Content uploaded by Jesper Roine
Author content
All content in this area was uploaded by Jesper Roine on Dec 13, 2013
Content may be subject to copyright.
1
Top Incomes in Sweden over the Twentieth Century
Jesper Roine* Daniel Waldenström**
SSE/EFI Working Paper Series in Economics and Finance No. 602
2005-08-15
Preliminary version, please do not quote without checking with the authors.
Comments are most welcome.
Abstract
This paper presents homogenous series of top income shares in Sweden from 1903 to 2003
using individual tax returns data. We find that Swedish top incomes have developed more
similarly to the US, Canada and the UK than to other continental European countries when
capital gains are included. The top income shares are U-shaped over time, falling steadily un-
til around 1980 when they start increasing again. Around 2000 they reach levels similar to
those found around 1950, before the expansion of the Swedish welfare state. However, unlike
the Anglo-Saxon countries, where the recent increases were mainly driven by increased wage
earnings inequality, Swedish top income shares have risen almost exclusively due to capital
gains, a finding consistent with relatively high marginal wage taxes and internationally high
price increases in financial and real estate markets since 1980. When excluding capital gains
the increase in top income shares since 1980 almost disappears and the Swedish experience
looks more like that of continental Europe. Furthermore, we also find that the largest decrease
of top income shares happens between 1935 and the beginning of the 1950s, but not (as in the
US and in France) during the war years, but before 1939 and after 1945 suggesting that the
Swedish development was more driven by policy than by exogenous shocks.
* Department of Economics, Stockholm School of Economics, P.O. Box 6501, SE-11383
Stockholm, Ph: +46-8-7369000, e-mail: Jesper.Roine@hhs.se
** Department of Economics, Stockholm School of Economics, P.O. Box 6501, SE-11383
Stockholm, Ph: +46-8-7369000, e-mail: Daniel.Waldenstrom@hhs.se
2
Table of Contents:
1 Introduction ..................................................................................................................................................3
2 Methodology and data..................................................................................................................................6
2.1 Reference Total for Income...................................................................................................................9
2.2 Reference Total for Population ...........................................................................................................12
3 Results..........................................................................................................................................................13
3.1 Main trends in Swedish top income shares over the twentieth century...............................................13
3.2 Composition of top incomes ...............................................................................................................15
3.3 Other possible explanations for the trends in Swedish top income shares..........................................20
3.3.1 Changes in the tax legislation.........................................................................................................20
4 International comparisons .........................................................................................................................20
4.1 Comparing top income shares.............................................................................................................20
4.2 Comparing the composition of top incomes........................................................................................22
5 Concluding remarks ...................................................................................................................................24
References.............................................................................................................................................................24
Appendix A: Income inequality data..................................................................................................................29
A1: List of sources ............................................................................................................................................29
A1.1: Sources for the income data................................................................................................................29
A1.2: Sources for other data .........................................................................................................................29
A2: Details of Swedish income data..................................................................................................................30
A2.1: Definitions of income concept and income sources............................................................................30
A2.1.1: Concept of income ......................................................................................................................30
A2.1.2: Definitions of sources of income ................................................................................................31
A2.1.3: Estimating the income composition using post-1990 data..........................................................32
A2.1.4: Estimating the top shares of capital income, 1912–2003............................................................33
A2.2: Concept of income earners .................................................................................................................34
A3: The Swedish income tax system, 1903–2003.............................................................................................35
A3.1: General overview................................................................................................................................35
A3.2: Marginal and average income tax rates in Sweden, 1903–2003.........................................................36
A3.3: Changes in capital gains taxation and its effect on our results............................................................36
A3.3.1: Rules for capital gains taxation, 1903–2003 ...............................................................................36
A3.3.2: The effect of changes in capital gains taxation on our study ......................................................37
A3.4: Lowest taxable income threshold........................................................................................................37
A3.5: Tax evasion and tax avoidance ...........................................................................................................38
Appendix B: Construction of reference totals ...................................................................................................39
B1: Reference total population..........................................................................................................................40
B2: Reference total income ...............................................................................................................................42
Appendix C: Sensitivity analysis: various adjustments, changing reference totals etc..................................44
C1: Sensitivity analysis of choosing using different reference totals ................................................................44
C2: Sensitivity analysis of choosing either individuals or families as tax units ................................................44
C3: Age adjustments and effects of censoring the youngest income earners ....................................................45
Appendix D: Estimation theory ..........................................................................................................................46
D1: Pareto interpolation: description and examples..........................................................................................46
D2: Robustness issues of Pareto interpolation...................................................................................................47
3
1 Introduction
Over the past couple of years there has been an increased interest in income inequality at the
top of the distribution, both academically and in the policy debate. The well documented long-
run trends, with clear increases in income inequality in the US (Piketty and Saez, 2003) and
the UK (Atkinson 2002) over the past decades, has made some conclude that policy has
shifted so as to tolerate widening gaps in society. Others see signs of “technological change”
and a new “Industrial revolution” which, in line with Kuznets’ hypothesis, initially leads to
increased inequality.
In the Swedish context recent corporate scandals and increased income gaps between “the
elite” and “the average worker” have also caused debates about top income earners and to
what extent the Swedish welfare state is being dismantled. These discussions of inequality
should be viewed in light of the fact that Sweden, at least until the early 1990s, had one of the
most equal distributions of income among industrialized countries (see e.g. Atkinson, Rain-
water and Smeeding (1995) and Gottshalk and Smeeding (1997)). However, as pointed out by
Björklund and Palme (2000), “when and how Sweden achieved its equal distribution of in-
come is more of an open question. Is it a rather recent outcome of the growth of the welfare
state during the 1960s and 1970s? Or is it a historical inheritance from rather a long time
ago?”(p. 115). In this paper we try to contribute to answering this type of question, focusing
on the top of the income distribution. What has been the development of top income shares in
Sweden over the twentieth century? What were the effects of the World Wars and the depres-
sion in the 1930s? What was the effect of the expansion of the welfare state in the 1960s and
1970s? Has there been an increase in top income shares over the past decades as in the Anglo-
Saxon countries, or is the development in Sweden closer to that found in France and the Neth-
erlands, where top income shares have been rather stable in the recent past? And what does
this recent development look like in relation to the rest of the Twentieth Century?
To be able to shed light on questions like these we need long-run, homogenous, series of in-
come inequality, which ideally should be comparable across countries. In this paper we con-
struct such series of pre tax income for Sweden over the one hundred years between 1903 and
2003. Using data from tax returns combined with National Accounts statistics and population
statistics we construct series for the share of “total income” going to the top decile, the top
five percent, the top percent, etc., down to the top 0,05 percent of the adult (tax) population.
4
As far back as possible we also calculate shares of different sources of income (wages, capital
income, business income and capital gains income) for each fractile.
There are three main contributions in the paper. First, we construct, for the first time a long-
run series of income inequality in Sweden spanning the whole of the Twentieth century. In the
absence of micro data few have ventured further back than to 1967. A recent exception is a
study by Björklund and Palme (2000) who focus on the period 1951–1973. Before them, e.g.,
Spånt (1976, 1979) reports on the period 1920-1976, Spånt and Selander (1969) 1951–1966,
Lydall (1968) 1920-1960, Bentzel (1953) 1930–1948, Lindstrand (1949) 1935–1947, Quensel
(1944), 1930–1941 and Hagstroem (1944, 1949) covers the period 1912-1947. Bentzel (1953)
stands out as a particularly interesting reference since his method is very similar to the one we
use.
1
In addition, Gustafsson and Johansson (2003) have carried out a study of the income
distribution between 1925 and 1958. Their study is, however, limited to the City of Göteborg.
To the best of our knowledge no one has previously constructed longer series than the ones
mentioned above and no one has studied income inequality in for the very first years of the
Twentieth Century. Furthermore, none of the above focus on the very top of the income dis-
tribution as we do in this paper.
Second, our series are constructed so as to be homogenous over the whole period. Previous
studies all use slightly different methods and concepts, and are not immediately comparable.
Throughout this paper, we focus on building homogenous series, and to the extent that this
means abstracting from some detail which may be of importance at one point in time but for
which we have no information at other points in time, we chose to have comparability over
time. Our series are hence, best suited for studying the long-run trends in the data (even
though, as we shall see, our series are also in line with previous studies based on more de-
tailed information).
Third, we focus on the development of the top income earners including data for fractions of
the top percent. We know of only one previous study on Swedish data with this focus by
Magnusson (2004) which is confined to the period 1985-2002. Most previous studies on
Swedish data have focused on welfare and especially on disposable incomes amongst those in
1
To the extent that Piketty and Saez (2001, 2003) extend and generalize the work by Kuznets on US data in the
early 1950s, we extend and generalize the work by Bentzel (1953) on Swedish data for top incomes, both by
going further back in time, as well as constructing series for the 50 years that has past since Bentzel’s publica-
tion.
5
the lower parts of the distribution. Some would even say that it does not matter what happens
at the top, and certainly not the very top of the distribution, since this (at least in a Swedish
context) concerns such a small number of individuals. We do not think this is the case. There
are a number of reasons for why it is interesting to know more about the development of top
incomes. First of all, the development of the top incomes gives important indications of the
structure of long-term economic development, as suggested by Kuznets (1953). Second, high
income individuals typically control large assets, and make decisions which potentially affect
the economy at large. What happens in this group is therefore of interest for the whole econ-
omy. Third, those with the highest incomes also constitute a potentially important source of
tax revenue. Last, but not least, not going beyond the income shares of the top decile or at best
the top five percent, one runs the risk of completely missing what is really driving changes in
the top of the distribution. As we shall see, much of the development of the top income shares
in Sweden over the past one hundred years actually stem from what has happened in the top
percent of the distribution. As pointed out by Atkinson (2004) changes at the very top can also
have a substantial impact on the Gini coefficient and, hence, not carefully studying the devel-
opment of the top may lead to incorrect inferences about what is driving changes in inequal-
ity.
Internationally the use of income tax data for the construction of long time series of income
distribution is far from new. Besides Kuznets’ classic work on the “Shares of Upper Income
Groups in Income and Savings” (1953), Champernowne (1973) has studied data based on
“super-tax” records which was introduced in 1908 in the UK, Hartog and Veenbergen (1978)
construct income distribution estimates for the period 1914-1972 using income statistics data
for the Netherlands, Sörensen (1993) has made use of Danish income statistics data for the
period 1870-1986. Morrison (2000) provides an overview of historical data for Europe. Not-
withstanding the many previous studies, it is fair to say that the approach of Kuznets has only
recently been fully revived, starting with the work of Piketty (2000, 2003) on France, Piketty
and Saez (2001, 2003) on the US, and Atkinson (2004) on the UK. In the past couple of years
a number of studies have used very similar methods to construct series for a number of coun-
tries, e.g. Atkinson and Salverda (2005) who compare top income shares in the Netherlands
and the UK, Atkinson and Leigh (2004) who compare top income shares in a number of An-
glo-Saxon countries, Banerjee and Piketty (2004) who study top income shares in India over
the Twentieth Century, Dell (2003, 2005) who constructs series for Germany and Switzerland,
and Saez and Veall (2003) who construct series for Canada. In this paper we do the same for
6
Sweden, using similar methods as those used in the recent papers mentioned above, so as to
allow comparisons between our series and those for other countries.
In the following section we briefly describe our methodology and the data used. All details
concerning adjustments, alternative specifications and sources have been placed in a number
of appendices. In Section 3, we present our main results, which we then compare to the find-
ings from other countries in Section 4. Section 5 concludes.
2 Methodology and data
This section briefly describes our method and the data used. Details concerning the data, the
sources, and the adjustments made are placed in a number of appendices. For some of the
most important concepts, however, Table 1 gives a brief overview.
Table 1: Overview of income data used, adjustments and reference totals
a
.
Income
years
Main income concept in tax statis-
tics [Swedish term]
Adjustments to
get comparable
income concepts
Reference total
income
Reference total
population
1903–1909 Total income [Taxerad inkomst] No adjustments
1910–1942
Total income plus a share (–1937:
1/60; 1938–1942: 1/100) of tax-
able wealth [Taxerat belopp]
Removal of
wealth shares
using data on
wealth distribu-
tion per income
class
Share of personal
sector total income
(in National ac-
counts) based on
estimates of items
not included in
preferred def.
(1903–1942)
1943–1950 No adjustments
Adult population
(16 yrs and above)
minus married
women (–1951)
1951–1970
Total net income = Total income
minus deficits at source [Samman-
räknad nettoinkomst]
Adult population
(16 yrs and above)
less housewives
(1951–1971)
1971–2003
Total income [Sammanräknad
inkomst (1971–1990), Samman-
räknad förvärvs- och kapitalin-
komst (1991– )]
Excluding tax
returns from
those under the
age of 16 in
each income
group.
Tax statistics in-
come plus esti-
mates of non-taxed
items included in
preferred def. plus
estimated incomes
of “non-filers”
(1943–2003)
Adult population
(16 yrs and above)
(1971–)
a
All concepts referred to in the table are elaborated upon in the Appendix. The below refers to our preferred
versions of the reference totals. To check the robustness of our results we have calculated a number of different
series showing that our main conclusions and the trends in the series do not depend on the precise choice of
reference total.
Our basic concern is to analyze the share of total income before tax (from all sources) earned
by the top decile (P90-100), the top 5 percent (P95-100), the top 1 percent (P99-100) and so
on. Furthermore, we distinguish between series including and excluding capital gains respec-
tively and we will also decompose the series according to the source of income. The shares
are constructed in relation to a reference total for the (tax) population as well as for income. It
is important to note that this is different from measuring the share of “total taxed income”
7
received by some share of the taxpaying population. Such a measure would fluctuate with
changes in tax policy regarding who should pay taxes and what incomes are taxable, without
there being any underlying changes in the actual income distribution. Our measures are in-
stead top shares in relation to estimates of 1) the total income as if all income (our preferred
definition of it) was subject to taxation and 2) the total tax population as if everyone filed a
tax return. The reason for this being the appropriate way to construct homogenous income
shares over time is of course that we do not want changes in tax policy regarding, for exam-
ple, exemption levels and what sources of income are subject to taxation, to affect our income
shares.
Given our focus on the top income earners, there are two basic requirements for our series to
be correct. First, income according to the tax return should be equal to the actual income of
everyone in the top decile. If the taxable amount in the income tax returns contains something
which we do not want to include, or if it does not contain something we would like to include,
we must have information on the distribution of this item over income groups so as to be able
to deduct or add the right amounts to the right group. Similarly, it must be the case that every-
one who actually belongs to the top decile also files a return. The main problem regarding this
first requirement is of course potential underreporting of income. Second, we must be able to
estimate aggregate numbers for all incomes and all potential tax units in the economy. Impor-
tantly, however, we do not need information on the distribution of excluded items as long as
they do not belong in the top decile.
Our calculations are based on data from income tax returns starting when the modern income
tax was introduced in Sweden in 1902. The series are based on pre-tax total income, that is
market income from all sources (labor, capital, business and capital gains, the last also being
treated separately). The relevant concept in the Swedish income statistics is, hence, Samman-
räknad inkomst (or the equivalent, see Table 1 above). To make the income data consistent
with our definition of reference total population we have excluded all income earners below
the age of 16 in each income class.
2
2
To be precise we have excluded the number of income earners below the age of 16 in each income class, but we
have not been able to deduct their incomes. The income shares of those below 16 are, however, insignificant and
much smaller than their share of the total number of income earners. See Appendix C3 for details.
8
Needless to say, there are a number of problems with tax statistics data; it is collected as part
of an administrative routine in which individuals have incentives to underreport income, it
tells us nothing per se about the welfare of individuals, tax units do not necessarily correspond
to the units we are interested in, initially it only covers a small part of the population, etc.
Nevertheless, as long as we think that tax statistics, at least for the top income earners, ap-
proximate actual incomes, and as long as the problems with the statistics have not changed
systematically over time, they are a useful source. And, importantly, it is the only available
source for the whole of the Twentieth Century.
In terms of data availability it is useful to distinguish between two periods; one between 1903
and 1942, the other from 1943 and onwards. In the first period, taxation data divided into in-
come groups are only available for the years 1903 (only the very top), 1907, 1911, 1912,
1916, 1919 and 1920, but the tax returns also formed the bases for income distribution esti-
mates calculated in the Census (Folkräkningen) in 1920, 1930, 1935 and 1945, as well as for
some studies of income distribution made at the time (e.g. SOU 1936:18 for the year 1934,
and Quensel, (1944) for 1937).
3
Combining these sources and making adjustments so as to get
homogenous series we have sufficient data to be able to study the development over the whole
period, with the very unfortunate exception of the 1920s for which there is only income distri-
bution data for the income year 1920. Appendices A-C provide more details on the data, the
adjustments made and sensitivity of our final results with respect to these adjustments. Start-
ing in 1943 taxation data for income classes have been compiled yearly and, hence, data is
available for every year after that. The statistics are not however completely comparable for
the whole period after 1943 since a number of changes have occurred in definitions and meth-
ods of compiling data. Again details of these changes have been confined to the appendices.
The form of the basic data is typically tabulations for a given year of the total number of tax
units in a given income class and the sum of their incomes. For example, according to the tax
statistics for the income year 1912, there were 245,792 tax units (i.e. singles or married cou-
ples, who at the time were treated as one tax unit) with a taxable income between 800 and 900
Swedish kronors (SEK). Their taxable income sums up to SEK 185,247,200. Similarly, there
were 148,464 tax units with income between 900 and 1,100 SEK earning together SEK
154,744,400, etc. all the way up to 698 tax units having above SEK 80,100 and a total of SEK
3
In the Swedish statistics one makes a distinction between the Taxation year and the Income year, the latter
being one year before. Throughout this study all years refer to the Income year.
9
130,830,500 in taxable income.
4
In 1912 taxable income, however, included a wealth share of
1/60 of taxable wealth. Fortunately we have data on the wealth shares per income group,
which allows us to subtract the wealth share for each income group so as to get the actual in-
comes. Taxable income also differed from actual income in a number of other ways, but this
difference mainly concerned incomes in the lower part of the distribution, and to the extent
that they affected top income earners their effects were negligible. Based on this data we can
construct cumulative numbers of incomes and tax.
The main income concept in the official Swedish income statistics is total income, which is
the gross income of labor, capital, business and capital gains before taxes and most deductions
(yet with slight variations over time as shown in Table 1).
To calculate the top income shares we now want to relate these numbers to total income in the
economy as well as to the total number of tax units. For the year 1912 the total taxed income
(adjusted for the wealth shares) was 1,353,046 thousand SEK but according to our preferred
estimate the total (individual) income was 2,126,596 thousand SEK. In that same year the
total number of tax filers was 714,919 tax units but the total number of tax units in the econ-
omy was according to our preferred estimate 2,888,302. Relating these numbers to the cumu-
lative number of tax filers and their total income divided into income classes, we get that
0.099 percent of all tax units (2,851/2,888,302) had 8.968 percent
(190,704,625/2,126,596,000) of all income, 0.047 percent (1,353/2,888,302) had 6.704 per-
cent (142,566,868/2,126,596,000) of all income, etc. Based on these shares and the assump-
tion that the distribution is of Pareto form, we can then use standard interpolation techniques
to get the income share of the different desired shares of the population (see Appendix D for
details on the inter- and extrapolation techniques used).
2.1 Reference Total for Income
In principal there are two ways of constructing the reference total income. Either, one starts
with data on “Total Personal Sector Income” reported in the National Accounts and deduct
items not included in the preferred definition, or one can start from the “Tax Statistics In-
come” and add items not included in the tax base and income estimates for individuals not
included in the tax statistics. As outlined in Atkinson and Salverda (2003);
4
See also Table A1 for another example (1945) of how the income tables appear in the published sources.
10
Personal sector total income
− Non-household income (e.g. charities)
− Items not included in preferred definition of income (like employers’ social security
contributions).
= Preferred Reference Total Income Definition
− Items not included in the tax base (like certain social security benefits)
− Taxable income of those not included in tax statistics (“non-filers”)
= Tax statistics income
We have calculated the reference total income from “both sides” and our final preferred series
is a combination of the two for reasons described below (see Appendix B for details and com-
parisons of different reference totals). The problem, in the Swedish case, with starting with the
National Accounts data for Personal sector income is that there are no homogenous series
which are detailed enough for the whole period. The most recent and up-to-date National Ac-
counts series have been published by Edvinsson (2005) from which we use data for “Wages
and salaries (including social benefits)” and “Imputed labor income of self-employed (includ-
ing social benefits)” from 1903-2000. Edvinsson (2005) does not, however, publish separate
figures for Property income and Capital income. What we do have is aggregate tax statistics
for these incomes starting in 1922. We therefore add these aggregate numbers to wages and
salaries and the labor income of the self-employed (implicitly assuming that all incomes from
these sources were included in the tax statistics). In the period before 1922, for which we have
no data on property- and capital income, we add a fixed share of labor income based on the
ratio of property- and capital income to labor income in 1922. The resulting series is on aver-
age 0.7 times the GDP (calculated from the expenditure side) reported in Edvinsson (2005)
with a standard deviation of 0.03 and forms our estimate for total personal income (which still
includes some items, such as employers’ social security contributions, which we would like to
exclude in our preferred definition of reference total income).
Starting from the other end, with the total income according to the tax statistics, gives a rela-
tively precise picture in recent years, as a majority of the adult population file tax returns (ac-
tually close to 100 percent (!) since 1978, when government institutions, employers, banks,
etc. started to send income statements, so called “kontrolluppgifter”, to the tax authorities)
and as most social security benefits are taxable. Hence, the tax statistics income is close to our
preferred total income definition for the past couple of decades. However, going back in time,
both the number of “non-filers” and the number of items not included in the tax base increase.
To get a homogenous total reference income series we must add the incomes of non-filers and
11
items not included in the tax base but which form part of total income. The major reform in
terms of including social security benefits in the tax base was the 1974 tax reform. This year
sick leave benefits, unemployment benefits and parental leave compensation all became part
of taxable income. We therefore add the aggregate government outlays for all of these items
to the tax statistics income in the period until 1973. We have not included a number of smaller
items which were tax free before, and remain tax free after 1974 (such as study grants, child
allowances, etc.). When it comes to estimating the incomes of “non-filers” there are two basic
approaches which have been used. Piketty and Saez (2001, 2003) and Atkinson and Salverda
(2003) assume that non-filers have a fixed share of the average income of those who pay tax
(like 0.3 times the average income). Bentzel (1953) and Spånt and Selander (1969) instead
assume that non-filers have 0.5 times the threshold income. That is, if everyone with income
above SEK 600 should pay taxes (as was the case in Sweden between 1919 and 1951) they
assume that everyone not filing a tax return had an income of 300 SEK.
5
We have calculated
series in both ways and believe that using a share of the threshold income is the best way to
proceed in the Swedish case. Even though this at times gives quite low estimates (and one
could argue that it would not be possible to survive on such a small income) we think these
same individuals probably get parts of the (at the time tax free) social welfare benefits. There-
fore, imputing larger incomes to them we would risk double counting items which we have
already added when adding tax free social benefits to the reference total. We think however,
that the average income for non-filers is closer to the threshold than 0.5. In our preferred se-
ries we impute 0.8 times the threshold income.
Having calculated an estimate for total personal income (which we know to be an upper limit
to the preferred income total since it contains items which are not included in our preferred
total income definition) and an estimate for the “Preferred Reference Total Income Defini-
tion” starting with the tax statistics income, adding income for non-filers and items not in-
cluded in the tax base we can compare the two series. This leads us to the following preferred
specification for the reference total income: 1) For the period 1903-1942 it is 0.89 times the
estimated total personal income and 2) for the period 1943-2003 it is the income according to
tax statistics, plus an average income of 0.8 times the tax threshold for non-filers, plus social
benefits that were not part of the tax base before 1974 (see Table 1). Appendix B contains all
calculations and compares the different methods, which in the end does not change the results
5
See Figure AX showing the lowest taxable income and its share of average income 1903–2003.
12
very much, nor does it change the trends at all (as is shown in Appendix C). Figure 1 shows
the different series as shares of GDP.
Figure 1 (from the xls. File “Finalreftotals1903-2003”)
2.2 Reference Total for Population
Just as we do not want to construct top income shares in relation only to taxed income, but to
an estimate of “total income”, we do not want to define the top shares of the population as a
share of those who pay taxes, but as a share of total number of potential tax units, i.e. as if
everyone would have filed a tax return. The major problem in constructing such a total in
Sweden is that the concept of tax unit has changed from a “family based” system, where mar-
ried couples filed a joint tax return, to an individually based system where individuals
(whether married or not) file separate returns. In the tax legislation this shift took place in
1971, with an option to file separate returns from 1966. In terms of tax statistics, however,
this change occurred (at least to some extent) already in 1951. Before this, tax statistics were
based on the entire tax population and figures referred to “tax units” i.e. married couples
counted as one income earner.
6
Before 1951 the obvious reference population is therefore the
adult population (which we take to be everybody aged 16 or above) less married women
(since a married women formed one tax unit together with her husband). After 1951, however,
statistics changed to being based on a representative sample (ten percent) of the population,
with married couples where both had an income, now treated as two income earners in the
statistics even though they were still taxed as one. The problem is that in cases where the
women did not work, or had low income, she was not necessarily counted (as is pointed out in
many of the sources). This means that income statistics between 1951 and 1971 is a mix be-
tween a family based system and an individually based system including some women but not
all. Starting 1971, the reference total is again relatively unambiguous, now obviously being
the adult population.
In terms of choosing the appropriate reference population in Sweden, the period 1903-2003
can, hence, be divided into the following three periods: 1) 1903-1950, the total population
aged 16 or above minus married women, 2) 1951-1970, the total population aged 16 or above
minus women likely to be excluded in the statistics, 3) 1971-2003, the total population aged
16 or above. In our preferred specification we define the reference total for the population to
6
Note that this is the case for tax statistics before 1951 but not income figures in the Census (Folkräkningen).
13
be the total population aged 16 or above minus married women in until 1951, the population
aged 16 or above minus the number of housewives until 1967, the population aged 16 or
above minus a declining share of housewives 1967-1971, and the population aged 16 or above
from 1971 and onwards. (We also make some minor adjustments for individuals who have
died during the year depending on how they are treated in the statistics). The reason for choos-
ing this particular way of defining the reference total for the population is that it makes for the
smoothest transitions at years when statistics (and the number of tax returns) change, without
being an ad-hoc adjustment. Figure 2 shows the ratio of the total number of tax returns for a
number of alternative specifications of the reference total population. As with the reference
total for income we have calculated top income shares using a number of different reference
totals for the population and the results can be found in Appendix C. Appendix B contains all
details of how the different reference totals for population have been constructed.
Figure 2 (from the xls. file “FinalIncShares1903-2003”)
3 Results
We have organized the presentation of our results as follows: We first present the main trends
in Swedish top incomes – with and without capital gains - over the Twentieth Century, then
we study the composition of income for the different fractiles. All results presented here have
been calculated based on our preferred series for adjusted taxed income and our preferred es-
timates of the reference totals, but in Appendix C we present the main results using different
reference totals, which show that while point estimates and levels may change somewhat, the
main trends hold true for all our different specifications. We have also checked whether the
results are sensitive to changing the age limit for the relevant population from 16 to 20. This
does not change the results in any significant way.
3.1 Main trends in Swedish top income shares over the twentieth century
In short, the Swedish top income shares over the twentieth century can very broadly be char-
acterized as 80 years, between 1903 and around 1980, of relatively steady decline, followed
by 20 years, between around 1980 and 2003, of increasing top income shares when including
capital gains (after 1980 the series peak in the year 2000 after which the top income shares
have fallen quite substantially). However, this increase over the past twenty years almost dis-
appears in the data when capital gains income is excluded.
14
The most rapid period of decline seems to be between the mid 1930s and the beginning of the
1950s. There are not, however, any clear signs that top shares would have declined especially
dramatically during the war years as seems to have been the case in France and the US (see
Section 4 in this paper). If anything the most rapid decline seems to have been the period just
after 1945, suggesting that the decrease in top income shares in Sweden had more to do with
policy than exogenous chocks. Seen over the whole period, most of the decline happened be-
fore the early 1950s. Figure 3 shows the top income shares of the top ten percent (P90-100),
the top five percent (P95-100) and the top one percent (P99-100) including capital gains for
the whole period, and also excluding capital gains for the period 1967-2003. The latter series
show that when excluding capital gains there is no clear upward trend in the top income
shares, suggesting that most of this development is a result of differences in capital gains
rather than other types of income.
Figure 3 (P90-100, P95-100, P99-100) (fig3 in the excl-file)
Most of the changes in the top income shares come from changes in the very top of the distri-
bution. The income share of those between the 90
th
and 95
th
percentile (P90-95) is almost con-
stant over the past one hundred years(!) The income share of those between the 95
th
and 99
th
percentile is also relatively constant for the period after 1950 (before that there is a relatively
steady decline in P95-99). This suggests that studying the very top of the distribution is of
great importance to understand what is driving changes in measures of the top decile. Again
most of the action over the past decades seems to come from capital gains.
Figure 4 (P90-95, P95-99, P99-100, from the fig4 xls. file “FinalIncShares1903-2003”)
Looking at the top percent and the top 0.5 percent of income, we see a similar pattern of rela-
tively steady decline, again with most of the decline before the early 1950s and with the
sharpest drops in the years 1946-1952 and before the Second World War.
Figure 5 (P99-100, P99.5-100, fig5 from the xls. file “FinalIncShares1903-2003”)
For the very top of the distribution, the top 0.1 percent and the top 0.05 percent, the pattern is
again similar, but the magnitude of the changes are even larger than for the other groups.
Roughly, the share for the top 0.1 and top 0.05 percent of income earners decreased by a fac-
15
tor of 10 between the beginning of the 1900s and 1980. It then increased by a factor of about 5
between 1980 and 2000. (For the top 10 percent of income earners the drop between 1903 and
1980 was less than a factor 2 and the subsequent increase between 1980 and 2000 about 1.3).
Figure 6 (P99.9-100, P99.95-100, fig6 from the xls. file “FinalIncShares1903-2003”)
Another way of looking at the data is to calculate the income thresholds (in constant prices)
for the various fractiles and to express these as multiples of average income. Figure 7 illus-
trates the development of the income needed to be in the respective fractiles of the distribution
expressed in 2003 prices.
Figure 7 (income thresholds Fig7 from the excel file).
Figure 8 illustrates these numbers but expressed as multiples of average income. This again
suggests that the difference between an individual just at the top ten (or five) percent threshold
and the average income earner has not been very large and has not changed much over the
past one hundred years, while the income of a top 0.1 percent compared to the average has
first dropped dramatically but then also increased substantially in the recent decades (when
including capital gains).
Figure 8 (income thresholds as multiples of average income)
3.2 Composition of top incomes
Examining the composition of top incomes is important for the understanding of the devel-
opment of top income shares. For example, shocks to capital income during World Wars I and
II explain much of the secular decline in French top incomes (Piketty, 2003) while drastic
increased wage and salary earnings for the already rich has been the prime factor behind the
tremendously increased income inequality in the U.S. during the 1980’s and 1990’s (Piketty
and Saez, 2004). The composition of Swedish top incomes also changes significantly during
the twentieth century, but not always in tandem with other Western economies.
16
The sources of income which are distinguished in Swedish tax laws are labor (wages and sala-
ries), capital (mainly interest earnings and dividends), business income and capital gains.
7
Apart from capital gains, these are the standard income sources used by most countries’ au-
thorities reporting tax and income statistics, and hence the ones analyzed in previous studies
of top incomes (see, e.g., Piketty 2003; Piketty and Saez 2004; Saez and Veall 2003, 2005;
Dell 2003, 2005). Capital gains are somewhat different from the other income sources, pri-
marily because they are lumpy outcomes dependent on the timing of their realization. Capital
gains can be seen as the result of an accumulated stream of wealth increases (or decreases)
over some period of time, and their appearance on the tax return may not coincide with the
period when they actually emerged. Therefore, unlike labor income, income from capital
gains at time t may already have been capitalized by the income earner through past consump-
tion or other economic activities. In practice, however, the other income sources can be well
as lumpy as capital gains (consider the income chock of losing or getting a job!) and some-
times capital gains are left out of the compositional analyses simply because they are not in-
cluded in the country’s income statistics (as in France where they were not subject to taxation
before the 1980’s). As for Sweden, capital gains have for most of the twentieth century been
an integral part of income taxation and the official income statistics. A more detailed discus-
sion of the compositional data and the construction of the series can be found in section
A2.1.2 in the Appendix. Important to note, however, is that all composition series from 1991
onwards are substantially less reliable than earlier years for two reasons. First, the main in-
come concept used in the tables shifted to earned income (excluding capital income and capi-
tal gains) along which the incomes in some composition tables are sorted. Second, the post–
1991 concept of capital income also includes capital gains, and though we have estimated the
shares of the two sources for the last years they are still indirect observations.
Figures 7, 8 and 9 show compositional cross-sections for Swedish top income fractiles in
three years, 1945, 1978 and 1997, both when capital gains are included (solid lines) and ex-
cluded (broken lines). Prior to 1945 there are no comprehensive data on income composition
(except for capital income, see below) and after 1997 data on all income sources are not freely
available from Statistics Sweden (see further Appendix A2). As can be seen in Figure 7 the
top incomes in 1945 were dominated by wages in all fractiles except for the top 0.05 percent,
7
As discussed in Appendix A, the Swedish income statistics reported six different sources of incomes until 1990
and only three thereafter. Using available data and mainly the latter definitions, we are able to construct consis-
tent and continuous series of the four above-mentioned sources for the entire postwar period.
17
although its share was falling in the level of income. While having much lower shares, both
business and capital income increase in income. Capital gains, finally, are completely insig-
nificant at all income levels. In 1978, the dominance of labor income is augmented at all in-
come levels, representing around 90 percent or higher for all income earners below the abso-
lute top (top 0.1 percent and above). Compared with 1945, the increasing shares of labor in-
come and capital gains are the most significant changes. Looking at the composition in 1997,
the increase in the share of capital gains is greatly augmented. When including capital gains,
the share of wages decrease rapidly as one increases the level of income. For the top 0.05 per-
cent, capital gains represent three quarters of total income while wages only one seventh. Ex-
cluding capital gains, however, the picture looks more similar to previous years. Also interest-
ing is the decline in business income, which only represents 1–2 percent of total income at all
income levels in 1997, i.e., the same share for above-median and the top 0.05 percent income
earners.
Figure 7: Income composition across top fractiles: 1945
Figure 8: Income composition across top fractiles: 1978
Figure 9: Income composition across top fractiles: 1997
The long-run evolution of income composition for fractiles P90–100, P99–100 and P99.95–
100 for the period 1945–2003 is shown in Figures 10a-b, 11a-b, 12a-b. In all cases, we present
series with and without capital gains (although they are always part of the underlying total
income concept). While wages and salaries are still relatively important for the top 10%, they
are only about two thirds of the income when capital gains are excluded (about half when they
are included). Specifically, after 1980 the share of capital income and capital gains increase
significantly for the highest fractiles.
8
For example, when including capital gains, their in-
come share for P99–100 increase from an average of 3.7 percent in the 1970’s to 12.5 percent
in the 1980’s and 30.4 percent in the 1990’s. This represents a percentage increase of 243 per-
cent and 143 percent, respectively. For P99.95–100, the average shares for the same periods
were 12.1 percent, 39.9 percent and 62.3 percent, reflecting increases of 226 percent and 58
percent, respectively. Interestingly enough, P99–100 increased their capital gains share from
1.1 percent to 3.8 percent and 13.2 percent, which represent percentage changes of 238 per-
8
It should be noted that capital gains were actually included in labor income before 1967 why their exact shares
are unknown in the early years. Census data in 1945 and 1951 suggests fairly modest shares, however, with point
estimates of 1.23 percent and 0.42 percent of total net income, respectively.
18
cent and 250 percent. In other words, in terms of wealth payoffs there was a clear conver-
gence during the 1990’s within Swedish top income groups. The exact types of capital gains
earned are not disclosed in the data sources we use, but special investigations by Statistics
Sweden (Statistics Sweden 2001, 2002, 2003) suggest that the main source is sales of finan-
cial assets, and especially so among top income earners. Swedish financial and real estate
markets experienced a well-documented boom during the 1980’s and 1990’s, coming in two
waves separated by the financial crisis of 1991–1993. Over the whole twenty-year period,
however, the stock prices in Stockholm increased 30 times compared to only ten times in New
York, London and Paris.
9
Another finding is that labor income represented a substantial part of Swedish top incomes
during most of the postwar period, since 1970 about 80–90 percent for the top 1 percent and
about half for the top 0.05 percent when capital gains are excluded. Over time, however, its
share exhibits an inverted U-shaped tendency, peaking around 1980 and then decreasing faster
in the 1990’s, especially when capital gains are counted. This development could be closely
related to the variance in top marginal income tax rates, which when they reached their high-
est postwar levels in the 1970’s and 1980’s may have induced top income earners to work
less.
10
The tax reform in 1991 also seems to have mattered by its more comprehensive defini-
tion of taxable capital income.
The share of business income displays perhaps the most uniform time pattern of all income
sources. It declines constantly throughout the postwar period. From initial levels of about 30
percent of total income down to a mere couple of percentage points in the 1990’s (after 1997
we have no data on business income specifically) for both the top 1 percent and top 0.05 per-
cent groups. This marked decay is partially explained tax rule changes after 1991, when self-
employed small-firm owners became increasingly employed by their own firms and thereby
able to choose whether to get their income in the form of wages, dividends or capital gains. A
more economic-structural explanation, however, could also be that the Swedish economic
postwar institutions have systematically generated incentives for people to be wage earners,
and thereby enjoy the full set of public insurances, instead of being self-employed entrepre-
neurs, with less security in case something goes wrong but still only reaping a fraction of
eventual profits due to high marginal taxes (e.g., Henrekson 1996).
9
These numbers are based on stock price indexes in the International Financial Statistics published by IMF.
10
Klein (1995) finds the tax revenue elasticities with respect to tax rates to be the highest during ...
19
Figure 11a: Income composition (excluding capital gains) for P90–100, 1945–2003
Figure 11b: Income composition (including capital gains) for P90–100, 1945–2003
Figure 12a: Income composition (excluding capital gains) for P99–100, 1945–2003
Figure 12b: Income composition (including capital gains) for P99–100, 1945–2003
Figure 13a: Income composition (excluding capital gains) for P99.95–100, 1945–2003
Figure 13b: Income composition (including capital gains) for P99.95–100, 1945–2003
Over an even longer time period, 1912–2003, Figure 14 shows the share of capital income of
total income in the top.
11
Just before World War I, capital income represented 30–55 percent
of top 10 percent–0.05 percent incomes, and this share increased in 1916 to 50–80. While the
high prewar levels correspond well with the kind of wealth accumulation witnessed in many
European countries during the era of intense industrialization and peace (see, e.g., Piketty et
al. 2005 on the French wealth distribution during this time), the remarkable Swedish war
boom 1914–1917 with greatly increased corporate profits and stock prices is less commonly
observed.
12
After the war, Sweden experienced a severe financial and monetary crisis which
depressed markets, and capital incomes, significantly. The build-up during the 1920’s resulted
again in relatively higher capital incomes in 1930 but then world depression and, especially
for the Swedish top wealth, the so called Kreuger Crash in 1932 represented a major blow to
Swedish top capital incomes. The late 1930’s saw another economic upswing and, interest-
ingly, which for Sweden lasted until after World War II. World War II does not seem to have
had any particular impact on capital income shares for Swedish top income earners. During
the early postwar years, however, the shares declined substantially (if one is to believe these
isolated point estimates) and the decrease was even larger than during the turbulent early
1930’s. Between 1950 and about 1980, top capital income shares were relatively stable at
their lowest level during the entire century. Thereafter, however, they started to increase and
by the late 1990’s they had reached approximately the same levels as during the pre-World
War I period, indicated a long U-shaped pattern of Swedish top capital income shares.
Figure 14: Capital income share of total income (excl- capital gains), 1912–2003.
11
Before 1945, we compute this share as a flat rate of return of the personal wealth estimated for different in-
come classes (see further appendix A2).
12
See, e.g., Östlind (1945, pp. 258ff) or Schön (2000, pp. 275ff).
20
3.3 Other possible explanations for the trends in Swedish top income shares
3.3.1 Changes in the tax legislation
There are many instances through which changes in tax laws could significantly affect the
income of top income earners. For example, drastic changes in the highest marginal income
taxes have significant incentive effects on the work efforts of people with high incomes, but
also on their propensity to spend resources on avoiding or evading taxation. In the first case, a
registered change in top income shares would reflect a true income share change in the spe-
cific top fractile whereas in the latter case our observed changes would not correspond to
changes in actual income shares but merely the ones reported to tax authorities.
13
Clearly the
latter variant is much more problematic for a study of before-tax incomes like ours.
In several of the appendices below, we describe some of the most significant changes in the
Swedish tax laws during the twentieth century in order to detect possible correlations with the
trends in top income shares. While there are specific times when influences of the tax system
on top income shares are apparent, we find no systematic, and especially not one-directional,
relation between the tax system and top income shares.
4 International comparisons
4.1 Comparing top income shares
Comparing our results to the evolution of top income shares in other countries reveals a num-
ber of interesting differences and unexpected similarities between the Swedish experience and
that of Anglo-Saxon countries on the one hand, and “Continental Europe”, on the other.
14
These differences and similarities may not only reflect what has happened to top income
shares, but also point to more fundamental economic variation across countries.
Figure 15 shows the income share for the top five percent of the population, P95-100, for the
Netherlands, the UK (both from Atkinson and Salverda, 2003), the US (from Piketty and
Saez, 2003), Canada (from Saez and Veall, 2003) and France (from Piketty, 2003) together
with our figures for Sweden. The broad patterns seen here go through for the top ten percent
as well as for the top one percent. For Sweden, Canada and the US we distinguish series in-
13
For a discussion on the endogeneity of efforts and incomes with respect to the tax system, see Bergh (2005).
14
These are admittedly broad terms given that we only make the comparison between Sweden and the US, UK,
the Netherlands and France. However the points we will refer to hold true when including Canada, New Zealand
and Australia to the Anglo-Saxon group and Switzerland to “Continental Europe”.
21
cluding and excluding capital gains. In the case of France this distinction is, according to
Piketty (2001, p. 20n), not very important as the capital gains share is very small even for the
top income earners. For the UK and Netherlands there is no separate data including capital
gains.
Figure 15 from excel file Internatcomp2, FigTop5
This figure reveals the most striking aspect about Swedish top incomes namely the fact that
Sweden seems to show an Anglo-Saxon pattern over the past decades when including capital
incomes but a continental European pattern when capital gains are excluded. Even if the
Swedish income share for the top five percent around 1980 starts at a much lower level in
Sweden than the corresponding shares in the US, the UK and in Canada, the pattern 1980-
2003 is unmistakably Anglo-Saxon when including capital gains. However, when excluding
capital gains the Swedish series are as flat as the French and the Dutch ones. This is not the
case for any of the Anglo-Saxon countries where the difference with and without capital gains
are not very dramatic. (This pattern becomes even more pronounced looking at the top one
percent and the top 0.1 percent, see below). Schematically the different experiences can be
grouped as shown in table below.
Table 2: Development of top income shares in the period 1980-2003
Increase Flat
Including
capital gains
US, Canada, UK, Sweden France, Netherlands (?)
Excluding
capital gains
US, Canada, UK (?) France, Netherlands, Sweden
Development of top income shares in the period 1980-2003 for different countries when including and excluding
capital gains.”?” denotes that we do not have the exact information on the treatment of capital gains in the cases
of the UK and the Netherlands.
Apart from the development over the past decades there are a number of other interesting
comparisons to be made. In chronologic order we can first note that the spike during the First
World War is there in the Dutch data as well, suggesting that the First World War economy
increased top income shares (there is also a sharp increase in the top one percent share for the
22
US and France during these years as can be seen below). As pointed out above we do not,
unfortunately, have sufficient data for Sweden to see any detailed effects of the 1920s, nor the
effects of the 1930s depression, which in the US, France and the Netherlands was a time of
increasing top shares in the 1920s and then a decrease in the early 1930s. A first major inter-
esting difference between Sweden and other countries comes during World War II. The period
1939-1945 is one of very sharp decline in top income shares in the US as well as in France.
This does not seem to be the case in Sweden. If anything the war years in Sweden seem to
constitute a slowdown in the decline of top income shares which started in the mid 1930s and
then accelerated again after the war. (Holland and the UK are difficult to judge since there is
not enough data for the war years). After 1945, however, when the fall in the US and France
stops and top shares even recover slightly, they drop sharply in Sweden until the early 1950s.
After the early 1950s there is a slow gradual decrease in top shares in Sweden, but not very
different from other countries. Our series suggest that, to the extent that Swedish top income
shares were different from other countries in the late 1970s, this was a difference established
already established in the beginning of the 1950s.
As noted above, top income shares in the US, Canada and in the UK start increasing around
1980. Looking first at the figures for the top five percent of income earners including capital
gains, there is not much of an upturn in Sweden until about 1985, (perhaps not until 1990).
However, if we look at the top one percent, and the top 0.1 percent of the population, we see
that something seems to be happening, starting at the very top of the distribution, around
1980.
Figure 16 from excel file Internatcomp2, FigTop1
Figure 17 from excel file Internatcomp2, FigTop0.1
4.2 Comparing the composition of top incomes
The composition of Swedish top incomes does not fully match the top income composition of
any of the other Western countries analyzed previously. For example, while French (in Piketty
2001, 2003), U.S (Piketty and Saez 2001, 2003) and Swedish capital income shares all display
marked declines since World War I, they are seemingly more related to the wars (and war-
fare?) in the two former cases whereas the Swedish decline, especially in the late 1940’s and
early 1950’s, seems more policy-related. Also while U.S. and French top capital income
23
shares continue decreasing throughout the postwar period, they start increasing again in Swe-
den around 1980 and by the mid-1990’s they reached pre-WWI levels.
As for the distribution of sources of income in the different top fractiles, Sweden is rather
similar to most other countries. For example, the share of wages and salaries of total income is
strictly decreasing in income in Canada (Saez and Veall, 2003, 2005), France (from Piketty,
2003), and the United States. Similarly, the share of capital income is increasing in the level
of total income. Business or entrepreneurial (or self-employment) income is less homogenous
across countries, and also less monotonically related to income. In Canada and France it in-
creases up to a certain point (about the top 1 percent–top 0.1 percent) and then starts to de-
cline for the highest fractiles. In the U.S. and Sweden, however, the relation is varying over
time between being either quite flat or weakly increasing in income.
Linking top income composition to the overall development of top income shares of the total
income in the economy, i.e., changes in income inequality over the twentieth century, has
received considerable attention in the past studies of top incomes. While authors explain most
of the increased inequality in Anglo-Saxon countries (Great Britain, USA, Canada, Australia
and New Zealand) with wage increases for the “working rich” (for a discussion see, e.g., At-
kinson and Leigh, 2004) the secular decline many continental European countries (France,
Netherlands, Germany) has been explained by both war-related shocks to capital income and
relatively compressed distribution of wage increases (see, e.g., Piketty, 2003; Atkinson and
Salverda, 2005 and Dell, 2005).
As noted in the previous section, Sweden is similar to continental European countries when
capital gains are not counted as part of income. This means that despite observing an in-
creased share of capital income in Swedish top incomes since 1980, this was of sufficiently
limited magnitude not to affect overall income shares. When including capital gains in total
incomes, however, Sweden jumps over to the Anglo-Saxon group of countries experiencing
sharp increases in income inequality since 1980 which for Sweden is predominantly driven by
a boost in capital gains. As was argued in section 3.2, this Swedish “anglo-saxon pattern” may
be related to the formidable upswing on Swedish financial and housing markets during the
two last decades of the century. Between 1980 and 2000, Swedish stock prices increased by a
24
factor of fifty, compared to about a factor ten for the U.S. and U.K.
15
Furthermore, since
French stock prices also increased substantially (though not as much as in Sweden - approxi-
mately a factor of fifteen) it seems as if Swedish top income earners also were more market-
oriented than the French ones who only experienced marginal increases in capital gains over
this period (Piketty, 2001, p. 20n).
5 Concluding remarks
So what conclusions about the development of top incomes in Sweden over the twentieth cen-
tury can be drawn at this stage? Three broad findings strike us as especially important. First,
Swedish top income shares have decreased relatively steadily between 1903 and around 1980.
Even though the most rapid decline in Sweden was between 1930 and 1950, as in the US and
in France, it does not seem to be as clearly related to the Second World War as in many other
countries. If anything, the war years seem to be a period of slowdown in the fall in top income
shares, which in Sweden was most pronounced just after 1945. This suggests that policy was
driving this trend rather than exogenous shocks. Second, Swedish top income shares have –
when including all sources of income – increased since the beginning of the 1980s. The pat-
tern of development in Sweden is in this respect close to that in Anglo-Saxon countries, but
different from continental Europe. The main reason for this increase in top income shares in
Sweden is increases in capital gains rather than wage earnings as in the US and in the UK.
Third, the composition of income in the top of the distribution has changed dramatically in
Sweden over the past decades. The share of wages and business income has decreased while
capital income and especially capital gains have become steadily more important.
References
Atkinson, Anthony B. (2003) “Measuring Top Incomes: Methodological Issues”, Working
Paper, Nuffield College: Oxford.
Atkinson, Anthony B. (2004), “Top Incomes in the UK over the twentieth Century”, Journal
of the Royal Statistical Society A 168(2), 325–343.
Atkinson, Anthony B., Lee Rainwater and Timothy M. Smeeding (1995), “Income Distribu-
tion in OECD Countries: Evidence from the Luxemburg Income Study (LIS)”, Social Pol-
icy Studies 18. Paris: OECD.
Atkinson, Anthony B. and Andrew Leigh (2004), “Understanding the Distribution of Top
Incomes in Anglo-Saxon Countries over the Twentieth Century”, Working Paper, Nuffield
College, Oxford.
15
Data on stock prices were take from IMF’s database International Financial Statistics.
25
Atkinson, Anthony B. and Wiemer Salverda (2003), “Top Incomes in the Netherlands and the
United Kingdom over the Twentieth Century”, Working Paper, Nuffield College.
Atkinson, Anthony B. and Salverda, Wiemer (2005), “Top Incomes in the Netherlands and
the United Kingdom over the Twentieth Century”, Journal of the European Economic As-
sociation 3(4): 883–913.
Banerjee, Abhijit and Thomas Piketty (2005), “Top Indian Incomes, 1922–2000”, World Bank
Economic Review 19(1): 1–20.
Bentzel, Ragnar (1953), Inkomstfördelningen i Sverige. Uppsala: Almqvist & Wiksell.
Bergh, Andreas (2005), “On the counterfactual problem of welfare state research: How can
we measure redistribution?”, European Sociological Review 21(4): .
Björklund, Anders and Mårten Palme (2000), “The Evolution of Income Inequality during the
Rise of the Swedish Welfare State 1951 to 1973”, Nordic Journal of Political Economy
26(2), 115–128.
Champernowne, D. G. (1953), “A Model of Income Distribution.” Economic Journal
63(250): 318–351.
Dell, Fabien (2003), “Top Incomes in Germany throughout the Twentieth Century”, forth-
coming in A.B. Atkinson and T. Piketty (eds.) Oxford: Oxford University Press.
Dell, Fabien (2005), “Top Incomes in Germany and Switzerland over the Twentieth Century”,
Journal of the European Economic Association 3(2–3): 412–421.
Edvinsson, Rodney (2005), Growth, Accumulation, Crisis. With New Macroeconomic Data
for Sweden, 1800–2000. Stockholm: Almqvist & Wiksell International.
Feenberg, Daniel and James Poterba (1993), “Income Inequality and the Incomes of Very
High Income Taxpayers: Evidence from Tax Returns”, in J. Poterba (ed.) Tax Policy and
the Economy, Cambridge: MIT Press.
Feenberg, Daniel R. and James M. Poterba (2000). “The Income and Tax Share of Very High-
Income Households, 1960–1995.” American Economic Review Paper & Proceedings
90(2), 264–270.
Flodström, Isidor (1906) Redogörelse för statistisk utredning i och för utarbetande af förslag
till förändrade bestämmelser dels angående bevillning af fast egendom dels ock om in-
komstskatt, Stockholm.
Flodström, Isidor (1909), Taxeringen till inkomstskatt år 1907 samt taxeringen till bevillning
s. å. för inkomst under 1,000 kronor af kapital och arbete, Ministry of Finance, Stockholm.
Flodström, Isidor (1914), Taxeringen till inkomst- och förmögenhetsskatt år 1912, Ministry of
Finance, Stockholm.
Flodström, Isidor (1915), Taxeringen till inkomst- och förmögenhetsskatt år 1913 jämte upp-
gifter om däri ingående förmögenhetsdelar, Ministry of Finance, Stockholm.
Gottschalk, Peter and Timothy M. Smeeding (1997), “Empirical Evidence on Income Inequal-
ity in Industrialized Countries” in A.B. Atkinson and F. Bourgignon (eds.), The Handbook
of Income Distribution.” London: North Holland Press.
Gustafsson, Björn and Mats Johansson (2003), “Steps toward equality: How and why income
inequality in urban Sweden changed during the period 1925–1958” European Review of
Economic History 7(2), 191–211.
26
Hagstroem, K. G. (1944), “Inkomstutjämningen i Sverige”, Skandinaviska Bankens Kvartals-
tidsskrift, April, pp. 35–40.
Hagstroem, K. G. (1949), “Inkomstinflation och inkomstnivellering i Sverige”, Skandinaviska
Bankens Kvartalstidsskrift, April, pp. 39–41.
Hartog, Joost and J. G. Veenbergen (1978), “Long-Run Changes in Personal Income Distribu-
tion”, De Economist 126: 521–549.
Henrekson, Magnus (1996) Företagandets villkor. Spelregler för sysselsättning och tillväxt.
Stockholm: SNS.
Jakobsson, Ulf and Göran Normann (1974), Inkomstbeskattningen i den ekonomiska politiken.
En kvantitativ analys av systemet för personlig inkomstbeskattning 1952–1971, Stockholm
och Uppsala: Industriens Utredningsinstitut och Almqvist & Wiksell.
Klein, Paul (1996), “Swedish Income Taxes 1951–90”, mimeo, Stockholm University.
Krantz, Olle (1987), Historiska Nationalräkenskaper för Sverige. 6, Husligt arbete 1800-
1980. Ekonomisk-historiska föreningen; Lund.
Kuznets, Simon (1955), “Economic Growth and Economic Inequality”, American Economic
Review 45(1), 1–28.
Lindstrand, M. (1949). Inkomstfördelningen före och efter kriget. Meddelande från Konjunk-
turinstitutet, Ser B:10, 89–100. Stockholm.
Lydall, Harold (1968), The Structure of Earnings, Oxford: Clarendon Press.
Löwnertz, Susanne (1983), De svenska skatternas historia [en artikelserie tidigare publicerad
i RSV Info åren 1982-83], Stockholm: Riksskatteverket och Liber.
Magnusson, Katarina (2004), “Utvecklingen av de högsta inkomsterna i Sverige, 1985–2002”,
TCO granskar No. 13/04.
Morrison, Christian (2000), “Historical Perspectives on Income Distribution: The Case of
Europe”, in Handbook of Income Distribution, Vol. 1, Elsevier.
Piketty, Thomas (2001), “Income Inequality in France, 1900–98”, CEPR Discussion Paper
No. 2876.
Piketty, Thomas (2003), “Income Inequality in France, 1900–1998”, Journal of Political
Economy 111(5): 1004–1042.
Piketty, Thomas and Saez, Emmanuel (2003) “Income Inequality in the United States, 1913–
1998”, Quarterly Journal of Economics 118(1): 1–39.
Quensel, Carl-Erik (1944) Inkomstfördelning och skattetryck, Lund: Sveriges industriförbund.
Saez, Emmanuel and Michael L. Veall (2003), “The Evolution of High Incomes in Canada,
1920–2000”, NBER WP #9607.
Saez, Emmanuel and Michael L. Veall (2005), “The Evolution of High Incomes in Canada,
1920–2000”, American Economic Review 95(3): 831–849.
Schön, Lennart (2000), En modern svensk ekonomisk historia. Tillväxt och omvandling under
två sekel. Stockholm: SNS.
SOU 1936:18 (1936), Undersökningar rörande det samlade skattetrycket i Sverige och utlan-
det, Ministry of Finance, Stockholm.
27
SOU 1949:9 (1949), Betänkande angåendebeskattning av realisationsvinster, m.m. samt ac-
kumulerade inkomster, Stockholm: Isaac Marcus.
SOU 1951:51 (1951), Betänkande angående den statliga direkta beskattningen. Stockholm:
K. L. Beckmans.
SOU 1975:53 (1975), Beskattning av realisationsvinster. Betänkande avgivet av Realisations-
skattekommittén, Stockholm: Liber.
SOU 1986:37 (1986), Reavinst, aktier och obligationer. Slutbetänkande av kapitalvinstkom-
mittén. Stockholm: Gotab.
Spånt, Roland (1976), Den svenska inkomstfördelningens utveckling, Ph.D. dissertation in
economics No. 4, University of Uppsala, Uppsala: Almqvist & Wiksell International
Spånt, Roland (1979), Den svenska inkomstfördelningens utveckling 1920–1976, background
material to Långtidsutredningen 1979:4, Ds E 1979:4.
Spånt, Roland and Pebbe Selander (1969), “Inkomstfördelningens utveckling i Sverige 1951–
66”, mimeo, Uppsala University.
Statistics Sweden, Statistical Yearbook, various years.
Statistics Sweden (1923) Taxeringen till inkomst- och förmögenhetsskatt år 1920. Stockholm:
Ivar Häggströms AB.
Statistics Sweden (1927) Folkräkningen den 31 december 1920. V. Yrkesräkningen, 2: yrke,
inkomst och förmögenhet kombinerade inbördes samt med kön, civilstånd och ålder,
Stockholm: P. A. Norstedt & Söner.
Statistics Sweden (1937) Folkräkningen den 31 december 1930. VII. Folkmängden efter yrke,
inkomst och förmögenhet: 2 avd., Stockholm: P. A. Norstedt & Söner.
Statistics Sweden (1938) Folkräkningen den 31 december 1930. VIII. Folkmängden efter yrke,
inkomst och förmögenhet: 3 avd., Stockholm: P. A. Norstedt & Söner.
Statistics Sweden (1940) Särskilda folkräkningen 1935/36. VIII. Partiella folkräkningen i
mars 1936: Yrke. Yrkesväxling. Skol- och yrkesutbildning. Inkomst och förmögenhet.
Stockholm: P. A. Norstedt & Söner.
Statistics Sweden (1951) Folkräkningen den 31 december 1945. VIII:1. Partiella undersök-
ningar (bottensamplingen). Behandlar delar av statistiken över inkomst. Sveriges Officiel-
la Statistik: Folkmängden och dess förändringar. Stockholm: K.L. Beckmans Boktryckeri.
Statistics Sweden (1956) Folkräkningen den 31 december 1950. VII. Urvalsundersökningar.
Statistiken över inkomst. Sveriges Officiella Statistik: Folkmängden och dess förändringar.
Stockholm: Statistiska Centalbyrån.
Statistics Sweden (2001), Reavinster och reaförluster 1999, Rapport 2001. Statistics Sweden:
Programmet för ekonomisk välfärdsstatistik.
Statistics Sweden (2002), Rekordåret 2000. Reavinster och reaförluster. Förmögenhet, Rap-
port 2002. Statistics Sweden: Programmet för ekonomisk välfärdsstatistik.
Statistics Sweden (2003), Minskade vinster 2001. Kapitalvinster och kapitalförluster. Förmö-
genhet, Rapport 2003. Statistics Sweden: Programmet för ekonomisk välfärdsstatistik.
Swedish National Tax Board (2000), Skattestatistisk årsbok 2000, Stockholm: Fritze.
Swedish National Tax Board (2004), Skattestatistisk årsbok 2004, Stockholm: Fritze.
28
Söderberg, Johan (1991). ”Wage differentials in Sweden, 1725-1950, in Y. S. Brenner, H.
Kaelble and M. Thomas Income distribution in historical perspective, Cambridge, Cam-
bridge University Press: 76-95.
Sörensen, R. S. (1993), “Changes in the Personal Income Distribution 1870-1986” in: Atkin-
son, A. B. (ed.) The Economic Development of Denmark and Norway since 1870, Chelten-
ham: Edward Elgar.
Tengblad, Åke (1994) “Beräkning av svart ekonomi och skatteundandragandet i Sverige
1980–1991” in Malmer, Håkan, Annika Persson and Åke Tengblad, Århundradets skatte-
reform. Del II. Stockholm: Fritze.
Östlind, Anders (1945), Svensk samhällsekonomi 1914–1922: Med särskild hänsyn till indu-
stri, banker och penningväsen, Stockholm: Svenska bankföreningen.
29
Appendix A: Income inequality data
A1: List of sources
A1.1: Sources for the income data
The income data used in this study come from official sources published by tax and statistics
authorities as well as a few specific investigations led by the Ministry of Finance. Before
1943, there were no regularly published annual statistics of the Swedish income distribution.
Therefore, we are confined to investigations of the Minstry of Finance and and Census mate-
rial. In particular, for 1903–1912, we use tabulated income distributions made by Isidor Flod-
ström and the Minsitry of Finance. For years 1916, 1919 and 1920, we use income tables pub-
lished in the Statistical Yearbook by Statistics Sweden. For 1930, 1935 and 1940, we use
numbers from Bentzel (1952) who in turn uses Census material (i.e., individual incomes)
which he has adjusted to make comparable with later tax returns-based material in which tax
units are analyzed. For 1934 and 1937, we use special investigations by the Ministry of Fi-
nance and adjusted by Quensel (1944). From 1943 onwards, we take the main tables on in-
come distribution for the whole tax population published annually by Statistics Sweden in
Skattetaxeringarna samt inkomstfördelningen inom yrkesgrupper (later renamed).
16
All in-
come sources for all years are listed in Table A1.
Specific data on the sources of income, e.g., labor or capital income, are not as good as for
total income. Prior to 1967, there exist no systematic annual publications on the composition
of incomes divided in income levels. On the aggregate level, there are annual figures in the
Statistical Yearbook starting in 1921. In the Censuses from 1945 and 1950 (Statistics Sweden,
1951, 1956), however, we have nevertheless located two point observations covering all
sources of income. As for capital income, we have been able to construct estimates all the way
back to 1912 using the fact that a share of personal wealth was included in the taxable income
and that Censuses from that time tabulated amounts of wealth for various levels of total in-
come. We simply assume a fixed rate of return flowing from this wealth to the income earners
and take this as capital income (not including capital gains). Similar estimates were made by
Flodström (1915, p. 46–47) and Statistics Sweden in the Statistical Yearbook of 1929. We use
the same rates of return used to compute the capital income as Flodström and Statistics Swe-
den, and as we argue in our main text in this paper these rates are in line with market rates at
each point in time. Unfortunately, after 1991 the compositional data becomes less detailed in
the income statistics, mainly because of changes in the publication strategy of Statistics Swe-
den (see further in section A2 about income sources).
Table A1: Sources on income distribution used in the study, 1903–2003.
Table A2: Example of tabulated income data from 1945.
A1.2: Sources for other data
Beside the income data we also need population data to calculate our reference totals for
population and incomes respectively. Here the main sources are Statistics Sweden’s popula-
16
In Sweden, the taxation year, i.e., the year when tax returns are filed at the tax authorities, is always the year
after the income year, i.e., the year when the income was made. Hence, 1943 was the first income year for which
the new income statistics was used but 1944 was the first taxation year.
30
tion statistics, Programmet för befolkningsstatistik, and the Statistical Yearbook of Sweden,
Statistisk Årsbok, for various years starting 1914. Our main source for historical national ac-
counts data is Edvinsson (2005).
A2: Details of Swedish income data
A2.1: Definitions of income concept and income sources
A2.1.1: Concept of income
The concept of income used in our main income series is total income, as defined in the
Swedish income statistics throughout the period as listed in Table 1. Total income is a gross
income concept, recorded before deductions and taxes. In terms of specific sources of income
(discussed in detail below), total income is the sum of wages, capital and business income and
capital gains, although we present calculations of top income share both with and without
capital gains included in the total income to make the series internationally comparable.
Above from these main features, the Swedish tax-based assessed total income has also under-
gone some significant, but yet noteworthy, adjustments over the twentieth century. For exam-
ple, during 1910–1942 the assessed income from the standard sources were also including a
fixed share of the personal net wealth. This “wealth share” entered the taxable income by 1/60
in 1910–1937 and 1/100 in 1938–1948. In the income statistics starting in 1943, however, this
share was never counted and for earlier years we have subtracted it from the rest of the in-
come for all income levels based on estimation from Census material on income and wealth.
Since the introduction of the annual income statistics of Statistics Sweden in 1943, the main
concept of income has undergone some adjustments. During the period 1943–1970, the main
concept used in all tables was total net income (in Swedish, sammanräknad nettoinkomst),
defined as the total income less deductions of deficit in any income source (underskott i
förvärvskälla), mainly interest payments. After a smaller reform of the tax system in 1971, a
new main income concept was introduced by Statistics Sweden: total income (sammanräknad
inkomst). This concept equaled the former except that deductions of deficits in sources were
no longer subtracted. In terms of the time consistency of our series, however, this adjustment
had no significant impact since the deductions were in general quite small. A somewhat more
important change occurred in 1974 when the Swedish government ruled that social benefits,
e.g., unemployment insurance, social security transfers and state pensions, were made taxable.
Hence, the incomes filed on the tax return changed in structure from one year to another, but
this also has little effect on our study of top incomes since the social benefits had a small
share of total income in the higher income classes. Moreover, our reference total income is
defined so as to include social benefits ever since the 1940’s to ensure comparability over
time.
Lastly, in 1990–1991 there was the most recent major Swedish tax reform, and along this
event Statistics Sweden changed its main concept of income in the income statistics to total
earned income (sammanräknad förvärvsinkomst), which is the sum of labor and business in-
come, hence excluding capital income and capital gains. Fortunately, however, Statistics
Sweden continued to publish some summary tables in which they use total income (now
called sammanräknad förvärvs- och kapitalinkomst) as concept of income and hence we have
been able to use the same concept since 1971.
31
A2.1.2: Definitions of sources of income
As already mentioned in previous appendices, the Swedish tax laws and income statistics de-
fine different sources of income which are supposed to be specified on the tax returns. During
the twentieth century, these definitions have been remarkably stable although there have been
one major change, in 1991. In the official income statistics, however, compositional data have
not been published for different classes of incomes until 1967, with exception from two early
Census observations in 1945 (Statistics Sweden, 1951) and 1950 (Statistics Sweden, 1956).
The definition of the sources of income before 1991 rested on a division of six different
kinds:
17
labor income (inkomst av tjänst), mainly wages and salaries; capital income (inkomst
av kapital), mainly interest earnings and dividends; entrepreneurial income (inkomst av
rörelse), mainly firm profits and royalties; farm income (inkomst av jordbruksfastighet),
mainly of sales of agricultural and forestry products and leases; real estate income (inkomst av
annan fastighet), mainly rents and in-kind payments and capital gains (inkomst av tillfällig
förvärvsverksamhet) from sales of real estate and securities.
18
After the tax reform in 1991, the number of official income sources was reduced to three: la-
bor (inkomst av tjänst), business (inkomst av näringsverksamhet) and capital (inkomst av kapi-
tal (överskott)). Compared with the earlier period, labor income was defined in basically the
same way. Business income, however, included not only the previous entrepreneurial income,
but also all of farm incomes and a small part of real estate income emanating from rental
apartments. In the new concept of capital income, the previous capital income was included
but also most of former real estate income coming from private rental and, notably, all forms
of capital gains.
To ensure that our compositional analysis is consistent for the whole period, we have defined
four main income sources that are used in our analyses. These primarily follow the definitions
of the post-1991 for computational reasons, and are constructed as follows:
• Labor: Includes wages and salaries and is basically defined in the same way both be-
fore and after 1991.
• Capital: Includes interest earnings, dividends and real estate income. In the period
before 1991, we add “capital income” (interests and dividends) and “real estate in-
come” together.
19
After 1991, estimate capital income from the “new capital in-
come”, which includes both the old concept and capital gains. Hence, we break out
interest earnings and dividends (called inkomst av ränta in the income statistics),
private rental income (inkomst av uthyrning av privatbostad) and special rental in-
come (inkomst av positiv räntefördelning) as is described in detail in section A2.1.3.
17
In the late 1960’s, there was also a specific entry for income from partnerships (inkomst av delägarskap i
vanligt handelsbolag etc), but this was included in entrepreneurial income from the 1970’s onwards and we do
this also for these years when it was reported separately.
18
Detailed descriptions of the income sources are found in, e.g., Statistics Sweden (1945, pp. 50–67) and Statis-
tics Sweden (1975), Inkomst och förmögenhet 1973, pp. 25–26.
19
Formally, one part of the real estate income was also included in business income after 1991, namely income
from public rental buildings. However, this only concerned so-called “physical persons” (private individuals) and
not “judicial persons” (public and private companies) which instead had to report all of their income (including
that from real estate) as entrepreneurial income and which was the largest part of the two incomes. Leif Johans-
son at Statistics Sweden (from a discussion on June 15, 2005) also would believe that the absolute majority of
the real estate income before 1991 should refer to what would after 1991 have been included in capital income.
For these reasons, we place all of real estate income in the capital income in our long-run series.
32
• Business: Includes mainly income from privately held firms. Before 1991, we add
together “entrepreneurial income” and “farm income”. After 1991, we use “business
income”.
• Capital gains: Includes net gains from sales of real estate and other assets. Before
1991, capital gains were reported as a separate source of income. After 1991, we
have to estimate it from the new concept income from capital as described in section
A2.1.3.
A2.1.3: Estimating the income composition using post-1990 data
Unfortunately, the compositional data after 1990 are much less detailed than and also not fully
comparable with earlier years. This is both because of the tax reform of 1990–1991 and the
redefinition of the sources of income. A more tangible cause, however, is the fact that Statis-
tics Sweden started listing all compositional data according to different classes of earned in-
come, instead of total income which was the case for the period up to 1990. Remarkably, this
income numeraire is also used when listing various kinds of income of capital despite the fact
that capital income was not even included in earned income.
20
Hence, there is a large portion
of income earners with positive capital income but zero earned income (e.g., children with
significant wealth). To get comparable series with earlier years we therefore have to two ad-
justments: First, we must map these compositional tables into classes of total income. Second,
we must separate out the capital gains part of the new income of capital-concept. This cannot
be done with perfect precision since we do not have the primary individual taxation material.
To make sure our series are roughly consistent approximations, we carry through a number of
checks against data from the years just before 1991.
We deal with these problems as follows. We start by transforming the three new income
sources (labor, business and capital, including capital gains), listed in classes of earned in-
come, into classes of total income. This is done by summing wages and business income to-
gether for each class of earned income. Recognizing that earned and total incomes are classi-
fied in the exact same nominal amounts, we then subtract the earned income classwise from
the total income. The resulting difference should then be roughly the total capital income in
each class listed in classes of total income. A slight correction is made to ensure that the sum
of these differences is the same as the reported aggregate capital income.
21
This gives us in-
come sums for all three sources in classes of total income, adding up to their correct income
totals.
The next step is to separate out capital gains from the new capital income-concept and achieve
them listed in total income classses. Unfortunately, it is not straightforward to do this since
capital gains and capital income are listed in gross terms (incomes and losses) whereas the
new capital income-concept (inkomst av kapital (överskott)) shows the sum of all individual
surpluses, i.e., what is left after incomes and losses are netted out against each other whenever
possible.
22
Since 1991, the tax laws do not differentiate between different kinds of capital
losses and gains why there is no way to tell from the publications which kind has been used
for what. Instead, we calculate the share of capital gains from their share of the total gross
20
Table 6.1 and 6.2. in SM, called “Inkomsttagarna med inkomst av capital fördelade på inkomster och avdrag
efter kön, ålder, sammanräknad förvärvsinkomst”.
21
We use the reported capital income (inkomst av capital (överskott)) as true total. The correction is very small,
never larger than roughly one percentage point of the total and often practically zero.
22
The existence of a surplus (total capital income) is due to the fact that all individuals with capital gains did not
have corresponding capital losses which they could use to net out the taxable gains. The gross income is bigger
than total capital income since it contains incomes which were later netted out by losses.
33
capital income (summa kapitalinkomster) and assume the net gains to be the same of corre-
sponding net capital income. We then multiply these shares with the capital income (including
capital gains) for each income class of total income using the approximated series derived in
step one.
The third, and final, step is to compute the capital income and capital gains. This is simply
done by multiplying the share of capital gains of total capital income (and one minus that
share for capital income) with the reported total capital income for each income class.
An obvious problem with this approach is the unobserved individual variation that potentially
could bunch together income earners with different total income in the same total income
class. For example, our estimate gives the same capital income to all persons with some
earned income, say almost zero, despite the fact that they can be either practically poor and
unemployed or wealthy rentiers. These two groups would end up in different classes of total
income, but this is nothing our method can take into account. For our approach to be roughly
adequate, we need individuals with similar income composition to end up in approximately
their correct income class. To “test” the plausibility of the approach, we compare the shares of
capital income across income classes between two years before the tax reform, when this
problem did not exist, and the first two years after the reform. We do not wish to use more
than two years since the income composition varied quite drastically during these years. In-
specting the two-year average ratios of earned to total income 1989-90 and 1991-92, they
seem to match remarkably well, especially from the median income and above. Using a Kol-
mogorov-Smirnov test to test formally for similarity between these two distributions of capital
income over total income, we get a D-statistic of 0.3012 and a p-value of 0.265, i.e., we can-
not reject the null hypothesis that they are drawn from the same distribution.
A2.1.4: Estimating the top shares of capital income, 1912–2003
Thanks to early wealth data in the tax statistics for income earners in different classes of total
income, we are able to construct shares of capital income of total income as far back as 1912
and for some more years until the postwar period when we use the compositional sources de-
scribed previously.
Specifically, the shares before 1945 are computed by assuming that capital income is a fixed
rate of return flowing from the individuals’ net wealth. Information about net wealth in differ-
ent classes of income is available from the tax-based income statistics due to the fact that 1/60
of that wealth was to be added as taxable income until 1938 when the share was reduced to
1/100 and 1943 when it was removed altogether (recall Table 1). The approach was previ-
ously used by, e.g., Flodström (1915, pp. 46–47) and Statistics Sweden (1927). Capital in-
come is then computed as the annual rate of return from this wealth. We assume that the yield
is flat and the same for all income earners, being 5 percent in the years 1912 and 1919, 5.5
percent in 1920, 4.5 percent in 1930 and 3 percent in 1935. These are the same rates that Flod-
ström and Statistics Sweden use (except for 1920 when they use 5 percent).
23
Unlike them,
however, we can also motivate our choice of these rates by referring to three other reference
interest rates from the same particular years. Specifically, the yearly averages of the minimum
lending rate (diskontot) set by the Swedish central bank, the average deposit rate at Swedish
savings banks and the effective Swedish Government bond yield were in 1912: 4.81, 4.35 and
4.80; in 1916: 5.23, 4.76 and 5.09; in 1919: 6.38, 5.08 and 5.71; in 1920: 6.92, 5.16 and 7.00;
23
Unfortunately, no income data were collected in the Census of 1940, why we have no information about
wealth shares in different classes of income.
34
in 1930: 3.71, 5.22 and 4.18; and in 1935: 2.50, 3.59 (in 1933) and 3.30 (Svensk Sparbank-
stidskrift 1934, p. 825).
Finally, the share of capital income of total income across the various top fractiles is com-
puted using Pareto interpolation in the same way as in the rest of the compositional analysis.
A2.2: Concept of income earners
The concept of income earners is problematic in most countries’ tax-based income statistics,
often associated with redefinitions and changes in the underlying demographic structure of the
country. Sweden is no exception although the situation seems not more complicated than
elsewhere. First of all, income earners in the Swedish income statistics refer to physical per-
sons who lived in Sweden during the income year (January 1–December 31 the year before
hading in the tax return) and who also filed a personal tax return.
24
We use the definition of income earners according to published income statistics, which is
mostly, but not always, identical with what is written in the tax legislation. Until 1971 Sweden
had a family-based tax system, which means that families were the tax units and all incomes
of family members added together and filed under a single tax return. Under this system, mar-
ried couples hence counted as one income earner. While before 1954 the wife’s income was
automatically assessed as a part of her husband’s income, however, between 1954 and 1965
spouses filed separate tax returns after which their incomes were lumped together and taxed as
one tax unit according to a specific rate of “joint taxation” (sambeskattning). In the period of
1966 to 1970, the system was further adjusted so that married couples could choose whether
to have their income being taxed separately or as one couple according to the specific scale.
Finally, in 1971 the Swedish tax system changed to being individual-based altogether. Mar-
ried couples were from that on hence treated as two income earners.
In the period 1943–1950 the income statistics followed the tax system by being family-based,
using the total number of filed tax returns as primary material. Due to processing constraints,
however, only a few variables could be collected for each tax unit and therefore it was decided
to switch to a sample-based system that allowed more background information to be collected
and analyzed. During 1951–1966, the income statistics was hence based on a representative
10 percent sample of the tax population. In practice this meant that the income statistics be-
came individual-based while having a family-based tax system since all persons with positive
income had to file an individual tax return regardless of whether they were eventually taxed
jointly with their spouses or parents.
25
The 10-percent sample was drawn from the population
of all adults aged 16 years or above and born on either the 5th, 15th or 25th in each month.
26
To avoid undersampling of income earners in the highest income classes, a full sampling of
these groups was made.
27
The sample-based income statistics lasted until 1967 when Statistics
Sweden returned to basing the income statistics on the complete tax population with the help
of new data processing techniques.
24
Formally, unfinished death estates and family foundations are also counted as income earners, but they only
represent about 1 percent of the total number of income earners.
25
The switch to using a population sample followed the instructions of a governmental statute (kungörelse den
21 december 1951, No. 832).
26
Having in fact 365.25 days calendar year, the chosen sample was actually smaller than 10 percent of the popu-
lation and instead of multiplying each income earner with 10 (for those jointly assessed 5) it should have been
10.146 (and 5.340). As noted by Statistics Sweden in Inkomst och förmögenhet 1968, p. 26 (see Appendix
sources), this could have some minor effects on the comparability of the data before and after 1967.
27
The definition of high income was SEK 30,000 or above during 1951–59 and with income above and SEK
50,000 or above in 1960–66.
35
Above from these major changes in the income earner definitions, there have been several
smaller adjustments and related changes that have affected the income earner concept in par-
ticular. For example, in income years 1972 and 1973 all retirees receiving public pension only
(folkpensionärer) were granted extra large tax deductions to avoid paying taxes.
28
Further-
more, from 1978 both employers and employees were required to report all incomes paid and
received, which in itself increased the tax liable population by a couple of hundred thousand
income earners who were most likely previously avoiding taxes altogether.
A3: The Swedish income tax system, 1903–2003
A3.1: General overview
The modern Swedish income tax system was introduced in 1902, with the first income year
being 1903. During the twentieth century there have been some major reforms and numerous
minor adjustments of the income tax system. This section lists some of the most important
ones.
The Partial Income Tax Act (lagen om partiell inkomstskatt) of 1902 was the first comprehen-
sive progressive tax on income in Sweden.
29
The word partial referred to the fact that Sweden
already had an “income tax” since 1812, called bevillning. This was a proportional tax (gener-
ally 5 percent of all incomes) and fulfilled other functions than merely being a source of reve-
nue to the state. For example, the income registers based on the bevillning were used to decide
income weights for the adult men who were allowed to vote for government in income-
weighted electoral system. The new tax in 1903, did not only introduce progressivity but also
a requirement that everyone with an annual income of SEK 1,000 or above had to file a per-
sonal tax return (deklaration) with detailed description of amount and type of incomes. The
progressivity of the tax was though fairly moderate, with marginal tax rates from 0.2 percent
to 5 percent. In the first tax year 191,515 families and individuals filed taxes, representing
about seven percent of the entire tax population (see Table A1). Soon after the introduction of
the partial income tax, inconsistencies arouse in the tax system by having two parallel income
taxes. Decision-makers therefore decided to keep the new income tax (which generated the
most revenues) while abrogating parts of the bevillning.
30
In 1910, a new state income tax came, the Income and Wealth Tax Act (lagen om inkomst-
och förmögenhetsskatt).
31
This law increased progressivity, with marginal tax rates between
0.4 and 6 percent. Furthermore, the concept of taxable income was expanded, adding to the
taxable income (from labor, business, farming, capital, and capital gains) also a share (one
sixtieth) of the tax filer’s net personal wealth. This law was kept intact until 1947 with the
exception that the wealth share was lowered to 1/100 in 1938 when another wealth tax was
introduced.
28
See, e.g., Statistics Sweden, Inkomst och förmögenhet 1973, p. 15.
29
However, almost a hundred years earlier, in 1810–1811, an early progressive income tax was introduced but
removed after only two years of practice (Löwnertz, 1983).
30
Bevillning was also launched on people’s estate. A brief description of these matters (in Swedish) can be
found in Löwnertz (1983).
31
Law (förordning) of October 28, 1910.
36
In 1928, the Municipal Tax Act of 1928 (kommunalskattelagen) extended the taxation rights
of the regions and in fact shifted the major part of the overall income tax burden from the state
level to the municipalities.
32
The next reform of the state income taxes came in 1947, with the State Income Tax Act (lagen
om statlig inkomsskatt).
33
This reform separated the taxation of income and wealth by the
state and removed the addition of a wealth share to the taxable income. Other changes in the
reform was that it introduced tax collection at source which made employers responsible to
deduct taxes before paying wages and salaries.
The most recent change of the Swedish tax system was the tax reform of 1990–1991, which
changed the structure of the entire tax system. In particular, tax bases were broadened to in-
crease the uniformity of treatment of different sources of income. At the same time, tax rates
were lowered to prevent a general increase in the overall taxation.
A3.2: Marginal and average income tax rates in Sweden, 1903–2003
Marginal taxes were gradually increased between 1910 and 1940, from levels of 0.4–6 in
1910, to 3–15 percent in 1919 and 2–31 percent in 1938.
34
With the new law of 1947, mar-
ginal taxes were further raised to between 10 and 70 percent. This level was kept until 1971
when the highest marginal taxes were lowered to 65 percent.
35
In the 1980’s, marginal tax
rates reached their highest levels of the entire century, peaking on 85 percent during several
years [More to come...].
A3.3: Changes in capital gains taxation and its effect on our results
36
A3.3.1: Rules for capital gains taxation, 1903–2003
Capital gains have been taxed as income in Sweden since the 19th century, although the forms
and rates of taxation in different periods have varied substantially. Only to define what capital
gains really are has been a delicate matter for tax authorities and legislators, e.g., concerning
the differences between capital gains and various kinds of business or capital income. During
our period of analysis, one can discern two main periods of capital gains taxation: 1903–
1990, during which capital gains were added to and taxed together with labor income, and
1991– , when capital gains were taxed separately at a flat rate different from other income tax
rates. Within these two periods, however, the taxation of capital gains have undergone several
smaller adjustments that are described in the following.
Before 1910, Swedish tax laws did not explicitly mention capital gains but only that the “in-
come from non-professional sales” should be added to labor income and tax accordingly.
Capital losses were generally deductable from labor income in the same way. From 1910, new
laws defined capital gains (inkomst av tillfällig förvärvsverksamhet) as the gains from non-
professional sales of property (agricultural or urban real estate) or other assets (e.g., securities)
acquired through purchase or barter and owned for no more than 10 years (for property) or 5
years (for other assets). This definition implied that gains from sales of previous gifts or be-
32
SFS 1928:370.
33
SFS 1947:576.
34
SOU 1964:25 (1964), p. 251.
35
Jakobsson and Normann (1974), Table 59, pp. 191–194.
36
This section relies on discussions in SOU 1949:9 (1949), pp. 20–52; SOU 1975:54 (1975), pp. 47–70; SOU
1986:37 (1986), pp. 53–66, Statistics Sweden, Inkomst, skatter och bidrag 1997, If 20 SM 9901, pp. 10–11, and
Swedish National Tax Board (2000), pp. 82–83.
37
quests were not taxable. Capital losses could be deducted against gains only if they emanated
from the same type of assets (i.e., losses from stock sales could not be deducted against gains
from real estate or lottery sales). The tax law of 1928 did not change the capital gains taxation
except through further restricting the deduction possibilities of losses through, this time in
terms of where they were made (i.e., losses from real estate sales in Stockholm were not de-
ductible against gains from real estate sales in Gothenburg).
In 1951, the laws were significantly changed in that the holding period of the property or as-
sets that were sold with a gain was shortened to qualify for taxable capital gains. In a falling
scale, possessions of less than two years rendered taxability of 100 percent of the gains, be-
tween two and three years 75 percent of the gains, between three and four years 50 percent of
the gains, between three and four years 25 percent of the gains and, finally, longer than five
years gains were not taxed at all. Moreover, the tax base was extended to include also gifts
and bequests. Between 1967 and 1976, a further tax base extension was to make taxable 10
percent of all gains from selling assets held longer than five years and rendering at least a five
percent profit. Between 1977 and 1990, the falling scale was replaced by a division into
“younger” (owned less than two years) “older” (otherwise) assets, with full taxability for the
former and 40 percent for the latter.
The tax reform in 1990–1991 separated the taxation of capital gains from the rest of the in-
come taxation. The tax rate was flat and for real estate only half the gains were due to tax until
2001 when two thirds were made taxable. The rates have varied since 1991, from 30 percent
in 1991–1992, 25 percent in 1994, and finally 30 percent since 1995.
37
A3.3.2: The effect of changes in capital gains taxation on our study
[To be written]
A3.4: Lowest taxable income threshold
Since the introduction of the Swedish income taxation in 1903, there have been rules defining
the smallest annual income that requires the income earner to file a personal tax return (in
Swedish deklarationspliktgräns or “skattestreck”). As shown in Figure A1, the lowest taxable
income was SEK 1,000 in 1903 and then lowered to SEK 800 in 1910 and SEK 600 in 1919,
which remained the case until 1951 when it was doubled. In the postwar period, this nominal
threshold has been raised at a few occasions until it was linked to an inflation-index in the
1990’s. Figure A1 also shows how the threshold relates to the average annual income, defined
as ratio between the reference total income and the reference total population. Over time, the
share of the lowest taxable income threshold of the average income has declined sharply,
meaning that lower incomes were made taxable income. As for our analysis, this affects the
size of our reference total income concept, which includes both the total assessed income (i.e.,
the income reported by all tax filers) and a fixed fraction of the lowest taxable income thresh-
old for all those outside the tax population (see below). The reference total population, how-
ever, is not sensitive for these changes since it is based on total population counts rather than
number of tax filers. For that reason, not even the discrete threshold increases in 1952, 1962
37
More correctly, the 15 percent tax on real estate-related capital gains is a 30 percent tax on half the gains. The
variation in tax rates across income years corresponds to switching political colors of the government: being
Social Democratic before 1991 and after 1995 and right-wing in between.
38
and 1971 which reduced the number of tax filers by at most up to three percent, has any sig-
nificant bearing on the top income shares analyzed by us.
38
Figure A1: Lowest taxable income and its share of average total income, 1903–2003.
A3.5: Tax evasion and tax avoidance
Among the graver problems with using tax returns when analyzing income distribution is the
presence of tax avoidance and/or tax evasion. While tax avoidance is the minimization of tax
liability by legally exploiting the tax regime, tax evasion denotes efforts by individuals who
evade tax payments by breaking the law. Arguably both these activities are endogenous to the
tax system as a whole, and countries with relatively high marginal and average tax rates, such
as Sweden, would be expected to have relatively more of them. Empirically, however, both
are difficult to observe and hence measure, why we know little about their extent both today
and in the past.
The presence of tax avoidance and tax evasion primarily affect our results when they differ
systematically in extent across different time periods as well as levels of income. While both
reduce the overall assessed amount of income, and hence also our reference total income
which is used as denominator for measuring income shares, they do not affect the top income
shares if all income earners in all income classes practice them to the same extent (e.g., as
percentage of their incomes). Only if there is a systematic difference across income classes we
may have problems. If, say, only the below-median income earners evade taxes by working
substantially in the “black market”, the reference total income is reduced by a large amount
while keeping top incomes intact and hence leading to an overestimation of their income
shares. Similarly, if only the top 10% practiced significant tax avoidance and evasion, they
would get smaller shares of total income than would be if all their “true” income was as-
sessed. As the anecdotal evidence from the history of Sweden shows, it is difficult to conclude
anything about significant differences across income earners with different income both due
to lacking data and the use of various measurement techniques. On the whole, however, it
seems like both low- and high-income earners evade and avoid taxes but in different ways.
Whereas the former more often work on the “black market” where they get paid in cash or in
kind, the latter more often hide their sales of foreign financial assets and use advanced legal
tax planning to avoid taxes.
The earliest statement found on tax evasion is in 1919 when a special inquiry into the extent
of past evasion was carried through (Statistics Sweden, 1923, p. 13*). It was found that 331
income earners had evaded taxes partly or fully and that their (eventually) taxable incomes
amounted to a sum of were SEK 3,640,720, i.e., an average taxable income (including wealth
shares) of SEK 11,700. Compared with the overall average income (excluding wealth shares)
in 1919, SEK 2054.5, the tax evaders were hence people earning about five times the average
household which actually corresponds to somewhere around the 97–98th percentiles of all
income earners (see Table AX – “Income thresholds, as multiples of average income). While
being distributionally skewed towards the top, the tax evaders were not so significant in rela-
tion to the total number of top income earners. In 1919, there were about 3,000 income earn-
ers in the top 1% group and hence...
38
The doubling of the threshold in 1962 was estimated to decrease the number of income earners by about
125,000, representing about 3 percent (Statistics Sweden, Skattetaxeringarna samt fördelningen av inkomst och
förmögenhet taxeringsåret 1963, p. 21).
39
For the late 1940’s and early 1950’s, estimates of the aggregate evaded, or “under-assessed”,
income point at levels of 8–11 percent of all personal income, of which half was labor income
and the rest business (entrepreneurial) income (Bentzel 1952: pp. 64–65, 156–157). In a pub-
lic investigation of direct taxation from the same period, further anecdotal evidence comes
from surveys among tax collection officials. They report extensive evasion among practically
all sources of income, but especially business income, where wages are replaced by in-kind
payments, and the amounts written on invoices reduced while the rest is paid in cash (SOU
1951:51, pp. 352–361). If one would assume entrepreneurs to be more often top income earn-
ers than the rest of the population, this evidence hence suggests that evasion of top incomes
around 1950 was significant.
For most of the postwar period, wage earners have a harder time avoiding, or evading, taxes
than have self-employed since Sweden after 1947 practiced income taxation at source (see
section A3.1). Estimations from the late 1960’s suggest that tax evasion in Sweden was mod-
erate (see, e.g., Spånt, 1976: pp. 27–29). For example, local tax collection bureaus (taxerings-
nämnderna) made complaints on deductions in over 600,000 tax returns corresponding to on
average taxes of SEK 1,100, this would if true only result in about 0.8 percent higher total
income of the population. For top income earners, however, the addition would be somewhat
larger as about every forth such tax return was filed by a top 5% income earner. When consid-
ering all tax returns actually checked by the tax authorities, the increased assessed income was
only about SEK 20 million, which corresponds to or 0.02 percent of total taxed income).
After the tax reform of 1990–1991, a series of estimations of the size of the aggregate tax eva-
sion has been made. In Tengblad (1994), the estimated amount of income evaded from taxa-
tion was SEK 60 billion, i.e., approximately five percent of the total assessed personal in-
come. Later this figure was lowered, and for 1997 the amount was taken to be about SEK 20–
40 billion, or about 1.5–3 percent of total assessed income (Swedish National Tax Board,
2004: p. 218). In interview surveys during 1997–1998, it was found that between 11 and 14
percent of the respondents, representing between 650,000 and 800,000 individuals, had them-
selves worked on the “black market”. Their average hourly wage of was SEK 112, which is
only slightly higher than the average private sector wage and hence indicates that very few, if
any, of these respondents belonged to the top 10% income earners.
39
Income from capital and capital gains in the 1980’s and 1990’s has also been hidden from
taxation, but at quite varying degrees. Domestic interest and dividend payments were hardly
hidden at all due to extensive control mechanisms and double reporting rules.
40
As for foreign
holdings, however, one estimate suggests that 40 percent of all individuals receiving foreign
capital income did not assess them on their tax returns. Realizations of sales of stocks were
also improperly assessed, leading in 2001 to reduced tax income of approximately SEK 900
billion (Swedish National Tax Board, 2004: p. 222).
Appendix B: Construction of reference totals
In this Appendix we explain in greater detail exactly how our reference totals have been con-
structed. The different reference totals are used in Appendix C to test the robustness of our
series to the choice of reference total.
39
The average hourly wage for private sector blue-collar workers was SEK 106.8 in 2000 (data from Statistics
Sweden: Lönestrukturstatistik, privat sektor, genomsnittlig timlön (totallön), arbetare).
40
Estimates suggest that the amounts hidden represented about one percent of total capital income during the
1980’s and a tenth of the for the 1990’s (Swedish National Tax Board, 1994: p. 221).
40
B1: Reference total population
As described above, there has been one major change in Swedish tax legislation in the Twen-
tieth Century which has fundamentally changed the concept of tax unit, namely the 1970 tax
reform shift from a family based tax unit to an individually based concept. In terms of tax
statistics, however, this change occurred (at least to some extent) already in 1951. Before this
tax statistics were based on the entire tax population and figures referred to “tax units” i.e.
individuals as well as married couples counted as one income earner.
41
Before 1951 the obvi-
ous reference population is therefore the adult population (which we take to be everybody
aged 16 or above) less married women (since a married women formed one tax unit together
with her husband). After 1951, however, statistics changed to being based on a representative
sample (ten percent) of the population with married couples, where both had income, now
treated as two income earners in the statistics even though they were still taxed as one unit.
The problem is that in cases where the women did not work, or had low income, she was not
necessarily counted. This means that income statistics between 1951 and 1971 when the indi-
vidually based system was fully introduced (for labor income, tax on capital income remained
family based) is a mix between a family based system and an individually based system in-
cluding some women (those with substantial income) but not all. Starting 1971, the reference
total is again relatively unambiguous, now obviously being the adult population.
Apart from the quantitatively more substantial decisions discussed above there are a number
of smaller adjustments which can be considered. Over the course of a year individuals move
in and out of the country, some die, some turn 16 after the population count but before taxes
are filed, etc. Based on recent years when we believe that the coverage in the tax statistics is
close to complete we have concluded that correcting for deaths is most important. The tax
statistics before 1951 contain tax returns for those who died during the previous year (the in-
come year), in the period 1951-1973 these are not present in our data, but from 1974 and on-
wards they are again part of the statistics. We have therefore added deaths to our reference
total for the population before 1951 and after 1973.
42
For these periods we therefore add the
number of deaths during the year when calculating the reference total population.
In terms of choosing the appropriate reference population the period 1903–2003 can, hence,
be divided into the following three periods: 1) 1903–1950, the total population aged 16 or
above minus married women, 2) 1951–1970, the total population aged 16 or above minus
women likely to be excluded in the statistics, 3) 1971–2003, the total population aged 16 or
above.
For the period 1903-1950 the reference total population is:
The population aged 16- (from Statistics Sweden, Population statistics, SCB
Programmet för befolkningsstatistik)
– married women (from Statistics Sweden, Statistical Yearbook of
Sweden, Statistisk Årsbok, various years)
41
Note that this is the case for tax statistics before 1951 but not income figures in the Census (Folkräkningen).
42
To be precise, deaths are not in the statistics 1951-1966 (though they are taxed) while they are separetly ac-
counted for in the period 1967-1973 and hence we can exclude them from our tables. References for the treat-
ment of deaths are e.g.: for the period before 1951, Statistics Sweden, Inkomst och förmögenhet 1969, p. 11, for
the period 1951-1966, Statistics Sweden, Skattetaxeringarna…1966, p. 32,for the period 1967-1973 Statistics
Sweden, Inkomst och förmögenhet 1969, p. 13-15, 20-21, and after 1974 Statistics Sweden,SCB SM N 1976:4
(p.2) and SCB OE 21 SM 0501.
41
+ deaths during the year (from Statistics Sweden, Statistical Yearbook of
Sweden, Statistisk Årsbok, various years)
For the period 1951-1971 the preferred reference total population is:
The population aged 16- (from Statistics Sweden, Population statistics, SCB
Programmet för befolkningsstatistik)
– “housewives” (before 1950 data is from Krantz (1987) and after
1950 the difference between men and women in paid
work (see Edvininsson (2005), p. 140)). All “house-
wives” are subtracted before 1967 (when individual
taxation became voluntary) there after a linearly de-
clining share between 1967-1971.
For the period 1972-2003 the preferred reference total population is:
The population aged 16- (from Statistics Sweden, Population statistics, SCB
Programmet för befolkningsstatistik)
+ deaths during the year (added after 1973 since they reappear in the statistics
in 1974, from Statistics Sweden, Statistical Yearbook
of Sweden, Statistisk Årsbok, various years)
To check the robustness of our results we have calculated a number of alternatives which dif-
fer mainly in the period 1951-1971. These are sometimes not “alternatives” in the sense that
we may know that they are clear over-, or underestimations, but rather they serve the purpose
of giving bounds to our estimates. Figure XYZ shows the population aged 16 and above, the
number of tax returns and the different alternative specifications. The alternative specifica-
tions are the following:
• Preferred series = (Pop 16-)-Married W + deaths for 1903-1950, (Pop 16-)-
(Housewives for 1951-1966), and from 1967- Pop 16-, subtracting declining share of
housewives 1967-1971 and adding deaths after 73 (1974-).
• Alt tax units 1 = (Pop 16-)-Married W for 1903-1950, and from 1951- Pop16-
• Alt tax units 2 = (Pop 16-)-Married W for 1903-1950, (Pop 16-)-Housewives for 1951-
1966, and from 1967-, Pop 16-
• Tax units 3 = (Pop 16-)-Married W + deaths for 1903-1950, (Pop 16-)-(0,5 x House-
wives for 1951-1966), and from 1967- Pop 16-, subtracting declining share of house-
wives 1967-1971 and adding deaths after 73 (1974-)
• Tax units 4 = (Pop 20-)-Married W for 1860-1950, and from 1951- Pop20-
Figure Diagreftotpop1 from FinalIncShares1903-2003June2805
Looking at the behavior of the ratio between the number of tax returns and our reference se-
ries, especially around the critical years when there are changes in the definition of tax unit,
i.e. 1951, 1967 and 1971, indicates which series seem best. Put simply, we do not want there
to be any sudden jumps in the ratio unless there are underlying real changes in the tax base.
To exemplify, in 1919 the tax threshold was dropped from SEK 800 to SEK 600 leading to a
real major expansion of the tax base. Here we expect the ratio to go up sharply. In 1951, how-
ever, the change was only in the type of statistics, not in the actual underlying number of tax
eligible individuals (units), so here we should not expect a break in the ratio. To the extent
42
that the number of returns increase, this should be compensated by an increase in the refer-
ence total. At the same time, we do not, of course, wish to make ad hoc adjustments to keep
the ratio fixed, since there are real changes in the number of tax filers. Figure ABC shows the
ratio between the number of tax returns and our preferred series with indications of critical
breaks (the ratios between the number of filed tax returns and the alternative series above are
shown in Figure 2, in Section 2.2 above).
Figure Diagreftotpop3 from FinalIncShares1903-2003June2805
B2: Reference total income
In constructing our reference total income we have used three basic approaches. The first two
are based on that we can (as explained in Section 2 above) arrive at the “Preferred Total In-
come Definition” either by 1) starting with “Total Personal Sector Income” and deducting
items not included in our preferred definition, or 2) by starting from the “Tax Statistics In-
come” and add items not included in the tax base and income estimates for individuals not
included in the tax statistics. The third - which is mainly included as a point of reference - is
based on the assumption that our preferred income total can be approximated as a fixed share
of GDP.
Starting with the first approach, we need homogenous estimates of “Total Personal Sector
Income” from which we want to deduct items not included in our preferred definition of total
income. The best homogenous National Accounts series which span the whole period which
we study are those by Edvinsson (2005). These, however, contain only aggregate series for
Wages and salaries of employees (including social benefits) and Imputed labor income of self-
employed (including social benefits). To these we have added aggregate capital income and
property income reported in the tax statistics giving us an estimate of “Personal sector total
income”.
43
This, hence, becomes:
Wages and salaries of employees (including social benefits) (from Edvinsson, 2005)
+ Imputed labor income of self-employed (incl. social benefits) (from Edvinsson, 2005)
+ individual capital income (from Taxeringarna…,
1922-1988, and corre-
sponding sources thereaf-
ter, and estimated before
1922).
+ individual property income (same as for capital in-
come above)
= Estimated “Personal sector total income”
This estimate fluctuates around 0.7 times GDP (calculated from the expenditure side, reported
in Edvinsson 2005) with a standard deviation of 0.03.
Starting from the tax statistics income we use the following method to get at our preferred
“Reference total income”:
43
These are available from the aggregate taxation statistics Taxering till inkomst- och förmögenhet 1922-1988,
for the years before we add shares based on the observations 1922, and after 1988 we add the corresponding
figures in the new tax statistics.
43
Tax statistics income (the aggregates from the same sources as the income
statistics described above in A2 (sometimes corrected
for wealth shares))
+ items not included in the tax base (we make the assumption that all important sources
of income including certain social security benefits
are included in the tax base after 1974 (hence ab-
stracting from child allowances, allmänt barnbidrag,
and study grants, studiebidrag, which are tax free)
and add aggregate government expenditures for un-
employment benefits (arbetslöshetsersättning), pay-
ments for sickleave (sjukpenning) and payments for
mothers (moderskapsförsäkring, which in 1974 was
replaced by “parenthood insurance”, föräldrarförsäk-
ring, which was taxed) based on figures in the Statis-
tical Yearbook of Sweden from 1948- (before that
they are not listed but can be assumed to be a small
share).
+ estimated income for “non-filers” (in our preferred specification we take (reference
population - tax filers) x (0,8 times the tax threshold).
As an alternative specifications we use 0.25 times the
average income of tax filers).
= “Preferred reference total” (starting from the tax statistics income)
Figure DEF shows the alternative specifications over the whole period as shares of GDP, as
well as 0.63 times GDP. What we can say with some certainty is that the estimate of “Per-
sonal sector total income” is an over estimate of our preferred reference total. We can also say
with some certainty that at least since 1974 the tax statistics income is relatively close to our
preferred reference total since most people file taxes and everything we wish to include as
income is included in the tax base. We can also note that in the period 1930-1990 our “Pre-
ferred reference total” calculated starting with the tax statistics income follows the estimated
“personal sector total income very closely. In fact, taking 0.89 times the latter, yields numbers
which follow the former with very small deviations.
44
We also note that for the early years
(1903-1920) imputing 0.8 times the threshold (or 0.25 times average income) clearly yields
over estimates of reference income. This is to be expected since when most individuals are
below the threshold small changes in assumptions about their average income make a big dif-
ference and at this point in time the average income amongst tax payers was certainly much
higher then later implying that imputing similar shares to non-filers as later means overesti-
mating their income a lot.
Figure 1 in Finalreftotals
Given the behavior of these series we have chosen to use 0.89 times our estimated “personal
sector total income” as our reference total for the period 1903-1942 and then (as tax statistics
become yearly) our calculated reference total income starting with tax statistics income. As
with the reference total population we have calculated top income shares using a number of
alternatives as well. The resulting series can be found in Appendix C.
44
The standard deviation is 0.02 and the maximum deviation is 0.05.
44
Appendix C: Sensitivity analysis: various adjustments, changing reference totals etc
Adjustments and comparisons to other studies of the period before 1950
Sensitivity of the results with respect to changing reference totals.
C1: Sensitivity analysis of choosing using different reference totals
Figure RST shows the top income shares when using all the alternative specifications of refer-
ence total population discussed in Appendix B1, while Figure UVX shows the top shares us-
ing the different reference totals for income.
Figure RST
Figure UVX
Again, we would like to stress that these are not all “alternatives” some series (like the refer-
ence population 1 is known to be incorrect but it creates a bound for the maximum deviation
if the estimate was maximally incorrect.
C2: Sensitivity analysis of choosing either individuals or families as tax units
Our income series are computed from the tax returns-based income statistics for most years,
and as we describe in Appendix A2 this implies that we use two different concepts of income
earners over the twentieth century. Before 1951, the income earner in our data is the family (or
household), i.e., married couples with, or without, children, single men 16 years and older,
and single women 16 years or older. From 1951 onwards, our income earner is the individual,
meaning all men and women 16 years or older. Hence, while we in the first period count mar-
ried couples as one income earner, they are counted as two income earners in the latter period.
This section offers some partial explorations of how this switch of income earner concept may
influence the overall results of our study. As our historical data were chosen largely due to
availability constraints, we cannot make a fully-fledged comparison as there are simply no
parallel datasets based on tax data available. What we can do, however, is to compare our
family-based series with the series in which individuals are the basis. This can be done from
the years from which we use the Census material (the years 1920, 1930, 1935 (partial census),
1945 (partial census), and 1950) when the primary material is individual-based but adjusted
by us and others (especially Bentzel, 1952) to be consistent with the family-based series from
the years before 1920 and in between the other years (1934 and 1937).
Figure C1 shows the income shares of the top fractiles (from top 10 percent to the top 0.05
percent). Solid lines represent our main family-based income series used in our analysis
(called “Family”) whereas the broken lines are the unadjusted, individual-based census series
(called “Individual”). Note that since we use different concepts of income earners in the two
cases, we must also use two different reference total populations to calculate the correct popu-
lation shares. In our family-based series, we use the adult population 16 years and above mi-
nus married women and in the individual-based series the adult population 16 years and above
is used. For this reason, the level of the shares may not be fully corresponding to each other
although as Figure C1 shows they as a matter of fact are to quite some extent. As for the
changes in shares over the period, they are pretty much coinciding in all cases for all fractiles,
and importantly there is no systematic tendency in some direction of either series. For exam-
ple, whereas the individual-based series produce slightly larger declines between 1935 and
1950 for the top 10 percent to top 0.5 percent income earners, the family-based series do it for
45
the top 0.1 to top 0.05 percent fractiles. Altogether, we feel confident with our choice of in-
come earner concepts and have not found any systematic biases when contrasting them with
alternative definitions.
C3: Age adjustments and effects of censoring the youngest income earners
Similar to the top income earners analyzed in previous studies for other countries, we use an
age cutoff to remove young people from our data and thereby ensure that the series are con-
ceptually consistent over the years. Specifically, we use an age cutoff of 16 years meaning
that all income earners aged 16 and older are included in the series. We choose 16 years for
several reasons. First, this was the lowest allowed age by Statistics Sweden in the sampling of
income earners when constructing the income statistics 1951–1966 (see section A.3). Second,
ever since the late 1970’s this has been the lowest reported age level followed by 20 years.
Third, 16 years has for long also been the first year after the Swedish compulsory secondary
education ended. For robustness purposes we have also run the analysis excluding all income
earners under 20 years of age and present some comparative results in other sections of this
paper. As Atkinson (2003) and Atkinson and Leigh (2004) show, however, the effects on the
results from varying the age cutoff a few years are rather modest.
In practice, our age cutoff means that we subtract the number of income earners age 15 or less
from our reference total population and from the main top income series. It should be noted
that the age data only includes the number of income earners in each age-income class without
disclosing any income sums for each such group. This means that our age adjustments in prac-
tice is a subtraction of the number of income earners below 16 years in each income class
while leaving the total income as it stands. As shown in the following discussion, however,
this has no effect whatsoever on the overall results of our study since these very young in-
come earners practically had no significant incomes at all.
Another data constraint is that before 1951 there exist no age data in the tax material (Statis-
tics Sweden only collected data on income and occupation). This also should have no effect
our study since before that year the Swedish income statistics was family-based and the tax
units counted were mainly family fathers and single-living households (men and women). The
ones below 16 years of age in these two groups should be quite few.
A more intriguing data problem is that the income statistics in some years report a higher low-
est age group, 0–16 years during 1957–1966 and 0–17 years during 1971–1977. To get num-
bers for the (hence unobserved) 0–15 group we use interpolation based on the continuously
observed 0–19 group and the relation between this group and the 0–15 group during the years
just before and after the missing data periods. This minor bridging of the series seems to have
no unwarranted effects.
The effect of removing the youngest income earners from the population of income earners
seems to be insignificant on our analysis of top income earners. Figure C1 shows how the
number and incomes of income earners under 16 and 20 years as share of the total population
developed since 1951 (the first year for which age data were collected). The youngest group,
0–15 years, was clearly insignificant as income earners as their share of 0.1 percent of all in-
comes shows. Their share of the number of income earners, however, increased dispropor-
tionately in 1978 and 1992. In 1978, new tax collection routines required employers to submit
income statements for all employees and in the same year the number of income earners be-
low 16 years increased fivefold. In 1992, the drastic increase in the share of young income
earners was related to new rules in the tax reform. More specifically, all capital income above
46
SEK 100 was made taxable and hence almost one million children, roughly one ninth of the
entire Swedish population, became tax units overnight.
45
Notably, the number of income earn-
ers between 16 and 20 years old hardly increased at all (about 3 percent, compared to 571
percent for those 15 years old and younger). In other words, by excluding the youngest in-
come earners we avoid some unwarranted heterogeneity in the income earner shares caused
by the tax reform of 1990–1991.
Figure C1: Shares of tax population and total income of young income earners (below 16 and
20 years), 1951–2003.
(add section showing results with different ref totals (figure B1 and others)
Appendix D: Estimation theory
D1: Pareto interpolation: description and examples
All our income data come from tables showing the number of income earners and their
summed incomes in classes of total income (see Table X for an example of these tables, show-
ing the year 1945). Since the groups of income earners most often represent uneven shares of
the population whereas we wish to analyze income shares for a given share of the population,
e.g., the top 1%, we need to use interpolation techniques. This is a standard methodological
approach used in all previous studies of top incomes, although the specifics of the methods
may differ somewhat. One of the most commonly methods is to assume that the top incomes
(when lined up for each individual in a row) are Pareto distributed (see, e.g., Feenberg and
Poterba 1993, 2000; Piketty and Saez, 2003; Dell, 2005), and this is also the basic assumption
used in this study. As described in, e.g., Atkinson (2003), if H(y) is the cumulative proportion
of income earners with income y or higher, then for some arbitrary constant A and model pa-
rameter α Pareto’s formula states that
H(y) = A y
–α
. (1)
The cumulative share of total income for persons with income y or higher is
G(y) = [α/(α–1)] A y
–(α–1)
. (2)
Dividing equations (1) and (2) to get the average income for those income earners who have
an income equal to y or higher, we get the following expression,
G(y)/H(y) = [α/(α–1)] y. (3)
Now, in order to calculate income shares for specifically defined top shares of the population,
we define the specific share i (say 0.01, or the top 1 percent) of the population H
i
(y) and that
their corresponding share of total income is G
i
(y). We eliminate y in (1) and (2) by substitu-
tion and get G
j
(y)
α
as a function H
j
(y)
α–1
times a constant. The Pareto distribution is then used
for interpolation between two population shares surrounding the specific share (say, top 1% or
45
Formally, the new rules were in practice already in 1991 but in that year’s income statistics Statistics Sweden
made an adjustment to exclude the new bulk of very young income earners. They excluded all income earners
below 18 years of age with labor income less than SEK 12,000 (Statistics Sweden, Inkomst- och skattestatistik
1991, Be 20 SM 9301, p. 9).
47
top 0.1%) for which we wish to compute the share of total income. Using these two points to
achieve relative income shares G
i
(y)/G
j
(y) and relative population shares H
i
(y)/H
j
(y) we get
our main expression for computing top income shares
G
i
(y)/G
j
(y) = [H
i
(y)/H
j
(y)]
(α–1)/α
. (4)
From this expression, we compute the value of α and then the income share of specific popu-
lation shares.
As an example, we take the incomes in 1945 and compute shares of the top 1% of all income
earners. According to that year’s tabulated incomes, the classes representing shares of popula-
tion surrounding the 1%-level are those earning SEK 15,000 or above, having a share of
population of 1.53 percent and of total income of 12.38 percent, and those earning SEK
20,000 or above, being 0.83 percent of all income earners earning 8.86 percent of all incomes.
Using equation (4) in logarithmic form, we get that α/(α –1) =
ln(0.0153/0.0083)/ln(0.1238/0.0886) = 1.8439, and α = 2.185. Then to compute the income
share of the top 1%, again use equation (4) and set H
i
(y) = 0.01, H
j
(y) = 0.0083, G
j
(y) =
0.0886, but leave G
i
(y) (the income share of the top 1%) undecided. Then using the value of α,
we get that G
i
(y) = 0.0886 · [0.01/0.0083]
1/1.8439
= 0.0983, or 9.83 percent, which is hence the
income share of the top 1% income earners.
D2: Robustness issues of Pareto interpolation
The Pareto interpolation methodology does not come without problems, as many authors have
shown (see, e.g., Atkinson 2003, 2005). In particular, the income distributions may in fact not
be Pareto distributed, not even in the top levels where they are generally conceived to be so.
However, robustness checks against micro-level data in Feenberg and Poterba (2000) and
Piketty (2003) suggest that in at least for U.S. and French top incomes the Pareto approxima-
tion seems generally adequate. Moreover, Atkinson (2003) shows that the Pareto interpola-
tions between two reference points (G
i
(y) and G
j
(y)) may give different final shares depending
on which of the two points one inserts in the formula (4) for computing the wanted income
share. The standard practice is to use the information from the share being closest to the
wanted share, and in our example above we used the numerical information associated with
the top 0.86% to interpolate the income share for the top 1% since this was closer located than
the top 1.53%. If we instead would have used the latter share, however, we would have gotten
the exact same share for the top 1% (9.82 percent).
A more serious problem with using this method arises if the tabulated income intervals in the
top are too low to admit interpolation of the top income shares. For example, whereas the
share of income earners located in the highest (open-ended) income interval in 1945 was
0.0048 percent, which is sufficient to interpolate the income shares of the P99.95–100 fractile,
the same group in 1976 represented 1.09 percent of the total population, which instead means
that we need to extrapolate the income shares of most top fractiles. These problems arise if
tax or statistics authorities do not recurrently update their tabulated nominal income brackets
and if there is significant inflation in the economy. Although there have been no systematic
quantifications of how severe the effects of extrapolations some comparisons made by Atkin-
son (2003) indicates that they may be considerable. When we make some simple checks on
the Swedish data, however, the effects seem not so radical at all. We start by censoring the
1945 income data (same as above) at the income level SEK 20,000 (instead of the true highest
level SEK 200,000), representing the top 0.83% instead of the 0.005%, and extrapolate the
shares of the higher fractiles we get small deviations. For the top 0.5%, the share increases
48
from 6.73 percent to 6.75 percent, and for the top 0.05% (our highest fractile), the share in-
creases from 1.84 percent to 1.93 percent of total income. Similarly, when comparing the
years 1976 and 1977, when the highest income level was raised from SEK 120,000 (contain-
ing the top 1.09%) to SEK 1,000,000 (containing the top 0.002%) the income shares for the
top fractiles change only marginally and in the direction as the overall decreasing trend would
suggest.
To at least know the extent of extrapolations in our analysis of Swedish top incomes, Figure
D1 depicts both the highest tabulated income boundary in the income statistics during 1903–
2003 as well as the share of those income earners having this income or higher of all income
earners. As the Figure shows, the need to extrapolate the share of top incomes has not been
overly common throughout the period. For the top 1%, we only did it in 1976 when the top
group represented 1.087 percent of the whole population. For the top 0.5%, we extrapolated
the share six times, (1971–1975, 2000), for the top 0.1% 32 times and for the top 0.05% 44
times, out of a total of 74 observations. From inspecting the final top income series, we do not
trace many unwarranted variations of shares that seem attributable to the extrapolations. Quite
the contrary, in the 1970’s when the potential extrapolation problems are the worst we have
remarkably stable shares of top 0.1% and top 0.05% and in particular consistent with later
years when the highest reported income level was raised and no extrapolation needed.
Figure D1: Highest reported income boundary (right scale) and the share of income earners in
the highest income class of reference total population (left scale), 1903-2003
49
Table A.2: List of sources, 1903–2003.
Income
year
Main source
a), b)
Tables in
source
Pages
Series
c)
1903 Flodström (1906) 1, 3
1907 Flodström (1909) XI–XII FU
1911 Flodström (1914) 11 FU
1912 Flodström (1915) 13º FU
1916 Statistics Sweden, Statistical Yearbook 1929. 286–287 SOS
1919 Statistics Sweden, Statistical Yearbook 1929. 286–287 SOS
1920 Statistics Sweden, Statistical Yearbook 1929. 286–287 SOS
1930 Bentzel (1952)
1934 SOU 1936:18 10 47 SOS
1935 Bentzel (1952)
1937 Quensel (1944)
1940 Bentzel (1952)
1943 Skattetaxeringarna (1) ... taxeringsåret 1944 L 31* SOS
1944 Skattetaxeringarna (1) ... taxeringsåret 1945 Q 43* SOS
1945 Skattetaxeringarna (1) ... taxeringsåret 1946 P 42* SOS
Statistics Sweden (1951), Census of 1945 4 2–3 SOS
1946 Skattetaxeringarna (1) ... taxeringsåret 1947 R 47* SOS
1947 Skattetaxeringarna (1) ... taxeringsåret 1948 V 51* SOS
1948 Skattetaxeringarna (1) ... taxeringsåret 1949 Q 48* SOS
1949 Skattetaxeringarna (2) ... taxeringsåret 1950 R 48* SOS
1950 Skattetaxeringarna (2) ... taxeringsåret 1951 S 51* SOS
1951 Skattetaxeringarna (2) ... taxeringsåret 1952 Å, 8 63*, 26–27 SOS
Statistics Sweden (1956), Census of 1950 7 20–21 SOS
1952 Skattetaxeringarna (2) ... taxeringsåret 1953 Z, 8 53º, 26–27 SOS
1953 Skattetaxeringarna (2) ... taxeringsåret 1954 Z, 8 49º, 26–27 SOS
1954 Skattetaxeringarna (2) ... taxeringsåret 1955 Z, 8 47º, 26–27 SOS
1955 Skattetaxeringarna (2) ... taxeringsåret 1956 Z, 8 46º, 28–29 SOS
1956 Skattetaxeringarna (2) ... taxeringsåret 1957 Z, 8 47º, 28–29 SOS
1957 Skattetaxeringarna (2) ... taxeringsåret 1958 Y, 8 47º, 28–29 SOS
1958 Skattetaxeringarna (2) ... taxeringsåret 1959 Å, 8 50º, 34–35 SOS
1959 Skattetaxeringarna (2) ... taxeringsåret 1960 J, 8 28º, 32–33 SOS
1960 Skattetaxeringarna (2) ... taxeringsåret 1961 I, 10 28º, 32–33 SOS
1961 Skattetaxeringarna (3) ... taxeringsåret 1962 I, 10 28º, 34–35 SOS
1962 Skattetaxeringarna (3) ... taxeringsåret 1963 J, 10 29º, 34–35 SOS
1963 Skattetaxeringarna (3) ... taxeringsåret 1964 J, 10 43º, 36–37 SOS
1964 Skattetaxeringarna (3) ... taxeringsåret 1965 K, 10 44º, 36–37 SOS
1965 Skattetaxeringarna (3) ... taxeringsåret 1966 J, 10 43º, 116–117 SOS
1966 Skattetaxeringarna (3) ... taxeringsåret 1967 L, 9 43º, 118–119 SOS
1967 Inkomst och förmögenhet 1967 2, 7 44–45, 58–61 SOS
1968 Inkomst och förmögenhet 1968 2, 7 50–51, 64–67 SOS
1969 Inkomst och förmögenhet 1969 2, 7 50–51, 64–67 SOS
1970 Inkomst och förmögenhet 1970 2, 7 48–49, 62–65 SOS
1971 Inkomst och förmögenhet 1971 3, 12 68–69, 90–93 SOS
1972 Inkomst och förmögenhet 1972 1, 3, 14 54–55, 70–71,
102–105
SOS
Inkomst- och förmögenhetsfördelningen år 1972 7 19 SM N 1973:94
1973 Inkomst och förmögenhet 1973 3, 14 68–69, 100–103 SOS
1974 Inkomst- och förmögenhetsfördelningen år 1974 1, 7 11, 33 SM N 1976:4
1975 Inkomst- och förmögenhetsfördelningen år 1975 1, 7 13, 35 SM N 1976:23
50
1976 Inkomst- och förmögenhetsfördelningen år 1976 1, 7 18, 41, 43 SM N 1977:24
1977 Inkomst- och förmögenhetsfördelningen år 1977 1, 7 22, 46–47 SM N 1978:22
1978 Inkomst- och förmögenhetsfördelningen år 1978 1, 4.1, 4.2 29, 38, 41 SM N 1980:9
1979 Inkomst- och förmögenhetsfördelningen år 1979 1, 4.1, 4.2 20, 27, 30 SM N 1981:9.1
1980 Inkomst- och förmögenhetsfördelningen år 1980 1, 4.1, 4.2 7, 14, 17 SM N 1976:4
1981 Inkomst- och förmögenhetsfördelningen år 1981 1, 4.1, 4.2 7, 14, 17 SM N 1976:4
1982 Inkomst- och förmögenhetsfördelningen 1982 1, 4.1, 4.2 14, 21, 24 SM Be 1984:6.1
1983 Inkomst- och förmögenhetsfördelningen 1983 1, 4.1, 4.2 14, 21, 24 Be 20 SM 8501
1984 Inkomst- och förmögenhetsfördelningen 1984 1, 3.1, 3.2 15, 19, 22 Be 20 SM 8601
1985 Inkomst- och förmögenhetsfördelningen 1985 1, 2.1, 2.2 15, 18, 21 Be 20 SM 8701
1986 Inkomst- och förmögenhetsfördelningen 1986 1, 2.1, 2.2 17, 20, 23 Be 20 SM 8801
1987 Inkomst- och förmögenhetsfördelningen 1987 1, 2.1, 2.2 17, 20, 23 Be 20 SM 8901
1988 Inkomst- och skattestatistik 1988 1, 2.1, 2.2 16, 19, 22 Be 20 SM 9001
1989 Inkomst- och skattestatistik 1989 1, 2.1, 2.2 16, 20, 23 Be 20 SM 9101
1990 Inkomst- och skattestatistik 1990 1, 2.1, 2.2 15, 20, 23 Be 20 SM 9201
1991 Inkomst- och skattestatistik 1991 1, 4.1, 4.2,
6.1, 6.2
15, 22–23, 28–29 Be 20 SM 9301
1992 Inkomst- och skattestatistik 1992 1, 4.1, 4.2,
6.1, 6.2
14, 21–22, 27–28 Be 20 SM 9401
1993 Inkomst- och skattestatistik 1993 1, 4.1, 4.2,
6.1, 6.2
14, 21–22, 27–28 Be 20 SM 9501
1994 Inkomst- och skattestatistik 1994 1, 4.1, 4.2,
6.1, 6.2
15, 22–23, 28–29 Be 20 SM 9601
1995 Inkomster, skatter och bidrag 1995 1, 4.1, 4.2,
6.1, 6.2
19, 26, 28, 36, 38 Be 20 SM 9701
1996 Inkomster, skatter och bidrag 1996 1, 4.1, 4.2,
6.1, 6.2
20, 27, 29, 37, 39 If 20 SM 9801
1997 Inkomster, skatter och bidrag 1997 1, 4.1, 4.2,
6.1, 6.2
20, 29, 33, 45, 49 If 20 SM 9901
1998–
2003
–
d)
a) Some publications titles are abbreviated. Skattetaxeringarna (1) = Skattetaxeringarna samt inkomstfördelning-
en inom yrkesgrupper; Skattetaxeringarna (2) = Skattetaxeringarna samt fördelningen av inkomst och förmögen-
het inom yrkesgrupper; Skattetaxeringarna (3) = Skattetaxeringarna samt fördelningen av inkomst och förmö-
genhet taxeringsåret.
b) The publications since 1982 also have the subtitle Totalräknad statistik.
c) “FU” denotes Finansstatistiska utredningar (Fiscal Surveys) and “SOS” Sveriges officiella statistik (Swedish
Official Statistics).
d) During 1998–2003, Statistics Sweden stopped publishing income data in the form of earlier years. Instead we
had to purchase data for these years directly from Statistics Sweden.
51
Table A.1: Example of how tax data appears in Swedish income statistics: Income year
1945.
Income earners earning at least SEK... Number of income earners Sum of incomes (SEK)
600 2,955,890 10,758,167,000
1,000 2,666,978 10,524,967,000
1,500 2,264,436 10,021,845,000
2,000 1,917,702 9,417,016,000
2,500 1,609,106 8,724,037,000
3,000 1,327,728 7,951,446,000
3,500 1,076,653 7,136,837,000
4,000 861,737 6,332,420,000
4,500 683,508 5,576,476,000
5,000 541,999 4,905,747,000
6,000 351,915 3,870,010,000
7,000 244,620 3,177,222,000
8,000 178,803 2,686,052,000
10,000 108,483 2,061,673,000
12,000 74,035 1,686,006,000
15,000 47,127 1,327,306,000
20,000 25,614 958,934,000
30,000 10,901 606,652,000
50,000 3,771 339,106,000
100,000 682 89,564,000
200,000 174 55,913,000
Sum 2,955,890 10,758,167,000
Source: See Table XX.
52
Figure 1
Different Reference Total Incomes as shares of GDP 1903-2000
0,2
0,3
0,4
0,5
0,6
0,7
0,8
0,9
1
1,1
1903
1906
1909
1912
1915
1918
1921
1924
1927
1930
1933
1936
1939
1942
1945
1948
1951
1954
1957
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
Pref RefTotInc/GDP
0.89xTotPersInc/GD
P
TaxStatInc+0.8xtaxth
reshold+notinclintaxb
ase/GDP
TaxStatInc+0.25xavg
taxpayerinc+notinclint
axbase/GDP
Total Personal
Income
0.63xGDP
53
Figure 2
Different reference total populations as share of tax returns
0
0,2
0,4
0,6
0,8
1
1,2
1903
1907
1911
1915
1919
1923
1927
1931
193
5
193
9
194
3
194
7
195
1
195
5
195
9
196
3
196
7
197
1
197
5
197
9
198
3
198
7
199
1
199
5
199
9
200
3
(N. ret 16-)/TU1
(N. ret 16-)/TU2
Pref. series (N. ret 16-)/TU3
(N. ret 16-)/TU4
(N. ret 20-)/TU5
Preferred series
54
Figure 3
Income shares for the top decile in Sweden 1903-2003
0
10
20
30
40
50
60
1903
1906
1909
1912
1915
1918
1921
1924
1927
1930
1933
1936
1939
1942
1945
1948
1951
1954
1957
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
P90-100
P95-100
P99-100
P90-100 (excl CG)
P95-100 (excl CG)
P99-100 (excl CG)
P90-100
P95-100
P99-100
55
Figure 4
Income shares for P90-95, P95-99 and P99-100 in Sweden 1903-2003
0
5
10
15
20
25
30
1903
1906
1909
1912
1915
1918
1921
1924
1927
1930
1933
1936
1939
1942
1945
1948
1951
1954
1957
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
P90-P95
P95-P99
P99-100
P90-P95 (excl CG)
P95-P99 (excl CG)
P99-100 (excl CG)
P 99-100
P 90-95
P 95-99
56
Figure 5
Income shares for the top percentile in Sweden 1903-2003
0
5
10
15
20
25
30
1903
1907
1911
191
5
1919
192
3
1927
193
1
1935
193
9
1943
194
7
1951
195
5
1959
1963
1967
1971
1975
1979
198
3
1987
199
1
1995
199
9
2003
P99-100
P99.5-100
P99-100 (excl CG)
P99.5-100 (excl CG)
P 99-100
P 99.5-100
57
Figure 6
Income shares of the top 0.1 percent in Sweden 1903-2003
0
2
4
6
8
10
12
14
1903
1907
191
1
1915
191
9
1923
1927
193
1
1935
193
9
1943
1947
195
1
1955
195
9
1963
1967
197
1
1975
197
9
1983
1987
199
1
1995
199
9
2003
P99.9-100
P99.95-100
P99.9-100 (excl CG)
P99.95-100 (excl CG)
P 99.9-100
P 99.95-100
58
Figure 7
Income thresholds (including capital gains) for different fractiles, 2003 prices
0
1 000 000
2 000 000
3 000 000
4 000 000
5 000 000
6 000 000
1903
1906
1909
1912
1915
1918
1921
1924
1927
1930
1933
1936
1939
1942
1945
1948
1951
1954
1957
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
SEK
P90
P95
P99
P99.5
P99.9
P99.95
P 99.95
P 99.9
P 99.5
59
Figure 8
Income thresholds as multiples of average income 1903-2003
0
10
20
30
40
50
60
70
80
90
190
3
1906
190
9
1912
191
5
1918
192
1
1924
192
7
1930
193
3
1936
193
9
1942
194
5
1948
1951
1954
1957
1960
1963
196
6
1969
197
2
1975
197
8
1981
198
4
1987
199
0
1993
199
6
1999
200
2
Multiples of average income
P90
P95
P99
P99.5
P99.9
P99.95
P 99.95
P 99.5
P 99.9
P 95
P 99
P 90
60
Figure 9: Income composition 1945 (solid: including capital gains, broken: excluding capital gains)
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
P50-90 P90-95 P95-99 P99-99.5 P99.5-99.9 P99.9-99.95 P99.95-100
Share of total income
Wages Capital Business Business CG
Wages CG Capital CG Capital gains
61
Figure 10: Income composition 1978 (solid: including capital gains, broken: excluding capital gains)
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
P50-90 P90-95 P95-99 P99-99.5 P99.5-99.9 P99.9-99.95 P99.95-100
Share of total income
Wages Capital Business Wages CG
Business CG Capital CG Capital gains
62
Figure 11: Income composition 1997 (solid: including capital gains, broken: excluding capital gains)
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
P50-90 P90-95 P95-99 P99-99.5 P99.5-99.9 P99.9-99.95 P99.95-100
Share of total income
Wages Wages CG Capital Capital CG
Business Business CG Capital gains
63
Figure 12a-b: Long-run income composition of the top 10 percent, 1945–2003
P90-100 (incl. capital gains)
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1945 1949 1953 1957 1961 1965 1969 1973 1977 1981 1985 1989 1993 1997 2001
Wages Business Capital Capital gains
P90-100 (excl. capital gains)
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1945 1949 1953 1957 1961 1965 1969 1973 1977 1981 1985 1989 1993 1997 2001
Wages Business Capital
64
Figure 13a-b: Long-run income composition of the top 1 percent, 1945–2003
P99-100 (incl. capital gains)
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1945 1949 1953 1957 1961 1965 1969 1973 1977 1981 1985 1989 1993 1997 2001
Wages Business Capital Capital gains
P99-100 (excl. capital gains)
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1945 1949 1953 1957 1961 1965 1969 1973 1977 1981 1985 1989 1993 1997 2001
Wages Business Capital
65
Figure 14a-b: Long-run income composition of the top 0.05 percent, 1945–2003
P99.95-100 (excl. capital gains)
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1945 1949 1953 1957 1961 1965 1969 1973 1977 1981 1985 1989 1993 1997 2001
Wages Business Capital
P99.95-100 (incl. capital gains)
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1945 1949 1953 1957 1961 1965 1969 1973 1977 1981 1985 1989 1993 1997 2001
Wages Business Capital Capital gains
66
Figure 15:
Capital income as share of total income (excludin
g
capital
g
ains) in Swedish top
incomes, 1912-2003.
1935
1930
1950
1967
1991
1945
1920
1916
0%
10%
20%
30%
40%
50%
60%
70%
80%
1912 1922 1932 1942 1952 1962 1972 1982 1992 2002
Capital income share of total income
P99.95-100
P99.9-100
P99.5-100
P99-100
P95-100
P90-100
P50-100
67
Figure 16:
International comparison of P95-100, 1903-2003
0
5
10
15
20
25
30
35
40
45
50
1903
1906
1909
1912
1915
1918
1921
1924
1927
1930
1933
1936
1939
1942
1945
1948
1951
1954
1957
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
SWE (inkl CG)
FR (Piketty)
NL (A&S)
US (P&S inkl CG))
UK (A&S)
Can (S&W incl CG)
Can (S&W excl CG)
US (P&S excl CG)
SWE (excl CG)
68
Figure 17:
International comparison of P99-100, 1903-2003
0
5
10
15
20
25
30
1903
1906
1909
1912
1915
1918
1921
1924
1927
1930
1933
1936
1939
1942
1945
1948
1951
1954
1957
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
SWE (incl CG)
FR (Piketty)
NL (A&S)
UK (A&S)
US (P&S inclCG)
US (P&S excl CG)
Can (S&W incl CG)
Can (S&W excl CG)
SWE (excl CG)
69
Figure 18:
International comparison of P99.9-100
0
2
4
6
8
10
12
14
1903
1906
1909
1912
1915
1918
1921
1924
1927
1930
1933
1936
1939
1942
1945
1948
1951
1954
1957
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
SWE (incl CG)
FR (Piketty)
NL (A&S)
UK (A&S)
US (P&S incl CG)
US (P&S excl CG)
Can (S&W excl CG)
Can (S&W incl CG)
SWE (excl CG)
70
Figure A1: The lowest taxable income and its share of average income (Reference total income/Reference total population).
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
1903 1913 1923 1933 1943 1953 1963 1973 1983 1993 2003
Lowest taxable income (SEK, current prices)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Ratio between lowest taxable income and
average income
Lowest taxable income (left scale)
Lowest taxable income divided by Average total income (right scale)
71
Figure B:
Population, tax returns and tax units
0
1 000 000
2 000 000
3 000 000
4 000 000
5 000 000
6 000 000
7 000 000
8 000 000
1903
1907
191
1
191
5
191
9
192
3
1927
1931
1935
1939
1943
194
7
195
1
195
5
195
9
1963
1967
1971
1975
1979
1983
198
7
199
1
199
5
199
9
2003
N. taxret (16-)**
Alt tax units 1
Alt tax units 2
Preferred series
Alt tax units 3
Alt tax units 4
Pop 16-
1951: Shift in income
statistics leading to the
inclusion of married
women with own income
1967: Shift in income
statistics and individually
based taxation is introduced
on a voluntary bases
1978: Employers start sending
income statements to the tax
authorities
Tax returns
Pop 16-
72
Figure B:
Tax returns, tax units and the ratio between the two 1903-2003
0
1 000 000
2 000 000
3 000 000
4 000 000
5 000 000
6 000 000
7 000 000
8 000 000
190
3
1907
1911
1915
191
9
192
3
1927
1931
193
5
193
9
1943
1947
1951
195
5
195
9
1963
1967
197
1
197
5
1979
1983
1987
199
1
1995
1999
2003
0,0000
0,2000
0,4000
0,6000
0,8000
1,0000
1,2000
N. taxret (16-)**
Preferred series
Pref. series (N. ret 16-)/TU3
1903-1920: Major
real increases
in the tax base
1951: Shift in income
statistics leading to the
inclusion of married
women with own income
1967: Shift in income
statistics and individually
based taxation is introduced
on a voluntary bases
1978: Employers start sending income
statements to the tax authorities (Real
increase in the number of returns)
Tax Returns / Tax Units
(right scale)
Number of Tax Returns
(left scale)
Number of Tax Units
(left scale)
73
Figure B:
Different Reference Total Incomes as shares of GDP 1903-2000
0,2
0,3
0,4
0,5
0,6
0,7
0,8
0,9
1
1,1
1903
1906
1909
1912
1915
1918
1921
1924
1927
1930
1933
1936
1939
1942
1945
1948
1951
1954
1957
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
Pref RefTotInc/GDP
0.89xTotPersInc/GD
P
TaxStatInc+0.8xtaxth
reshold+notinclintaxb
ase/GDP
TaxStatInc+0.25xavg
taxpayerinc+notinclint
axbase/GDP
Total Personal
Income
0.63xGDP
74
Figure B:
Top income shares with alternative specifications of reference population
0
10
20
30
40
50
60
190
3
190
6
190
9
191
2
191
5
191
8
1921
192
4
1927
193
0
1933
193
6
1939
194
2
194
5
194
8
1951
195
4
1957
196
0
1963
196
6
1969
197
2
1975
197
8
1981
198
4
1987
199
0
1993
199
6
1999
200
2
P90 pref
P95 pref
P99 pref
P99.5 pref
P99.9 pref
P99.95 pref
P90:ref pop 4
P95:4
P99:4
P99.5:4
P99.9:4
P99.95:4
P90: ref pop 5
P95:5
P99:5
P99.5:5
P99.9:5
P99.95:5
P90:ref pop 6
P95:6
P99:6
P99.5:6
P99.9:6
P99.95:6
P90:ref pop 7
P95:7
P99:7
P99.5:7
P99.9:7
P99.95:7
75
Figure B:
Top income shares with alternative specifications of reference income
0
10
20
30
40
50
60
190
3
190
6
1909
191
2
1915
191
8
192
1
1924
192
7
1930
193
3
193
6
1939
194
2
1945
194
8
195
1
1954
195
7
1960
196
3
196
6
1969
197
2
1975
197
8
198
1
198
4
198
7
1990
199
3
1996
199
9
200
2
P90 pref
P95 pref
P99 pref
P99.5 pref
P99.9 pref
P99.95 pref
P90:ref inc 2
P95:2
P99:2
P99.5:2
P99.9:2
P99.95:2
P90:ref inc 3
P95:3
P99:3
P99.5:3
P99.9:3
P99.95:3
76
Figure C1: Sensitivity analysis: Comparison of shares when income earner concept is either individuals or households
0
5
10
15
20
25
30
35
40
45
1920 1925 1930 1935 1940 1945 1950
Share of total income (%)
P90-100 Individuals P90-100 Families P95-100 Individuals P95-100 Families
P99-100 Individuals P99-100 Families P99.5-100 Individuals P99.5-100 Families
P99.9-100 Individuals P99.9-100 Families P99.95-100 Individuals P99.95-100 Families
P99.95-100
P95-100
P99.9-100
P99.5-100
P99-100
P90-100
77
Figure C2: Shares of total income and income earners of the youngest age groups, 1951–2003
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
20%
1951 1958 1965 1972 1979 1986 1993 2000
Share of all income earners or total income
Earner_share 0-19 yrs
Income_share 0-19 yrs
Earner_share 0-15 yrs
Income_share 0-15 yrs
78
Figure D1: Highest report income level and share of income earners in highest income class.
0.0%
0.2%
0.4%
0.6%
0.8%
1.0%
1.2%
1903 1913 1923 1933 1943 1953 1963 1973 1983 1993 2003
Share of all income earners of highest income class
0
200,000
400,000
600,000
800,000
1,000,000
1,200,000
Highest reported income bound (SEK)
Highest reported income
boundary (right scale)
Share of income earners in
highest income class (left scale)
