Volscho Thomas - The Rise of The Super-Rich

458508
erican Sociological ReviewVolscho and Kelly

2012
ASRXXX10.1177/0003122412458508Am
American Sociological Review
The Rise of the Super- 77(5) 679699

American Sociological
Association 2012
Rich: Power Resources, Taxes, DOI: 10.1177/0003122412458508
http://asr.sagepub.com
Financial Markets, and the

Dynamics of the Top 1 Percent,
1949 to 2008
Thomas W. Volschoa and Nathan J. Kellyb
Abstract
The income share of the super-rich in the United States has grown rapidly since the early
1980s after a period of postwar stability. What factors drove this change? In this study, we
investigate the institutional, policy, and economic shifts that may explain rising income
concentration. We use single-equation error correction models to estimate the long- and short-
run effects of politics, policy, and economic factors on pretax top income shares between 1949
and 2008. We find that the rise of the super-rich is the result of rightward-shifts in Congress,
the decline of labor unions, lower tax rates on high incomes, increased trade openness, and
asset bubbles in stock and real estate markets.
Keywords
economy, inequalities, neo-liberalism, social class
The distribution of income in the United cent increase. This is a dramatic change that
States has grown markedly unequal in the puts income concentration on par with levels
past 30 years (Atkinson, Piketty, and Saez not seen since the late 1920s. Referring to
2011; Harrison and Bluestone 1988; McCall conspicuous patterns in the data, authors of
and Percheski 2010; Morris and Western one of Citigroups Plutonomy memos
1999; Nielsen and Alderson 1997). Based on remarked that such data, show that the rich
data from Piketty and Saez (2003), between in the U.S. continue to be in great shape
1913 and the end of World War II, the top 1 (Kapur, Macleod, and Singh 2006:3). Some
percent accrued between 11.3 and 23.9 per-
cent of income (see Figure 1).1 Aside from a
a
spike in the late 1920s, the general trajectory City University of New York-College of Staten
Island
is downward from 1913 to the mid-1970s. b
University of Tennessee
The income share held by the top 1 percent
fell from a high of nearly 24 percent (in 1928) Corresponding Author:
to its lowest point of 8.9 percent (1975 to Thomas W. Volscho, Department of Sociology,
1976), a decline of 63 percent. This pattern Anthropology, and Social Work, City University
of New York, College of Staten Island, 2800
dramatically reversed after 1980, with income Victory Boulevard, Building 4S-203, Staten
concentration rising from just over 10 percent Island, NY 10314
in 1981 to 23.5 percent by 2007, a 135 per- E-mail: Thomas.Volscho@csi.cuny.edu
Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

680 American Sociological Review 77(5)
30
25
Percent of Income
20
15
10
0
1900 1920 1940 1960 1980 2000 2020
Year
Figure 1. The Top 1 Percents Share of Income, 1913 to 2008
have characterized this rising inequality as do not necessarily translate into conclusions
the beginning of a new Gilded Age (Bartels about income shares of the super-rich. But
2008; Hacker and Pierson 2010). answering this question is also important for
What accounts for the rise of the super- other reasons. First, concentration at the top
rich? In this study, we attempt to answer this of the income distribution is qualitatively dif-
question by drawing on an accumulating ferent from inequality in the middle and lower
body of research in sociology and political portions. Normative objections to income
science emphasizing politics and public pol- concentration at the top are much easier to
icy as important determinants of poverty and make than objections to general inequality
inequality (Bradley et al. 2003; Brady 2009; levels because the latter are driven by changes
Brady and Leicht 2008; Hacker and Pierson just above and below the median. Second,
2010; Hibbs and Dennis 1988; Hicks 1999; much of the rise in general inequality in the
Kelly and Witko 2012; Kenworthy 2004; United States appears to be driven by changes
Moller, Alderson, and Nielsen 2009; Moller at the very top (Atkinson et al. 2011). The
et al. 2003). Such studies suggest that power most important variation to analyze, then,
resources of labor and capital can affect the may be variation in income shares of the
distribution of income in a society (Korpi super-rich.
1983; Kristal 2010; Stephens 1979). Further-
more, recent time-series research on the post-
war United States demonstrates that overall Explaining the Rise of
income inequality and redistribution can be the Super-Rich: Power
explained, in part, by changes in policy liber- Resources and Public
alism and left party power (Bartels 2008; Policy
Kelly 2005, 2009). But do politics and policy
influence income shares of the super-rich in Our analysis of top income shares is rooted in
the United States? a substantial body of work in the social sci-
This remains an open question because ences devoted to explaining economic
patterns of general inequality and the share of inequality. We emphasize the connection
income held by the top do not track perfectly between politics, policy, and income inequal-
over time (Hacker and Pierson 2010). Exist- ity (Bartels 2008; Hacker and Pierson 2010;
ing analyses of public policy, partisan poli- Irvin 2008; Krugman 1997; Phillips 2002).
tics, and general levels of income inequality Given this focus, the primary theoretical

Volscho and Kelly 681
foundation of our work is Power Resources stage is driven by markets and produces vary-
Theory (henceforth PRT), one of the most ing degrees of inequality. As discussed ear-
influential social science theories linking lier, the central PRT factor influencing this
class-based political power with income dis- stage of the distributional process is labor
tribution. PRT was developed to explain dif- unions. As labor union membership increases,
ferences in welfare states (Hicks 1999; Huber unions gain greater bargaining power, and the
and Stephens 2001; Stephens 1979). The market distribution of income becomes more
general insight is that welfare states are more equal (Bradley et al. 2003; Freeman 1984;
sophisticated, egalitarian, and advanced in Gustafsson and Johansson 1999; Kelly 2009;
countries where left parties and labor unions Moller et al. 2003). The second stage of the
are stronger. Left parties and union strength distributional process relates to partisan poli-
are important in PRT because these resources tics. After the market has produced a particu-
can alter the a priori asymmetrical bargaining lar level of economic inequality, the state
power of labor and capital. PRTs main becomes involved in the process through a
assumption is that working and middle classes variety of redistributive programs that operate
have different distributional preferences than through taxes and benefits. The expectation
do owners of capital, with lower classes hav- of PRT is that left party government strength
ing more egalitarian distributional prefer- increases redistribution, and recent studies
ences than do those at the top. suggest as much (Bradley et al. 2003; Brady
The PRT model emphasizes two major and Leicht 2008; Kelly 2009).
spheres within which the working and middle We add to this literature in four ways. The
classes can organize to achieve progressive first two contributions are theoretical. Our
redistribution: politics and the market. These analysis focuses on the first stage of the dis-
groups can affect government policy through tributional process (the market), which is
election outcomes by supporting left parties where unions, rather than partisan politics,
that pursue redistribution and by affiliating affect distributional outcomes according to
with labor unions to affect the market distri- existing presentations of PRT. We suggest
bution of income. that the effect of political power resources is
Scholars have traditionally applied PRT to not restricted exclusively to the realm of
explain levels of welfare generosity in redistribution. Our more extensive view of
advanced capitalist democracies (Esping- political power resources suggests that politi-
Andersen 1990; Hicks 1999; Hicks and cal factors should not only influence redistri-
Swank 1992; Huber, Ragin, and Stephens bution through taxes and transfers but also
1993). In recent years, social scientists have alter economic outcomes produced in the
used PRT to explain redistribution and eco- market. Following Kelly (2009), we call this
nomic inequality directly (Bradley et al. 2003; market conditioning. It is not controversial to
Brady 2009; Brady and Sosnaud 2010; Kelly suggest that government actions shape deci-
2005, 2009; Moller et al. 2003). That is, sions made in markets. The recent meltdown
analyses have moved away from studying the of the financial sector, which had far-reaching
size and generosity of welfare states and consequences for a variety of market out-
toward the actual distributional outcomes comes, was undoubtedly influenced by gov-
achieved by welfare states. Bradley and col- ernment regulation and policy drift, that is,
leagues (2003) and Moller and colleagues government created no new regulations to
(2003) found that unions reduce market- cope with complex financial innovations
generated income inequality and poverty and (Hacker and Pierson 2010; Johnson and Kwak
that left parties increase redistribution (the 2010). Firms decisions in hiring and com-
direct effect of explicit taxes and transfers). pensating their employees are influenced by
In its original development, PRT suggested many government activitiesfrom payroll
a two-stage distributional process. The first taxes, to government contracts, tax credits,

workplace safety rules, and environmental Our second contribution relates to examin-
regulation. Public education changes working the distributional impact of specific poli-
force skills, which undoubtedly also affects cies falling under the category of market
wages. Clearly, government actions affect conditioning. Policies are not power resources,
markets. but they do flow from constellations of politi-
We ask whether market-conditioning cal power that are central to PRT. Existing
activities affect income concentration. Exam- applications of PRT to the U.S. context pay
ining the effect of political power resources limited attention to the idea that political
on the distribution of income produced by the power resources, in part, generate policy.
market is our first theoretical contribution. Although scholars view policy as a result of
We argue that political dynamics affect distri- power resources and a cause of inequality,
butional outcomes by changing the distribu- research has not assessed specific policies.
tion of market income in predictable ways Kelly (2005) found that market-generated
anticipated by PRT. Just as political power inequality decreases in response to a liberal
resources of the middle and lower classes shift in policy, but his focus is general trends
reduce overall levels of inequality through in the ideological direction of aggregate pol-
redistribution under current conceptions of icy rather than effects of specific market-
PRT, we hypothesize that these power conditioning policies. We make an initial
resources reduce the amount of inequality effort to examine distributional consequences
through market conditioning. If lower-class of specific policies by analyzing tax and
political power resources are mobilized to interest rates.
reduce inequality, and government action has Of course, numerous policies might affect
the potential to change the income distribu- the market distribution of income, and we are
tion produced by the market, examining the limiting ourselves to a very small subset for
connection between political power resources this initial effort. Moreover, it is somewhat
and market inequality is a straightforward surprising to suggest that tax rates would
extension of PRT. affect market inequality, because taxes more
In terms of specific mechanisms, left party obviously generate explicit redistribution
occupancy may influence the distribution than market conditioning. In part, we start
accruing to the super-rich via administrative here because it is a tough test for the idea of
office-holding (Brady and Leicht 2008). A market conditioning and because tax rates
greater share of Democrats in Congress can have the potential to affect both individual
limit the rent-generating effects of property and firm behavior that subsequently affect the
rights benefiting big business and limit other market distribution of income.
governance structures used to generate prof- We specifically examine top marginal
its. Administrators appointed by Democrats income tax rates and capital gains rates. In the
may more stringently enforce labor laws such postwar era, top tax rates have fluctuated dra-
as the minimum wage, union election rules, matically, ranging from a high of 91 to a low
overtime pay, and other forms of labor-related of 28 percent. The Economic Tax Recovery
compensation. Additionally, the overall ideo- Act of 1981 is one of the more significant
logical influence of Democratic governments turning points in U.S. taxation because it
may operate via the ability to affect policy severely cut top marginal tax rates (Slemrod
disputes and reflect a lower level of collusion 1990). Importantly, the profits of most corpo-
between political leaders and big business rate businesses (roughly 90 percent of corpo-
(Brady and Leicht 2008).2 Our expectation, rations) are taxed as individual rather than
then, is that Democrats presence in national corporate income (Fisher 2009:30), underlin-
government (political power resources) and ing the importance of changes in income tax
union strength (market power resources) rates. Top rates on capital gains income have
reduce market inequality. also fluctuated substantially in the postwar

era, from a high of 39 percent between 1976 earning capacity of those lower in the income
and 1978, to a low of 15 percent currently. distribution.
Tax progressivity is typically understood in In addition to tax rates, we examine inter-
an explicit redistribution context, where a est rates potential effects on top income
more progressive system reduces inequality in shares. Interest rates are set by the Federal
after-tax income, but recall that we are exam- Reserve, but Republicans favor monetary
ining market-generated (pretax) inequality in policy that controls inflation while Democrats
this article. How would these tax rates influ- are concerned about keeping lower unem-
ence market income concentration? One ployment (Alesina and Rosenthal 1995). Our
possibility is a straightforward behavioral expectations regarding the effect of interest
response of individual earners: higher rates of rates on top income shares are somewhat
taxation may generate disincentives to accrue mixed. The most straightforward prediction is
more earned income and deter investment, that higher interest rates will increase top
thereby lowering capital income. Although a incomes because the rich have large amounts
standard argument based in microeconomic of savings that earn interest (Galbraith 1998;
reasoning, there is certainly debate on this Palley 1998). However, the effect may not be
point (Atkinson 2004; Auerbach and Hassett quite so simple. Because the rich engage in
1990; Roine, Vlachos, and Woldenstrom high-volume securities trading (that contrib-
2009). Changes in tax rates may also affect utes heavily to capital gains income) and
how income is reported, corporations legal interest rate hikes depress bond prices (Can-
charters, and how compensation is timed terbery 2002; Henwood 1997), an interest-
(Slemrod 1992). Pretax income concentration rate increase may lower the top income share
may respond to tax rates due to incentives for by depressing capital gains that accrue
high-income units to shift income to a time through bond trading. In any case, interest
when it will be taxed at a lower effective rate rates are closely watched by financial markets
or to disperse income across sources to mini- and therefore are a potentially salient policy
mize taxable income. This is especially impor- that may affect the top 1 percent.
tant when it comes to capital gains income Our remaining contributions are primarily
(Minarik 1981; Piketty and Saez 2003; Slem- empirical, but we will see as the analysis
rod 1996), but it could also play a role in top unfolds that they also have implications for
marginal income tax rates. Higher income tax theory testing. The first empirical contribu-
rates might generate demands by highly paid tion is the use of U.S. time-series data. Most
CEOs for deferred compensation or for com- studies applying PRT to the study of distribu-
pensation in the form of stock options. Both tional outcomes utilize cross-national data,
forms of compensation make it easier to shift focusing especially on Europe. Only a few
income temporally to take income during peri- studies have used U.S. time-series data to
ods of lower taxation; to the extent that this study income inequality.4 Kellys (2005,
happens, inequality is likely to rise as top tax 2009) study of market-generated income
rates decline.3 Although executive pay is ulti- inequality and post-tax/transfer redistribution
mately determined by firms, top executives stands as one of the sole applications of PRT
have the ability to negotiate individual com- to the United States.
pensation packages based on personal incen- The second empirical contribution is an
tives. Finally, taxes collect revenue that fund analysis of the top 1 percents pretax pretrans-
government appropriations, and to the extent fer income share. Our focus on the super-rich
that higher tax rates spur spending on human departs from existing power resources studies
capital formation (education and health) and that examine redistribution or the amount of
other programs that help lower- and middle- social spending (Huber et al. 1993; Kelly
class individuals compete economically, tax 2004). Research that explicitly examines
rates may reduce top shares by building the political influences on income inequality gen-

erally analyzes broad distributional outcomes been accrued from the financial sector (Foster
measured by statistics like the Gini coeffi- and Holleman 2010; Henwood 1997; Kaplan
cient (Bradley et al. 2003; Kelly 2005; Ken- and Rauh 2010), supplanting the once domi-
worthy and Pontusson 2006). Even these nant oil and gas sector. The rapid rise in
studies are typically based on data that inad- finance, insurance, and real estate profits sug-
equately capture top incomes because they gests successful rent-extraction from the non-
rely on Census income data that top-code financial sector (Bakir and Campbell 2010;
high incomes, thereby ignoring variation at Tomaskovic-Devey and Lin 2011).
the very top of the distribution. Therefore, we The stock markets performance in the lat-
use the updated Piketty and Saez (2003) top 1 ter half of the 1990s likely contributed to top
percent share based on income tax returns income concentration because of the rapid
that better capture high incomes.5 increase in stock prices and dividend payouts
In summary, PRT predicts that top income (Baker 2009; Shiller 2005). Buying stock at a
shares should decrease in response to low price and selling it at a higher price
increases in lower-class power resources. We enhances capital gains income. In the 1980s,
thus predict that top income shares will rapid increases in stock prices coincided with
exhibit a negative adjustment in response to changes in managerial incentive structures.
Democratic presidential administrations, The shareholder value movement shifted
Democratic strength in Congress, and union executive compensation from a system in
membership. We also expect specific policies, which managers were paid salaries (to make
such as top tax rates, to put downward pres- decisions in the best long-term interest of
sure on top income shares. Other factors also their firms) to a system in which most execu-
deserve attention as determinants of the tive compensation is tied to short-term fluc-
income gap, and we discuss these factors in tuations in stock prices (DiPrete, Eirich, and
the next section. Pittinsky 2010; Dumenil and Levy 2004;
Fligstein 1990). Thus, short-term fluctuations
in stock prices likely have a greater impact on
Financial Markets managerial decisions than long-term eco-
and Macroeconomic nomic health. As noted earlier, top tax rates
Explanations of on individual incomes and capital gains have
Inequality been cut. These changes in public policy
coincided with, and likely increased incen-
Our main focus is the effect of politics and tives for, a shift toward compensation via
policy on the concentration of income. Clearly, stock options (Fligstein 1990).7
however, numerous other explanations should After the precipitous drop in stock prices
be considered. Probably the most important is in 2000, the next bubble occurred in real
the financialization of the economy and the estate. As a result, the financial sector saw a
performance of financial markets (Epstein and vast increase in the mortgage market and
Jayadev 2005; Foster and Magdoff 2009; securitization of home mortgages into mort-
Henwood 1997; Krippner 2011). This is likely gage backed securities that could be bundled
important for explaining the rise of the super- and traded as derivatives (Johnson and Kwak
rich because of the shift of income and profits 2010; Tomaskovic-Devey and Lin 2011).
toward the financial sector (Dumenil and Levy Given the concentration of stock and bond
2004; Tomaskovic-Devey and Lin 2011) and ownership among the wealthiest income
the fact that ownership of stocks and other units, prices of stocks and homes should dis-
securities are highly concentrated among top proportionately benefit the richest 1 percent.
wealth holders (Kennickell 2009).6 Looking at Aside from the stunning shifts in financial
various rankings of top income earners, much markets, other macroeconomic shifts such as
of the new money since the early 1980s has trade openness, economic growth, and busi-

ness cycles may also have consequences for noted earlier, comes from Piketty and Saez
top income shares. (2003), updated to 2008. The measure is the
A recurrent question is whether and how share of pretax income (including capital
trade is associated with an increase in top gains) accruing to the top 1 percent of tax
incomes. Roine and colleagues (2009) discuss units. Tax units refer to either a married couple
the Heckscher-Ohlin theory of trade, suggest- living together with dependents or a single
ing that trade openness of capital-rich countries adult with dependents (Piketty and Saez
should be associated with higher top incomes, 2003). The income definition is individually
but they find some evidence that increasing pretax, but it is net of employer-paid payroll
trade openness reduces top income shares in and corporate income taxes. Income sources
OECD countries. Alderson and Nielsens include wages and salaries, small business and
(2002) study of 16 OECD countries found that farm income, partnership and fiduciary
imports from less developed countries and income, interest, rents, dividends, royalties,
investment outflows explained the longitudinal capital gains, and other miscellaneous sources.
rise in income inequality. Trade openness may
weaken workers bargaining power and wages Political power resources. We include
because of increasing labor competition (Tonel- two measures related to political power
son 2000; Wood 1994). In the United States, resourcespresidential and congressional par-
trade openness may benefit the rich because of tisan power. The presidential measure is a
the abundance of capital and the propensity to dummy variable taking on values of 1 in years
export capital-intensive goods and import with Democratic presidents and 0 when Repub-
labor-intensive goods, thereby weakening licans hold the presidency. The measure of
labors bargaining power over wages. We thus Democratic congressional power is the per-
predict that top income shares should respond centage of seats held by Democrats. In addition,
positively to increased trade openness. we include a control for divided government to
Economic growth and business cycles have account for the idea that partisan effects should
been linked to inequality. The Kuznets curve be estimated while holding unified versus
and the Great U-Turn are influential theories divided government constant. We coded this
of growth and inequality. Taken together, the variable 0 when the House, Senate, and presi-
fundamental argument is that at moderate lev- dent were all of the same party and 1 otherwise.
els of development, growth increases equality, Although this is not a traditional power
while at low and very high levels of develop- resources variable, it is an important control
ment, growth induces inequality (Harrison variable given the goal of obtaining accurate
and Bluestone 1988; Nielsen 1994, 1997; coefficient estimates of the two primary politi-
Nielsen and Alderson 1995, 1997; Roine et al. cal power resources variables. Information
2009). In addition to economic growth, ana- used to create these variables is available in
lysts have argued that higher unemployment various years of the U.S. Census Bureaus Sta-
rates may increase inequality (Blank and tistical Abstract of the United States.
Blinder 1986; Blinder and Esaki 1978; Jantti
1994). We expect that increases in unemploy- Market power resources. As an indicator
ment, which represent a cyclical downturn in of market power resources, we include a mea-
the economy, are associated with greater sure of union density (the percentage of private
income concentration because top earners are sector workers in labor unions) that comes
less likely to be affected by unemployment. from Hirsch and Macpherson (2009).8 As is
well-known, union membership peaked in the
1950s and declined from a high of 35 percent
Data and Methods to just over 20 percent by the late 1970s. The
We measured all variables on an annual basis steepest decline occurred in the early through
from 1949 to 2008. The dependent variable, as mid-1980s with a steady but less rapid decline

into the 2000s. By 2008, private sector union Estimation Strategy

membership was 7.6 percent.
In time-series analysis it is essential to iden-
Public policy. We include three measures tify whether each series is stationary or non-
of policy. The first is the top marginal tax rate. stationary (also known as integrated or unit
These data come from the Urban Institute and root). This step is needed because of the well-
the Brookings Institutions Tax Policy Center.9 known problem in which a regression of one
The second policy measure is the top capital nonstationary series on another nonstationary
gains tax rate.10 Third, we consider the short- series can produce the spurious inference that
term interest rate. The federal funds rate is the two are related (Granger and Newbold
available back to 1955, but our analysis begins 1974). We conducted a series of tests to deter-
in 1949. Therefore, we use the nominal mine whether each of our variables contain a
3-month T-bill rate (the interest rate on trea- unit root, and results are reported in an online
sury bills that mature in three months), which supplement (http://asr.sagepub.com/supple-
goes back to the beginning of our series (U.S. mental).13 These findings indicate that most
Executive Office of the President 2011, Table variables in our analysis, including the depen-
B-73). For overlapping years, this series mir- dent variable, contain a unit root. This is not
rors other interest rates such as the federal surprising; the presence of a unit root sug-
funds and prime rate. gests a series with permanent memory such
that any disturbance to the series persists
Financial markets and economic permanently into the future and most of the
conditions. Finally, we include several mea- variables in our analysis are expected to
sures related to financial markets and behave in this way. Shocks to inequality,
economic conditions. Our measure of stock unemployment, or tax policy do not naturally
market valuation is Standard & Poors 500 diminish over time.14 In contrast, a few vari-
composite stock market index (U.S. Execu- ablesincluding Democratic president,
tive Office of the President 2011, Tables B-95 Democratic share in Congress, and divided
and B-96). We deflated the index to 2008 governmentwere either stationary or pro-
prices using the CPI-U series and rescaled it duced mixed results that could not lead to a
so that one unit in our measure represents a firm identification of the series as either sta-
10-point change in the real S&P 500. Our tionary or unit root.
measure of the housing market is Shillers In early work dealing with unit-root pro-
(2005) real historical home price index.11 As cesses, the solution was to transform any
with the S&P 500, we rescaled the Shiller nonstationary variables into stationary data by
index so that one unit in our measure repre- calculating the first difference and analyzing
sents a 10-point change in the index. change in the variable of interest rather than
We use variables standard in the literature its level (Granger and Newbold 1974). This
to measure other economic factors. Trade transformation converts a unit-root variable in
openness is defined as imports and exports as levels to a stationary variable in differences.
a percentage of GDP (Roine et al. 2009; U.S. Once unit-root variables are transformed into
Bureau of Economic Analysis 2011: Table stationary, differenced variables, regression
1.1.5). The next economic variable is the analysis can proceed without concern about
unemployment rate.12 We also include the log spurious regression due to nonstationary data.
of real GDP (in 2005 inflation-corrected dol- In our analysis, this strategy would simply
lars) as a measure of general economic condi- entail differencing each variable suspected of
tions (U.S. Bureau of Economic Analysis containing a unit root and then using the dif-
2011:Table 1.1.6). ferenced version of the variable in the analysis,

along with other variables that have been income share capitalizes on this flexibility by
identified as stationary in their raw form. estimating ECMs with a mix of stationary and
This strategy removes concerns about spu- nonstationary data (Kristal 2010).
rious regression but restricts the type of rela- As this discussion shows, ECMs have three
tionship that can be uncovered to those in characteristics that make them particularly
which the effect of an explanatory variable is appropriate in the context of this analysis.
constrained to a single point in time. Engle First, and most importantly, ECMs model
and Granger (1987) point out a different type long-run equilibrium relationships that likely
of relationship that can exist between two exist between the variables, although the
variables, in which one variable sets a target model also captures more immediate effects.
level to which the other adjusts over time. Second, because several variables contain a
Ignoring the potential for such an equilibrium unit root, ECMs prevent the spurious regres-
relationship in the context of our analysis sion problem that can arise when analyzing
would be problematic because it is unlikely nonstationary data (Engle and Granger 1987).
that effects of most, if not all, of our explana- In essence, ECMs test whether cointegration
tory variables are constrained to a single point exists between two unit-root variables. Third,
in time. More likely, tax policy and power ECMs accommodate stationary and integrated
resources variables maintain a long-run equi- variables, which is useful because our analysis
librium relationship with inequality. We thus has a mix of both data types. In summary, the
need to utilize a method that captures this ECM is a very general model that is easy to
long-run relationship, and an analysis in dif- implement and estimate, does not impose
ferences fails in this respect. assumptions about cointegration, and can be
The type of relationship we expect to applied to both stationary and nonstationary
observe in models of top income shares is data (Banerjee et al. 1993; De Boef and Gra-
known as an error correction relationship nato 1999; De Boef and Keele 2008).16
deviations from the long-run relationship In this study we estimate single-equation
(errors) are eliminated over time through an ECMs, which are among the most flexible
adjustment process (error correction). Engle models of the error correction process.17 A
and Granger (1987) originally (and later Ban- bivariate version can be represented as follows:
nerjee and colleagues [1993] and Enders
[2011]) discuss error correction models Yt = 0 + 1Yt1 + 1Xti + 2Xt1 + t
(ECMs) in the context of cointegrationthat
is, two unit-root variables that maintain a This specification allows for a test of both
long-run error correction relationship. Recent short- and long-run effects. The immediate
discussions of error correction make it clear, short-term effect of X is captured by 1. The
however, that cointegration is a special case of error correction rate is captured by 1 and
error correction and cointegration is not indicates the rate at which discrepancies
required. When evidence of error correction is between Y and X are recalibrated to their
found in an analysis of integrated variables, equilibrium state. Importantly, if the error
this demonstrates a cointegrating relationship. correction rate is not significant, it indicates
But Bannerjee and colleagues (1993) also that a long-run relationship does not exist (for
point out that error correction can occur in integrated variables this is a cointegration
stationary data, and De Boef and Keele test). An increase in X can have an immediate
(2008:199) clarify this point, concluding that impact on Y and a long-run impact that is
the ECM is useful for stationary and inte- distributed over time (dictated by the error
grated data alike, [and] analysts need not enter correction rate) such that Y readjusts to the
debates about unit roots and cointegration to long-run equilibrium between X and Y. The
discuss long-run equilibria and rates of equili- total long-run impact, known as the long-run
bration.15 One recent study of laborcapital multiplier effect, is calculated by 2/1.

Results (although we will see that the short-term effect

does not hold in later models). When capital
Table 1 presents five specifications of the top gains rates are high, the super-rich are more
1 percent income share. Each specification likely to strategically defer or altogether avoid
considers short- and long-run effects. The capital gains income. Finally, higher treasury
first two models focus on two sets of explana- bill rates decrease inequality. This may at first
tory variables separately. Because the primary seem counterintuitive, but as rates increase,
purpose of these models is to serve as a base- the price of outstanding notes decreases (Can-
line for comparison with later models, we terbery 2002). This means that a rise in rates
briefly discuss the results before moving on would lead to lower returns for these invest-
to more fully specified models. ments in secondary markets. However, this
Model 1 presents a power resources model result does not hold in some later models and
of income concentration, including partisan- should be viewed with caution.
ship of the president and Congress and union Model 3 can be described as a politics and
strength. We include divided government as a policy model, including the significant effects
control. Our results are mostly supportive of from the first two models. A few aspects of
traditional power resources hypotheses. Dem- results from this model are particularly note-
ocratic strength in Congress and union worthy. First, primary results from the two
strength decrease the share of income held by portions of the model estimated separately in
the top 1 percent. The presidents party has no Models 1 and 2 still find support. Second,
statistically significant effect on top income however, we see some important differences
shares. This result is inconsistent with some between the results here and the two previous
recent analyses of income inequality gener- models. Note in particular that the size of the
ally (Bartels 2008; Kelly 2009). However, it effect of congressional partisanship declines.
echoes Kenworthys (2010) assessment that This suggests that a portion of the effect of
Democratic presidents have little to no effect this variable is absorbed by the specific pol-
on distributional outcomes at the very top of icy measures, which is consistent with the
the income distribution. In terms of top idea that congressional partisanship affects
incomes, it appears that neither partys presi- distributional outcomes via specific policy
dents have achieved differential outcomes. mechanisms. We do not include a comprehen-
But there is evidence that political and market sive set of market-conditioning policies, so a
power resources are associated with income direct effect of congressional partisanship
concentration. remains. In addition, several effects are
Model 2 focuses solely on effects of spe- remarkably stable between this model and the
cific policies. In this model, we see evidence two preliminary models. The effects of union
that policy matters. Specifically, the top mar- membership, the capital gains rate, and bond
ginal tax rate has the expected effect on rates remain about as large in Model 3 as in
income concentration in the long run, decreas- previous models. We also see that in terms of
ing inequality even when income shares are overall model fit, the combined politics and
measured prior to taxes and transfers. This policy model is an improvement over the ear-
effect is likely present not only because higher lier models. The adjusted R2 rises substan-
marginal tax rates deter income accumulation tially to .35 and the Bayesian Information
at the top of the income distribution, but also Criterion drops to about 193.
because higher tax rates provide funding for Models 4 and 5 represent more complete
programs such as education and health care models that account not only for politics and
that broaden economic opportunities among policy, but also for a variety of economic fac-
middle- and lower-income households. The tors. Model 4 includes the full complement of
capital gains tax rate also influences top economic factors, and Model 5 focuses on the
income shares in both the short and long term statistically significant variables from the pre-

Table 1. Models of Top 1 Percent Income Share, 1949 to 2008
Top 1% Share
(1) (2) (3) (4) (5)

Explanatory Variables OLS OLS OLS Prais Prais
Top 1% Sharet1 .363*** .362*** .506*** .731*** .648***
(.094) (.083) (.09) (.187) (.10)
Democratic Presidentt1 .138
(.563)
Democratic Presidentt1 .201
(.364)
% Congressional Democratt1 .050
(.043)
% Congressional Democratt1 .125** .072* .032 .052**
(.039) (.029) (.021) (.016)
Divided Governmentt1 .111
(.496)
Divided Governmentt1 .932* .441 .039 .368*
(.416) (.301) (.223) (.170)
% Union Membershipt .285
(.284)
Union Membershipt1 .113*** .108* .261* .277***
(.028) (.048) (.130) (.066)
Top Marginal Tax Ratet .041
(.037)
Top Marginal Tax Ratet1 .057*** .025 .031* .032**
(.013) (.020) (.014) (.011)
Cap. Gains Tax Ratet .142* .091 .020 .027
(.054) (.053) (.036) (.032)
Cap. Gains Tax Ratet1 .115*** .098** .097*** .064***
(.033) (.032) (.025) (.016)
3-Month Treasury Bill Ratet .154
(.105)
3-Month Treasury Bill Ratet1 .153* .182** .013 .006
(.067) (.065) (.049) (.039)
Trade Opennesst .270* .201*
(.119) (.098)
Trade Opennesst1 .117
(.092)
Unemployment Ratet .183
(.225)
Unemployment Ratet1 .218
(.153)
Log Real GDPt 16.639
(10.031)
Log Real GDPt1 5.002* 5.038***
(1.932) (1.436)
Real S&P 500 Composite Indext .057*** .063***
(.008) (.007)
Real S&P 500 Composite Indext1 .026* .033***
(.011) (.008)
Shiller Home Price Indext .104
(.155)
(continued)

Table 1. (continued)
Top 1% Share
(1) (2) (3) (4) (5)

Explanatory Variables OLS OLS OLS Prais Prais
Shiller Home Price Indext1 .311** .283***
(.105) (.071)
Constant 15.046*** 12.034*** 18.270*** 57.834** 58.994***
(3.830) (2.563) (3.195) (20.910) (14.272)
Adj. R2 .197 .269 .354 .761 .759
Rho .519 .532
Breusch-Godfrey Test, p-value .238 .242 .308
BIC 209.142 197.656 193.155 168.194 155.777
N 60 60 60 60 60
Note: Regression coefficients with standard errors in parentheses. The null hypothesis of the Breusch-
Godfrey test is no serial correlation. Prais is the Prais-Winsten (GLS) estimator.
*p < .05; **p < .01; ***p < .001 (two-tailed tests).
vious models. We focus our discussion of sive of any indirect effects that they have via
substantive effects on Model 5, which is economic variables, for example, via stock
clearly preferred to the other models reported market valuation (Davis 2009; Krippner
in the table based on the measures of model 2011). In this way, our final model and the
fit reported at the bottom of the table (Bayes- substantive effects reported below provide a
ian Information Criterion and Adjusted R2). conservative estimate of the effect of politics
With all the variables included, our model and policy on distributional outcomes. But
explains 76 percent of variance in top income even in the presence of controls for a host of
shares. In the final model, congressional par- policy and economic effects, partisan politics
tisanship, divided government, union mem- matters.
bership, top marginal tax rates, and the capital It is also worth noting that nearly all of the
gains rate all have the expected effects on effects we identify come via the error correc-
income concentration. With regard to the esti- tion component of the model, indicating the
mated effects of the economic factors, we see presence of a long-run equilibrium relation-
that trade openness, stock market valuation, ship. For variables with long-term effects
and home prices increase top income shares. distributed over time, we must take account
Economic growth decreases top income of the error correction rate to explain both the
shares. size and the temporal dynamics of the effect.
A careful examination of results in Table 1 When we compare the error correction rate to
indicates that the effect sizes of some of the the early models containing just political and
political and policy variables are lower in policy variables, we see a fairly slow error
Model 5 than in earlier models. We pointed correction rate of between .36 and .51. The
out that some of the decline in the effect of rate increases substantially in models that
congressional partisanship between Model 1 include economic effects and is .65 in the
and Model 3 is likely due to indirect effects final model. This means that disequilibria
that this variable has via specific policies. It is generated by a shift in an explanatory varia-
possible that this political variable has addi- ble are corrected at the rate of 65 percent per
tional indirect effects via other explanatory year. An error correction rate of 65 percent
variables. Our model captures only the direct corresponds to an effect that is distributed
effects of political and policy variables exclu- over a few years, with 99 percent of the total

% Congressional Democrats Union Membership

.45 .45
Reducon in Top Share %

.40 .40
.35 .35
.30 .30
.25 .25
.20 .20
.15 .15
.10 .10
.05 .05
.00 .00
0 1 2 3 4 0 1 2 3 4
Period Period
Top Marginal Tax Rate Capital Gains Rate

.45 .45

.40 .40
.35 .35
.30 .30
.25 .25
.20 .20
.15 .15
.10 .10
.05 .05
.00 .00
0 1 2 3 4 0 1 2 3 4
Period Period
S&P 500 Shiller

.45 .45
Increase in Top Share
Increase in Top Share
.40 .40
.35 .35
.30 .30
.25 .25
.20 .20
.15 .15
.10 .10
.05 .05
.00 .00
0 1 2 3 4 0 1 2 3 4
Period Period
Figure 2. Lag Distributions of a One-Unit Increase in Key Explanatory Variables
effect in place within four years of the initial ables and we focus only on the last model
shock. This suggests the adjustment rate is when calculating substantive effects. This
faster for economic variables than for politi- ensures a conservative estimate of the long-
cal and policy variables, which fits well with run effect of the political and policy variables.
intuition. More important is the overall size of Figure 2 illustrates how effects of a one-
the impact. We calculated the long-run multi- unit change in six of our central explanatory
plier effect by dividing the coefficient for the variables are distributed over time, based on
lagged level of each independent variable by results from Model 5 in Table 1. This figure
the error correction rate. Faster error correc- presents the lag distribution of the effect at a
tion rates yield smaller long-run multiplier particular point in time (bars), along with the
effects. Although results suggest that the cumulative effect of each variable at each
adjustment process may be slower for politi- time point (line). We see that the effect of
cal and policy variables than for the estimate each variable grows over time, rather than
in the final model, ECMs constrain the error being restricted to a single period. The annual
correction rate to be the same across all vari- effect on the top 1 percent in response to a

Part A: Total Effect

Divided Government .568
Union Membership .428
% Congressional Democrat .080
Top Marginal Tax Rate .050
Capital Gains Rate .099
Trade Openness (short) .201
Shiller Home Price Index .437
S&P 500 Valuaon .051
.000 .100 .200 .300 .400 .500 .600
Part B: Parally Standardized Effect
Divided Government .069

Union Membership 1.084
% Congressional Democrat .122
Top Marginal Tax Rate .274
Capital Gains Rate .152
Trade Openness (short) .324
Shiller Home Price Index .223
S&P 500 Valuaon .503
Log Real GDP 1.114
.000 .200 .400 .600 .800 1.000 1.200
Figure 3. Total Effects on Top Income Shares
Democratic congressional shift fades after similar in magnitude to the effect of partisan-
three years, whereas the initial union density ship in Congress. These numbers at first seem
effect is stronger and remains notable even quite small, but given that national income in
after a few years. Looking at the top marginal 2008 (estimated by Piketty and Saez) was
tax rate, the top 1 percent adjusts in the first more than $7.8 trillion, an increase of only 1
two years after the tax rate increase and percent in Democratic seat share ( just over
effects for the capital gains rate are similar. A five seats), would decrease the income of the
unit increase in the real S&P and Shiller indi- top 1 percent by nearly $6.6 billion. That
ces has a substantial long-run effect on the top equates to about $6,600 per tax unit in the top
1 percent. The stock market shock is felt 1 percent. These are not trivial effects and sug-
immediately whereas the effect of home gest that campaign contributions are a useful
prices begins the year after the shock. investment for the super-rich.
To give a sense of the relative impact of When we look at relative effects of the vari-
each variable, we also report the effect of a ables in our model (Figure 3, Part B), we see
standard deviation shift in each of the explana- that union strength stands out as an explanation
tory variables (Figure 3, Part A). Black bars for top income shares. Our evidence is consist-
indicate positive effects and gray bars indicate ent with the argument that unions are able to
negative effects. Just a one percentage point extract concessions from management that
increase in union membership is associated increase workers relative earnings. It is impor-
with more than a .40 point decline in the tant to note, however, that although union
income share of the super-rich. The impact of strength might directly decrease CEO and top
a one percentage point increase in the share of management compensation, top income shares
seats held by Democrats in Congress decreases are influenced by what happens lower in the
the top income share by about .08. The income distribution as well. If unions affect the
unstandardized effects of a percentage point economy such that income growth flows
increase in capital gains and income taxes are toward the middle class, then top income shares

would decline even without directly reducing Importantly, the link between politics and
executive compensation. Although there are inequality is not merely due to redistribution.
important political and policy effects on top In this article, we set aside the states redistri-
income shares, other factors matter even more. butional effects to focus on how political and
A standard deviation change in economic economic variables affect the distribution of
factors such as economic growth and stock income produced by the market. Our results
market valuation has substantial effects that are imply that the pretax pretransfer distribution
larger than the impact of politics and public of income is shaped by electoral outcomes
policy. The message here is that economic fac- and policy decisions. Democrats are more
tors are powerful predictors of pretax, pretrans- favorable than Republicans toward social
fer top income shares. However, politics and programs that redistribute income, but the
policy fit quite nicely alongside economic indi- parties also differ over what the economic
cators as predictors of income concentration. rules of the game should be. Based on our
Moreover, this effect of politics is not occurring analysis, Democrats appear to favor an eco-
through traditional redistributive mechanisms, nomic system that produces more egalitarian
because our measure of inequality is prior to outcomes even before any redistribution
taxes and transfers. Rather, we are observing a occurs. In essence, the market is not com-
substantial market-conditioning effect of gov- pletely beyond the influence of politics and
ernment on the incomes of the super-rich. policy, and it is not just in the realm of
explicit redistribution that political parties
produce divergent distributional outcomes.
Discussion and Political decisions in part make the market
Conclusions (Hacker and Pierson 2010:44).
We found evidence that congressional shifts to Our results also shed some light on which
the Republican Party, diminishing union aspects of politics matter for distributional
membership, lower top tax rates, and financial outcomes. Both Bartels (2008) and Kelly
asset bubbles played a strong role in the rise of (2009) emphasize the presidents role in deter-
the super-rich. From 1980 to 2008, these mea- mining the level of income inequality in the
sures saw major shifts, after relative stability United States. Other researchers such as
in Democratic dominance of Congress, union Hacker and Pierson (2010) place more empha-
membership, tax rates, and prices of stocks sis on congressional party control in promot-
and real estate during the postwar era of the ing top income shares. Kenworthy (2010)
late 1940s to the late 1970s. recently suggested that the impact of presiden-
These results have several important tial partisanship on distributional outcomes
implications. First, inequality is in part an has waned since the 1980s. Essentially, he
outcome of political contestation. A common argues that Democratic and Republican presi-
refrain holds that inequality has risen substan- dents differed dramatically in the distribu-
tially but is merely the result of natural mar- tional outcomes they achieved prior to 1980
ket forces that are in large part out of our but have produced fairly similar outcomes
control. By this logic, policy and partisan since. Our results suggest a stronger role for
politics are unimportant players in rising Congress than the president in the rise of the
inequality. But the evidence does not support top 1 percent. This points to the central role
this idea. Both specific policies and the parti- that Congress has in the legislative process.
san balance of Congress are associated with The president has limited ability to make the
distributional outcomes. Conservative shifts sort of legislative changes necessary to affect
in policy and Republican strength in Con- top shares without the support of Congress,
gress are associated with higher levels of making Congress the central actor here. The
inequality. Political outcomes have implica- politics of the labor market is also important.
tions for distributional outcomes. As union membership has decreased, a greater

share of income has shifted toward the top 1 long noted the rising dominance of the finan-
percent. With a decrease in union member- cial sector (Foster and Magdoff 2009; Kripp-
ship, workers wage bargaining power dimin- ner 2011; Tomaskovic-Devey and Lin 2011),
ishes and this can increase firms market value and our study demonstrates its connection to
and their profitability (Hirsch 1991). A higher the concentration of income.
market value often translates into higher stock Our analysis covers 1949 to 2008, but the
prices and executive compensation, thereby economic crisis that began in 2007 to 2008
shifting income toward the top. likely had significant consequences for the top
We also find evidence that the top 1 per- 1 percent. The steep drop in stock prices and
cents share of income responds to changing the accelerated deflation of the housing bub-
income and capital gains tax rates. Note that ble likely lowered the capital gains, interest,
we analyzed pretax income, so our result executive compensation, and dividend income
implies a market-conditioning effect of taxes of the top 1 percent. In the last year of our
through a behavioral response in addition to analysis (2008), the top 1 percents share fell
the explicit redistribution that happens by by approximately 2.6 percentage points (an 11
definition via taxes. One interpretation is that percent relative decline from 2007). Likewise,
members of the top 1 percent may choose the Shiller home price index fell 11 percent
leisure over labor if their tax rate increases, and the S&P fell 21 percent between 2007 and
and the concentration of income among the 2008. But this downturn is likely to have been
top 1 percent will fall as they work or invest transitory, in part because the $700 billion
less (Slemrod 1990, 1996). A more mechani- Troubled Asset Relief Program (TARP) bail-
cal interpretation is that lower tax rates reduce out and Federal Reserve loans included pur-
incentives for high income units to shift chases of toxic assets (e.g., mortgage backed
income or engage in tax avoidance (Feenberg securities). This program injected liquidity
and Poterba 1993), thereby increasing their into the financial sector but without conditions
pretax income. That tax rates influence the attached on how financial firms could distrib-
pretax income shares of the top 1 percent ute the funds for executive compensation
strongly suggests that government influences (Baker 2010; Blinder 2009). By 2010, the real
market outcomes in ways that have predicta- S&P 500 registered a rebound of 18 percent
ble effects on distributional outcomes. over 2009 and another 8 percent in 2011 over
Financialization of the economy has also 2010. Trade openness (imports and exports as
played a significant role in the rise of the top a percent of GDP) decreased in 2008 but
1 percent. We found that stock and home increased to over 30 percent in 2011, which
prices had substantial effects on the super- likely helped the top 1 percent recover. Save
rich. Because the rich receive substantial for a small reduction in the top capital gains
income from dividend payouts and capital rate (phased in during 2010) initiated by the
gains from stock trading, it is not surprising 2003 Bush tax cuts, there were no major
that high stock prices appear to have helped changes in top tax rates. Private sector union
concentrate income. Similarly, the wave of membership stayed the same or decreased
mortgage backed securities rooted in mort- during the economic crisis, and the 2010 mid-
gages derived from rising home prices appears term elections favored the Republican party in
to have contributed to income concentration. the House, which our model suggests will
And even these factors, which we treat as increase the top 1 percents share.
market conditions, may have been partially This article has answered many questions,
induced by public policy. The governments but there is of course much left to do. We have
failure to regulate innovations in the financial laid out an argument linking politics and pol-
sector reflects policy drift that facilitated the icy to distributional outcomes, but the details
proliferation of financial asset bubbles of exactly how these variables are connected
(Hacker and Pierson 2010). Scholars have have not been fully tested. There are clear and

fairly strong relationships between political (accessed August 18, 2011). All income data in this
variables on one hand and distributional out- article refer to Piketty and Saezs series inclusive of
capital gains.
comes on the other, but we have only scratched 2. As applied to the United States during the past 30
the surface regarding the specific policies that years, the only caveat is that Democrats (beginning
link politics to market inequality. Future in the early 1980s) have become more cooperative
research should focus more clearly on identi- with big businessespecially with the increasing
fying the many policies that might influence importance of corporate money in politics (cf.
Hacker and Pierson 2010).
market outcomes and test for the relative 3. In our discussion of financial markets, we note that
effects of these various policies on distribu- changes in taxation coincided with the shareholder
tional outcomes. We have also set aside the value movement that encouraged CEOs to focus on
question of what produced the political out- short-term stock price increases (Fligstein and Shin
comes that contributed to the rise in inequal- 2007), which is also related to increasing reliance
on stock options as compensation.
ity. Such questions are the focus of recent 4. For exceptions based on datasets extending into the
work by Hacker and Pierson (2010) and Kelly late 1970s and early 1980s, see Devine (1983),
and Enns (2010), who argue that economic Hibbs and Dennis (1988), Raffalovich (1993), Raf-
inequality is in some ways self-reinforcing. falovich, Leicht, and Wallace (1992), and Wallace,
For Hacker and Pierson (2010), economic Leicht, and Raffalovich (1999).
5. As in other recent studies, we observe an increase in
inequality generates political inequality that top shares during the mid-1990s that is not fully
prevents redistributive policy change from explained by changes in measurement techniques
occurring. For Kelly and Enns (2010), the (Piketty and Saez 2003; Raffalovich, Monnat, and
heart of the story is that the public becomes Tsao 2009).
more conservative as inequality rises, support- 6. However, Raffalovich and colleagues (2009) argue
that wealth is a relatively unimportant source of
ing the very policies that produce inequality. income for the rich, which would mitigate any
Although the causal underpinnings of elec- effects of financial factors on top income shares.
toral politics and public policy are beyond the 7. Changes in accounting rules made it more profitable
scope of this article, these are important ques- for firms to use stock options as compensation at the
tions that deserve attention as the literature on same time that changes in tax policy increased execu-
tives incentives to bargain for a shift toward
U.S. inequality continues to develop. compensation in this form (Hacker and Pierson 2010).
Both of these shifts were policy related and driven by
Authors Note the same underlying political dynamics. This suggests
that even the role of stock market valuation in the rise
The dataset and supplemental materials for this article
of income concentration is not divorced from policy
can be found at Nathan Kellys Dataverse: http://dvn.
(Davis 2009; Krippner 2011), and this is an ideal
iq.harvard.edu/dvn/dv/nkellydata.
example of market conditioning. We do not include
every market-conditioning policy in our models
Acknowledgments because it would likely be impossible. This is why it is
important to assess more general political dynamics
We thank Luke Keele, Lane Kenworthy, Matt Lebo, Jana such as partisan power. Effects of partisan power on
Morgan, members of the political science department at market income concentration capture the broad mar-
Washington University in St. Louis, and the ASR editors and ket-conditioning effects of politics that are not
reviewers for providing constructive feedback on this article. captured in our limited set of policy variables.
8. Series available from http://www.unionstats.com,
accessed March 26, 2011.
Funding 9. Available online at http://www.taxpolicycenter.org/
Thomas Volscho was supported by PSC-CUNY Grant taxfacts/displayafact.cfm?Docid=213, accessed
#60047-4041 for this research. Nathan Kelly was sup- March 26, 2011. This measure does not capture the
ported by a faculty development leave at the University effective tax rates paid by the richest Americans.
of Tennessee. Ideally, we would be able to measure top effective
tax rates, but such data are not available over a suf-
ficient time span. Although top effective tax rates
Notes would be a preferred indicator of tax policy, our use
1. Updates to 2008 are from Emmanuel Saezs web- of marginal tax rates provides a conservative test of
site: http://elsa.berkeley.edu/~saez/TabFig2008.xls the impact of tax policy on income concentration.

10. Available online at http://www.ctj.org/pdf/regcg. References

pdf, accessed March 26, 2011.
11. Available online at http://www.econ.yale. Alesina, Alberto and Howard Rosenthal. 1995. Partisan
edu/~shiller/data.htm, accessed March 26, 2011. Politics, Divided Government, and the Economy.
12. From the Federal Reserve Economic Database, New York: Cambridge University Press.
accessed March 23, 2011 (http://research.stlouisfed. Alderson, Arthur S. and Francois Nielsen. 2002. Glo-
org/fred2/series/UNRATE?cid=12). balization and the Great U-Turn: Income Inequality
13. Numerous tests are available to help determine Trends in 16 OECD Countries. American Journal of
whether a time series is stationary. We used Dickey- Sociology 107:124499.
Fuller (1979) and KPSS (Kwiatkowski et al. 1992) Atkinson, Anthony B. 2004. Top Incomes in the UK
tests. Dickey-Fuller tests the null hypothesis of a over the Twentieth Century. Journal of the Royal
unit root and KPSS tests the null of stationarity. Statistical SocietyA 168:32543.
Consistent results of stationarity or nonstationarity Atkinson, Anthony B., Thomas Piketty, and Emmanuel
from both tests is strong evidence in favor of a clear Saez. 2011. Top Incomes in the Long Run of His-
identification of the series. tory. Journal of Economic Literature 49:371.
14. Unit-root processes also require infinite variance. Auerbach, Alan J. and Kevin Hassett. 1990. Investment, Tax
Several of our variables cannot technically meet Policy, and the Tax Reform Act of 1986. Pp. 1340 in
this criterion for nonstationarity because they have Do Taxes Matter? The Impact of the Tax Reform Act of
an upper bound (e.g., tax rates cannot really exceed 1986, edited by J. Slemrod. Cambridge, MA: MIT Press.
100 percent). But in practice, the permanent Baker, Dean. 2009. Plunder and Blunder: The Rise and
memory characteristic is the one most relevant for Fall of the Bubble Economy. Sausalito, CA: Poli-
the analysis of unit-root processes. Point Press.
15. De Boef and Keeles (2007) recent work shows that Baker, Dean. 2010. False Profit: Recovering from the
single-equation ECMs can be applied to both sta- Bubble Economy. Sausalito, CA: Poli-Point Press.
tionary and integrated data, but they do not directly Bakir, Erdogan and Al Campbell. 2010. Neoliberalism,
address the possibility of fractional integration. the Rate of Profit, and the Rate of Accumulation.
Although their work implies that single-equation Science and Society 74:32342.
ECMs can be applied in a situation of fractional Banerjee, Anindya, Juan Dolabo, John Galbraith, and
integration, other scholars argue that applying David F. Hendry.1993. Co-Integration, Error Correc-
ECMs to fractionally integrated data could lead to tion, and the Econometric Analysis of Non-Stationary
biased estimates due to over-differencing the series Data. New York: Oxford University Press.
(Box-Steffensmeier and Tomlinson 2000; Clarke Bartels, Larry M. 2008. Unequal Democracy: The Politi-
and Lebo 2003; Lebo and Moore 2003; Lebo, cal Economy of the New Gilded Age. Princeton, NJ:
Walker, and Clarke 2000; Lebo and Young 2009). Princeton University Press.
We estimated all models in this analysis using an Blank, Rebecca M. and Alan S. Blinder. 1986. Macro-
alternative fractionally-differenced ECM and found economics, Income Distribution, and Poverty. Pp.
that results were substantively similar. These results 180208 in Fighting Poverty: What Works and What
are reported in supplemental materials available on Doesnt, edited by S. H. Danziger and D. Weinberg.
Kellys Dataverse (http://dvn.iq.harvard.edu/dvn/ Cambridge, MA: Harvard University Press.
dv/nkellydata). Blinder, Alan S. 2009. Crazy Compensation and the Cri-
16. Sociologists have often applied a variant of ECMs sis. Wall Street Journal, May 28, p. A15.
known as partial adjustment differential equation Blinder, Alan S. and Howard Y. Esaki. 1978. Macro-
models (Tuma and Hannan 1984; for applications, economic Activity and Income Distribution in the
see Raffalovich et al. 1992; Wallace et al. 1999). Postwar United States. Review of Economics and
The single-equation ECM is an Autoregressive Dis- Statistics 60:604609.
tributed Lag Model (ADL), but the ECM Box-Steffensmeier, Janet M. and Andrew R. Tomlinson.
representation is easier to interpret and preferred in 2000. Fractional Integration Methods in Political
empirical work (De Boef and Keele 2008). Science. Electoral Studies 19:6376.
17. The Vector Error Correction Model (VECM) is a Bradley, David, Evelyne Huber, Stephanie Moller,
version of the ECM that allows all of the variables Francois Nielsen, and John D. Stephens. 2003.
in the model to be endogenous, much like a VAR Distribution and Redistribution in Post-Industrial
model. We considered the VECM but found that the Democracies. World Politics 55:193228.
conditions necessary to calculate substantive effects Brady, David. 2009. Rich Democracies, Poor People:
from such models could not be met. Given this, we How Politics Explain Poverty. New York: Oxford
opted to estimate and report results from standard University Press.
ECMs. Although VECMs have a variety of useful Brady, David and Kevin Leicht. 2008. Party to Inequal-
characteristics, large VECM models are more easily ity: Right Party Power and Income Inequality in Rich
estimated in longer time series than we could ana- Western Democracies. Research in Social Stratifica-
lyze in this article. tion and Mobility 26:77106.

Brady, David and Benjamin Sosnaud. 2010. The Politics Foster, John Bellamy and Hannah Holleman. 2010. The
of Economic Inequality. Pp. 52141 in Handbook of Financial Power Elite. Monthly Review 62:120.
Politics: State and Civil Society in Global Perspec- Foster, John Bellamy and Fred Magdoff. 2009. The Great
tive, edited by K. Leicht and C. Jenkins.New York: Financial Crisis. New York: Monthly Review Press.
Springer. Freeman, Richard B. 1984. What Do Unions Do? New
Canterbery, E. Ray. 2002. The Theory of the Bondhold- York: Basic Books.
ing Class. Journal of Economic Issues 36:36572. Galbraith, James K. 1998. Created Unequal. Chicago, IL:
Clarke, Harold and Matthew Lebo. 2003. Fractional University of Chicago Press.
(Co)integration and Governing Party Support in Brit- Granger, Clive and Paul Newbold. 1974. Spurious
ain. British Journal of Political Science 33:283301. Regressions in Econometrics. Journal of Economet-
Davis, Gerald F. 2009. Managed by the Markets. New rics 2:11120.
York: Oxford University Press. Gustafsson, Bjorn and Mats Johansson. 1999. In Search
De Boef, Suzanna and Jim Granato. 1999. Testing for of Smoking Guns: What Makes Income Inequality
Cointegrating Relationships with Near-Integrated Vary Over Time in Different Countries? American
Data. Political Analysis 8:99117. Sociological Review 64:585605.
De Boef, Suzanna and Luke Keele. 2008. Taking Time Hacker, Jacob S. and Paul Pierson. 2010. Winner-Take-
Seriously. American Journal of Political Science All Politics: Public Policy, Political Organization, and
52:184200. the Precipitous Rise of Top Incomes in the United
Devine, Joel A. 1983. Fiscal Policy and Class Income States. Politics and Society 38:152204.
Inequality: The Distributional Consequences of Gov- Harrison, Bennet and Barry Bluestone. 1988. The Great
ernmental Revenues and Expenditures in the United U-Turn: Corporate Restructuring and the Polarizing
States, 19491976. American Sociological Review of America. New York: Basic Books.
48:606622. Henwood, Doug. 1997. Wall Street: How It Works and
Dickey, David A. and Wayne A. Fuller. 1979. Distribu- For Whom. New York: Verso.
tion of the Estimators for Autoregressive Time Series Hibbs, Douglas A. and Christopher Dennis. 1988.
with a Unit Root. Journal of the American Statistical Income Distribution in the United States. American
Association 74:42731. Political Science Review 82:46790.
DiPrete, Thomas, Greg Eirich, and Matthew Pittinsky. Hicks, Alexander M. 1999. Social Democracy and
2010. Compensation Benchmarking, Leapfrogs, and Welfare Capitalism. Ithaca, NY: Cornell University
the Surge in Executive Pay. American Journal of Press.
Sociology 115:16171712. Hicks, Alexander M. and Duane H. Swank. 1992. Poli-
Dumenil, Gerard and Dominique Levy. 2004. Capital tics, Institutions, and Welfare Spending in Industri-
Resurgent: Roots of the Neoliberal Revolution. Cam- alized Democracies, 196082. American Political
bridge, MA: Harvard University Press. Science Review 86:65874.
Enders, Walter. 2011. Applied Econometric Time-Series. Hirsch, Barry T. 1991. Union Coverage and Profitability
New York: John Wiley and Sons. among U.S. Firms. Review of Economics and Statis-
Engle, Robert F. and Clive W. J. Granger. 1987. Co-inte- tics 73:6977.
gration and Error Correction: Representation, Estima- Hirsch, Barry T. and David A. Macpherson. 2009. Union
tion, and Testing. Econometrica 55:25176. Membership and Earnings Data Book: Compilations
Epstein, Gerald and Arjun Jayadev. 2005. The Rise of from the Current Population Survey. Washington,
Rentier Income in OECD Countries: Financialization, DC: Bureau of National Affairs.
Central Bank Policy, and Labor Solidarity. Pp. 4674 Huber, Evelyne, Charles Ragin, and John D. Stephens.
in Financialization and the World Economy, edited by 1993. Social Democracy, Christian Democracy,
G. Epstein. Northampton, MA: Edward Elgar Press. Constitutional Structure, and the Welfare State.
Esping-Andersen, Gosta. 1990. The Three Worlds of Wel- American Journal of Sociology 99:71149.
fare Capitalism. Princeton, NJ: Princeton University Huber, Evelyne and John D. Stephens. 2001. Develop-
Press. ment and Crisis of the Welfare State. Chicago: Uni-
Feenberg, Daniel and James Poterba. 1993. Income versity of Chicago Press.
Inequality and the Incomes of Very High Income Irvin, George. 2008. Super Rich: The Rise of Inequality
Taxpayers: Evidence from Tax Returns. Pp. 14577 in Britain and the United States. Malden, MA: Polity
in Tax Policy and the Economy, edited by J. Poterba. Press.
Cambridge, MA: MIT Press. Jantti, Markus. 1994. A More Efficient Estimate of the
Fisher, Patrick. 2009. The Politics of Taxing and Spend- Effects of Macroeconomic Activity on the Distribution of
ing. Boulder, CO: Lynne Rienner. Income. Review of Economics and Statistics 76:37278.
Fligstein, Neil. 1990. The Transformation of Corporate Johnson, Simon and James Kwak. 2010. 13 Bankers: The
Control. Cambridge, MA: Harvard University Press. Wall Street Takeover and the Next Financial Melt-
Fligstein, Neil and Taekjin Shin. 2007. Shareholder down. New York: Vintage Books.
Value and the Transformation of the U.S. Economy, Kaplan, Steven N. and Joshua Rauh. 2010. Wall Street
19842000. Sociological Forum 22:399424. and Main Street: What Contributes to the Rise in

the Highest Incomes? Review of Financial Studies McCall, Leslie and Christine Percheski. 2010. Income
23:10041050. Inequality: New Trends and Research Directions.
Kapur, Ajay, Niall Macleod, and Narendra Singh. 2006. Annual Review of Sociology 36:32947.
Revisiting Plutonomy: The Rich Getting Richer. Minarik, Joseph J. 1981. Capital Gains. Pp. 25565 in How
Citigroup, Equity Strategy, Global Markets, March 5. Taxes Affect Economic Behavior, edited by H. Aaron and
Kelly, Nathan J. 2004. Does Politics Really Matter? Policy J. A. Pechman. Washington, DC: Brookings Institution.
and Governments Equalizing Influence in the United Moller, Stephanie, Arthur S. Alderson, and Francois
States. American Politics Research 32:26484. Nielsen. 2009. Changing Patterns of Income
Kelly, Nathan J. 2005. Political Choice, Public Policy, Inequality in U.S. Counties. American Journal of
and Distributional Outcomes. American Journal of Sociology 114:10371101.
Political Science 49:86580. Moller, Stephanie, David Bradley, Evelyne Huber, Francois
Kelly, Nathan J. 2009. The Politics of Income Inequality Nielsen, and John D. Stephens. 2003. Determinants
in the United States. New York: Cambridge Univer- of Relative Poverty in Advanced Capitalist Democra-
sity Press. cies. American Sociological Review 68:2251.
Kelly, Nathan J. and Peter K. Enns. 2010. Inequality Morris, Martina and Bruce Western. 1999. Inequality
and the Dynamics of Public Opinion: The Self-Rein- in Earnings at the Close of the Twentieth Century.
forcing Link between Economic Inequality and Mass Annual Review of Sociology 25:62357.
Preferences. American Journal of Political Science Nielsen, Francois. 1994. Income Inequality and Indus-
54:85570. trial Development: Dualism Revisited. American
Kelly, Nathan J. and Christopher Witko. 2012. Feder- Sociological Review 59:65477.
alism and American Inequality. Journal of Politics Nielsen, Francois and Arthur S. Alderson. 1995. Income
74:41426. Inequality, Development, and Dualism: Results from
Kennickell, Arthur B. 2009. Ponds and Streams: Wealth an Unbalanced Cross-National Panel. American
and Income in the U.S., 1989 to 2007. FEDS Work- Sociological Review 60:674701.
ing Paper 2009-13, Federal Reserve Board. Nielsen, Francois and Arthur S. Alderson. 1997. The
Kenworthy, Lane. 2004. Egalitarian Capitalism: Jobs, Kuznets Curve and the Great U-Turn: Income
Incomes, and Growth in Affluent Countries. New Inequality in U.S. Counties, 19701990. American
York: Russell Sage Foundation. Sociological Review 62:1233.
Kenworthy, Lane. 2010. How Much Do Presidents Palley, Thomas I. 1998. Plenty of Nothing: The Down-
Influence Income Inequality? Challenge 53:90112. sizing of the American Dream and the Case for
Kenworthy, Lane and Jonas Pontusson. 2006. Rising Structural Keynesianism. Princeton, NJ: Princeton
Inequality and the Politics of Redistribution in Afflu- University Press.
ent Countries. Perspectives on Politics 3:44971. Phillips, Kevin. 2002. Wealth and Democracy: A Political
Korpi, Walter. 1983. The Democratic Class Struggle. History of the American Rich. New York: Broadway.
London: Routledge and Keegan Paul. Piketty, Thomas and Emmanuel Saez. 2003. Income
Krippner, Greta. 2011. Capitalizing on Crisis: The Politi- Inequality in the United States, 19131998. Quar-
cal Origins of the Rise of Finance. Cambridge, MA: terly Journal of Economics 118:139.
Harvard University Press. Raffalovich, Lawrence E. 1993. Structural Sources of
Kristal, Tali. 2010. Good Times, Bad Times: Post- Change in Earnings Inequality: Evidence from the
war Labors Share of National Income in Capital- Current Population Survey, 19671981. Research in
ist Democracies. American Sociological Review Social Stratification and Mobility 12:11344.
75:72963. Raffalovich, Lawrence E., Kevin T. Leicht, and Michael
Krugman, Paul. 1997. The Age of Diminished Expecta- Wallace. 1992. Macroeconomic Structure and
tions. Boston: MIT Press. Labors Share of Income: The United States, 1950
Kwiatkowski, Denis, Peter C. B. Phillips, Peter Schmidt, 1980. American Sociological Review 57:24358.
and Yongcheol Shin. 1992. Testing the Null Hypoth- Raffalovich, Lawrence E., Shannon M. Monnat, and Hui-
esis of Stationarity Against the Alternative of a Unit Shien Tsao. 2009. Family Income at the Bottom and
Root. Journal of Econometrics 54:15978. at the Top: Income Sources and Family Characteris-
Lebo, Matthew J. and Will Moore. 2003. Dynamic For- tics. Research in Social Stratification and Mobility
eign Policy Behavior. Journal of Conflict Resolution 27:301309.
47:1332. Roine, Jesper, Jonas Vlachos, and Daniel Woldenstrom.
Lebo, Matthew J., Robert W. Walker, and Harold D. 2009. The Long-Run Determinants of Inequality:
Clarke. 2000. You Must Remember This: Dealing What Can We Learn from Top Income Data. Journal
with Long Memory in Political Analyses. Electoral of Public Economics 93:97488.
Studies 19:3148. Shiller, Robert. 2005. Irrational Exuberance. Princeton,
Lebo, Matthew J. and Everett Young. 2009. The Com- NJ: Princeton University Press.
parative Dynamics of Party Support in Great Britain. Slemrod, Joel. 1990. Optimal Taxation and Optimal
Journal of Elections, Public Opinion, and Parties Tax Systems. Journal of Economic Perspectives
19:73103. 4:15778.

Slemrod, Joel. 1992. Do Taxes Matter? Lessons from of Income: A Quarterly Analysis of the United States,
the 1980s. American Economic Review 82:25056. 19491992. Social Science Research 28:26588.
Slemrod, Joel. 1996. High Income Families and the Tax Wood, Adrian. 1994. NorthSouth Trade, Employment,
Changes of the 1980s: The Anatomy of a Behavioral and Inequality: Changing Fortunes in a Skill-Driven
Response. Pp. 16988 in Tax Progressivity and World. New York: Oxford University Press.
Income Inequality, edited by J. Slemrod. New York:
Cambridge University Press.
Stephens, John D. 1979. The Transition from Capitalism
to Socialism. London: Macmillan. Thomas W. Volscho is Assistant Professor of Sociology at
Tomaskovic-Devey, Donald and Ken-Hou Lin. 2011. CUNY-College of Staten Island. His research interests are in
Income Dynamics, Economic Rents, and the Finan- political economy, racism studies, and quantitative methods.
cialization of the U.S. Economy.American Socio- His work has been published in journals such as Research in
logical Review 76:53859. Social Stratification and Mobility, Critical Sociology,
Tonelson, Alan. 2000. The Race to the Bottom: Why a Wicazo Sa Review, and Social Science Quarterly. With
Worldwide Worker Surplus and Uncontrolled Free Angie Beeman he is working on a sociological study of
Trade Are Sinking American Living Standards. Boul- racial and gendered disparities in depression. With Nathan
der, CO: Westview Press. Kelly he is currently working on a multiple time-series study
Tuma, Nancy and Michael T. Hannan 1984. Social of the link between public policy, politics, and top incomes.
Dynamics: Models and Methods. Orlando, FL: Aca-
demic Press. Nathan J. Kelly is Associate Professor of Political Science
U.S. Bureau of the Census. Various years. Statistical at the University of Tennessee. He conducts research in
Abstract of the United States. Washington, DC: Gov- American politics, public policy, political economy, and
ernment Printing Office. public opinion. He is author of The Politics of Income
U.S. Bureau of Economic Analysis. 2011. National Inequality in the United States (Cambridge University
Income and Product Accounts. Available at http:// Press) and numerous articles in journals including the
www.bea.gov/national/nipaweb/DownSS2.asp. American Journal of Political Science, Journal of Politics,
U.S. Executive Office of the President. 2011. Economic and Political Analysis. His active projects include work on
Report of the President. Washington, DC: Govern- cross-national economic inequality, the effects of inequal-
ment Printing Office. ity on political processes and outcomes, aggregate policy
Wallace, Michael, Kevin T. Leicht, and Lawrence E. Raf- production in the United States, and the influence of social
falovich. 1999. Unions, Strikes, and Labors Share movements on the American political system over time.

Volscho Thomas - The Rise of The Super-Rich

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Volscho Thomas - The Rise of The Super-Rich

Uploaded by

Copyright:

Available Formats

458508

erican Sociological ReviewVolscho and Kelly

American Sociological Review

The Rise of the Super- 77(5) 679699

Financial Markets, and the

Thomas W. Volschoa and Nathan J. Kellyb

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

Figure 1. The Top 1 Percents Share of Income, 1913 to 2008

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

into the 2000s. By 2008, private sector union Estimation Strategy

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

Results (although we will see that the short-term effect

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

Table 1. Models of Top 1 Percent Income Share, 1949 to 2008

(1) (2) (3) (4) (5)

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

(1) (2) (3) (4) (5)

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

% Congressional Democrats Union Membership

Reducon in Top Share %

Reducon in Top Share %

Top Marginal Tax Rate Capital Gains Rate

Reducon in Top Share %

S&P 500 Shiller

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

Part A: Total Effect

Part B: Parally Standardized Effect

Divided Government .069

Figure 3. Total Effects on Top Income Shares

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

10. Available online at http://www.ctj.org/pdf/regcg. References

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

Downloaded from asr.sagepub.com at ASA - American Sociological Association on October 1, 2012

You might also like

Reducon in Top Share %

Reducon in Top Share %

Reducon in Top Share %

Part B: Parally Standardized Effect