Search the Site

An Inequality Tax Trigger: The Brandeis Ratio Explained

On Monday, Aaron Edlin and I published a cri de coeur op-ed in the New York Times calling for a Brandeis tax, an automatic tax that would put the brakes on income inequality.  This is the second in a series of posts (the first post is here) explaining more about our rationale and providing more details on how a Brandeis tax might be implemented.

An Inequality Tax Trigger
By Ian Ayres and Aaron Edlin

A central idea behind our Brandeis tax proposal was to have a tax that is triggered by increases in inequality. Our Brandeis tax does not target excessive income per se; it only caps inequality. Billionaires could double their current income without the tax kicking in — as long as the median income also doubles. The sky is the limit for the rich as long as the “rising tide lifts all boats.” Indeed, the tax gives job creators an extra reason to make sure that corporate wealth does in fact trickle down.

But in crafting an inequality trigger we might have chosen a more traditional measure of income inequality – such as the Gini coefficient or the Herfindahl-Hirschman Index (HHI).  Any standard of inequality would show a sharp increase over the last 25 years. 

In part, we choose the Brandeis ratio because we think it is more transparent. We could try to explain that the Gini number represents the ratio of two areas, including crucially the area above the Lorenz curve. The Lorenz curve is the inequality measure that Robert Shiller and coauthors use in their 2006 paper on this topic of inequality of tax triggers; they account for behavioral responses to such changes and consider how to optimally balance the benefits of lower risk and higher economic growth with the traditional negative incentive effects.  For those interested in the details, Shiller, one of the paper’s authors, has also written a compelling book on the costs of high risk finance and even proposed “inequality insurance.”  Shiller should get credit for coming up before us with the general idea of an inequality-contingent tax system.

Or alternatively, we might explain that a slight modification of the HHI represents the probability that two randomly selected items come from the same source.  For example, in antitrust analysis, if the modified HHI for the lawnmower market is equal to .43, that means there is a 43% chance that two randomly selected lawnmowers would have been produced by the same company.  A market is more concentrated if this probability approaches 100%. As applied to income concentration, the modified HHI would tell you the probability that two random dollars of income would be received by the same person (or possibly the probability that the two randomly selected dollars would have been earned by people of the same economic class).  But notwithstanding this nifty interpretation, it remains unclear why this probabilistic concentration measure is related to the Brandeis concern that excess inequality and democracy are incompatible.

In contrast, the Brandeis ratio as a measure of concentration is immediately graspable, and is more closely tied to the specific concern that a sliver of plutocrats with gargantuan wealth could distort the political process.  It doesn’t take a Ph.D in economics to understand that something seismic has occurred when the average one percenter goes from earning 12.5 medians to 36 median incomes.  It’s true that not all these measures of inequality are about income inequality; some might rightfully cite the Nobel Laureate Amartya Sen in arguing that income alone doesn’t capture human welfare, hence the multidimensional indices often used in global development.  Indeed, part of Brandeis’s concern was not income equality for its own sake but rather the consequences of income inequality on democracy.

More subtly, the Brandeis ratio intentionally ignores what is happening to two other parts of the income distribution.  Unlike the Gini coefficient, the Brandeis ratio does not take into account how many Americans are unemployed or living below the poverty line.  This is not a weakness.  The purpose of our inequality tax is not to respond to short- or even medium-term business cycle fluctuations.  We want the tax to respond to long-term structural shifts in inequality.  We consider it a strength of the measure that the denominator (the median household income) is relatively uninfluenced by the unemployment rate.  And using the Brandeis trigger similarly is independent of the impact of illegal immigration on the economy – so we don’t have to worry that in the shadow of a Brandeis tax that Congress would have particularly different incentives to include or exclude new lower paid households from entering the economy.  There are important debates and reforms needed to respond to the separate questions of structural unemployment and undocumented workers, but the Brandeis is geared toward another – to our minds – even more important policy issue. 

The Brandeis ratio as a measure of income inequality also doesn’t pay attention to the relative success of 2- or 3-percenters over time.  Focusing just on the relative income of the richest one percent is appropriate if we are concerned with the deleterious impacts of inequality on our democratic institutions because one-percenters (those currently making more than $330,000) disproportionately fund our political campaigns.  As emphasized by Lawrence Lessig in Republic, Lost (presaged somewhat in Ayres’ book with Bruce Ackerman, Voting With Dollars), the bulk of campaign finance dollars comes disproportionately from not just the 1% club, but the richest one-half of one-percenters.  Focusing on the average income of one-percenters is a good proxy for the rising political power of plutocrats.

But at the end of the day, we’re not wedded to the idea that there needs to be any single sufficient statistics.  We could imagine a world in which a Brandeis tax was contingent on a different inequality measure or even upon multiple measures.

For example, one might argue that rising wealth inequality could be an even worse problem for democracy than income inequality.  Accordingly, one could imagine a wealth tax or an estate tax that was contingent on some measure of wealth inequality.  But given the current political environment, an income-contingent wealth tax is bridge too far.

Steve Silberstein has been promoting an interesting way to make the corporate income tax for specific corporations contingent on an analogous inequality ratio.  As mentioned in the New Republic:

[A]nother proposal, put forward by investor Steve Silberstein, would adjust the corporate tax rate based on the ratio of CEO pay to the average worker. A company with a ratio at the 1980 level of 50:1 would pay tax at the current rate of 35 percent, with the rate rising
for companies with a higher ratio and lower for those with a narrower pay gap.

We had briefly thought about modifying our proposal to allow one percenters to avoid a trigger Brandeis tax if they could show that their income was less than 36 times the median income of workers who produced it, but concluded that personalized Brandeis ratios would be an administrative nightmare.  The Silberman corporate tax proposal is by comparison elegantly straightforward.

While we proposed a Brandeis tax based on the 2006 pretax Brandeis ratio of 36, it would be more natural to use a trigger based on an after-tax Brandeis ratio which went from something like 8 medians in 1980 to about 25 medians in 2006.1  Accordingly, the IRS might each year calculate the after-tax ratio and trigger the Brandeis tax if the ratio exceeds 25 medians.

Our proposal of a 36-median cap was doubly conservative.  The first reason is because the tax only asked that the after-tax ratio not exceed the pretax 2006 ratio and the second is because the tax used the 2006 year as a trigger, a year that probably had a higher Brandeis ratio than we would find today. The Brandeis ratio is likely to decline during a recession because the average income of one-percenters (think hedge-fund managers) is more sensitive to the recession than the median U.S. income.   Indeed, another debate which we would like to promote is about the question of the appropriate trigger size. 

Our proposal starts with an out-of-the-money status quo inequality trigger as a way to promote political common ground.  You can vote for a contingent Brandeis tax without voting to necessarily raise taxes.  Ours is a “tax more tomorrow” idea where the relevant tomorrow may never come.  Our trigger avoids the concern that we’re engaged in crude “class warfare.” It doesn’t take away any of existing inequality, it just tries to make sure that 99% share in prospective future gains of the 1%.  But reasonable people could argue for either higher or lower triggers – for example, returning to a simpler time when rich people only earned 20 medians.   Perhaps like with carbon emissions we could seek to lower inequality to 1990 levels by 2020. 

1 Our after-tax estimates of the Brandeis ratio are estimates, because we do not have comparable information on the after-tax median household income.  As a proxy, we used CBO after-tax data and divided the after-tax average income of one-percenters by the average income of the third quintile.