Monday, 28 April 2014

Mind the gap: Overestimating income inequality

How much money do you think you would have to make each year to land yourself in the infamous One Per Cent of salary earners?

According to a new study, your answer is very likely to be wrong. Research conducted by John Chambers of St. Louis University and colleagues at the University of Florida reveals that – at least where Americans are concerned – people are actually significantly likely to overestimate how much money is earned by the richest people.

This finding directly contradicts recent research coming from other labs, including a widely-publicised 2011 study conducted by Michael Norton and Dan Ariely reporting widespread underestimation of American wealth inequality. According to Norton and Ariely, many Americans who responded to their survey seemed to have a very poor idea of how drastic the current state of wealth inequality really is, estimating that the bottom 40 per cent of American society accounts for a much higher percentage of overall American wealth than they really do, and not grasping the enormity of wealth held by the top 20 per cent of Americans.

Chambers and his co-authors, on the other hand, are arguing that Norton and Ariely’s original conclusion does not actually reflect how people truly think about American society. Rather, they think that the original results may have been biased by the specific way in which these questions were asked.

In the original Norton and Ariely survey, participants were asked to estimate how much wealth is accounted for by the top 20 per cent of Americans. However, Chambers and his colleagues believe that asking the question this way is fairly abstract, and people might be more likely to respond by using quick, snap-judgment types of answers – which might not reflect how they actually think or feel.

In contrast, Chambers and colleagues asked Americans to estimate the percentage of US citizens whose annual incomes fell into three categories: Under $35,000, $35,000-$74,999, and $75,000 or higher. These numbers roughly represent the cutoff points for the bottom, middle, and top third of American earners. Yet, on average, the respondents in this survey greatly overestimated the percentage of Americans who fall into the lowest category (estimating 48 per cent) and underestimated the percentage of Americans who fall into the highest category (23 per cent).

The researchers also found similar results when they specifically tried to get respondents to provide estimates of how big the income gap itself happens to be. On average, people thought that incomes earned by the 'top 20 per cent' in 2010 were about 31 times greater than those earned by the bottom 20 per cent. Yes, there is a significant degree of income inequality in the United States – but it’s not quite that high (in reality, the top 20 per cent earns incomes that are 15 times greater than the bottom 20 per cent). Further probing into this inaccuracy reveals that people aren’t exactly equal in their misperceptions, either. In fact, this effect is largely caused by severe overestimations of the incomes of top earners. Although the average income for someone in the top 20 per cent was around $170,000 in 2010, participants estimated that this figure was actually closer to $2 million.

As the authors simply asked participants to provide their own numerical estimates, I was concerned and curious if the high average could be explained by a few participants providing outrageously large estimates and skewing the mean, so I asked Chambers about the distribution of the responses that they received. Although Chambers confirms that the distribution was positively skewed and there were some of these outrageously high estimates in the mix (10 per cent of the sample estimated that the average income of the top 20 per cent was over $10 million each year), the median estimate was $287,500 – still much higher than the true amount ($169,000). About 40 per cent of the participants gave estimates of $400,000 or greater, and the data were 'Winsorized', a procedure that minimises the impact of outliers when computing an average. So, these results cannot simply be explained by a few participants skewing the numbers!

It is important for Americans to recognise the extent of income inequality in the United States. However, if we perpetuate the idea that this inequality is significantly worse than it actually is, it could end up having severe consequences for rectifying the actual problems that exist. If people have reason to suspect that these claims are being exaggerated, they may doubt there is any problem at all.

Chambers JR, Swan LK, & Heesacker M (2014). Better off than we know: distorted perceptions of incomes and income inequality in America. Psychological science, 25 (2), 613-8 PMID: 24317422

Post written for the BPS Research Digest by guest host Melanie Tannenbaum, UIUC Social Psych PhD Candidate and Scientific American Blogger.


Anonymous said...

Does this actually contradict the study by Norton and Ariely? It seems their study was to do with wealth inequality, whereas this has to do with income. Obviously, there are going to be relationships between these two (income can be readily converted to wealth, after all), but I don't see why it couldn't simply be the case that people overestimate income inequality and underestimate wealth inequality.

Melanie Tannenbaum, M.A. said...

That's a great question!

You are absolutely right that the Norton & Ariely paper asked about "wealth," whereas the new Chambers et al. paper asked about "income." However, there are a couple of things to consider.

First of all, it might be worth providing the explanation given by Chambers and colleagues as to why they use income instead of wealth: "For this investigation, we concerned ourselves narrowly with Americans’ perceptions of incomes and income inequality, adopting the assumption that most Americans probably are more familiar with and therefore have an easier time conceptualizing relative incomes than net worth."

But, that wasn’t quite your question (which raised a valid conceptual question).

So, secondly, regarding the possibility that these two findings are reconcilable if Americans systematically underestimate wealth inequality and overestimate income inequality, Chambers and colleagues point out in the introduction of their paper: "Eriksson and Simpson (2012) compared Norton and Ariely’s method — asking study participants to grasp the abstract concept of aggregated net worth across households (i.e., the percentage of the country’s overall wealth owned by each quintile) — with a method involving a logically equivalent but computationally simpler criterion — the average wealth of individual households within a given quintile. They found that the latter (average) method yielded dramatically higher estimates of wealth inequality than the former (percentage) approach."

It appears that prior research actually has found that simply altering the way in which the question is asked (by making it about more concrete, easy-to-grasp metrics, like individual household averages rather than overall quintile ownership) can reveal overestimates of *wealth* inequality as well. For me, that makes it easier to believe that this is more of a measurement issue (and a difficulty in grasping abstract things like “the top 20%”) than a systematic difference in how (in)equality is perceived.

I hope that addressed your question — I would love to hear other thoughts as well, however!

- Melanie

Mark L. said...

It seems like the differences between this study and Norton and Ariely are more than just income vs. wealth, but also the kind of reasoning they ask for or assume subjects are doing. Asking about income seems like it supposes numerical reasoning, while asking about wealth seems to suppose proportional reasoning. This might be easy and similar enough at quantities I'm familiar with from my own or my family's finances, but once we're talking about the top earners or wealthiest in the country, the numbers start getting very abstract, and I'm not sure a lot of people would use much of any mathematical reasoning at all. Once we get to 1 million a year, or 500,000 a year, or a net worth of 15 million, those quantities get less numerical and get more intuitively just "big-feeling." In other words, I might be able to tell you that triple of 5 million is 15 million, but if we're talking about money a household has, I can't tell the difference between very large numbers. They all feel the same to me: more money than I (or anyone) would ever, ever, ever need.

What do you think? What kind of effect do other kinds of reasoning have on this study? Did any of these authors look into it?

Katie Mateo said...

In America there is a much bigger wealth gap than there is an income gap. So it would make sense that people would tend to conflate the two, and over-estimate the income gap, and under-estimate the wealth gap.


Post a Comment

Note: only a member of this blog may post a comment.