Delighted to release this paper on Auten & Splinter’s work, painstakingly going through their code & assumptions We uncover key mistakes—and show that once fixed, their methodology implies the same rise of inequality as in Piketty-Saez-Zucman 🧵Since the first version of their work, AS has been a moving target Key aspects of the methodology changed from one version to another, though the bottom line results never change Now that their paper is published we can finally see what’s behind all thatThis table summarizes the key issues in AS, which can be classified in three buckets - Empirical errors - Conceptual errors - Questionable assumptions Let's start with the largest empirical error: excess fiscal depreciationRemember Trump’s “I love depreciation”, back in the 2016 debate with Clinton? Partnerships claim enormous amounts of depreciation, allowing them to reduce their taxable income below their true economic income This untaxed income has been boomingIn AS, this untaxed income is allocated not to partnerships owners… but to sole proprietors That’s about $400 billion earned by owners of big private businesses (in oil, real estate, etc) in 2019 ... that AS give to Uber drivers, small farmers & other self-employed individualsSame problem with the untaxed income of “S-corporations”: These are private businesses overwhelmingly owned by the rich, with sizable untaxed income — which AS allocate to poorer individuals, including retirees with IRAs and 401kTax evasion Most tax evasion is undetected by the IRS AS simply assume that undetected tax evasion is done primarily and increasingly by the poor The problems with that assumption are discussed in this great comment by @danhreck & John IselinThe bottom line is this: The observable income of private businesses—car dealerships, real estate firms etc.—is highly and increasingly concentrated But AS assume that the untaxed income of these businesses is much more equally distributed with no trend, for no valid reasonTurning to capital There’s a lot of untaxed investment income in 401k, etc AS allocate this income to beneficiaries of funded & ~unfunded~ pensions The public school teachers of Illinois get a fraction of Alphabet’s profit even if the Illinois state pension system has no assetThis, along with similar unfounded assumptions on untaxed capital income, explains why in AS it looks like capital inequality has declined When we know from many different data sources that wealth inequality has increased a lotAnd then there are conceptual issues, starting with sales taxes Consider a tax on bread: In AS, the pretax income of the poor is inflated by the amount of the bread tax they pay The higher the tax, the richer the poor become, and the less inequality there is!Government deficits In AS the equalizing effect of deficits is double-counted: A first time because gov. sending is equalizing (which is true) & a second time because the deficit itself is assumed to be notionally “paid” disproportionately by the richAlso did you know there’s a big “universal basic income” in the United States? The US government spends around 17%–18% of national income on defense, police and prison, education, infrastructure...In AS, half of this spending is allocated as a lump sum per capita Like a big UBI of $4,500 per person in 2019 See, there's much less inequality that we thoughtOnce these issues are addressed, the AS top 1% income share becomes similar in level and trend to the one we estimate There are many other methodological differences between AS and us, but they sum to nearly zeroAS’s claim that inequality did not increase much never passed a basic smell test: All available sources — income in tax returns, Social Security wage data, CEO pay, wealth surveys, rich lists...— show that income & wealth have become massively more concentrated since the 1980sNow we know exactly what's wrong with their approach Will they fix it? /end

Delighted to release this paper on Auten & Splinter’s work, painstakingly going through their code & assumptions We uncover key mistakes—and show that once fixed, their methodology implies the same rise of inequality as in Piketty-Saez-Zucman 🧵 ... Since the first version of their work, AS has been a moving target Key aspects of the methodology changed from one version to another, though the bottom line results never change Now that their paper is published we can finally see what’s behind all that ... This table summarizes the key issues in AS, which can be classified in three buckets - Empirical errors - Conceptual errors - Questionable assumptions Let's start with the largest empirical error: excess fiscal depreciation ... Remember Trump’s “I love depreciation”, back in the 2016 debate with Clinton? Partnerships claim enormous amounts of depreciation, allowing them to reduce their taxable income below their true economic income This untaxed income has been booming ... In AS, this untaxed income is allocated not to partnerships owners… but to sole proprietors That’s about $400 billion earned by owners of big private businesses (in oil, real estate, etc) in 2019 ... that AS give to Uber drivers, small farmers & other self-employed individuals ... Same problem with the untaxed income of “S-corporations”: These are private businesses overwhelmingly owned by the rich, with sizable untaxed income — which AS allocate to poorer individuals, including retirees with IRAs and 401k ... Tax evasion Most tax evasion is undetected by the IRS AS simply assume that undetected tax evasion is done primarily and increasingly by the poor The problems with that assumption are discussed in this great comment by @danhreck & John Iselin ... The bottom line is this: The observable income of private businesses—car dealerships, real estate firms etc.—is highly and increasingly concentrated But AS assume that the untaxed income of these businesses is much more equally distributed with no trend, for no valid reason ... Turning to capital There’s a lot of untaxed investment income in 401k, etc AS allocate this income to beneficiaries of funded & ~unfunded~ pensions The public school teachers of Illinois get a fraction of Alphabet’s profit even if the Illinois state pension system has no asset ... This, along with similar unfounded assumptions on untaxed capital income, explains why in AS it looks like capital inequality has declined When we know from many different data sources that wealth inequality has increased a lot ... And then there are conceptual issues, starting with sales taxes Consider a tax on bread: In AS, the pretax income of the poor is inflated by the amount of the bread tax they pay The higher the tax, the richer the poor become, and the less inequality there is! ... Government deficits In AS the equalizing effect of deficits is double-counted: A first time because gov. sending is equalizing (which is true) & a second time because the deficit itself is assumed to be notionally “paid” disproportionately by the rich ... Also did you know there’s a big “universal basic income” in the United States? The US government spends around 17%–18% of national income on defense, police and prison, education, infrastructure... ... In AS, half of this spending is allocated as a lump sum per capita Like a big UBI of $4,500 per person in 2019 See, there's much less inequality that we thought ... Once these issues are addressed, the AS top 1% income share becomes similar in level and trend to the one we estimate There are many other methodological differences between AS and us, but they sum to nearly zero ... AS’s claim that inequality did not increase much never passed a basic smell test: All available sources — income in tax returns, Social Security wage data, CEO pay, wealth surveys, rich lists...— show that income & wealth have become massively more concentrated since the 1980s ... Now we know exactly what's wrong with their approach Will they fix it? /end