiplied by home values. See description in text and
section A.8.
6. Tenant occupied housing, 2021 total = $589 B. Total estimated from
property type specifc price indexes by real estate values. See description
in text and section A.8.
7. Household debt, capital gains not included.
Nominal capital gains for asset class j during year t are converted to real
gains using the formula:
j
· (1 +
Real gainsjt = M Vt−1

Nominal gainsjt
1
j
)·
− M Vt−1
,
j
1 + πt
M Vt−1

(A.1)

where M Vtj−1 is the market value of the asset at the end of year t − 1, πt is
infation for year t using the Net National Product price index (calculated by the
percent change of the index from December t − 1 to December t). The formula
multiplies the market value (end of year t − 1 prices) by the real price growth to
yield real gains. Finally, to put gains in midyear t prices, we multiply the gains
by NNP infation during the frst half of the year.
To form fve year moving average for capital gains, for every year we calReal gainsjt
culate the real yield as
, where gains and wealth totals are in end of
MV j
t−1

year t − 1 prices. We then take a geometric average of the yield centered around
year t, and mutliply the real yield by M Vtj−1 . To put gains in midyear t prices,
we multiply the gains by NNP infation during the frst half of the year.

A.3 Ordinary factor income
We estimate ordinary factor income following the methodology of PSZ: national
income components are apportioned to individuals in proportion to taxable income and estimated wealth categories. Some of our wealth categories are different than PSZ, which leads to differences in the income distribution methodology.
A.3

1. Labor and labor component of mixed income, 2021 total = $13,904 B.
Apportioned as in PSZ.
2. Corporate equity income directly held (other than S-corp income), 2021
total = $947 B. Total corporate equity income distributed to three separate wealth variables using equal returns: public corporate equity directly
held, public corporate equity held through pension funds, private C corps
directly held. We compute the C corp yield as aggregate C income / aggregate C wealth, then multiply the yield by the three components. Note that
because the retained earnings is contained in capital gains, we strip them
out from the numerator here.
3. Fixed income directly held, 2021 total = $594 B. Total fxed income
apportioned to directly held vs pension in proportion to asset totals.
4. Pension, 2021 total = $1,701 B. Pension component of fxed income and
corporate income.
5. Private business, 2021 total = $1,676 B.
• S-corporation, 2021 total = $940 B.
• Partnership, 2021 total = $543 B. Partnership and soleprop distributed in proportion to asset values.
• Sole proprietorship, 2021 total = $192 B.
6. Owner occupied housing 2021 total = $1,168 B. Apportioned proportional to owner occupied housing using net imputed rent. Rent is imputed
for owner occupiers by multiplying housing wealth by area specifc net
rental rates.
7. Tenant occupied housing, 2021 total = $398 B. Apportioned proportional to tenant occupied net rental income.
8. Debt payments, 2021 total = $-735 B.
• Owner and tenant occupied mortgage interest, 2021 total = $-422
B. Allocated in proportin to owner and tenant occupied mortgages.
• Other debt interest, 2021 total = $-313 B. Allocated in proportion
to non mortgage debt.
9. Nonproft and government income, 2021 total = $-261 B. Apportioned as
in PSZ.

A.4

A.4 Replication and use of Distributional National Accounts and Smith, Zidar, and Zwick variables
We replicate the DINAs of PSZ using the 2020 vintage of their fles, using the
Stata programs provided on https://gabriel-zucman.eu/usdina/ and directly from
the authors. We extend analysis for several additional years, through 2021. This
requires (i) using additional CPS data for the non fler sample, from NBER data
(ii) using the 2019 and 2022 vintage of SCF fles for use of constructing tabulations from Fed Reserve fles. Our replication matches closely the key and
income and wealth inequality series from the DINAs.
We use the DINAs through two channels.
1. To capitalize a number of our wealth variables, as described in section A.1.
The wealth totals for these categories will be slightly different, however,
as we use updated Financial Accounts data.
2. To distribute ordinary factor income for most income categories, with the
exception of private business income, tenant occupied housing income,
and owner occupied housing income. The totals will exactly equal the
DINA total, as we derive the income totals from the parameters.xlsx fle
provided in the replication fle.
We use the exact methodology of SZZ to estimate bond mutual fund wealth
and taxable corporate equities. To estimate taxable fxed income wealth, we use
capitalization rates provided by SZZ in their ExhibitData.zip, wealth excel exhibits.xlsx,
DataFig3CD sheet, provided on Zidar’s website. SZZ provide capitalization for
four wealth groups: 0-99%, 99-99.9%, 99.9-99.99%, and top .01%. We rank individuals into quantiles based upon non-interest fnancial wealth, and capitalize
taxable interest using the appropriate cap rate to the wealth group.

A.5 Analysis of Financial Accounts data
Financial accounts data is used for (i) wealth totals (ii) capital gains (iii) comparison with our bespoke estimates. Our use of this data is similar to PSZ and
SZZ; we compare our variables below.
The FA contains estimates of wealth, fows, and capital gains by asset class
for households, nonprofts, and all other sectors of the economy. We use data
from the Z1 release, https://www.federalreserve.gov/releases/z1/. We use the
following category of variables: (i) FL, LM: levels, used for wealth totals (ii)
FR: revaluations, used for capital gains.
We manipulate FA data to form asset classes that correspond to income fows
on tax returns. In addition, for pooled assets such as mutual funds, ETFs, and
closed end funds, we separate out the equity from the fxed income. This is
necessary because we do not include fxed income capital gains in our baseline
estimates.

A.5

The primary tables we use are B.101 Balance Sheet of Households and Nonproft Organizations, B.104 Balance Sheet of Nonfnancial Noncorporate Business, B.101.n Balance Sheet of Nonproft Organizations, S.3.a Households and
Nonproft Institutions Serving Households.
We use additional Investment Company Institute (ICI) data on the composition of IRA mutual funds, as in PSZ and SZZ. This is necessary to fully strip out
fxed income capital gains from IRA mutual funds.
We form seven mutually exclusive categories
1. Taxable dividend wealth: money market non-munis, directly held stocks,
ETFs, mutual funds. We separate this category into two groups: assets
that pay qualifed dividends (money market funds, bond mutual funds,
bond ETFs), and those that pay non-qualifed dividends (stocks and equity
funds). This is in line with SZZ, but differs from PSZ who do not include
indirectly held bonds in this category.
2. Currency: follows PSZ and SZZ.
3. Taxable fxed income: deposits, bonds directly held, loans. Does not include bonds held through mutual funds / ETFs. Follows SZZ, but not PSZ
who includes these indirectly held fxed income assets.
4. Munis: through money market, mutual funds, ETFs. Follows PSZ and
SZZ.
5. Real estate: includes owner occupied as well as tenant occupied. Differs
slightly from PSZ and SZZ in that we include nonresidential (LM115035035)
as well as residential in this category. PSZ and SZZ include nonresidential
in noncorporate business.
6. Pension: includes defned beneft, defned contribution, life insurance,
IRA. Our construction follows the FA, which is somewhat different than
PSZ, who do not include unfunded defned beneft pension plans. SZZ
likewise do not include unfunded DB, but use estimates from Sabelhaus
and Volz (2019) to estimate this component.
7. Private business wealth: non corporate business wealth (excluding tenant
occupied real estate), plus private S corp and private C corp wealth. Follows PSZ with the exception of the exclusion of commercial real estate.
The household balance sheets do not breakdown the holdings of household
mutual funds and ETFs between fxed income and equity assets. In addition,
they do not separately break out IRAs from other assets. To estimate these breakdowns, we utilize the following detail tables.
1. L.117 Private and Public Pension Funds
2. L.123 Closed-End Funds
A.6

3. L.124 Exchange-Traded Funds
4. L.224 Corporate Equities
5. L.122 Mutual Funds
6. L.229 Pension Entitlements
7. L.118.b Private Pension Funds: Defned Beneft Plans
8. L.118.c Private Pension Funds: Defned Contribution Plans
9. L.119 Federal Government Employee Retirement Funds
10. L.119.b Federal Government Employee Retirement Funds: Defned Beneft Plans
11. F L.119.c Federal Government Employee Retirement Funds: Defned Contribution Plans
12. F L.120 State and Local Government Employee Retirement Funds
13. F L.120.b State and Local Government Employee Retirement Funds: Defned Beneft Plans
14. L.120.c State and Local Government Employee Retirement Funds: Defned Contribution Plans
The total for our seven categories equals the total from Table B.101, FL152000005,
and total liabilities likewise equal B.101 liabilities FL152090005A.
To separate household and nonproft assets, we subtract totals from the nonproft balance sheets on Table B.101n. This differs from PSZ, who have alternative methods of estimating nonproft wealth.
Total Financial Accounts capital gains are calculated using the revaluation
variables FR for the above assets classes.
The totals for our wealth variables (described in section A.1) will generally
match FA totals, with the exception of private business wealth and tenant occupied housing, which we estimate separately. To equal FA wealth, take our
starting estimates, subtract private C-corp, S-corp, partnership, sole proprietorship, tenant occupied real estate, and add in FA private C and S corp wealth,
partnership and sole proprietorship wealth, and tenant occupied real estate.

A.5.1 Private business wealth in the Financial Accounts
Nonfnancial noncorporate business wealth is measured in table B.104. Table
B.104 combines together many disparate asset types: partnership and sole proprietorship business assets, tenant occupied real estate, and fnancial assets held
on balance sheets. The elements are as follows:

A.7

1. Residential real estate (LM115035023). This is estimated by a perpetual
inventory type method, where fows of BEA residential investment are
combined with a capital gains price index.
2. Nonresidential real estate (LM115035035), again measured using perpetual inventory type method from BEA nonresidential investment.
3. Mortgages (FL113165005)
4. Fixed assets of businesses (LM115015205 + LM115013765 + LM115020005).
Measured using BEA fxed assets. I break out partnership from sole proprietorship assets using the underlying BEA data. Total partnership assets
equal partnership fxed assets of equipment, IPP, and inventories (k1ntot17eq00
+ k1ntotl7ip00 + estimated partnership share of LM115020005).
5. Net non-mortgage fnancial assets (FL114090005 - (FL114190005-FL113165005)).
Our EV/EBITDA and EV/SA estimates of private noncorporate business values are estimated on transaction data that do not include fnancial or real estate
assets. The best apples to apples comparison is thus a comparison to the Financial Accounts fxed assets of businesses. To yield the total market value of private
noncorporate business, we add to our capitalized estimates the net non-mortgage
fnancial assets from Table B.104.
Financial accounts private corporate wealth equals S-corp (LM883164133)
+ private C-corp (LM883164135) market values, which are directly comparable
to our estimates on an apples to apples basis. The total comparison between
our estimates and Financial Accounts private business wealth is shown in fgure
A.17.
One additional difference of our analysis is that the FA implicitly assumes
all noncorporate businesses (and tenant occupied real estate) are owned by U.S.
residents. As seen in fgure 4, this is not the case, and our inequality estimates
refect lower wealth totals than the aggregate.
We compare our tenant occupied housing values to those on Table B.104: our
residential estimates are directly comparable to LM115035023, and our commercial to LM115035035. This comparison is shown in fgure A.6.
We estimate the value of commercial/industrial real estate by capitalizing
property taxes from rental properties. This is potentially missing industrial or
commercial properties that are used for own-use by businesses (for example, a
restaurant that owns its premises through the same llc). This is a current drawback of our method, however our aggregates in fgure A.6 are still in line with
the FA and SCF.

A.8

Table A.1: Private business wealth
2002

2007

2012

2017

2021

15.85
17.37
8.48
6.00
12.63
4.46
5.29
1.91
0.96

23.63
25.78
11.62
8.47
18.66
6.28
6.70
4.53
1.16

Estimated Values (trillions of $)
Enterprise Value Total
Market Value Total
Partnership Total
S-corp total
Market Value Individuals
Partnership Individuals
S-corp individuals
Private C-corp individuals
Sole prop individuals

6.25
6.10
2.54
2.54
4.32
0.91
2.39
0.84
0.19

11.97
12.39
5.54
4.11
8.95
2.45
3.76
1.91
0.83

11.97
12.56
5.83
4.16
8.91
2.62
3.72
1.78
0.78

Financial Accounts Private Business (trillions of $)
Total Market Value
Partnership
S-corps
Private C
Sole prop

2.02
0.34
1.06
0.36
0.26

4.45
1.14
1.88
1.14
0.29

4.75
1.45
1.99
1.02
0.29

8.61
2.50
4.04
1.75
0.32

12.45
3.32
6.42
2.32
0.38

A.6 Estimation of capital gain and Haig-Simons tax rates
To compute capital gain tax rates, we use the following data series:
• S1, Macro taxes paid on capital gains: aggregate data from U.S. Treasury
and Tax Foundation historical series. Directly comparable to our micro
estimates below.
• S2, Macro capital gain realizations: same sources as S1.
• S3, Micro taxes paid on capital gains: data at the tax return level estimated
using Taxsim for short and long term capital gains.
• S4, Micro capital gain realizations: 1040 Line 6, total capital gains from
Schedule D. Includes capital gain distributions and supplements.
• S5, Macro capital gains: Aggregate Financial Accounts household + nonproft revaluations, plus adjustments for our custom private business, tenant occupied housing, and owner occupied housing.
• S6, Macro household capital gains: Macro capital gains minus nonproft
capital gains.
• S7, Household taxable gains: these include gains that are taxable in principle: household owned share wealth (stocks, mutual funds, ETFs), private
businesses, owner occupied real estate, tenant occupied real estate. Does
not include nonproft capital gains or pension/IRA gains. Note that owner
occupied capital gains are included even though in practice they are untaxed due to large exclusions.
A.9

• S8, Capital gain exclusions: estimates of total capital gains that are not
subject to tax due to exclusions, such as those for selling a primary residence, 1031 exchanges when selling tenant occupied real estate, and small
business sale exemptions. Described in section 4.
We then compute the following tax rates: (i) Macro = S1
, Macro hh = S1
,
S5
S6
S1
S1
S1
HH txable = S7 , KG in agi + excl = S2+S8 , KG in agi = S2 .
To compute Haig-Simons tax rates, we follow PSZ (2021)’s construction of
tax variables and tax incidence assumptions. In particular,
• Payroll taxes: paid by labor.
• Individual income taxes: paid by individual taxpayers.
• Corporate income tax: falls on all capital except housing.
• Property taxes: business property taxes borne by all capital excluding
housing, residential property tax are borne by the owners of housing assets.
• Sales and excise taxes: proportional to disposable income less savings.
For comparison with PSZ tax rates, we make additional modifcations to income to match what PSZ refers to as “pre-tax” income. In particular, we add
in Social Security, unemployment benefts, and private pension benefts, and exclude the contributions to Social Security, private pensions, and unemployment
insurance.

A.7 Estimation of owner occupied housing
A.7.1 Housing valuation
For households that are itemizers, we estimate home values by scaling up property tax payments listed on deductions. We assume that the tax unit that lists the
property tax on their return is the homeowner, and allocate the entire value of
the home to the tax unit.
In order to go from property tax payments to home values, we use the zip
code address listed on the tax form to match the individual to their county of
residence, using a zip-county crosswalk from the Department of Housing and
Urban Development (see here).
We estimate average tax rates using data from the American Community
Survey and Decennial Census. The ultimate analysis relies on a complicated
combination of datasets, necessary due to the fact that there is incomplete coverage of counties in any one data set.
The primary source of property tax rates come from county-level aggregates
of the ACS, accessed through the Census API. To account for as many counties
as possible, the 5-year fles are used. The second source of property tax rates
A.10

come from county-level aggregates of the 1990 and 2000 census, again accessed
through the census API.
The third source of property tax payments come from micro-data from the
ACS and 1990 and 2000 decennial censuses, as well as the 2001-2018 ACS micro data. We restrict the fles to homeowners, and estimate property tax payments
using the midpoints from the categorical variables, with top-coded households
imputed at 1.5 times the threshold. Similarly for housing values, we assume topcoded housesholds have housing value 1.5 times the threshold. We collapse the
data to the PUMA level, and link PUMAs to counties using the crosswalk from
the Missouri Census Data Center.
County of residence for itemizers are available for 99.9% of all tax returns.
Scaling up property tax payments to housing value, we are able to account for
a substantial percentage of total housing wealth, as measured in the Financial
Accounts. Figure A.1 (a) shows around 80% of housing wealth is accounted for
by capitalizing property tax payments.

(a)

(b)

Figure A.1: (a) Comparison of capitalized itemizer owner-occupied housing
wealth with Financial Accounts totals (b) Comparison of average tax payer
house price growth with FHFA national index.
Following PSZ, we scale the value of itemizers’ housing wealth to equal 80%
of Financial Accounts owner-occupied housing wealth, and allocate the rest to
nonitemizers/nonflers using averages from the SCF.

A.7.2 Computation of capital gains
We estimate capital gains using housing price appreciation data from the FHFA.
The FHFA has estimates of house appreciation at the 5-digit zip code, county, 3digit zip code, and state level. These are annual price indexes, which in practice
capture price appreciation in year t for both year t and t-1. To better capture an

A.11

annual index change during the course of year t, we estimate price growth in
year t through an average of (index t/index t-1) and (index t+1/index t).
Due to limited data on home sales, the FHFA does not have price data for
all zip codes. For each tax unit, we try to estimate house price using the fnest
geography available frst. If this is missing, we then proceed to use larger geographies, proceeding from 5 digit zip codes to counties, 3-digit zip codes, and
states. For itemizers, 80% of tax-units have 5 digit zip code house price data,
15% have county-level price data, while the rest have state data.
Figure A.1 (b) gives the weighted average real housing return for itemizers,
and compares it to the FHFA annual house price growth index. In general the
two align fairly closely, although they should not be expected to exactly match,
given our average is for itemizers only and have a more detailed geographical
breakdown than the FHFA index, which is a combination of state-level price
indexes.

A.7.3 Analysis of property tax rates using the ACS/Census
We complement our analysis of differential property tax rates through a comparison of our results to those from a different data set: the ACS/Decennial
census microdata. Every year we rank individuals into percentiles by household income, and calculate weighted average property tax rates by percentile,
with weights corresponding to the value of individuals’ houses. An advantage
of ACS data is the microdata on both (self-reported) property tax payments and
housing value. A disadvantage is that housing values and incomes are top-coded,
potentially biasing our estimates for the top of the distribution. This is particularly problematic given our fnding of substantial tax rate variation at the very
top of the distribution. For our analysis of ACS data, we exclude households that
have top-coded housing values.
Figure A.3 (a) plots average property tax rates by income percentile for our
two sources: both show that top percentiles have lower property taxes than those
lower in the distribution. Figure A.3 shows the time series. Overall, the patterns
for the top 10% and top 1% match very closely between the two data series,
again showing substantially lower property tax rates for the upper parts of the
distribution.

A.12

(a)

(b)

Figure A.2: Average property taxes, comparison ACS vs IRS data.

(a)

(b)

Figure A.3: Average property taxes, ACS vs IRS data comparison (a) 90-99th
percentile (b) top 1%

A.8 Tenant occupied housing
Tenant occupied real estate wealth is estimated by capitalizing property tax payments. For properties that are directly owned, this is present on Schedule E, line
16. For properties indirectly owned through partnerships and S-corps, this is on
form 8825 line 11. Each individual property is valued using a county-year-type
cell specifc property tax rate. Properties are sorted into three general classes:
homestead (single family), multifamily, and commercial.

A.13

A.8.1 Estimating property tax rates
We obtain data on effective property tax rates (ETRs) from annual reports published jointly by the Lincoln Institute of Land Policy and the Minnesota Center
for Fiscal Excellence.
The reports estimate ETRs for each property type for three sets of geographies: 1) the largest city in each state 2) the largest 50 cities in the country regardless of state, and 3) one rural jurisdiction in each state, defned as county seats
in non-metropolitan counties with population sizes between 2,500 and 10,000.
We assign these ETR averages to county data in two stages. First, using
Census Bureau data on county-place (incorporated places and Census Designated Places) geographic equivalencies, we directly assign ETRs in the Lincoln
Institute/Minnesota Center data to county equivalents in our county-level data.
For those counties not directly represented by a place (city or rural jurisdiction), we assign as ETRs simple averages of ETRs within state-geography-year
cells, where the geography is urban (from the largest cities in each state or the
50 largest cities in the U.S.) or rural (from the rural jurisdictions in each state).
Figure A.4 gives mean effective tax rates by property type for urban counties.
Mean Effective Tax Rates by Property Type - Urban

.015
.01

.01

.015

.02

Commercial-Industrial - Urban Geographies

.02

Homestead - Urban Geographies

2005

2010
$150k

2015
$300k

2020
ACS rate

2005

2010
$100k

2015
$1M

2020
$25M

.01

.015

.02

Apartments - Urban Geographies

2005

2010

2015

2020

$600k

(a)

Figure A.4: Mean effective tax rates, tenant occupied real estate, urban counties.
Figure A.5 shows estimated aggregate tenant occupied housing wealth by
owner and property type.

A.14

(a)

(b)

Figure A.5: Tenant occupied housing wealth (a) by owner type (b) by property
type

(a)

Figure A.6: Comparison of aggregate tenant occupied real estate: Survey of
Consumer Finances, Financial Accounts, our estimates.

A.8.2 Capital gain yields
Data on commercial and multifamily property price growth comes from two
principal sources. From the Federal Home Loan Mortgage Corporation (Freddie
Mac), we obtain a quarterly price index for multifamily rental properties for 25
metropolitan areas across the U.S., a component of their Apartment Investment
Market Index.
From the CoStar Group, Inc., we obtain their CoStar Commercial RepeatSales Indices (CCRSI), a set of quarterly price indices that include separate indices for offce, industrial, retail, and multifamily properties at the Census region
level (Northeast, South, Midwest, and West).
A.15

We estimate county level multifamily price growth in two stages. For the
counties that comprise the 25 metropolitan areas available in the Freddie Mac
data, we assign these directly. For other counties, we use data at the metropolitan area level available in the Freddie Mac data to estimate multifamily price
levels as a function of owner-occupied, single-family home price levels, and
then project that relationship onto our remaining county-level data.
To do so, we combine the Freddie Mac metropolitan area quarterly MFPI
with the Federal Housing Finance Agency’s (FHFA) quarterly House Price Index
for All Transaction (HPI) at the same geographic level and regress the MFPI
level in metropolitan area m in year t on HPI in the same area and year with
year-dummy indicator variables as described in Equation A.8.2.
MFPImt = α + βHPImt + δt + 

(A.2)

Predicted Metro MF Price Index
200
300

400

We form predictions of MFPI (\
MFPIct ) in our county-level data for those
counties without an exact match to the metropolitan areas in the Freddie Mac
MFPI using the coeffcients recovered from Equation A.8.2. Finally, we take
annual growth rates of \
MFPIct to obtain estimates of county-level multifamily
property price growth. Figure A.7 compares the predicted house price to the
actual for the counties we have available and shows a close correspondence.

100

Slope: 0.877 (SE: 0.013)

100

200
300
Actual Metro MF Price Index

400

(a)

Figure A.7: Multifamily house price vs predicted for exact county matches
To obtain county-level price growth rates for the commercial property classes
in the CCRSI, we perform a similar procedure to the one employed for the Freddie Mac MFPI, modifed to account for the fact that the CCRSI are available at a
much higher geographic level than the Freddie Mac MFPI. Like the MFPI procedure, we combine the CCRSI with the FHFA HPI, this time the quarterly HPI
for All Transactions for Census Divisions. As for multifamily properties, we
obtain county-level estimates of commercial property price levels by regressing

A.16

the commercial price index (CMPI) in the CCRSI on regional HPI and yeardummy indicator variables and using the recovered coeffcients to predict CMPI
\ ct ) from county-level HPI.
(CMPI

A.8.3 Returns
We estimate returns at the property level, using line items from Schedule E and
Schedule 8825. The total returns on tenant housing is the sum of rental returns
and capital gain yields. For property p, the rental return is equal to the gross
pt
pt
rental yield, GRYpt = Rent
, minus the cost yield, CYpt = Cost
. Costs
M Vpt
M Vpt
consist of maintenance, management, utilities, and other expenses. The capital
gain yield is the real increase in housing price at the county-year-type cell, as
described above.

B

Additional fgures

B.1 Owner occupied housing

(a)

(b)

Figure A.8: Share of owner-occupied housing wealth, by capitalization type

A.17

(a)

(b)

Figure A.9: Average real house price growth by income group, rank factor income.

(a)

(b)

Figure A.10: Share of capital gains, by capitalization type, by factor income
rank.

A.18

B.2 Tenant occupied housing

(a) Total

(b) Homestead

(c) Multifamily

(d) Commercial

Figure A.11: Average schedule E real estate capital gain returns, by income
group.

A.19

(a) Total

(b) Homestead

(c) Multifamily

(d) Commercial

Figure A.12: Average Schedule E real estate rental return, by income group.

A.20

(a)

(b)

Figure A.13: Average Schedule E t.o. property wealth, by capitalization type.
‘hom rent’ capitalizes rental income using homogeneous returns. ‘hom proptax’
capitalizes property taxes with homogeneous rates. ‘het prop tax’ represents our
baseline series.

(a)

(b)

Figure A.14: Share Schedule E t.o. property wealth, by capitalization type.
‘hom rent’ capitalizes rental income using homogeneous returns. ‘hom proptax’
capitalizes property taxes with homogeneous rates. ‘het prop tax’ represents our
baseline series.

A.21

(a)

(b)

Figure A.15: Average Schedule E t.o. capital gains, by capitalization type. ‘hom
cap hom ret’ capitalizes wealth using homogeneous property taxes, and estimates
capital gains with homogeneous returns. ‘het cap hom ret’ capitalizes property
taxes with heterogeneous returns and estimates capital gains with homogeneous
returns. ‘het cap het ret’ is our baseline series.

(a)

(b)

Figure A.16: Share Schedule E t.o. capital gains, by capitalization type. ‘hom
cap hom ret’ capitalizes wealth using homogeneous property taxes, and estimates
capital gains with homogeneous returns. ‘het cap hom ret’ capitalizes property
taxes with heterogeneous returns and estimates capital gains with homogeneous
returns. ‘het cap het ret’ is our baseline series.

A.22

B.3 Private business wealth

(a)

Figure A.17: Private business wealth estimates vs Financial Accounts totals.
Total private business market value equals S-corp + partnership + private Ccorp + sole proprietorship. Financial Accounts equal market value of C-corp +
market value S-corp + fxed asset values of partnership + sole proprietorships +
net non-mortgage fnancial assets of noncorporate businesses. See section A.5.1
for details.

A.23

(a) S-corp

(b) Partnership

(c) Sole prop

Figure A.18: Private business, average Enterprise Value to Sales ratio.

A.24

(a) S-corps

(b) Partnerships

(c) Sole prop

Figure A.19: Private business wealth, average business size.

A.25

(a)

(b)

Figure A.20: S-corp business industry composition

A.26

(a)

(b)

Figure A.21: Partnership industry composition.

A.27

Direct ownership value of partnership businesses
by partner type
Billions of Dollars
1,000 1,500 2,000 2,500

Full pass-through value of partnership businesses
by partner type

20%

31%

500

16%

0%

Partnerships

(a)

Indirect ownership

(b)

Figure A.22: (a) Direct ownership of partnership business (b) Indirect ownership
of partnership business

A.28

d
te

ie

is

tri

bu

tif
en
ot
N

m

id

en

gn

tF

en

un

d

d

y
tit

of
it
rie

N

on

-p
r

or

ps

es

SC

at
Es
t
&

C

Direct ownership

U
nr
ed

Not identified

ire

Foriegn entity

Retirement Fund

et

Non-profit

R

Trusts & Estates

S-Corps

Fo

C-Corps

Tr

In

di

4%

7%
Individuals

us
ts

vi

du

1%

-C

al

or
ps

s

0

14%

7%

B.4 S-corp comparisons

(a) Value

(b) Sales

(c) EBITDA

Figure A.23: S-corp pass through totals.

A.29

B.5 Capital gains tax rates

(a)

(b)

Figure A.24: (a) Comparison of aggregated capital gains, 1954-2021 (b) Measures of capital gains, 1954-2021. ‘Nominal KGs’ are total realized and unrealized macro capital gains. ‘Taxable’ gains exclude pension and nonproft gains.
‘KG in agi + Excl’ are capital gains reported on tax returns plus an estimate of
capital gains on certain categories that are excluded by law (described in section 4). ‘Nominal KGs’ estimated from Financial Accounts data. ‘KG in AGI’
estimated from individual tax fles.

(a)

(b)

Figure A.25: Average percent of macro capital gains realized, by income group.
Calculated as total realizations divided by macro household capital gains. Bars
display averages across years.

A.30

(a)

(b)

Figure A.26: Average percent macro capital gains that are taxable, by income
group. Calculated as total taxable capital gains divided by macro household
capital gains. Bars display averages across years.

(a)

(b)

Figure A.27: Average tax rate on realized capital gain income in AGI, by income
group. Calculated as total taxes paid on capital gains divided by realized capital
gains. Bars display averages across years.

A.31

---

Source: Frix Law Library, https://www.frixlaw.com/law-library/documents/agency%3Airs%3A5e6da15f26e80b15. Public record. Not legal advice.