American’s desire for less wealth inequality does not depend on how you ask them

Abstract

A large body of survey research offers evidence that citizens are not always fully aware of the economic and political realities in their respective countries. Norton and Ariely (2011) extended this research to the domain of wealth inequality, showing that Americans were surprisingly unaware of the shape of the wealth distribution in America. Using an alternative methodology, Eriksson and Simpson (2012) found that asking Americans to estimate the average wealth of quintiles, rather than the percent of wealth owned by each quintile, led to relatively more accurate estimates. We note, however, that the Eriksson and Simpson (2012) results do not challenge Norton and Ariely’s (2011) conclusion that Americans desire a much more equal distribution of wealth.

Inequality in wealth and income has increasingly become a topic of interest not only to academics (e.g., Alesina & Angeletos, 2005; Bartels, 2005; Picketty & Saez, 2003) but also to policymakers—as witnessed by the rise of the Occupy Wall Street movement and the role that inequality plays in defining issues such as taxation and universal healthcare. In our recent research (Norton & Ariely, 2011), we made three primary points in an effort to contribute to the academic literature and public policy debates on inequality: 1) People underestimate the current level of wealth inequality in America; 2) people’s ideal level of inequality is more equal than their estimates; and 3) Americans across the political and economic spectrum show consensus in stating that their ideal level is more equal than the current level.

Eriksson and Simpson (2012) use an interesting alternative methodology that sheds light on points 1 and 2 above (they do not offer data on point 3 and so we will not further address the issue of consensus). We will address points 1 and 2 in turn below, but our primary response can be summarized as: we believe there is more agreement than disagreement between the two papers, particularly in their joint demonstration of Americans’ desire for more equality in the distribution of wealth. We conclude by offering suggestions as to why the two methodologies produce different estimates, as well as directions for future research.

First, we define the differences between the two methodologies, using the terminology proposed by Eriksson and Simpson (2012).

The Percent Measure. The Norton and Ariely (2011) methodology asks respondents to estimate the percentage of wealth owned by each of the five wealth quintiles.

The Average Measure. The Eriksson and Simpson (2012) methodology asks respondents to estimate the average household wealth (in dollars) for each of the five wealth quintiles.

Accuracy of Estimates. A large body of survey research has documented people’s general lack of awareness of economic and political realities, such as the effects of taxes and other government programs on their own outcomes (Delli Carpini & Keeter, 1996; Gilens, 2001). For example, despite the enormous increase in income inequality in the United States in the 20 years between 1980-2000, one survey showed that fully 25% of Americans were unaware that incomes had become more unequal (Bartels, 2005). Indeed, recent research suggests that one reason that people may be unaware of the shape of the income distribution is that they are unaware of their own place in that distribution; in one survey, just 15% of Argentinians placed themselves in the correct income quintile (Cruces, Perez-Truglia, & Tetaz, 2013).

Norton and Ariely (2011) took the approach of asking Americans to estimate not their own position in the distribution of wealth in America, but rather the overall shape of that distribution—by estimating the percentage of wealth owned by each wealth quintile. One simple means of summarizing those estimates is to compare the estimated ratio of wealth between the richest and poorest quintiles. The Norton and Ariely (2011) Percent methodology led respondents to estimate a ratio between the poorest and richest quintiles of 1:19.8, a result closely replicated when Eriksson and Simpson (2012) used our Percent methodology (1:21.2). In contrast, Eriksson and Simpson’s (2012) Average measure led respondents to estimate a ratio of 1:1500.

Note that the actual ratio between the richest and poorest is on the order of 1:1000. We concede that the 1:1500 estimate generated by the Average measure is closer in spirit to 1:1000 than the roughly 20:1 ratio generated by the Percent measure. We point out, however, that estimating a 1:1000 ratio as 1:1500 remains far from accurate; students generating that answer to some arithmetic problem on an exam would fail just as readily as those who came up with 1:20. Again conceding that the Average measure reveals conceptually more accurate responses than our Percent measure, it is possible to view the two papers as offering different methodologies that cause respondents to either under- or over-estimate wealth inequality, with neither producing accuracy.

Desires for Ideals. As to ideal distributions of wealth, we believe that the Norton and Ariely (2011) and Eriksson and Simpson (2012) results—despite their differences—in fact point to a very similar conclusion. Eriksson and Simpson (2012) show that their Average measure elicits estimates of ideal inequality that are more unequal than those elicited by the Norton and Ariely (2011) Percent measure. However, it is important to note that both methodologies elicit ideal levels of inequality that are vastly more equal than respondents’ estimates of the current level. The Average measure in Eriksson and Simpson (2012) led respondents to express a desire for an ideal ratio between the poorest and richest quintiles of 1:50, which they correctly note is much less egalitarian than the ideal ratio produced by the Norton and Ariely (2011) Percent measure (1:4).

However, as noted above, the Eriksson and Simpson (2012) Average measure elicited an estimated ratio of 1:1500 and an ideal ratio of 1:50—note the large discrepancy—while the Norton and Ariely (2011) Percent measure elicited an estimate ratio of 1:20 and an ideal ratio of 1:4. In short, the Average measure reveals that people may prefer relatively greater inequality than the Percent measure, but both ratios (1:50 and 1:4) are far more equal than those same respondents’ estimates (1:1500 and 1:20)—suggesting a consensus desire for a more equitable distribution.

Future Directions. Despite some similarities, it is clear that future research is needed to explore why the two methodologies generate such different estimates. Certainly, there is evidence that laypeople approach the issue of inequality using both Percent (“we are the 99%” or “the 47%” of non-contributing Americans to whom Mitt Romney referred in leaked remarks) and Average mindsets (with debates about raising taxes on people making over $250,000 or the Buffett tax), suggesting that both methodologies hold promise for understating lay beliefs.

First, it is likely that the two methodologies prime different notions of inequality that then influence estimates. For example, one crucial difference is that the Percent measure is zero-sum (respondents must divide a fixed pie of wealth) whereas the Average measure is not bounded (respondents can make people as rich or poor as they like); the former may prompt more concerns about equality than the latter. Second, different methodologies may prime different notions of what variables should be considered in estimating inequality. For example, the Norton and Ariely (2011) and Eriksson and Simpson (2012) both defined wealth as the sum of an individual’s assets and debts, excluding other forms of wealth such as Social Security and Medicare (see Alesina, Glaeser, & Sacerdote, 2001); respondents with this social safety net in mind may express different estimates of the actual disparity in wealth. (Note that we do not suggest these are exhaustive or even the most fruitful avenues for future research.)

In sum, while Americans may be better at estimating the current level of wealth inequality than the Norton and Ariely (2011) results suggest, the Eriksson and Simpson (2012) results do not challenge Norton and Ariely’s (2011) conclusion that Americans desire a much more equal distribution of wealth.