Lawrence H. Summers is a professor at and past president of Harvard University. He was treasury secretary from 1999 to 2001 and an economic adviser to President Barack Obama from 2009 through 2010. Natasha Sarin is an assistant professor of law at the University of Pennsylvania Law School and an assistant professor of finance at the Wharton School.

Sen. Elizabeth Warren (D-Mass.) recently proposed a 2 percent “wealth tax” on those worth more than $50 million. Emmanuel Saez and Gabriel Zucman, economists at the University of California at Berkeley, have played a major role in developing and validating this proposal. They estimate that the tax would raise $187 billion in 2019 (Warren’s additional 1 percent “billionaire surcharge” brings their total revenue estimate to $212 billion). This represents a substantial sum and has been widely quoted in both academic and policy discussions of wealth taxation.

AD
AD

Saez and Zucman are perhaps the world’s leading experts on wealth statistics, but they have not previously been involved in tax-revenue estimation. One of us (Summers) has often been struck by how much his judgments regarding tax revenue based on published economic statistics have differed from those of professional revenue estimators. So it seemed a worthwhile exercise to estimate the proposed wealth tax revenue based on actual tax revenue data.

We reasoned as follows: The existing estate tax is a wealth tax levied at the time of death. If 2 percent of wealthy families experience a death and intergenerational transfer (rather than a spousal transfer) each year, then the current 40 percent estate tax should roughly be the equivalent of a wealth tax of 40 percent multiplied by 2 percent — or a 0.8 percent wealth tax — assuming equivalent definitions of wealth and the same threshold for taxation. Since most wealth is held by fairly elderly people, and the mortality rate of 70-year-olds is above 2 percent, we suspect that 2 percent mortality is a conservative estimate. So the actual wealth-tax equivalent of the estate tax is likely greater than 0.8 percent.

The IRS reports that for 2017, the most recent year for which data is available, the estate tax raised around $10 billion from estates over $50 million — and this included tax collected on the first $50 million of estate tax value, so it overestimates the conceptually appropriate figure. Therefore, if this is what the revenue yield would be from a 0.8 percent wealth tax, the implication is that a 2 percent wealth tax would raise a total of $25 billion. That’s around one-eighth of the Saez and Zucman estimate.

AD
AD

We recognize that the effective estate tax rate (defined as what estates over $50 million actually paid in taxes) is only 13.5 percent. This is primarily because of charitable contributions and bequests to surviving spouses. We imagine that under a wealth tax, the charitable transfers that now take place at death might take place during the lifetimes of the wealthy, through transfers to donor-advised-funds and the like. Our mortality probability is an estimate for households, so that already excludes inter-spousal bequests. But even being maximally optimistic about the wealth tax’s revenue potential by assuming that the estate tax is equivalent to a 0.27 percent wealth tax (the 13.5 percent effective estate tax rate multiplied by 2 percent mortality), we still get a revenue estimate for wealth taxation of only about $75 billion, or approximately 40 percent of the Saez and Zucman estimate.

What is going on? As a matter of arithmetic, a great deal of the wealth that Saez and Zucman assume will be hit by their wealth tax is escaping the estate tax. It’s plausible that people are more aggressive in their estate-tax planning than they might be with their wealth-tax planning; they might, for example, be more prepared to make gifts to charity after they die than they would during their lifetimes. And there are certainly ways in which those who are competently advised can move assets to their children while they are alive while avoiding gift and estate taxes and still maintaining substantial control.

But this discrepancy is pretty large. We suspect that to a great extent it reflects the myriad ways wealthy people avoid paying estate taxes that in some form will be applicable in any actually legislated wealth tax. These include questionable appraisals; valuation discounts for illiquidity and lack of control; establishment of trusts that enable division of assets among family members with substantial founder control; planning devices that give some income to charity while keeping the remainder for the donor and her beneficiaries; tax-advantaged lending schemes; and other complex devices known only to sophisticated investors. Except for reducing a naive calculation by 15 percent, Saez and Zucman do not seem to take account of these devices.

AD
AD

If our suspicion is correct, such a wealth tax will not yield the revenue that its proponents hope for, and that when actual scorekeepers score actual proposals, their estimates will disappoint advocates.

In response to this, advocates might insist on purity and not permit the wealth tax to be gamed as taxes typically are. Possibly. Experience suggests, however, that over time special benefits are almost always introduced into tax systems. No previous U.S. tax has existed in anything like this wealth tax’s aspirational pure form. And comprehensive wealth-tax collection has not been accomplished in even the most egalitarian European societies — which have moved away in droves from far less ambitious wealth taxes. Such a push in the United States would involve forcing the sale of many family-owned businesses and require vast audit resources at a time when the IRS is unable to audit even 10 percent of millionaires. And it will involve placing limits on the ability to be charitable or to establish trusts for the benefits of grandchildren.

We would be surprised if the $25-billion-a-year figure we suggest was not a significant underestimate of the revenue potential of a 2 percent wealth tax. But it is likely extremely premature to bank on anything like the $200 billion plus that Saez and Zucman estimate. And to be clear, this reality, while important in evaluating the wealth tax, is certainly not dispositive as to its merits, which we are sure will be the subject of continuing debate.

AD
AD

There is, however, an important, more general point here. Common-sense revenue estimates by economists who are not very deeply steeped in revenue estimation tend to be overly optimistic. Beyond the wealth tax, two further examples illustrate this.

When one of us (Summers) was in government, he recalls being extremely frustrated with revenue estimators’ scoring of proposals to tax carried interest at ordinary income-tax rates because they seemed so low. He did a version of the following calculation (adjusted to match recent data): U.S. private equity firms have approximately $3 trillion in assets under management. Rates of return for the industry average around 8 percent, 20 percent of which typically flow to general partners as carried interest, taxed at the capital gains rate (23.8 percent) rather than the income-tax rate (now 37 percent for top earners). Using these numbers, taxing carried interest as ordinary income this year should generate $6.3 billion. The Congressional Budget Office reports revenue estimates that are less than one-fourth this. This is because professional revenue estimators adjust naive calculations such as this one to account for a variety of legal tricks that decrease the revenue-raising potential of reforms.

Similarly, economists who are not professional estimators go wrong in estimating the impact of a value-added tax (VAT) using national accounts data. They reason that a 5 percent broad-based VAT should be able to collect on two-thirds of domestic consumption ($13.3 trillion in 2017), and so should generate 5 percent times 66 percent times $13.3 trillion. That totals $440 billion, or about $5 trillion between 2020 and 2028. Yet the CBO estimate is only $3 trillion over this horizon. A prudent person would bet on the CBO.

AD
AD