Piketty and U.S. Wealth Inequality

It seems crazy to have such a heated debate on facts that are so easily measurable. Here we use the SCF from 1992 to 2010 to show the pretty unambiguous rise in wealth inequality in the United States. The housing boom hid the rise in inequality to some degree, but the crash in house prices from 2007 to 2010 made inequality transparent.

At the end of this post we tell you where to get the data and we give you the Stata code. We did this pretty quickly, so happy to be told if we’ve made some mistake.

The SCF is not a huge sample, so we have chosen to look at the top 20% (80% to 100%) of the wealth distribution versus the middle 20% (40% to 60%) of the distribution every year. Below, we plot the ratio of wealth of the richest 20% versus the middle 20%. We took the averages within each of these quintiles.

houseofdebt_20140524_1

The graph shows that the top 20% of the wealth distribution had 15x the wealth of the middle 20% of the distribution in 1992. In 2010, the richest 20% has more than 25x the wealth of the middle 20%. How is that not a substantial increase in wealth inequality?

The pattern is especially stark if we exclude home equity. This is a useful exercise because we know from 2010 to 2013 financial assets have performed very well. Housing has performed decently as well, but much of the strong performance of housing is driven by investors who are likely part of the richest 20% of the distribution. We don’t yet have the 2013 SCF available.

To be clear, the split of households into quintiles is still based on total net worth. The graph below plots the ratio of non-home equity wealth of the richest 20% to the middle 20%:

houseofdebt_20140524_2That looks like a pretty sharp rise in wealth inequality to us. In 1998 the richest 20% had 25x the non-home equity wealth of the middle 20%. it has risen to almost 45x in 2010.

Just to complete the story, here is inequality using home equity wealth:

houseofdebt_20140524_3

The rise in wealth inequality focusing only on home equity is smaller in magnitude, but goes in the same direction overall. Interestingly, the middle 20% made gains on the top 20% from 2004 to 2007, which reflects the fact that house prices rose fastest in middle and lower income areas, something we have documented in our research.

We are happy to be corrected here. Like we said earlier, we did this quite fast. The data sets are available from the Fed. They are the Stata data sets based on the SCF bulletins. For example, here is the link to the 2010 data. And here is our Stata code — let us know if there is a mistake:

**5/24/2014: Piketty and SCF
foreach num of numlist 1992 1995 1998 2001 2004 2007 2010{
cd $scf
use rscfp`num’, clear
sort YY1
by YY1: keep if _n==1
xtile t1=networth [aw=wgt], nq(5)
foreach x of numlist 1/5{
gen networth`x’=networth if t1==`x’
gen networthhom`x’=homeeq if t1==`x’
gen networthnhm`x’=(networth-homeeq) if t1==`x’
}

collapse (mean) networth? networthhom? networthnhm? [aw=wgt]
gen year=`num’
cd $datapath
save t`num’, replace
clear
}
foreach num of numlist 1992 1995 1998 2001 2004 2007 2010{
append using t`num’
erase t`num’.dta
}
foreach y in networth networthhom networthnhm{
gen `y’rat=`y’5/`y’3
}
*graphs
# delimit ;
graph twoway line networthrat year,
scheme(s1mono) title(“Ratio of richest 20% to middle 20%” “Total wealth”)
xtitle(“”) ytitle(“”)  lc(maroon) lw(thick)
xlabel(1992 1995 1998 2001 2004 2007 2010) lw(thick)
caption(
“houseofdebt.org, @AtifRMian & @profsufi, Data source: SCF”,
size(vsmall)
) xsize(4.25) ysize(3);
# delimit cr
graph export “$output\houseofdebt_20140524_1.png”, replace
# delimit ;
graph twoway line networthnhmrat year,
scheme(s1mono) title(“Ratio of richest 20% to middle 20%” “Non-home equity wealth”)
xtitle(“”) ytitle(“”) lc(navy) lw(thick)
xlabel(1992 1995 1998 2001 2004 2007 2010) lw(thick)
caption(
“houseofdebt.org, @AtifRMian & @profsufi, Data source: SCF”,
size(vsmall)
) xsize(4.25) ysize(3);
# delimit cr
graph export “$output\houseofdebt_20140524_2.png”, replace
# delimit ;
graph twoway line networthhomrat year,
scheme(s1mono) title(“Ratio of richest 20% to middle 20%” “Home equity wealth”)
xtitle(“”) ytitle(“”) lc(dkgreen) lw(thick)
xlabel(1992 1995 1998 2001 2004 2007 2010) lw(thick)
caption(
“houseofdebt.org, @AtifRMian & @profsufi, Data source: SCF”,
size(vsmall)
) xsize(4.25) ysize(3);
# delimit cr
graph export “$output\houseofdebt_20140524_3.png”, replace

Bookmark and Share

9 Responses to Piketty and U.S. Wealth Inequality

  1. an interested economist on May 24, 2014 at 12:05 pmi

    I very much appreciate that you did this, and it’s an interesting and important fact that you document here, but this does not directly respond to most of the discussion. As the extreme ratios seen here (on the order of ~20) indicate, the middle 20% has very little wealth compared to the top 20%, and this has always been true. I don’t think many conservative critics are trying to argue one way or another on this front.

    The current discussion is more about the concentration of wealth at the very top, particularly the 1%. And there the SCF shows little to no evidence to support increased wealth inequality – only a minimal rise in the share of wealth held by the top 1%. This is what Kopczuk and Schrager’s article is referencing, and this is the most relevant question for the debate about Piketty’s (and Saez and Zucman’s) findings of higher wealth inequality at the top.

    You really need to look at *that* issue, and if you think this is impossible because “the SCF is not a huge sample” (though it does oversample at the top), you need to say so, rather than passing off an interesting but essentially distinct point as being a decisive response to critics – which, frankly, is what you’re doing in this post.

    • kjmclark on May 28, 2014 at 8:55 pmi

      Are you looking at some other graphs? The ones I’m looking at, in the blog post I would have thought we both read, pretty clearly show that back in the 90s, the middle 20% had a much higher portion of the wealth compared to the top 20%. And that’s only going back to the 90s, not the 60s or 50s. It most certainly hasn’t always been the case that the middle income quintile had so little wealth compared to the top quintile. You’re simply looking at some other information if you think that.

  2. Zack on May 24, 2014 at 12:28 pmi

    It looks like you have written 45% where you meant to write 45x (under 2nd chart)

  3. Zvi Goldstein on May 24, 2014 at 1:01 pmi

    I’m glad to see this here, as income inequality is finally become a topic for economists to take seriously.

    Given declining marginal utility of wealth, variance in wealth leads to less than the maximum utility possible for a given level of wealth. So, the topic warrants study.

    One point: rather than the standard comparing quantiles, we should aim to chart the standard deviation of wealth.

    It might yield a metric easier to compare across countries, using a sort-of sharpe ratio: mean_wealth/standard_deviation_of_income.

  4. Jay on May 24, 2014 at 4:32 pmi

    More financial illiterates that believe accrued Social Security, Medicare and pension benefits due are not assets.

    • Richard on May 28, 2014 at 10:11 ami

      The portion of companies supporting strong pensions has decreased drastically, and the benefits from SS and Medicare aren’t measurably more than 20 decades ago, so wouldn’t the increase in the gap look even worse if you counted all those?

  5. pete on May 28, 2014 at 1:19 pmi

    Re: Jay…there are no accrued SS or medicare benefits. For each individual there is the present value of future (expected) payments, which are approved by legislation and subject to change (e.g., as in the 1980s), and, unlike a 401K, disappear upon death (except for dependant minors).

    Otherwise interesting post, other than ignoring critically that it is the .01% that kicks everything. I think this is called cherry picking the numbers.

  6. Jon on May 28, 2014 at 2:04 pmi

    The SCF for 1989 – 2010 is up in an easy to look at and manipulate form at http://sda.berkeley.edu/cgi-bin36/hsda?harcscfcomb+scfcomb

    ASSETCAT reflect the total asset percentile group for each year, and assets by type can be summarized (Means comparison)by asset percentile group and year very easily.

  7. Patrick Haughey on May 29, 2014 at 11:14 ami

    math is awesome. Best most simple example I have seen yet in this ridiculous debate. I think all economic policy pundits and think-tanks should publish their sources and excel spreadsheets