Editor’s note: My colleague Arpit Gupta has kindly agreed to share his thoughts on a recent call for a more steeply progressive income tax code. This is the first of two posts on related themes.
Earlier, I discussed the work by Saez and Piketty on inequality. Though their research is valuable, its policy relevance is questionable given that much of the rise in inequality is concentrated among a very small fraction of Americans; and the welfare impact of this rise in upper-tail inequality on ordinary Americans has been limited.
Yet even if Saez and Piketty were right in pointing to a negative role of rising inequality, that fact by itself wouldn’t necessarily demand any public policy response. The higher progressive taxation needed to combat inequality may come with costs of its own. If the costs of taxation outweighed the benefits, than the ideal policy response to inequality, or the role of progressive taxation in general, would be limited.
Hence the level of attention garnered by a recent article by Emmanuel Saez, co-authored with Nobel Laureate Peter Diamond, arguing in favor of progressive taxation. While the headline result — that the top marginal tax rate “should” be 76% — has been widely broadcast, as always the estimate comes from a web of assumptions that are useful to unpack.
The “Optimal” Marginal Tax Rate
Saez and Diamond begin with a discussion of what the “optimal” marginal tax rate ought to be. From a progressive utilitarian standpoint, they assert that the private benefits of additional income among the rich are socially worthless; hence tax rates on the rich ought to be set at the rate which maximizes government tax revenues. This corresponds to the top of the Laffer Curve.
Under this perspective, the only limitation goverments face in setting tax policy comes from behavioral responses to taxes. Higher taxes result in individuals reporting less taxable income, responses which can be statistically captured in as tax elasticities. Beyond a certain point, higher taxes reduce reported income so much that the government makes less income overall (the supply-siders famously argued that America was already taxing too much by this standard). Below that point, higher taxes do result in greater income, but at a lower rate.
In arguing that top marginal tax rates should be set only in order to maximize revenues, Saez and Diamond gloss over several issues. There’s a curious absence of discussion of how the government actually spends the money, though surely that’s relevant in setting tax rates. Philosophically, we may have many reasons for wanting to keep more of our money rather than spending it on public goods. There are good reasons that tax rates are determined by representative institutions, rather than technocratic economists. The idea that the welfare of the rich does not matter at all also seems troublesome. However, even if you believe that, the responses of the rich to higher taxes frequently have impacts on others. For instance, another recent Saez paper documents how lower tax rates in Denmark resulted in rich immigration. To the extent wealthier Americans are providing specialized services — in job creation, or skilled services like medicine or law — policies that soak the rich may hurt ordinary Americans, and top marginal tax rates should be lower as a result.
For these reasons, it’s better to think of the Saez-Diamond estimates as estimating the highest feasible tax rates rather than the tax rates that are optimal. Still, this is a fruitful question on its own.
Saez and Diamond’s core analysis of top marginal rates comes in this section:
As an illustration using the different elasticity estimates of Gruber and Saez (2002) for high income earners mentioned above, the optimal top tax rate using the current taxable income base (and ignoring tax externalities) would be T*=1/(1+1.5 x 0.57)=54 percent, while the optimal top tax rate using a broader income base with no deductions would be T*=1/(1+1.5 x 0.17)=80 percent. Taking as fixed state and payroll tax rates, such rates correspond to top federal income tax rates equal to 48 and 76 percent, respectively. Although considerable uncertainty remains in the estimation of the long-run behavioral responses to top tax rates (Saez, Slemrod, Giertz, 2011), the elasticity e=0.57 is a conservative upper bound estimate of the distortion of top U.S. tax rates. Therefore, the case for higher rates at the top appears robust in the context of this model.
The basic idea here is to start with an estimate of tax elasticities — how much taxable income drops in response to higher taxes. This allows Saez and Diamond to work out the tax rates that would maximize revenue. While many commentators have seized upon this paper as representing new evidence; all Saez and Diamond here do is calculate the top of the Laffer Curve given estimates from prior papers (including one of their own — more on that later). They conclude that, taking the tax code as given, the top marginal income tax rate should be 48%; but it could be as high as 76% if the tax code was shorn of deductions.
Now consider Saez and Diamond’s discussion of their paper in the WSJ:
According to our analysis of current tax rates and their elasticity, the revenue-maximizing top federal marginal income tax rate would be in or near the range of 50%-70% (taking into account that individuals face additional taxes from Medicare and state and local taxes). Thus we conclude that raising the top tax rate is very likely to result in revenue increases at least until we reach the 50% rate that held during the first Reagan administration, and possibly until the 70% rate of the 1970s. To reduce tax avoidance opportunities, tax rates on capital gains and dividends should increase along with the basic rate. Closing loopholes and stepping up enforcement would further limit tax avoidance and evasion.
Here, Saez and Diamond present top marginal tax rates up to 70% as optimal in isolation, and closing loopholes as a further bonus. This strikes me (and Scott Sumner) as discordant with their own research — tax rates as high as 70%, without corresponding tax reform, would likely put America on the wrong side of the Laffer Curve by Saez’s work. This key distinction has largely been ignored in much of the commenary regarding Saez’s research.
Ultimately, the core empirical evidence used in this paper goes back to an estimate from Saez and Gruber in 2002, who calculated the behavioral responses to taxes that are critical in estimating how high government taxation can go.
Saez and Gruber calculate these by examining states that changed marginal tax rates during the 1980s, and calculating how individuals hit by tax changes altered changed overall income, as well as taxable income in response. Overall, Saez and Gruber find that for every dollar of tax increase, individuals lowered taxable income by 40 cents in response. Among incomes higher than $100,000 a year, people cut back reported income by 57 cents. However, individuals cut back overall income by less in response to tax hikes, suggesting that part of the picture is due to tax evasion, as well as due to cutbacks in real income.
The results do suggest that higher taxes result in tangible costs — individuals tend to cut back on income overall, suggesting that the broader economic pie shrinks due to higher taxation (Compare this finding to the headline the WSJ gave to an op-ed in which Saez and Gruber make their case: “High Taxes Won’t Slow Growth“). People also reduce reported taxation by even more, suggesting that there are limits to how much governments can raise taxes. Finally, these behavioral responses overall are most prounced among wealthier Americans.
A key caveat of the Saez and Gruber research is that they only examine responses to taxes that happen three years after a law change. While reported income changes over such a short interval are undoubtedly part of the story, taxes surely have broader effects on the makeup of the entire economy in the long-term. Yet Saez and Gruber’s key tax policy advice doesn’t take into account effects of taxes after three years.
A Broader Theory of Taxes
A fuller theory of how taxes matter focuses on focuses on lifecycle decisions made by households. Higher marginal tax rates affect after-tax returns on a number of economic decisions, and so have the potential to alter a wide variety of economic activity, only a small fraction of effects of which are captured by Saez and Diamond.
We expect taxes to have larger long-run effects on individual incomes than in the short-run. Over a period of a few years, individuals generally have a limited ability to alter their employment and lifestyle choices. They generally have a far greater ability to alter income and labor decisions in response to incentives on a time span of years or decades. Saez and Gruber find some evidence for this, in that responses to taxes were greater three years after the tax increase than two years, but they did not look at tax responses beyond three years.
Harvard Economist Raj Chetty has explored this issue, and found that incoporating long-run effects results in far higher estimates for how individuals respond to taxes. His work suggests that though individuals may not respond immediately to tax incentives, the impact may be greater over the long-run. That is, rather than the Saez and Gruber estimates being a “higher bound” on how responsive taxes may be; they may drastically understimate the overall costs of taxation — suggesting that a top marginal tax rate of 54% may, if anything, be too high.
Other evidence on the roles of taxation comes from analyzing other specific channels by which taxes change incentives and behavior. While studies like those by Saez focus on the behavioral responses of people already in the workforce; the decision to enter or leave work is a major behavioral choice affected by marginal tax rates. Research by Rogerson and other economists has emphasized that labor supply can be altered by tax rates, particularly among people already on the margin between working and not, such as married women with children or the elderly. Taxes also affect the decision to accumulate human capital and education, suggesting that higher marginal tax rates may lower economy-wide productivity.
Adding up the aggregate effects of these various channels is a challenge. Ed Prescott has argued that Europe-US differences in labor supply and income per capita can be explained by higher US taxes. It’s possible that this overstates the case, especially given the various other sources of Europe-US differences (such as unions, etc.). However, understanding the economy-wide impacts of taxation is critical and poorly understood.
Another interesting recent piece of evidence here comes from Britain, which has found that hiking the top marginal tax rate bracket from 45 to 50% led to negligible hikes in revenue — even as tax revenues were in general rising among other groups. While this captures the short-run response to taxes — tax evasion and migration — the long-run responses to taxes would likely be far greater as individuals face worse work incentives and respond by cutting back on work effort. These negative behavioral responses to taxes shape how much money taxing the rich will raise, as well as how much those higher taxes impact the rest of the economy.
The point here is mostly that things are complicated, and we still understand little about how taxes feed into the overall economy. There are good reasons to suspect that the behavioral responses to taxes could be quite high in the long run, suggesting a more limited scope for how high top marginal tax rates could or should go. Saez and Diamond obviously disagree, but it’s worth looking at their core pieces of supporting evidence, which involve far more assumptions and imprecision than suggested in the popular impression.