Gender Dimensions in Top Income Literature: Evidence from Sweden and Beyond

WOMEN IN TOP INCOMES EVIDENCE FROM SWEDEN OVER THE 20THCENTURY 3 May, 2018 LIS/LWS Users Conference 2018 The legacy of Tony Atkinson in inequality analysis Jesper Roine, SITE, Stockholm School of Economics (joint work with Anne Boschini, SOFI, SU, and Kristin Gunnarsson, UU)

MAIN RESEARCH AREAS CONNECTED TO THIS PROJECT (AND TO TONY S WORK) Top income literature Started by Tony and Thomas Piketty around 2000. Collected volumes by Atkinson and Piketty 2007, 2010 summarizes first wave of this literature). Women in top incomes Atkinson, Casarico and Voitchovsky (2016) Plan is to present main results from two on-going projects ( and maybe mention a third project) Evidence based detailed data since the 1970s, WP: Women in Top Incomes: Evidence from Sweden 1974 2013 Women in top incomes over the 20thcentury (ongoing) Couple-formation in the top of the distribution (cross country evidence using LIS)

INCOME INEQUALITY RESEARCH A TALE OF TWO LITERATURES Often focus on disposable income (adjusted for household size) Household is the income unit, Individual is the unit of analysis; includes all incomes and taxes and transfers and full population. Parallel literature on labour market outcomes (for individual adults) focusing on wages. Individual (adults) are the income unit and the unit of analysis, measures are often per time unit (pay/hour, monthly, full time salary).

TOP INCOME INEQUALITY A THIRD STRAND OF INEQUALITY LITERATURE Top income literature differs in important ways Focus on income from all sources (before taxes and transfers) Unit of analysis is the tax unit (mixed over time and across countries!) Main contributions: Provides long run series, high-lights importance of developments within the top, high-lights importance of all sources of income (in particular importance of capital incomes in the very top)

WHAT ABOUT GENDER DIMENSIONS IN THE TOP INCOME LITERATURE? A lot of work in labour economics focus on the gender dimensions in the top of the wage distribution (e.g. glass ceiling following Albrecht Bj rklund and Vroman, 2003, recent work by Guvenen, Kaplan and Song, 2014, on top earners in the US, 1981- 2014) Much less has been done in the top income literature Partly for good reasons. Many countries tax married couples jointly and most countries did so historically. Main exception is Atkinson, Casarico and Voitchovsky (2016) who studies shares of women in top groups and their income composition in 8 countries with individual taxation.

WHAT WE DO PART 1: Using detailed panel data (starting in the 1970s; individual taxation was introduced in 1971) we study thing such as: Development of the share of women in top groups Income composition Mobility of women in top groups Family associations, and characteristics of top income women PART 2: Using tabulated data from various historical sources we study long run developments Focusing on share of women in top groups But also relation income/wealth, age, occupation etc.

WOMEN IN TOP INCOMES IN SWEDEN 1974-2013 MAIN TRENDS IN REPRESENTATION, INCOME COMPOSITION, MOBILITY, AND CHARACTERISTICS

SHARE WOMEN IN TOP GROUPS 1974-2013 Strong continued growth of women in the top groups Still far from equal, fewer women the higher up we go Realized capital gains (RCG) make a difference in the top 1

MAINRESULTSONINCOME COMPOSITIONANDMOBILITY Women have a larger share of capital incomes (but difference is shrinking over time) Relative to men the share of capital rich women has been falling and the share of working rich women have increased In terms of mobility women are more likely to leave the top group (both from one year to the next and over longer periods)

DIFFERENCESININCOMECOMPOSITIONBETWEEN MENANDWOMEN (RANKEDEXCLUDING RCG, THENADDING RCG)

RELATIVE DEVELOPMENTOFDIFFERENT TYPES INTHETOP

CHARACTERISTICSOFTOPINCOMEWOMEN? No large differences (relative men) in terms of age, education etc. Women more likely to be single, divorced or widowed (in top 1). BUT, the really big difference is in terms of who top income women are married to

PARTNERS OF THOSE IN THE TOP 1 GROUP Share with partner in top 10 Fewer top women have top 1 men, but overwhelming majority have top 10 partner. Decreasing number of top 1 men have low income partner, but most are still below top 10.

WOMENIN TOPINCOMESIN SWEDEN 1974-2013 INTERNATIONALOUTLOOK

SIMILAR DEVELOPMENTS IN OTHER COUNTRIES If anything the Nordic countries have fewer women in the top groups (!)

WOMENIN TOPINCOMES OVERTHE 20TH CENTURY SOMEPRELIMINARY RESULTS

SHAREWOMENINDIFFERENTTOPGROUPS U-SHAPEDOVERTIME, WOMENWITH CAPITALMORENUMEROUSINFIRSTHALFOF CENTURYCOMPAREDTO 1960S

WHAT DRIVES THE MAJOR TRENDS? Capital incomes much more important as a source of income before WW II. The top at this time consisted of a much larger share of capital rich women (mainly widows) comparing to the 1950s 1960s. As capital falls in importance, so does the share of women since there are virtually no high labour income women. Since the 1970s the share of working women has increased and they have gradually become a larger share of the top group. Still far from equal and potentially lagging behind other countries in terms of the top (even though these countries are less gender equal in other dimensions).

CONCLUSIONS Share of women has increased steadily since the 1970s, from about 15 to 30 percent in top 10, from about 8-18 in top 1. Women rely more on capital, and especially RCG, and are more likely to exit top groups from one period to the next, but over time women look increasingly similar to men in top groups with the exception of civil status, if they are married, and especially in terms of who they are married to. Most top women have top income partners while the reverse is true for men. Interesting long run trends still to be explored...

THANKSFORLISTNING! JESPER.ROINE@HHS.SE

EXTRASLIDES

RELATEDLITERATURE Top incomes: Piketty (2001/2003), Piketty and Saez (2003), Atkinson (2005), etc. Collected volumes, edited by Atkinson and Piketty (2007 and 2010). Overviews by e.g. Atkinson, Piketty and Saez (2011) and Roine and Waldenstr m (2015) Gender inequality: Overviews by e.g. Bertrand (2011), Azmat and Petrongolo (2014), Ponthieux and Meurs (2015), Blau and Katz (2016), Glass-ceiling and differences in executive compensation, e.g. Albrecht, Bj rklund and Vroman (2003) Bertrand, Goldin and Katz (2010), Smith, Smith and Verner (2013), Keloharju, Knupfer and T g (2016); Women and wealth over the long-run, Edlund and Kopczuk (2009) Close to our study (but focus on earnings) is Guvenen, Kaplan and Song (2014) Gender in top incomes: Closest to our work: Atkinson, Casarico, and Voitchovsky (2015) on female shares in aggregate data

GENDERDIMENSIONSINTHE DIFFERENT APPROACHES Not obvious how to think of gender in the disposable incomein households -literature Depends on assumptions about intra-household bargaining. Much gender work has been done in studies focusing on labour market outcomes Relatively straight forward to study labour force participation, gender pay gap, glass-ceiling, discrimination, etc.

RESULTSINTHE GENDERINEQUALITY LITERATURE (FOCUSINGWAGESANDON SWEDEN)

WHATDOWEKNOWABOUTTHEGENDER DIFFERENCESINEARNINGS? General labour market feature that women earn less than men. Difference persists (most of the time) even when controlling for anything that could explain the difference. Difference have, in general, decreased over time but remain, especially in the top of the earnings distribution ( glass-ceiling result)

CROSS-SECTORALWAGEGAPIN SWEDEN, 2015 Women s wage (full time equivalent) as percent of men s Male dominated sector Female dominated sector Gender balanced sector Average wage

EVIDENCEOFA DECLININGGLASS- CEILINGIN SWEDEN Albrecht, Bj rklund, Vroman JOLE, 2003 Albrecht, Skogman-Thoursie, Vroman , 2015

GENDERINEQUALITYWITHRESPECTTO WEALTH (ASOURCEOFCAPITALINCOME) Focus is on analysing wealth differences between men and women. But the lack of individual data makes even that difficult. Atkinson (1971), Edlund & Kopczuk (2009) , Ruel & Hauser (2013))

WOMENINTOPINCOMESBASIC FINDINGSFROMOTHERCOUNTRIES Closest to our paper is Atkinson, Casarico, and Voitchovsky (2015). Basic findings on women s share in top income groups

TOPINCOMELITERATURE BASICMETHODOLOGY Key question: What share of total income (pre taxes and transfers) goes to the top 10, 5, 1, 0,1 etc. per cent of all potential income earners Reference total for income (all personal income not just what is taxed) Reference total for the population (not just those who file taxes) Data on top incomes from tax statistics (typically available over roughly the past 100 years) Changes in who and what is taxed => many adjustments to raw data (and many robustness checks)

MAINFINDING 1: TOP 10 % INCOME SHAREDROPSUNTIL AROUND 1980 THENDIVERSEDEVELOPMENT (HEREEXEMPLIFIEDBY THE US, FRANCEAND SWEDEN) Top? 10? income? share? 1900-2013? 50? 40? 30? Average? OECD? Sweden? 20? Sweden? (incl? realized? capital? gains)? France? 10? USA? USA? (incl? realized? capital? gains)? 0? 1900? 1910? 1920? 1930? 1940? 1950? 1960? 1970? 1980? 1990? 2000? 2010?

MAINFINDING 2: MUCHDRIVENBY WHATHAPPENSTOTOP 1 % INCOMESHARE (AND RCGSMATTERSALOTIN SWEDEN) Top? 1? income? share? 1900-2013? 30? Sweden? Sweden? (incl? realized? capital? gains)? 25? France? USA? 20? USA? (incl? realized? capital? gains)? 15? 10? 5? 0? 1900? 1910? 1920? 1930? 1940? 1950? 1960? 1970? 1980? 1990? 2000? 2010?

COMBINING 1 AND 2: TOP 1O % DROP DRIVEN MAINLY BY TOP 1 %, LOWER HALF (P90-95) OF TOP 10 % RELATIVELY STABLE Inkomstandel? f r? P99-100? j mf rt? med? P90-95? ver? 1900-talet? Income share for P99-100 compared with P90-95 30? 25? 20? Frankrike? P99-100? Frankrike? P90-95? %? 15? Sverige? P99-100? Sverige? P90-95? USA? P99-100? USA? P90-95? 10? 5? 0? 1909? 1915? 1918? 1921? 1927? 1939? 1945? 1948? 1951? 1957? 1969? 1975? 1978? 1981? 1987? 1999? 2005? 2008? 2011? 1900? 1903? 1906? 1912? 1924? 1930? 1933? 1936? 1942? 1954? 1960? 1963? 1966? 1972? 1984? 1990? 1993? 1996? 2002?

MAINFINDINGSSTUDYINGTOPINCOMES IN SWEDEN 1. U-shaped development of top 10 income share over the 20th Century Increase since the 1980s is significant in pecentage terms (but starts from comparatively low level. In levels Sweden is still relatively equal) 2. Much driven by what happens in top 1 group; so studying the very top is not a fringe phenomenon) 3. Capital incomes, (and especially realized capital gains after 1980s), make a big difference. so income from all sources is important to capture what is driving inequality

DIFFERENCESININCOMECOMPOSITIONBETWEEN MENANDWOMEN (EXCLUDING RCG)

DIFFERENCESININCOMECOMPOSITIONBETWEEN MENANDWOMEN (INCLUDING RCG)

DIFFERENCESININCOMECOMPOSITIONBETWEEN MENANDWOMEN (RANKEDEXCLUDING RCG, THENADDING RCG)

DIFFERENCESIN MOBILITYBETWEENMENAND WOMEN (RANKEDEXCLUDING RCG)

WHATSTHECOMPOSITIONOFDIFFERENT TYPES INTHETOP? Define working-rich and capital-rich individuals as those with >2/3 of total income from either source. What has happened to the relative importance of these types in the top?

RELATIVE DEVELOPMENTOFDIFFERENT TYPES INTHETOP

DIFFERENCESINAGE, EDUCATIONETC.? Education composition Age composition More old women when including RCG, but no big differences between men and women Women slightly more educated (as in the population) but no big differences

DIFFERENCESIN MARITALSTATUS? Women much more likely to be non-married. Relatively stable over time incl RCG, but falling share widows when excl RCG. Possibly matches the age pattern, Widows make one off sales when partner dies. For men increasing number single/divorced in top

WOMENIN TOPINCOMESIN SWEDEN 1974-2013 INCOMEDEFINITIONSAND DATA

BASIC DATAANDDEFINITIONS LINDA-data base, yearly panel of 3,35% representative sample (today about 300,000 individuals) of Swedish population for the period 1968-today. Income data from the same source (tax registers) as in top income papers on Sweden. Also Roine and Waldenstr m (2012) show that LINDA data yields very similar results for top shares as using full population. Focus on Total Income (gross income before taxes and transfers) by source of income In Sweden four sources after 1991; Total income = Earnings + Capital income + Business income + Realized Capital Gains Before 1991 Capital income was subdivided into interest and dividends, private rental income and special rental income. These are just added together to form Capital income for the period before 1991.

WHYTHETIMEPERIOD 1974-2013? Fully individual taxation of income introduced in 1971 (before that period of choice between individual and family based from 1965 and joint taxation (separate filing but jointly taxed), and before 1954 married womens income automatically added to husband s) In 1974 parental leave replaces maternity leave; major political push to individualize economic rights (rather than having hen family based). Major tax reform in 1974 made all social benefits (unemployment insurance, social security transfers, state pensions etc.) become taxable and thereby part of earnings. Consequently, starting in 1974 we have individually based income data and a homogenous total income concept.

SHAREOFWOMENINTHELABOURFORCE 1974- 2013 TAKE-OFFHAPPENSINTHE 1970S

DESCRIPTIVESOFTOPGROUPTHRESHOLDS