Edward Conard

Top Ten New York Times Bestselling Author

  • “Unintended Consequences offers deep and well-argued analyses on almost every issue.” - The New York Times
  • “…reminds us that inequality sends a signal of what society lacks most, in America’s case, entrepreneurship and risk taking.” - Lawrence Lindsey, CEO, The Lindsey Group, former Director of the National Economic Council
  • “…a fresh argument for the productive value of inequality.” - David Autor, Professor of Economics, Massachusetts Institute of Technology
  • “…challenges misconceptions that distort our economic debates.” - Arthur Brooks, President of the American Enterprise Institute
  • “…a fresh argument for the productive value of inequality.” - David Autor, Professor of Economics, Massachusetts Institute of Technology
  • “…serious thinking for serious thinkers. …a thought-provoking blueprint for growing middle- and working-class incomes.” - Mitt Romney, former Governor of Massachusetts
  • “…challenges misconceptions that distort our economic debates.” - Arthur Brooks, President of the American Enterprise Institute
  • “…a very valuable contribution.” - Larry Summers, former Secretary of the Treasury and director of the National Economic Council, president emeritus, Harvard University
  • “…a must-read for serious students of economic policy.” - Glenn Hubbard, Dean, Columbia Business School, and former Chairman of the Council of Economic Advisers
  • “…reminds us that inequality sends a signal of what society lacks most, in America’s case, entrepreneurship and risk taking.” - Lawrence Lindsey, CEO, The Lindsey Group, former Director of the National Economic Council
  • “…a comprehensive explanation of the modern economy.” - Julian Robertson, Founder, Tiger Management
  • “…a comprehensive explanation of the modern economy.” - Julian Robertson, Founder, Tiger Management
Upside of Inequality Oxford Unintended Consequences

AI Summary. The authors argue that a widely cited estimate of tax evasion overstates it by about $460bn, reaching $1,230bn in filer evasion versus roughly $770bn implied by national accounts. They attribute the gap to detection-controlled estimation multipliers that reallocate detected evasion disproportionately to top income brackets.

David Splinter Working Paper
Date Posted:
Is Database:
Database

@DavidSplinter Working Paper: Auten and Splinter argue that Zucman’s methodology overestimates tax evasion by approximately $460bn.

Are estimates of tax evasion by top earners overstated?

Core argument: Auten and Splinter argue that Zucman's methodology overstates tax evasion by roughly $460bn, with his estimate of $1,230bn in filer evasion exceeding the roughly $770bn reported in national accounts by more than half.

Auten and Splinter argue that Zucman's methodology overestimates tax evasion by approximately $460bn. They highlight that Zucman's approach, which includes detection-controlled estimation (DCE) multipliers, significantly inflates evasion figures by reallocating detected evasion disproportionately to the top income brackets. Their analysis suggests that national accounts report about $770bn in filer evasion, while Zucman's estimates reach $1,230bn, including capital gains not accounted for in national income. This discrepancy indicates that Zucman's figures exceed national income estimates by more than half, suggesting a need for more distributionally consistent methods to assess true top income shares and underreporting rates.

Takeaways by Macro Roundup® AI

  1. Auten and Splinter argue that Zucman's methodology overstates tax evasion by roughly $460bn, with his estimate of $1,230bn in filer evasion exceeding the roughly $770bn reported in national accounts by more than half.
  2. Auten and Splinter contend that Zucman's detection-controlled estimation multipliers inflate evasion figures by reallocating detected evasion disproportionately to top income brackets.
  3. Auten and Splinter note that Zucman's $1,230bn evasion estimate includes capital gains not counted in national income, and they call for more distributionally consistent methods to assess top income shares and underreporting rates.

Gerald Auten and David Splinter, "Comment: Tax Evasion at the Top of the Income Distribution: Theory and Evidence," Working Paper, May 5, 2021, http://www.davidsplinter.com/AutenSplinter-TaxEvasion.pdf

Their takeaway, “…While some increase in high-income audit rates may be appropriate, additional enforcement resources could also help audit selection by better identifying returns with lower reported incomes but higher likelihood of evasion. The combined efforts and increased cooperation of IRS divisions could lead to improved methods of selecting returns for audit and help maintain high ROIs for enforcement activities…. We note in this comment, however,that estimates of evasion not detected in NRP audits using simple DCE multipliers and allocating in proportion to reported income are not distributionally consistent. Moreover, we believe these approaches tend to overstate true top incomes and underreporting rates. We suggest alternative methods that would be more distributionally consistent. These suggest substantially less impact of evasion on top income shares. We are aware that there is considerable additional sophisticated evasion by some high-income individuals not found in NRP audits….”

They make the case this overstates the top income share, “.. Two-thirds of the estimated increase in top income shares results from applying DCE multipliers. This approach causes exaggerated upward re-ranking of returns with large amounts of detected evasion,shifting estimated evasion from the bottom and middle to the top of the income distribution. The second largest increase in top shares results from allocating additional passthrough evasion by reported income. However, returns with business losses appear to be allocated little additional evasion despite accounting for over one third of evaded income….While offshore evasion has received considerable attention, it accounts for only a small portion of the total effect on top income shares….”
Bottomline, “…National accounts explicitly break out proprietor evasion (combined amount for sole proprietors and partnerships) of $561 billion and wage evasion of $75 billion when averaging 2006-2013 (at 2012 dollars). Auten and Splinter (2019a, online data Table T1) provide estimates of likely evasion in national accounts for other income sources. For 2006-2013, there is annual evasion of about $44 billion for farms, $43 billion for rents, and $80 billion for S corporations. To compare with GLRRZ, there are smaller amounts for dividends and interest evasion that should be added and non-filer portions that should be removed from the values above. If we assume the former is $50 billion and the latter is 10 percent,this suggests about $770 billion in filer evasion in national income. In comparison, GLRRZ add $1,304 in audit-based evasion in 2012 (Table A6, exam and DCE columns). But this amount includes capital gains realizations that are not in national income. Excluding $70 billion in added evaded capital gains (5.3 percent of total based on Table A1) suggests that GLRRZ add audit-based filer evasion comparable to national income definitions of about $1,230 versus about $770 billion in national income.11 Therefore, the GLRRZ filer evasion estimate appears to exceed amounts in national income by $460 billion, or more than one half.This represents an additional 3 percent of national income…”

They describe Zucman’s methodology, “…GLRRZ’s analysis has four steps. First, it uses audit study micro data to add detected evasion. Second, GLRRZ imputes undetected evasion using a set of multipliers developed from earlier audit data. These multipliers are from detection-controlled estimation (DCE) and can nearly quadruple the amount of evasion in audit studies.Third, it adds imputations of additional evasion from passthrough entities (including partnerships, S corporations, and certain other sources). The amounts already in the NRP audits are removed and then this assumed total passthrough evasion is allocated by reported income, such that essentially all goes to the top of the distribution. Fourth, GLRRZ estimates how income from offshore wealth affects income distributions... Combining the four components of estimated evasion, GLRRZ argue that top 1% fiscal income shares increase by 1.5 percentage points (pp). This includes a 0.5pp decrease from detected evasion, 1.1pp increase from undetected evasion using DCE multipliers, 0.6pp increase from additional business evasion, and 0.3pp increase from income of offshore wealth..”

Ben and I both read Auten and Splinter arguing Zucman and his fellow travelers are likely overstating tax evasion at the top. However they don’t have a quantification table in the paper so we can’t exactly draw out the calculation you were interested in. AS do note a directional answer in the text, “…this suggests about $770 billion in filer evasion in national income. In comparison, GLRRZ add $1,304 in audit-based evasion in 2012 (Table A6, exam and DCE columns). But this amount includes capital gains realizations that are not in national income. Excluding $70 billion in added evaded capital gains (5.3 percent of total based on Table A1) suggests that GLRRZ add audit-based filer evasion comparable to national income definitions of about $1,230 versus about $770 billion in national income.11 Therefore, the GLRRZ filer evasion estimate appears to exceed amounts in national income by $460 billion, or more than one half…”

Ben Comment:“I did a deeper dive this weekend to work out a quantification like you asked for. I didn’t really see a good way to do that because AS don’t have access to the micro data. There’s no where in the paper where they say “we think Zucman et al overestimate things by X%”. That said, they do give some other papers to compare with. I’m laying that out below with some caveats.Steve, anything to add? AS highlight several issues with the Zucman methodology and suggest the Zucman estimate is too high. Their main 2 points are 1) Marco adjustments (DCE multipliers) are inappropriate for micro data and 2) allocating too much passthrough tax-evasion to the top end of the income distribution. That said - Section 2.b has a nice comparison:National accounts explicitly break out proprietor evasion (combined amount for sole proprietors and partnerships) of $561 billion and wage evasion of $75 billion when averaging 2006-2013 (at 2012 dollars). Auten and Splinter (2019a, online data Table T1) provide estimates of likely evasion in national accounts for other income sources. For 2006-2013, there is annual evasion of about $44 billion for farms, $43 billion for rents, and $80 billion for S corporations. To compare with GLRRZ, there are smaller amounts for dividends and interest evasion that should be added and non-filer portions that should be removed from the values above. If we assume the former is $50 billion and the latter is 10 percent, this suggests about $770 billion in filer evasion in national income. In comparison, GLRRZ add $1,304 in audit-based evasion in 2012 (Table A6, exam and DCE columns). But this amount includes capital gains realizations that are not in national income. Excluding $70 billion in added evaded capital gains (5.3 percent of total based on Table A1) suggests that GLRRZ add audit-based filer evasion comparable to national income definitions of about $1,230 versus about $770 billion in national income.11 Therefore, the GLRRZ filer evasion estimate appears to exceed amounts in national income by $460 billion, or more than one half. This represents an additional 3 percent of national income.So compared to the GLRRZ (Zucman paper) overshoots the NIPA estimates by almost 60% ($460Billion over estimate from NIPA estimate of evasion of 770B). As far as I understand, the first paragraph says the AS 2019a estimate it in line with the NIPA estimate. I would say these are the bounds. The Zucman estimate exists because they think evasion in the official accounts is understated - so I’d take this as a lower bound. However, AS make a compelling case that the Zucman number is overstated - so this would be our upper bound. If we split the difference, we come to 996.5B in evasion. Then there is the issue of attributing evasion appropriately. This is a tough issue and, while it seems clear that Zucman doesn’t do it right, AS don’t really suggest what the proper allocation is. I also want to note that on pg 13, AS point out that there is a Piketty Saez zucman (2018) paper that says there is about 637B in evasion - Lower than the official NIPA accounts. This paper allocates about 40% (250B) of this evasion to the top 1%.”

Showing 48 database articles primarily about Taxation