Content uploaded by Steven R. Beckman
Author content
All content in this area was uploaded by Steven R. Beckman on Jan 15, 2019
Content may be subject to copyright.
The Effects of Race, Income, Mobility and
Political Beliefs on Support For Redistribution.
Steven R. Beckman and Buhong Zheng*
Abstract
A questionnaire is used to assess the impact of race, current
past and future family income, as well as political beliefs on the
support for redistribution. Current income maximization predicts
those with above average income oppose redistribution. However
blacks support redistribution until income is well above average
and whites oppose redistribution even if income is well below
average. Those with incomes below average expect to move up and
this prospect of upward mobility reduces support for redistribution.
The rich are more likely to espouse arguments that protect their
wealth. Most intriguingly, as blacks become richer support for
redistribution falls especially rapidly.
JEL Codes: H23, C42
Keywords: POUM, redistribution, race, income, mobility
* Department of Economics, University of Colorado, Denver.
Campus Box 181, P.O.Box 173364, Denver Colorado 80217-3364
Steven.Beckman@cudenver.edu BZheng@carbon.cudenver.edu
1. Introduction
The skewed income distribution in the US implies the majority has less than average
income and raises the possibility the poor majority will use the democratic process to redistribute
income from the rich. Therefore it seems natural that a populist party would enjoy considerable
success - and yet the order of the day continues to be lean social services and tax reductions.
Why doesn’t the moderate-income majority vote their short-term self-interest and dominate
politics?
Many possible explanations have been offered. Benabou and Ok (2001) argue the middle
-income voters have longer-term horizons and may not support redistribution because they
harbor realistic dreams of becoming rich. Piketty (1995) believes mobility convinces workers
that effort is rewarded. Workers therefore are more likely to oppose taxes because they reduce
the marginal return to effort. In his model there are two potential equilibria – a mobile society
that rewards effort and therefore has low taxes and an immobile society with high taxes.
Roemer (1998, 2001) shows how racial divisions may force parties to moderate their
stance on redistribution. The party of the rich may be able to siphon off middle-income white
votes by protecting white people’s privileges even though they propose low taxes. In response,
the populist party has to appeal to rich supporters of affirmative action by reducing proposed
taxes. Evidence of such a link between race and redistribution is also provided by Alesina,
Glaeser and Sacerdote (2002), Corneo and Gruner (2002) and Fong (2001).
Louis Putterman (1997) provides an extensive list of other possible explanations and
some guesstimates of their relative importance. He attributes 35% of the tax forbearance to the
belief redistribution will in some way reduce growth, 20% to the belief that wealth is legitimately
acquired, 15% to subversion of the political process by the rich, 10% to greater participation in
politics by the rich, 10% to dreams of becoming rich and the remaining 10% to local histories
and social movements. We hasten to add that the Putterman scorecard is a small part of a deep
and interesting paper.
We approach the question by constructing a questionnaire. We ask respondents to
indicate if they support higher taxes on the rich if the proceeds are used to help the poor. We
collect their past, current and expected incomes reported on a 10-point scale with 5 as the
average. This allows a simple presentation of results. Do respondents reporting above average
income oppose additional taxes as self-interest suggests? Are respondents shortsighted or do
they consider expected income while formulating an opinion?
We follow up with a number of questions soliciting the respondent’s reasons for
supporting or opposing redistribution. Most of these questions are adapted from Putterman
(1997). Space to write in any reason we fail to cover is also provided. Finally, we collect data
on race, gender, age, occupation etc.
Briefly, the responses provide strong support for the theoretical foundations provided by
Roemer and Benabou and Ok. Race, current and future incomes play dominant roles but not
precisely in the ways these authors suggest. It appears that our respondents do not form separate
opinions about race, income, mobility and redistribution but that instead these are inextricably
linked. Research in social psychology will shed some light on the possible linkages.
Putterman’s guesses match our respondent’s opinions fairly well except that our respondents give
greatest weight to the moral legitimacy of wealth. Both Putterman, and our respondents,
underestimate the importance of race.
2
2. Why might moderate-income voters oppose redistributive taxes?
If people with below average incomes believe they will soon enjoy a higher income they
may rationally oppose redistributive taxes. Benabou and Ok (2001) provide a thorough formal
treatment. We present a simple illustration. For example, let the incomes be 20, 30 and 50. The
average is 33.3 and a majority benefits from taxes and redistribution providing 33.3 to all. If
there is a 20% chance that middle and high income earners exchange places middle’s expected
income is .8(30)+.2(50) = 34 and now a majority supports no redistributive taxes. The
required concavity of the mobility function is obvious from a plot of expected income on actual
income. Compared to the 45 degree line, low income is on the line, for the middle class
expected income exceeds actual and is above the 45 degree line, and for the rich expected
income is less than actual creating the characteristic concave shape.
Our primary motive for the questionnaire is to provide a simple clean test of the
possibility of upward mobility (POUM). Our questionnaire collects preferences on increasing
taxes to redistribute income and also collects information on current, past and expected income
on a 0 to 10 scale where 5 is average. Therefore we are able to relate preferences to the
respondent’s beliefs about their relative positions over time. This is the main advantage of our
survey and allows us to estimate not only if income and expected income affect support for
redistribution in the expected directions but also allows us to estimate the current income where
support for redistribution is 50%. Someone who maximizes current income will shift from
support to opposition at precisely average current income. The POUM hypothesis states that the
crossover occurs at a lower level of current income due to the concavity of expected mobility.
3
We will be able to directly check if expected mobility has the required concavity and show how
far the crossover point is shifted.
There are a number of existing studies relating current and expected income to
preferences for redistribution. Ravallion and Lokshin (2000) find that the coefficients linking
income and attitudinal variables to support for restricting the rich are quite different for those
who expect to move up. Corneo and Gruner (2002) use past mobility as a proxy for expected
mobility and find a significant effect. Fong (2001) uses a Gallup survey that asks the respondent
to rate their life on a 0 to 10 scale in the past, present and future and also finds that expected
future well-being reduces support for redistribution. While we believe each of these papers has
its virtues, our approach is more direct and will allow us to check the concavity of the mobility
function and the effect of mobility on support for redistribution.
Race and redistribution are linked in a number of ways. Roemer (1998, 2001) shows how
party politics may link race and redistribution. The literature in social psychology offers several
other potential explanations. If interdependencies within a group lead to a common fate then the
individual members benefit from the formation of the group. Membership may also confer status
and improve self-esteem if the group is unusually successful. Both of these ideas fit easily in the
utilitarian tradition but there is another mechanism of group identity that may seem less natural
to economists: “when individuals are committed to group membership, ingroup defeat produces
even higher levels of ingroup preference than does success.” (Gilbert et al. 1998 p. 564.)
Pettigrew (2002) provides an excellent review of both the evolution of race relations in
the US and its impact on social psychology. Slavery, lynchings, segregation, and the Great
Depression all create defeats that forged a strong group identity. The civil rights movement
created positive interdependencies and heightened group status. The strong group identity of
4
African-Americans is therefore no mystery. It is well established that groups are more likely to
ascribe failures by group members to events beyond the individual’s control (Gilbert et al p.
561). If a group is poor despite their own good efforts, then the group may see redistribution as
their right.
If whites also form a group, then the tendency to ascribe failures by outgroup members to
the individual’s personal failings tends to limit sympathy and assistance. However, the
mechanism for the creation of a white group identity is not altogether clear.1 Various theories
have been offered. It may be that the creation of a black group implies a white group in that
whites feel threatened by black advances. It may be that whites are more likely to come in
contact with upwardly mobile blacks and therefore underestimate the remaining difficulties poor
blacks continue to face.
This perspective enriches the discussion. If we extend Piketty’s argument to the racial
domain it seems natural to argue that immobile blacks doubt effort is rewarded and support
redistribution while mobile whites believe hard work is rewarded and oppose redistribution. Our
sample will allow a direct test of this hypothesis because we record race, expected mobility and
support for the idea effort is rewarded. On the other hand, we may use the social rivalry
hypothesis put forward by Corneo and Gruner (2002) where redistribution mixes up
neighborhoods and occupations threatening the prestige and self-esteem of whites. Poor whites
may therefore avoid redistribution to preserve high-status neighborhoods and jobs. There are of
course, a number of studies that already link redistribution to race.2
1 Social psychologists have been working in this area for over 100 years and we are familiar only with a few basic
reference works. If we missed something obvious we apologize.
2 A number of studies link race and redistribution. Alesina, Glacer and Sacerdote (2001) construct a racial
heterogeneity variable and find that countries with greater heterogeneity have less redistribution. Luttmer (2001)
shows that support for redistribution among people with poor neighbors rises or falls if the neighbors are the same or
a different race. Alesina and LaFerrara (2000, 2001) show that participation in social activities involving direct
contact between individuals is lower in racially fragmented communities in the US and trust is higher in more
racially homogeneous communities. Glaeser, Laibson, Scheinkman and Soutter (2000) document experimentally
5
The stage is now set. There are a number of theories for middle-income voters opposing
redistributive taxes and these have been cleverly tested with existing data sets. Our design,
however, will allow us to find the crossover point in voter support: at what income level will
voters have a 50/50 probability of supporting redistribution? Is it mean income as current
income maximization suggests? Will expected income be considered, will it have the form
required by the POUM hypothesis and will it significantly shift the crossover point? Issues of
race are also very interesting but designing questionnaires that uncover racial preferences is a
tricky task and we decided not to include racially explicit questions. We needn’t have worried -
the race issue is so strong it emerges quite naturally.
3. Results
The questionnaire, reproduced in the appendix, was administered to 1096 respondents at the
University of Alabama, North Carolina Agriculture and Technology, East Carolina University,
the Denver and Boulder campuses of the University of Colorado, Colorado State University at
Fort Collins and Guelph (in Ontario). We essentially imposed on our colleagues who used their
students as respondents. Therefore many of the respondents are business or economics majors.
The first question is: “Do you think rich (high income) people in the U.S. should be taxed
more with the money given to the less fortunate so that income and wealth are more equally
distributed? Circle your response.”3 There are 283 Yes responses, 700 No, 100 uncertain and 13
missing for a total of 1096. Therefore support is 26% with 65% opposed and 9% uncertain.
The questionnaire solicits information about the respondent which we coded as indicated
in table 1. There are a number of dichotomous variables that indicate whether the respondent
that people of different races are more likely to cheat one another. Fong (2001) shows that white males are less
likely to support redistribution.
3 The questionnaires in Canada referred to Canada, not the US.
self-identifies them self as White, African-American, Hispanic, Asian or Other.4 About 72% of
the sample is White with 13% African-American, and 8% Asian. Respondents are 53% male,
fairly tightly grouped around 21 years of age, with 60% that paid income taxes and 42% that
voted.5 They are also rather prosperous and politically moderate as the average current family
income is recorded as a 7 and liberalism averages 5 on a 10 point scale. The income questions
refer to current, past and future income. These have been coded to reflect current income and the
relative status of current to future and past income. Our respondents see themselves as being
upwardly mobile, having climbed 1.4 points on the 10 point scale and fully expecting to climb
one more step in the next 10 to 15 years.
The number of responses is recorded for each variable. We will make use of a procedure
provided by King, Honaker, Joseph and Scheve (2001) to handle missing data.
Some of the characteristics are correlated with each other. For example churchgoers tend
to be conservative southerners while Asians are more likely to identify themselves as
immigrants. Table 2 reports the correlation coefficients that rise above 15% in absolute value.6
The strongest correlations are among the income categories. Apparently those that have done
badly recently expect the future will be brighter. It is intriguing that white is correlated with high
current income but is also correlated with pessimism about upward mobility.
Our questionnaire collects the data we believe may be important; therefore all regressions
reported in this paper use all the variables listed in table 1. We have no additional variables to
check for the robustness of results, and sensitivity testing is confined to dropping variables or
altering functional form. We are fortunate that the higher correlations occur in the income group
4 The Canadian questionnaire allowed only White, Asian and Other. Given there is only one ‘other’ we do not see
this as a significant difference.
5 The Canadian questionnaire asks if the respondent voted in the most recent Federal election. (The US
questionnaire asks about the Presidential election.)
6 The correlations among the races are of course strong and negative. Someone that reports they are African-
American does not report they are white. These are obvious and unreported.
where theory leads us to include all three variables. The remaining correlations are not
particularly high but some will play a significant role.
We use ordered probits in our analysis. The response to question 1 is coded as a 1 if
“No”, 2 if “Unsure” and 3 if “Yes”. Ordered probits can be difficult to interpret but our
circumstances are straightforward. Let
XR*
represent the unobserved levels of support
for redistribution where
X
is the data matrix and
is a vector of coefficients. Then
and
are selected using standard maximum likelihood techniques to assign probabilities to the
responses 1, 2 or 3. The assignments are made as follows:
ii
ixFxRPR 1
),,|1(
(1)
iii
ixFxFxRPR 12
),,|2(
(2)
ii
ixFxRPR 3
1),,|3(
(3)
where
i
x
is the
th
i
row of
X
and
F
is the cumulative normal distribution function. In
general, it is difficult to relate even the sign of any coefficient to its influence on probabilities –
as can be clearly seen from equation (2) where the coefficients are given opposing influences on
the probability the respondent is unsure. However, equations (1) and (3) show a positive
coefficient reduces opposition and increases support. Therefore a positive coefficient does
indicate greater support for redistribution.
Table 3 presents the results with support for redistribution as the dependent variable.7
Columns (a) and (b) compare the list delete and multiple imputation methodologies. These are
7 We do not include a constant term because the
i
serve the function of an intercept within each category.
8
different only in the way they treat missing data. Column (b) ignores all the information from a
given respondent if any one necessary piece of information is missing. Column (a) employs a
methodology provided by King et. al. (2001) that uses all the responses from all respondents to
estimate missing values. In our case, the differences are small therefore we will proceed under
the assumption that missing data is not a serious problem.8 Column (c) adds those interaction
terms with black that are significant at conventional levels.
Race immediately emerges as a major determinant of redistribution preferences.
Compared to whites, Asians are more likely to support redistribution and black support is higher
yet.9 Figure 1 illustrates by presenting the estimated probabilities from column (c). The picture
that emerges is one of a highly polarized society, with poor blacks far more likely to support
redistribution than poor whites. As income increases, the gap between white and black support
levels narrows substantially reflecting the significant interaction affect between black and current
income. Asian support offers a convenient reference, as it is nearly what one would expect if
Asians maximized current income. Asians with income above five are likely to oppose
redistribution and Asians with incomes below 4 are likely to be supportive. 10 The crossover
8 King et. al. provide a program, Amelia, available at http://Gking.Harvard.Edu. Amelia uses all the data provided to
generates five complete data sets with missing data estimated. The analyst then conducts whatever exercise he is
interested in on all five data sets and then combines results according to rules provided in King et. al. (2001 p. 53).
In our case, the coefficient estimates are the average responses from the five data sets and the variances are the
average of the five variances plus 1.2 times the variance of the five coefficient estimates. In effect, we add the
variance within each imputation to the variance between imputations – weighted by a factor because imputations are
not infinite. Programs that correct for truncation or censoring biases are in widespread use. Missing data may
impart a similar bias if people of a particular race, age or income are less likely to respond. Knowing that - at least
up to our current ability to test - missing data does not bias results is reassuring. The disadvantages include a
significant investment in time and the fact that one quickly comes to the limits of known procedures with imputed
data. We follow Scheve and Slaughter (2001) and report none of the usual goodness of fit statistics for (a) because
none are currently available for imputed data.
9 If we drop Age, Female, Business, Migrant, Church, Vote and Tax the remaining coefficients and standard errors
are qualitatively similar except for the coefficient of Asian. This coefficient increases and becomes significant at the
5% level because Asian and Migrant are strongly correlated. Dropping only Migrant produces nearly the same
increase in the coefficient.
10 Support levels are calculated from column (c) in Table 3 with average values of variables other than race or
current income according to the formulas set out in (2) and (3) with the exception that the average used for migrant
is conditional on being Asian or non-Asian. Given someone is Asian, the average value of migrant is .53, if they are
non-Asian the average is .06. Race is set to 1 or 0 as appropriate and YC gradually rises from 0 to 10.
9
from support to opposition occurs for whites at an income level of about 2 and for blacks near an
income of 8. The fact our respondents are students make these results even more dramatic:
Race powerfully conditions attitudes even among our educated young people.
We consider the particularly strong effect of income on black support for redistribution to
be one of our most interesting results. The effect strongly suggests that group identity is heavily
conditioned by current economic status: if the declining support levels among whites measures
the normal effect of income within a group, the higher effect of income on black support may
reflect weakening group ties as blacks become rich.
Our questionnaire is also designed to measure whether expected income influences
redistribution preferences as Benabou and Ok (2000) surmise.11 They require not only that voters
take expected income into account but also that the mobility pattern influences the swing voters
most. As a first step we regress expected upward mobility on a constant, current income, current
income squared and current income cubed. The coefficient on current income is positive, the
coefficient on the square is negative and both are significant at the 1% level consistent with
concavity. The equation is then used to forecast mobility. Figure 2 presents support levels for
redistribution applying the equation in column (c). One estimated support level uses forecasted
mobility and the other uses average mobility while both use the Asian intercept.12 The
difference between the two support levels shows the effect of expected mobility and shifts the
crossover point from above four to below one. Support drops below 50% for all but the poorest
and remains below 50% for the rich demonstrating how concave mobility may decrease support
for redistribution. This three-step shift in the cross over point due to upward mobility is even
stronger than the effect of being white. However, our respondents are students and students may
11 Theory suggests that support for redistribution may shift dramatically around mean income. Therefore we
constructed an indicator variable that was 1 if current income was within one unit of 5. This proved insignificant.
We also tried a cubic in current income, this also failed to improve the fit of the equations reported in table 3.
12 We again use the conditional mean for migrant. The regression of YF-YC on YC, YC2 and YC3 has an R2=.52.
10
see themselves as more mobile than society at large. Therefore we should conclude only that the
POUM hypothesis is a powerful influence among students.
The remaining coefficients in table three are unsurprising and easily interpreted and we
move on to the rest of the questionnaire. Among other things, we will get some clues as to why
race and redistribution are so closely linked. For all questions, respondents are asked to rate the
importance of a series of statements on a 0 to 10 scale. Table 4 presents the questions and their
average responses.
Supporters of redistribution do not believe the rich became rich unfairly and tend not to
be interested in abstract equality. Instead, they point to the moral responsibility of the rich and
the possibility that inequality will lead to social strife. None of the responses to question 2 are
rated highly suggesting the main reasons may be in the 58 written responses. Of these 42 put
forward some version of a fairness argument. Respondents often claim the poor work hard and
are entitled to have basic needs met. Ten point out that the poor are more likely to spend income
and this will boost the economy. No one mentions race.13
The responses to question 3 indicate opponents of redistribution strongly believe the rich
are entitled to the fruits of their labor. The possible disincentive effects on saving or work effort
get moderate support as do the propositions that the poor may become rich and that the right to
private property is important for freedom. The written responses typically restate a version of
these arguments with differences in emphasis. 14 Again, no one mentions race.
13 A few specifically argue that a dollar to a poor person is worth more than to a rich person. Three claim the rich
subvert the political process and two are clearly envious. No one argues that taxes should be increased because they
personally will benefit.
14 65 of 117 written responses make some reference to efficiency or incentive effects. Often the statements blend
efficiency arguments with fairness – people that work hard and contribute to growth deserve to keep their earnings.
Fairness issues are the second most popular - with 41 responses, however the division between fairness and
efficiency is quite arbitrary for many of the remarks. Fairness takes many forms: some argue the poor are doing just
fine as it is, or the poor need to have consequences for their decisions, and a few even argue the poor want to be
poor.
11
The fourth question asks about each respondent’s perception of public opinion. Our
respondents strongly agree that middle income voters will support tax cuts skewed to the rich as
long as they get something as well. This mirrors experimental results reported in Beckman et al
(2002) where small side payments are enough to dramatically reduce opposition to payments to
the rich. Our respondents also give strong credit to the disproportionate influence the rich have
on politics through campaign contributions. The propositions that the rich are more likely to
vote and that globalization implies high taxes force firms offshore get moderate support while
the ideas that the rich spread the wealth to buy support or that taxes are determined in back room
deals are not popular. The written responses are the most varied once respondents speculate
about the motives of others. Several question the basic intelligence of voters but none mention
race. 15
It is interesting to contrast these results to the guestimates of Putterman. He estimates
that 35% of the opposition to taxes is due to fear of negative long run consequences for the
average citizen; 20% to the moral legitimacy of wealth - although he speculates this may be
more important in the US than Europe; 15% to the rich hijacking the political process through
campaign contributions; 10% to the possibility of upward mobility; 10% to the greater voting
proclivities of the rich and 10% to local social movements. Our respondents give their highest
rating to the moral legitimacy of wealth as captured in question 3a. The second highest rating
goes to the form of the tax cut in question 4f and efficiency arguments come in third in questions
3b, 3c and 4c. Our respondents agree with Putterman in that they rate hijacking the political
process (4a) next with POUM (3d) and the voting proclivities of the rich (4b) tied for last.
15 26 of 71 point out that people like more money and politicians are rewarded for giving it to them. It is interesting
that people believe this about others but not themselves. 10 claim voters do not understand the consequences of tax
reductions, 9 argue for small government, 8 assert the rich have a right to their property, 11 that tax cuts affect
growth, 6 that the rich have greater influence and three mention fairness. These mimic options already available
except the assertions that voters are uninformed and government is inefficient.
12
However the largest error Putterman makes is not even mentioned by our respondents. He rates
history and social movements at 10% while race relations in the US appear to be a powerful
influence even when set against current income. Overall then, Putterman appears to underrate
morality, the form of the tax cut and race relations at least for US respondents.
Tables 5, 6 and 7 report how these support levels break down across respondent
characteristics. The dependent variable in each table is the support level on a 0 to 10 scale and
we once again use ordered probits.
Perusing these tables we find that African-Americans give more importance to all the
reasons listed to support redistribution and less importance to all the reasons listed to oppose
redistribution. For example, African-Americans are more likely to believe that inequality breeds
social conflict, that the rich are more likely to vote and that taxes are settled by back room
politics. They are less likely to believe the rich obtained their wealth through the fruits of their
own labor. Such beliefs lend themselves to multiple equilibrium models similar to Piketty. If
politics are unfair, there is little reason to vote. If income is not earned then there is little reason
to exert effort. In an unfair society where democratic processes are ineffective social conflict
may seem a rational solution. While none of these ideas point to a particular multiple
equilibrium trap, they do suggest the multiple equilibrium methodology may be fruitful.
However our conjecture that whites see themselves as mobile and blacks see themselves
as immobile producing divergent views on how society rewards effort does not fare well. At
least among our students, blacks expect greater upward mobility. In our survey there are 132
black responses with YF-YC available. YF-YC > 0 for 99, YF-YC = 0 for 31 and YF-YC < 0 for
2. For non-blacks there are 915 observations divided as 492, 290 and 133 across the same
categories. The Pearson chisquare statistic is 26.9 and the probability of a larger statistic given
13
homogeneity is 0.000. Black support for redistribution does not require personal pessimism
about future income.
Income emerges as a powerful influence on opinion in quite predictable ways. Higher
income reduces ratings for all the reasons listed to support taxes and increases ratings for all the
reasons listed to oppose taxes. Once respondents speculate about the opinions of others, income
only increases the rating for globalization. These are all self-serving beliefs in that they shelter
income from taxes and remind us ideology may well follow social position.
Southerners and liberals have directly opposing tendencies for questions 2 and 3 but both
tend to agree that the rich have disproportionate influence through lobbying. We leave the
remaining sporadic influences to the reader. Given the large number of variables the possibility
of false positives is quite real and there is danger in interpreting every coefficient.
Conclusions
An important reason Americans oppose redistributive taxes is a reason Putterman – and
all of our subjects – fail to mention: race. Asians appear to maximize current income with
support for redistribution crossing the 50% level close to average income. For whites, the 50%
support level occurs two steps lower on a 10 step scale. Black support is nearly unanimous
among the poor but drops off rapidly as income rises so that opposition levels for rich whites and
blacks are similar.
The social psychology of groups offers some clues as to the reasons. The formation of a
black group that sees itself as necessary for mutual defense and progress is natural enough given
US history. Ingroup members are likely to be given the benefit of the doubt and failings are
attributed to external causes - therefore redistribution is fair. The reasons for the creation of a
white group identity are less obvious. However, once the group identity is created, whites are
14
likely to see blacks as personally responsible and therefore redistribution is not warranted. While
this is simple, it does not explain the opposition to redistribution by poor whites that benefit from
redistribution – unless somehow there is status and self-esteem to protect. We suspect this will
be an active area of research for some time.
The possibility of upward mobility also receives strong support. Respondents not only
consider expected income but it is precisely those with incomes just below average that expect to
move up the most. This effect is actually larger than the effect of being white. However, given
that our student respondents may be more mobile than society at large, we need to be cautious
and conclude only that both race and expected mobility are powerful affects on the preference for
redistribution.
The rich tend to form beliefs that rationalize tax relief. The rich are more likely to
believe they earned their rewards fairly, that redistribution reduces the incentive to work and
save, that property is a right and that the global economy limits the ability to tax. The rich are
less likely to believe they have a moral responsibility to the poor or that social stability is
affected by inequality. Perhaps the most interesting effect of income is the more rapid decline in
support for redistribution as blacks become richer. This raises the prospect that the group
identity of blacks may fracture.
We are also able to contrast our respondents’ opinions with Putterman’s. Our respondents
give higher priority to the moral legitimacy of wealth and the form of tax relief. They give less
importance to efficiency effects. Both our respondents and Putterman underestimate the
importance of race.
15
References
Alesina, A., Glaeser E., Sacerdote B., 2001. Why doesn’t the US have a European style welfare
system? NBER working paper #8524.
Alesina, A., La Ferrara, E., 2000. Participation in heterogeneous communities. Quarterly
Journal of Economics 115(3): 847-904.
Alesina, A., La Ferrara, E., 2001. Preferences for redistribution in the land of
opportunities. NBER Working Paper No. 8267.
Alvarez, M., Brehm, J., 1997, Are Americans ambivalent toward racial policies? American
journal of political science. 41: 345-374
Corneo, G., Gruner, H., 2002. Individual preferences for political redistribution. Journal of
Public Economics 83, 83-107.
Beckman S., Formby J., Smith W., and Zheng B., 2002. Envy, malice and Pareto efficiency: An
experimental examination. Social Choice and Welfare, 19: 349-367.
Benabou, R., Ok, E., 2001. Social mobility and the demand for redistribution The Quarterly
Journal of Economics, 116, 447-487.
Fong, C., 2001. Social preferences, self-interest, and the demand for redistribution 82, 225-246.
Gilbert, G., Fiske, S., Lindzey, G., 1998. The Handbook of Social Psychology. McGraw-Hill,
New York.
Glaeser, E., Laibson, D., Scheinkman, J., Soutter, C., 2000. Measuring trust. Quarterly
Journal of Economics, 115(3): 811-846.
Greene, W. 1997. Econometric Analysis, Prentice Hall, New Jersey.
King, G., Honaker J., Joseph A. and Scheve K. 2001. Analyzing incomplete political science
data: An alternative algorithm for multiple imputation. American Political Science
Review 95, 49-69.
Luttmer, E., 2001. Group loyalty and the taste for redistribution. Journal of Political
Economy 109(3): 500-528.
Pettigrew, T., 2002. Implications for Europe of the study of American race relations. At
http://www.eumc.at/events/measuring/doc/Pettigrew_American%20Race%20Relations.doc.pdf
Piketty, T., 1995. Social mobility and redistributive politics. Quarterly Journal of Economics
110, 551-584.
Putterman, L., 1997. Why have the rabble not redistributed the wealth? On the stability of
democracy and unequal property. In Roemer, J. E. (Ed), Property Relations, Incentives
and Welfare. London: Macmillan
Ravallion, M., Lokshin, M., 2000. Who wants to redistribute? The tunnel effect in 1990’s
Russia. Journal of Public Economics 76, 87-104.
Roemer, J., 1998. Why the poor do not expropriate the rich: an old argument in new garb.
Journal of Public Economics 70, 399-424.
Roemer, J., 2001. Political Competition, Harvard University Press, Cambridge MA.
Scheve K., Slaughter M., 2001. Globalization and the perception of American workers, Institute
for International Economics, Washington.
16
Table 1
Definitions of variables and descriptive statistics.
Variable Definition Obs Mean Std.
Dev.
Black 0, 1: if African-American 1063 0.124 0.330
Hispanic 0, 1: if Hispanic 1063 0.038 0.190
Asian 0, 1: if Asian 1063 0.077 0.267
Other 0, 1: if other. 1063 0.040 0.195
Female 0, 1: if female 1066 0.488 0.500
Business 0, 1: if major is business or economics 1052 0.634 0.482
Age Age in years 1054 20.658 3.826
Church 0, 1: Regular church goer 1068 0.379 0.485
Immigrant 0, 1: Self, or parents, are immigrants 1067 0.097 0.460
Vote 0, 1: if subject voted in 2000 Presidential election. 1056 0.439 0.497
Tax 0, 1: if subject paid taxes in 2000. 1061 0.622 0.485
Yc-Yp Difference in family’s economic status between now and 20
years ago. Both current (Yc) and past (Yp) status on 0 to
10 scale. 1056 1.355 2.210
Yc 0 to 10: Family’s current economic status. 1060 6.664 1.724
Yf-Yc 0 to 10: Family’s expected economic status 10 to 15 years
from now relative to current status. 1053 1.092 1.896
South 0, 1: respondents are from Alabama or Carolina 1096 0.523 0.500
Liberal 0 to 10: Political views from extreme conservative to
extreme liberal. 1044 5.269 2.269
Table 2
Correlations above |0.15|
Pair Correlation Pair Correlation
Black, Church 0.181 Black, Yf-Yc 0.167
Yc, Yc-Yp 0.465 Yf-Yc, Yc-Yp -0.369
Yf-Yc, Yc -0.721 Vote, Tax 0.167
Asian, migrant 0.298 Other, Migrant 0.158
Liberal, Church -0.184 Vote, Asian -0.169
Vote, Age 0.275 Tax, Age 0.200
Age, Migrant 0.251 South, Age -0.197
Black, South 0.279 South, Asian -0.189
South, Church 0.257 Yf-Yc, migrant 0.154
South, Tax -0.208 White, Yc 0.180
White, Migrant -0.238 White, Yf-Yc -0.168
Table 3
Ordered Probit: Dependent = 1, 2, 3
if oppose, unsure or support tax and transfer.
(a)
Multiple
Imputation
Obs. = 1096
(b)
List Delete
Obs. = 963
(c)
List Delete
Obs. = 963
BLACK 0.737** 0.672** 2.402**
(5.743) (5.014) (4.198)
HISPANIC 0.247 0.108 0.088
(1.197) (0.490) (0.398)
ASIAN 0.291 0.241 0.242
(1.861) (1.358) (1.363)
OTHER 0.230 0.299 0.281
(1.138) (1.288) (1.209)
YC-YP -0.025 -0.029 -0.029
-(1.161) -(1.323) -(1.332)
YC -0.190** -0.180** -0.158**
-(5.309) -(4.775) -(4.040)
YF-YC -0.151** -0.144** -0.145**
-(4.834) -(4.387) -(4.389)
AGE -0.010 -0.011 -0.013
-(0.863) -(0.881) -(1.074)
FEMALE -0.042 -0.050 -0.058
-(0.511) -(0.578) -(0.669)
BUS 0.015 0.002 -0.002
(0.177) (0.019) -(0.023)
CHURCH 0.069 0.083 0.088
(0.774) (0.895) (0.944)
MIGRANT 0.043 0.159 0.177
(0.427) (1.065) (1.184)
VOTE -0.125 -0.160 -0.165
-(1.446) -(1.747) -(1.798)
TAX 0.017 0.058 0.083
(0.194) (0.635) (0.902)
SOUTH -0.274** -0.280** -0.288**
-(2.925) -(2.774) -(2.843)
LIB 0.119** 0.122** 0.143**
(6.104) (6.018) (6.476)
BLACK*YC -0.160*
-(2.045)
BLACK*LIB -0.127*
-(2.235)
LR statistic 129.140 139.704
Probability 0.000 0.000
Pseudo-R2 0.079 0.085
*, ** indicate significance at the 5% and 1% levels.
z-Statistics are reported in the parentheses.
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
123 4 5 6 7 8 9 10
Probability of voting yes
Current income on a 10 point scale
Blacks
Whites
Figure 1 - Racial polarization in tax support
(includes half the uncertain vote)
Asians
.1
.2
.3
.4
.5
.6
.7
.8
123 4 5 6 7 8 9 10
Probability of supporting redistribution
Current income on a 10 point scale
Figure 2 - The Possibility of Upward Mobility
(includes half the uncertain vote)
Asian - average income increase
Asian - expected
income increase
Table 4: Responses to questions 2-4. All responses are coded on a 0 to 10 scale with 10
extremely important and 0 not important at all.
Statement Mean St.
Dev.
2. If you think the rich (high income) people should be taxed more so that
incomes become more equal, please indicate your reasons below. Obs = 273-
275
a) All men are created equal, and thus income and wealth -which
represent command over resources- should be distributed equally. 3.895 3.116
b) The rich got rich unfairly, thus income and wealth should be taken
away from the rich and given to the less fortunate. 2.015 2.581
c) The rich have a moral responsibility to help the less fortunate and the
poor. 5.821 3.127
d) The big gap between the rich and the poor may affect social stability
and narrowing the gap may reduce the tension between the rich and
the poor. 6.618 2.729
3. We would like to know why you think the rich should not be taxed more to
bring about more equality in income and wealth distributions. Obs. = 683-689
a) A person is entitled to do what he wishes with the fruits of his labor.
The wealthy earned their money and make fair tax contributions
already. 8.255 2.126
b) Higher taxes on the rich may reduce the incentive to save and thus
may reduce investment, growth and employment -which will only hurt
the poor. 6.410 2.527
c) If taxes go any higher the incentive to work will be too low and the
economy will grow too slowly. 5.732 2.701
d) I may not be rich today, but there is some chance that I, my children or
my grandchildren, will be rich in the future. So higher taxes on the
rich may mean I or my children have to pay more in the future. 5.319 3.289
e) Private property ownership allows personal freedom. Higher taxes
therefore restrict freedom and would trigger a taxpayer revolt. 5.709 2.888
4. Over the last two decades, tax rate reductions have been politically
popular. What do you believe is behind this fact? Obs. = 1041-1057.
a) The rich have more influence through their contributions to campaign
funds. 6.181 2.690
b) High income (rich) people are more likely to vote. Thus the outcomes
from voting result in lower taxes for the rich. 5.310 2.849
c) The global economy means we can not tax the rich too much or they
will take their investments and businesses offshore. 5.160 2.749
d) The rich spread the wealth just enough to get middle class support. 4.373 2.646
e) Voting is based on personality. Taxes are settled by back room politics. 4.034 2.910
f) Tax cuts may give more to the rich but middle income people will
support such tax cuts as long as their own taxes are reduced at least a
little. 7.040 2.311
Means and standard deviations are calculated from multiple imputed data sets and combined
according to the rules set out in King et al. (2001, p. 53).
Table 5
Importance of reasons 2a-2d for supporting redistribution.
Ordered Probit, multiple imputations, 288 observations
Dependent = responses to questions 2a-2d on a 0-10 scale.
a b c d
BLACK 0.594** 0.503** 0.450** 0.719**
(4.071) (2.910) (3.147) (5.139)
HISPANIC 0.391 -0.146 0.134 0.030
(1.766) -(0.505) (0.602) (0.133)
ASIAN 0.256 0.629** 0.205 0.253
(1.380) (3.300) (1.143) (1.436)
OTHER 0.428 0.236 0.326 0.301
(1.797) (0.882) (1.406) (1.290)
YC-YP -0.041 -0.070* -0.042 -0.024
-(1.662) -(2.526) -(1.779) -(1.026)
YC -0.172** -0.160** -0.143** -0.184**
-(4.256) -(3.551) -(3.654) -(4.742)
YF-YC -0.150** -0.123** -0.128** -0.130**
-(4.227) -(3.143) -(3.717) -(3.858)
AGE -0.022 -0.022 -0.001 -0.001
-(1.520) -(1.452) -(0.068) -(0.073)
FEMALE -0.163 -0.320** -0.224* -0.204*
-(1.685) -(2.894) -(2.405) -(2.240)
BUS -0.005 0.100 -0.083 -0.036
-(0.054) (0.883) -(0.877) -(0.384)
CHURCH 0.069 -0.039 0.156 0.076
(0.667) -(0.332) (1.551) (0.777)
MIGRANT 0.283 0.263 0.225 0.251
(1.838) (1.564) (1.516) (1.691)
VOTE -0.027 0.080 -0.096 -0.061
-(0.262) (0.703) -(0.982) -(0.637)
TAX 0.089 0.271* 0.108 0.056
(0.864) (2.244) (1.083) (0.571)
SOUTH -0.184 -0.274* -0.298** -0.329**
-(1.595) -(2.091) -(2.714) -(3.037)
LIB 0.130** 0.088** 0.160** 0.132**
(5.697) (3.372) (7.106) (6.110)
LR statistic (16 df) 104.931 94.517 121.555 129.458
Probability (LR stat) 0.000 0.000 0.000 0.000
LR index (Pseudo-R2) 0.061 0.077 0.062 0.065
*, ** indicate significance at the 5% and 1% levels.
z-Statistics are reported in the parentheses.
Table 6
Importance of reasons 3a-3e for supporting redistribution.
Ordered Probit, multiple imputations, 952-957 observations.
Dependent = responses to questions 3a-3e on a 0-10 scale.
a b c d e
BLACK -0.730** -0.451** -0.713** -0.486** -0.545**
-(5.775) -(3.664) -(5.698) -(3.857) -(4.345)
HISPANIC -0.104 0.089 -0.076 -0.356 0.013
-(0.527) (0.459) -(0.388) -(1.707) (0.067)
ASIAN -0.149 0.055 -0.118 -0.054 -0.125
-(0.903) (0.338) -(0.725) -(0.323) -(0.765)
OTHER -0.323 -0.477* -0.523* 0.117 -0.105
-(1.464) -(2.161) -(2.368) (0.545) -(0.492)
YC-YP 0.045* 0.021 0.009 0.011 0.039
(2.336) (1.167) (0.504) (0.578) (2.131)
YC 0.208** 0.100** 0.173** 0.224** 0.162**
(6.062) (3.040) (5.170) (6.470) (4.840)
YF-YC 0.162** 0.092** 0.148** 0.197** 0.147**
(5.450) (3.220) (5.088) (6.550) (5.056)
AGE 0.008 -0.001 0.005 -0.005 0.000
(0.702) -(0.130) (0.497) -(0.407) -(0.040)
FEMALE 0.009 -0.039 0.072 0.096 0.051
(0.124) -(0.549) (0.998) (1.305) (0.702)
BUS -0.008 0.045 0.002 -0.033 0.032
-(0.100) (0.611) (0.031) -(0.435) (0.429)
CHURCH -0.052 -0.035 -0.024 -0.093 -0.068
-(0.652) -(0.458) -(0.315) -(1.189) -(0.886)
MIGRANT -0.021 -0.182 0.054 -0.142 -0.043
-(0.153) -(1.350) (0.403) -(1.025) -(0.320)
VOTE 0.175* 0.179* 0.159* 0.130 0.085
(2.234) (2.351) (2.078) (1.676) (1.109)
TAX 0.016 -0.003 -0.036 0.036 -0.037
(0.204) -(0.042) -(0.469) (0.467) -(0.476)
SOUTH 0.279** 0.200* 0.323** 0.331** 0.180*
(3.197) (2.379) (3.814) (3.833) (2.120)
LIB -0.097** -0.077** -0.079** -0.073** -0.083**
-(5.593) -(4.646) -(4.743) -(4.330) -(4.945)
LR statistic 161.839 85.285 128.021 125.509 104.203
Probability 0.000 0.000 0.000 0.000 0.000
Pseudo-R2 0.049 0.023 0.034 0.034 0.028
*, ** indicate significance at the 5% and 1% levels.
z-Statistics are reported in the parentheses.
Table 7
Importance of reasons 4a-4e for supporting redistribution.
Ordered Probit, multiple imputations, 932-945 observations.
Dependent = responses to questions 4a-4f on a 0-10 scale.
a b c d e f
BLACK 0.103 0.396** -0.127 -0.010 0.196 -0.015
(0.887) (3.403) -(1.100) -(0.086) (1.680) -(0.128)
HISPANIC 0.104 0.212 -0.072 -0.107 -0.363* -0.210
(0.578) (1.180) -(0.400) -(0.586) -(1.958) -(1.165)
ASIAN -0.109 0.132 0.173 0.057 0.602** 0.043
-(0.736) (0.885) (1.163) (0.383) (4.005) (0.288)
OTHER -0.144 0.046 0.447* -0.228 -0.151 0.182
-(0.730) (0.229) (2.199) -(1.108) -(0.728) (0.877)
YC-YP -0.009 -0.026 -0.018 -0.035 -0.008 0.028
-(0.512) -(1.496) -(1.038) -(1.990) -(0.479) (1.588)
YC -0.001 0.035 0.104** 0.037 0.041 -0.028
-(0.028) (1.147) (3.383) (1.193) (1.310) -(0.905)
YF-YC -0.006 -0.018 0.071** 0.010 0.015 -0.009
-(0.228) -(0.672) (2.663) (0.373) (0.540) -(0.342)
AGE 0.002 0.002 0.007 -0.014 -0.026 -0.004
(0.172) (0.207) (0.693) -(1.267) -(2.227) -(0.370)
FEMALE 0.073 0.083 -0.205** -0.003 -0.122 0.022
(1.057) (1.206) -(2.992) -(0.045) -(1.747) (0.322)
BUS 0.014 -0.014 0.112 0.012 0.017 -0.046
(0.191) -(0.194) (1.587) (0.166) (0.237) -(0.640)
CHURCH 0.002 -0.086 -0.121 0.041 0.007 -0.069
(0.025) -(1.179) -(1.656) (0.558) (0.098) -(0.936)
MIGRANT 0.264* -0.118 -0.066 0.238 0.070 -0.147
(2.115) -(0.943) -(0.525) (1.914) (0.548) -(1.166)
VOTE 0.034 0.134 0.072 -0.071 -0.110 -0.049
(0.473) (1.847) (1.002) -(0.979) -(1.496) -(0.679)
TAX -0.032 0.033 0.079 -0.046 0.119 -0.055
-(0.433) (0.460) (1.088) -(0.631) (1.619) -(0.752)
SOUTH 0.243** 0.124 -0.013 0.123 -0.094 0.025
(3.018) (1.552) -(0.161) (1.519) -(1.157) (0.315)
LIB 0.084** 0.058** 0.008 0.040* 0.039* 0.012
(5.249) (3.652) (0.482) (2.503) (2.423) (0.747)
LR statistic 49.079 47.448 41.072 26.578 52.875 10.705
Probability 0.000 0.000 0.001 0.046 0.000 0.827
Pseudo-R2 0.012 0.011 0.010 0.006 0.013 0.003
*, ** indicate significance at the 5% and 1% levels.
z-Statistics are reported in the parentheses.
Questionnaire On Taxes, Income Redistribution and Wealth
In democratic countries such as the U. S., voters participate either directly or indirectly in
forming public policies on taxes and income distribution. In this set of questions, we are
interested in whether you believe taxes should rise or fall and your opinion about the forces that
shape tax policy. There are no right or wrong answers – all we want to know is your opinion.
Also please do not write your name anywhere on this questionnaire. We sincerely thank you for
your participation.
1. Do you think rich (high income) people in the U.S. should be taxed more with the money given to
the less fortunate so that income and wealth are more equally distributed? Circle your response.
a) Yes b) No c) Don’t know or not sure
If you answered “Yes” to question 1, please answer questions 2 and 4. Do not answer question 3.
If you answered “No” to question 1, please answer questions 3 and 4. Do not answer question 2.
If you answered “Don’t know or not sure,” please answer only question 4.
We also like you to answer the questions in the bottom of page 2.
2. If you think the rich (high income people) should be taxed more so that incomes become more
equal, please indicate your reasons below. Use a 0 – 10 scale to indicate the importance of each
of your reasons (0 – not important at all, …, 5 – somewhat important, …, and 10 – extremely
important).
______ a) All men are created equal and thus income and wealth – which represent people’s
command over resources – should be equally distributed.
______ b) The rich got rich unfairly, thus income and wealth should be taken away from the rich
and given to the less fortunate.
______ c) The rich have a moral responsibility to help the less fortunate and the poor.
______ d) The big gap between the rich and the poor may affect social stability and narrowing the
gap may reduce the tension between the rich and the poor.
______ e) Other reasons (please state):
3. We would like to know why you think the rich should not be taxed more to bring about more
equality in income and wealth distributions. Please indicate your reasons by assigning an
indicator of importance. Use a 0 – 10 scale for each of your reasons (0 – not important at all, …,
5 – somewhat important, …, and 10 – extremely important).
______ a) A person is entitled to do what he wishes with the fruits of his labor. The wealthy earned
their money fairly and make fair tax contributions already.
______ b) Higher taxes on the rich may reduce incentives to save and thus may reduce investment,
growth and employment – which will only hurt the poor.
______ c) If taxes go any higher the incentive to work will be too low and the economy will grow
too slowly.
______ d) I may not be rich today, but there is some chance that I, my children or my
grandchildren, will be rich in the future. So higher taxes on the rich may mean I or my
children have to pay more in the future.
______ e) Private property ownership allows personal freedom. Higher taxes therefore restrict
freedom and would trigger a taxpayer revolt.
______ f) Other reasons (please state):
4. Over the last two decades, tax rate reductions have been politically popular. What do you believe is
behind this fact? Please assign an indicator of the importance to each of your choices. Use a 0 – 10
scale for each of your reasons (0 – not important at all, …, 5 – somewhat important, …, and 10 –
extremely important).
______ a) The rich have more influence through their contributions to campaign funds.
______ b) High income (rich) people are more likely to vote. Thus the outcomes from voting
result in lower taxes for the rich.
______ c) The global economy means we cannot tax the rich too much or they will take their
investments and businesses offshore.
______ d) The rich spread the wealth just enough to get middle class support.
______ e) Voting is based on personality. Taxes are settled by back room politics.
______ f) Tax cuts may give more to the rich but middle income people will support such tax cuts
as long as their own taxes are reduced at least a little.
______ g) Others (Please state):
Thank you very much for answering our questions. Now we want to know something about you (but
not your identity, please do not write your name anywhere on this questionnaire).
I. Please circle your ethnicity:
a) White b) African American c) Hispanic d) Asian e) Other __________
II. Please circle your gender
a) Male b) Female
III. Please state your major or profession here: ____ ________
IV. Please state your age here: ____
V. Do you regularly go to church? Please circle your response.
a) Yes b) No
VI. Are you, your mother or your father immigrants? Please circle your response.
a) Yes b) No
VII. Please provide a summary indicator of your family’s economic status 20 years ago using
a scale from 0 to 10 (0 – extremely poor, 1 – poor, 2 – moderately poor, … 5 – about
average, …, 8 – well to do, 9 – rich and 10 – extremely rich):
VIII. Please provide a summary indicator your or your family’s current economic situation
using a scale from 0 to 10 (0 – extremely poor, 1 – poor, 2 – moderately poor, …, 5
– about average, …, 8 – well to do, 9 – rich and 10 – extremely rich): _______
IX. Please indicate your expectation about you or your children’s future economic status, say
10 or 15 years from now: (0 – extremely poor, 1 – poor, 2 – moderately poor, …,
5 – about average, …, 8 – well to do, 9 – rich and 10 – extremely rich): ___
X. Please provide a summary indicator of your political views. Use a 0 to 10 scale
(0 – extremely conservative, …, 5 – about average, …, and 10 – very liberal) : ______
XI. Did you vote in the 2000 Presidential election?
a) Yes b) No
XII. Did you personally pay income taxes in 2000?
a) Yes b) No
Thank you very much for your participation!
