Plans are afoot to charge drivers to enter Manhattan. But we need a fairer way to reduce traffic.
[I published this opinion piece in the ill-fated OZY in 2019. As the OZY website appears to be gone, and with it this piece, I’m reposting it here.]
March 31, 2019
By Ramsi Woodcock
Henry Ford’s dream was to democratize transportation by selling cars so cheaply that every American could own one. His shareholders sued him for it, but Ford eventually succeeded, and we owe today’s driver-friendly America in part to Ford’s insistence that the automobile be a mass-market item. While climate change has taught us that the car was the wrong route to transport democracy, it has done nothing to undermine the principle that there should be no class divide in American transportation.
But one of the most progressive states in the union, New York, is about to write such a class divide into law, in the form of congestion pricing for access to Manhattan. If the plan goes forward, drivers will be charged more than $10 to enter the island, and other major U.S. cities may follow the Big Apple’s lead.
Congestion pricing taxes car commuters. Advocates, including New York Gov. Andrew Cuomo, public transport groups and even a once reluctant Mayor Bill de Blasio, argue that the tax would reduce rush-hour traffic and raise billions of dollars to invest in improving the city’s decrepit subways, which in turn would increase subway ridership, further reducing traffic.
All that is true, but congestion pricing does something else: It puts the burden of decongesting the city entirely on the backs of poor and middle-class drivers, by politely but effectively making it impossible for some to drive into the city because they simply can’t afford the tax.
If there were no other way to reduce traffic, then, of course, New York should ask the poor and the middle class to shoulder this burden. But my research shows that there are other ways to reduce traffic that won’t write a class divide into law.
The Limits of the Price System
The premise of congestion pricing is that the best way to prevent overuse of an important resource is by charging for access to it. Those who value the resource more, the theory goes, will be willing to pay more for it, and so those who actually end up paying the price and using the resource will be those who value it the most. Those who decide not to pay—and therefore shoulder the burden of eliminating overuse of the resource—will be those who needed the resource the least.
In reality, however, willingness to pay is an imperfect indicator of need. A dollar, after all, is worth much less to a rich person than to a poor or middle-class person. So congestion pricing isn’t really reserving the city’s streets for those who need them most; it’s reserving them for the rich.
Ration with Tech, Not Price
But there is a way to reduce traffic in Manhattan without excluding the middle class or the poor: allocate access based on the rule of first come, first served. Instead of charging for access, simply close the island (or the parts of the island that are the focus of the current plan) once it has filled up.
First come, first served, like congestion pricing, strives to grant access based on need—the more you want to enter the city, the earlier you will line up—but unlike congestion pricing, it would be far harder for the rich to use their wealth to short-circuit the sorting mechanism. Sure, the wealthy could pay people to line up in their place, but New York could ban the practice, whereas under congestion pricing the city could never prohibit the rich from buying access when they don’t really need it.
First come, first served evokes visions of lines of vehicles stretching off into the distance along the approaches to the city, and 10 years ago that would have been true. But the internet has taken the effort out of waiting by allowing us to join virtual lines from the comfort of home, which means the first-come, first-served model is now a viable alternative to the price system.
Imagine that instead of congestion pricing, city leaders were to create a city access app. The app would know your location and how long it would take you to drive into the city, and therefore could inform you before you depart whether you will be granted entrance. Even better, the app could measure your need to drive, allowing those who live far from public transport to drive during rush hour but requiring those with public transport options to use them during busy periods. Either way, no physical line would be necessary.
Sure, it’s a bit more cumbersome than just driving into the city and having the tax deducted automatically from your bank account. But that’s only for those lucky enough to be able to afford congestion pricing. For those who would be priced out of the city, and thus their jobs, by congestion pricing, an app-based first-come-first-served approach would be a big improvement.
Some have suggested that New York’s congestion pricing plan should include an exemption for the poor, but any proposal to make the plan truly affordable is doomed to failure, because if everyone can still afford to drive into the city under the plan, then the plan won’t stop congestion. The proposal to exempt the poor should be seen for what it is: an attempt to obscure congestion pricing’s classist reality, at the expense of the middle class, which would not be exempt.
Tax Incomes, Not Drivers
First-come-first-served might be a more democratic way of rationing access to the city’s streets, but what about all the tax revenue congestion pricing will generate to fund the subways? First-come-first-served can’t generate that revenue, because it keeps city access free. Nor should it.
Economists have long argued that the best way to soak the rich is directly — through taxes on high incomes and capital gains, like the state’s Millionaire’s Tax and a proposed exaction on second homes — not by taxing behaviors. Behaviors, such as commuting, often cut across class lines, and you end up taxing the rest along with the rich.
America has a long tradition of preferring market-based solutions to public problems, which is what congestion pricing represents, but America also has a longstanding hatred of class privilege — epitomized by Ford’s desire to put a car in every garage. That may explain why a still-class-bound London, and authoritarian Singapore, have embraced congestion pricing, but American cities have not.
Ramsi Woodcock is a law professor at the University of Kentucky.
In a new paper, Glick, Lozada, and Bush have done both antimonopolism and the antitrust academy a service by making the first real attempt to put the movement in direct conversation with contemporary antitrust method.
GLB have a simple message: welfare economics long ago stopped using willingness to pay to measure consumer welfare, and antitrust should too.
What is more, welfare economics today pursues an eclectic set of approaches to measuring welfare. Some of them suggest that the dispersal of economic power and the availability of small businesses can make people happy.
It follows, argue GLB, that it is entirely consistent with contemporary welfare economics to take these things into account in evaluating mergers or prosecuting monopolies.
The Social Welfare Function
GLB start with the problem that welfare economists faced at the beginning of the 20th century: how to compare the value that different people—say a producer and a consumer—obtain from a transaction in the absence of some universal measure of value.
If the producer gets a profit of $2 and the consumer pays $5 for a bag of apples, did the transaction confer the same amount of value on the two? Are $2 worth the same to the producer as a-bag-of-apples-for-$5 is worth to the consumer?
If there were some universal measure of happiness—denominated in, say, “utils”—then we could answer that question.
We would look up the consumer’s change in pleasure associated with swapping $5 for apples and compare it to the producer’s change in pleasure associated with making a $2 profit. If the former were 50 utils and the latter 30 utils, then we could say that the transaction did not confer the same benefit on both parties.
Economists eventually decided that they would not be able to find a universal metric of happiness. But they hoped that they might be able to glean some information about happiness from the behavior of economic actors.
The first approach that they hit upon was the pareto criterion. It said: the only bad transactions are those into which the parties do not enter voluntarily, because those must make at least one party worse off (the party who would not voluntarily enter into the transaction).
Any transaction the parties do enter into voluntarily is, in contrast, good, because they wouldn’t be willing to enter into it unless the transaction made neither worse off.
It followed that voluntary transactions could be treated as welfare improving—or at least not welfare reducing. The parties were signalling, through their willingness to enter into them, that the transactions were at least not undesirable.
If the producer and consumer voluntarily transact in apples at $5, then welfare could be said not to have been reduced and indeed potentially to have increased. That was the pareto criterion.
It helped welfare economics a bit. But it also failed to answer an important question: what about people who are affected by a transaction but who are not entering into it themselves?
If, for example, two producers merge, and, as a result of the merger, they are able to charge a higher price, consumers are affected. But consumers have no choice over whether the merger takes place.
The pareto criterion tells us that the merger does not make the merging parties worse off. But it tells us nothing about whether the merger makes consumers worse off.
Some way of comparing the costs of the transaction to consumers with the benefits to the merging producers is needed, but the pareto criterion cannot provide it.
Willingness to Pay and Potential Pareto
The solution proposed by some economists in the early 20th century was to use willingness to pay as a measure of happiness.
The idea was that if a consumer would be willing to pay $10 for an apple, then that would be a measure of the pleasure the consumer would get from consuming the apple. By noting that a person should be willing to pay cash for cash on a dollar-for-dollar basis, one could proceed to do with dollars what economists had originally hoped to do with utils.
To return to our example of an apple purchased for $5, if the consumer were in fact willing to pay $10 for the apple, then the value to the consumer of the transaction would be the $10 the consumer would be willing to pay less the $5 price that the consumer actually paid for it.
And the value of the transaction to the producer would be the producer’s $2 profit. It would then follow that the consumer did better than the producer in the transaction because the consumer generated a “surplus” of $5 whereas the producer generated a profit (“producer’s surplus”) of only $2.
This willingness-to-pay approach made it possible to evaluate a merger of producers.
If producers were to merge and drive the price up to $7, then the producers (who, if their costs are as before, would now make a $4 profit) would end up better off than the consumers (who would now enjoy a surplus of $10 less $7, or $3). The merger would reduce the welfare of the consumer by $3.
If antitrust were to adhere to a consumer welfare standard—the rule that mergers that reduce consumer surplus are to be rejected—then this merger would fail the test and be rejected.
As GLB note, the willingness to pay concept made it possible to consider tradeoffs as well.
The merger might, for example, also reduce the costs of production of the merged firms from $3 to $0.50, thereby increasing the merging firms’ profits on the transaction from $4 to $6.50.
If one were to view the goal of the antitrust laws as the maximization of total welfare—meaning the maximization of the combined surplus of producers and consumers, however that surplus may be distributed between them—this cost reduction would justify the merger. It would expand the sum of producer and consumer surplus from $7 ($2 for the producers and $5 for the consumer) to $9.50 ($6.50 for the producers and $3 for the consumer).
Moreover, the merger might even be said to satisfy the consumer welfare standard if one were to adhere to the peculiar sophistry that any increase in total welfare should count as an increase in consumer welfare because the increase in total welfare could be redistributed to consumers.
Because the merged producers could be forced to pay the $2.50 increase in total welfare to the consumer, leaving the consumer with $5.50, which is more than the $5 he would have without the merger, the deal could, according to this peculiar sophistry, be classified as consumer welfare enhancing.
At least in potential. And if such a transfer were made, then the consumer and the producers alike would welcome the deal (the producers would be left with $4, which is more than the $3 in profit earned without the deal). Hence GLB refer to this as the “potential pareto criterion”. It is also called the Kaldor-Hicks efficiency criterion.
Economists should have, and, indeed, did, realize from the start that willingness to pay was a doomed approach because a person’s willingness to pay changes with his budget.
Two people who would be willing to pay the same amount for an apple if they had the same wealth would likely be willing to pay vastly different amounts if one were poor and the other rich. The rich person might be willing to pay much more for the apple than would a cash-strapped poor person.
One can avoid this problem by supposing that the poor man is willing to pay less for an apple because he in fact would derive less pleasure from it. He might have to deny his child meat in order to be able to afford the apple, and that might ruin his meal.
But viewing actual pleasure as perfectly consonant with willingness to pay amounts to shoehorning subjective feelings into budget constraints.
It is just as likely that the poor man who did make such a substitution would feel a great deal of guilty pleasure. His rational faculties might enable him to forego that pleasure and give his child meat. But that does not mean that his pleasure centers would not be the worse for it. They would be.
If wealth effects matter, however, then one cannot compare producer and consumer surpluses—or indeed the surpluses generated by any two people.
One cannot say, for example, that a merger that decreases cost by $2.50 is on net a good thing if it results in a price increase of only $2 because $2.50 is more than $2, so the total amount of pleasure generated by the economy has gone up. For if the producers are rich but the consumer poor, then the $2 cost to the consumer might inflict more pain on him than the $2.50 increase in profits for the producers.
Redistribution of those $2.50 in benefits to the consumer is now required for efficiency and not just to achieve distributive justice. If efficiency is about increasing the total amount of happiness generated by the economy, and those $2.50 make the consumer happier than the producers, then efficiency requires that the $2.50 go to the consumer.
If the only implication of wealth effects were that redistribution from rich to poor is required for efficiency, then wealth effects would not be particularly problematic for progressives.
But very often a policy change not only creates a benefit and raises a price, as in our merger example, but also inflicts an economic cost in the sense of precluding some production—or aspect thereof—that consumers value.
The merger might, for example, not only reduce apple production costs by $2.50 but also lead to slightly less tasty apples. Perhaps the merger saves on costs by enabling the sale of an orchard that produced particularly tasty apples but was also relatively costly to maintain.
If the consumer’s maximum willingness to pay falls by $2 because the apple is less tasty, then the willingness to pay measure suggests that the merger should go ahead. The benefits in terms of a reduction in costs of $2.50 exceed the costs in terms of a reduction in the value of the apple to consumers of $2. There is a net gain of $0.50.
To be sure, if the price again rises to $7 as a result of the merger, consumers find themselves even worse off than before. Their surplus falls to $1 (a maximum willingness to pay of $8 less a price of $7).
But the merging producers can, at least in theory, make up for this by transferring $2 to the consumer to offset the price increase and by transferring at least $2 of the cost reduction they enjoy as well, ensuring that the consumer ends up with at least the $5.00 in surplus the consumer would have enjoyed without the deal.
And the producers, who initially enjoyed an increase in profits of $5.50 ($2.50 in cost reductions plus $2.00 from the increase in price) end up better off so long as they do not pay more than $5.50 to the consumer.
So all parties can, in theory, end up better off.
That’s because the benefits created by the merger exceed the costs by $0.50. Once one uses transfers to correct for the resulting price increase and to compensate the consumer for his loss, which is smaller than the producers’ gains, there is necessarily some net gain left over that producers and consumer can divide up, leaving them all better off.
The potential pareto criterion is satisfied and, if the transfers are actually made, so is the consumer welfare standard.
If wealth effects matter, however, then one cannot reliably compare the $2.50 benefit in terms of production cost savings to the $2 loss associated with the reduced tastiness of the apple. If the consumer is poor, then the consumer may place a dollar value on the reduction in tastiness of the apple that is far below the actual loss of pleasure the consumer would suffer in consuming a less tasty apple.
If there were utils and we could compare the value of the production cost savings to the producers to the reduction in the consumer’s happiness associated with the less tasty apple, we might find that the producers’ gain is 100 utils and the consumer’s loss is 1000 utils, resulting in a net reduction in happiness due to the merger.
Wealth effects prevent the consumer from registering his dissatisfaction in terms of willingness to pay, however, and so the merger appears to offer a net gain when in fact it does not.
It follows that the producers will never be able fully to compensate the consumer for the loss without incurring a loss themselves, and so according to the potential pareto criterion the merger should be blocked.
If we nevertheless treat willingness to pay as a measure of welfare, however, the deal will appear to be welfare increasing and the deal will go through, reducing overall happiness.
Wealth effects cause willingness to pay to lead to bad policymaking.
Wealth effects also undermine the commensurability of values with respect to the same person.
To see why, let’s go back to the example in which the merger raises prices but doesn’t reduce the tastiness of apples.
If unwinding the merger would reduce the price of an apple from $7 to $5, it is clear that the consumer becomes $2 richer. He saves $2, which he can now spend on other things.
In order for willingness to pay to be a useful proxy for welfare, one would, then, like to be able to say that the consumer is made just as well off by the price reduction as he would have been had he been given $2 in cash in lieu of the price increase.
But if willingness to pay depends on wealth, we cannot say that a $2 cash payment would leave the consumer in the same position as the consumer would be had price fallen by $2.
If a consumer cares more for apples the richer that he is, then the consumer will prefer a $2 cut in the price of apples to a $2 cash payment. Given his stronger preference for apples, the consumer might want to plow the $2 savings on apples into buying more apples, and that money would buy more apples at the lower apple price than would a $2 cash payment used to purchase more apples at the higher price.
It follows that the consumer would require a cash payment in excess of $2 in order to be made as happy as he would be if the price of apples were reduced by $2.
Similarly, we might ask whether taxing away $2 from the consumer when prices are low would leave the consumer just as happy as the consumer would be were he to experience a $2 price increase.
Again the answer would be “no.”
When the price of apples increases, it is clear that the consumer becomes poorer; his wealth buys him less. If the consumer’s taste for apples decreases with poverty, however, then the consumer will prefer a $2 increase in the price of apples to having $2 of cash taxed away from him.
Because he prefers other things to apples as he becomes poorer, the consumer will place a higher value on cash, which he can use to buy things other than apples, than he places on the price of apples.
But if a tax of $2 makes him less happy than he would be under an increase in the price of apples of $2, then a tax of less than $2 is equivalent, from his perspective, to an increase in the price of apples of $2.
So, overall, we have the peculiar result that a $2 price reduction is equivalent to a cash payment of more than $2 but a $2 price increase is equivalent to a cash reduction of less than $2.
Commensurability would, of course, require that all these things be equal.
And so we see that wealth effects not only prevent us from saying that a $2 gain to the producers creates the same amount of pleasure as a $2 gain to a consumer, but also that a $2 gain to the consumer via a price reduction creates the same amount of pleasure as a $2 cash payment. And the same can be said of losses.
GLB don’t acknowledge that between- and within-person incommensurability both stem from the same problem of wealth effects. But they do a good job of discussing both.
They also spend considerable time refuting the arguments of mainstream economists that within-person incommensurability is small and can be ignored.
But even if it were small, and indeed, even if wealth effects were not a problem for commensurability between persons either, willingness to pay would remain a highly problematic measure of value.
There is no basis for supposing that, just because two people having the same wealth level are willing to pay the same amount for a particular good, they will get the same level of pleasure from it.
Indeed, it is possible that two people who place the same relative values on all goods, and so are willing to pay the exact same amount for each good, might experience very different levels of pleasure from consuming them.
One person might take almost no pleasure from any good. Another might be sent into fits of ecstasy by the smallest purchase.
So long as the relative pleasure conferred by each good vis a vis the other goods is the same for both people, each will be willing to pay the same amount for each good. They will divide their budgets between goods in exactly the same way despite deriving very different levels of pleasure from them.
The Return to the Social Welfare Function
As GLB relate, welfare economists responded to these limitations by giving up on what might be called the overall “revealed value” approach to measuring welfare embodied in the pareto criterion and potential pareto (i.e., willingness-to-pay-based) criterion.
These criteria took a common revealed value approach because they both tried to read value from the actions of economic agents.
Whether a transaction satisfied the pareto criterion could be determined by checking to see whether the parties entered into it voluntarily. If they did, then it followed that neither party was made worse off.
And if a consumer purchased an apple at $10 but not at $11, one could infer that the maximum the consumer was willing to pay for apples was $10 and use that number to determine by how much the consumer could be compensated, pursuant to the potential pareto criterion, for the loss of an apple.
Under both approaches, economic agents were assumed to reveal the pleasure they take in goods via their actions, enabling economists to identify changes in welfare associated with various policies without needing direct access to the pleasure centers in consumers’ brains in order to make those determinations.
With the demise of willingness to pay, welfare economists would no longer try to find a way to read the pleasure and pain of consumers through their economic behavior.
Instead, they would return to the direct approach that they had abandoned more than fifty years before; they would try to measure happiness directly.
They took a variety of approaches to this problem. They would ask people if they are happy or not in various situations; they would study health indicators such as longevity, freedom from disease, and so on, in various situations; they would consult psychologists and neurologists.
Based on the results of these inquiries, they would identify the material circumstances most likely to be conducive to happiness and recommend economic policies (such as antitrust cases) that produce those circumstances.
Medical inquiry might determine, for example, that spinach is good for consumers. Welfare economists would then respond by ranking policy choices that lead to more spinach consumption higher than those that lead to less.
This was a departure from the willingness to pay approach, according to which welfare economists would have given spinach consumption the ranking implied by the dollar value that consumers revealed themselves to be willing to pay for spinach relative to what they would pay for other things.
Now other branches of science, and not revealed preference, determined the ranking.
This takes us up to the present state of welfare economics.
And for GLB, this completes the argument for taking political power and small businesses into account in doing antitrust.
According to GLB, one can no longer argue that, because consumers are manifestly willing to pay high prices charged by dominant firms, consumers like big firms and like the influence they have over politics.
Consumers’ willingness to pay is no reliable measure of the pleasure they get from buying the products of politically influential, small-business-destroying monopolists.
Instead, as already mentioned, GLB point to studies that suggest that consumers are happier in democratic environments free of concentrations of economic power. And that consumers are happier when they have access to small businesses.
It follows, argue GLB, that it is perfectly reasonable, per current practice in welfare economics, to argue that mergers that increase consumer surplus in the willingness-to-pay sense nevertheless make consumers unhappy, and should therefore be targeted for antitrust enforcement.
The Willingness to Pay Measure Is about Choice, Not Happiness
GLB’s paper presents a powerful rejoinder to any antitruster who might have been under the misapprehension that willingness to pay is a good measure of happiness. There are surely some out there.
But I suspect that the paper will not win too many converts, because what attracts people to willingness to pay is not that it is a good measure of happiness, but instead that it is the best way of doing justice to consumer choice that we have.
Welfare economics embodies a tension felt throughout the modern human rights project regarding who decides what happiness means.
Do we study human beings as if they were complex robots, figure out what makes these machines happiest, and impose those conditions on them? Or do we let the machines decide what makes them happiest?
GLB tell the story of welfare economics as if the field has always been interested only in the first option: to figure out what makes people happy and then impose those conditions upon their economic lives.
Under this assumption, GLB’s conclusion follows immediately from the arc of welfare economics. Willingness to pay is not a good measure. Others must be found.
But, as GLB acknowledge, economists have known almost from the inception of the willingness to pay approach in the 1940s that it was unsound. Why hasn’t the field moved on?
GLB chalk it up to “zombie economics.”
The real reason is that many people want to preserve a space in which consumers can vote for what they want through their purchase decisions.
That is, these people don’t view economics as a descriptive science but rather as a democratic project. It is the project of empowering consumers to vote on the character and magnitude of production through their purchase decisions.
The willingness to pay measure is ultimately built upon such a foundation, because willingness to pay is measured by observing the prices at which consumers do and do not buy.
The measure is highly imperfect, even incoherent, but it is the only way economics knows to recommend policy changes that account for the votes consumers have cast in markets. It honors their choices.
Happiness surveys, public health information, and the like are based on consumer input, but they are not based on purchase decisions—they are not based on circumstances in which consumers are forced to put their money where their mouths are.
Of course, the question whether consumers should take direction from experts regarding what to buy, or make those choices themselves, has already been resolved in favor of consumer choice.
Neither GLB nor anyone else will be able to impose purchases on consumers unless consumers vote to elect political leaders who take the GLB approach.
If antitrust enforcers decide to follow GLB’s paper, but consumers don’t like it, consumers can always vote political leaders into office who will sack those enforcers or give them new legislative commands to follow.
The premise of the economic project of enabling consumers to vote through their purchase decisions is, however, that the electoral process is defective.
The assumption is that, at least with respect to industrial production, consumers are better able to choose by voting through purchase decisions than by voting for elected representatives to direct production.
That is the subject of public choice theory. It is the view that, at least with respect to some matters, markets are more democraticthan democracy.
People who hold this view won’t be swayed by GLB. In their view, markets are most likely to maximize happiness if they are structured to read it in consumers’ purchase decisions, not if they are structured by consumers’ elected representatives to achieve happiness according to any other measure.
Ultimately, the battle in antitrust over the consumer welfare standard, is, like all battles over regulation, a battle over the legitimacy of the electoral process.
And yet progressives have spent remarkably little time contesting the public choice view of the electoral process and government regulation as inherently vulnerable to capture.
I suspect that is in part because many progressives share the public choice intuition.
Indeed, distrust of government seems to be one of the major reasons for which some progressives have focused in recent years on strengthening antitrust instead of pursuing the projects that earlier generations of progressives thought were more likely to be effective, such as price regulation and taxation.
Even an antitrust that imposes an external standard of happiness on markets instead of trying to read a standard from consumer purchase decisions pays a certain amount of respect to those purchase decisions. It is oriented toward preserving markets and empowering consumer choice within them.
In contrast, taxation and price regulation are relatively indifferent to those goals. They represent a pure privileging of choice via the electoral process over choice via markets.
And to many people from both left and right operating in an essentially anti-statist culture, that’s scary.
The irony, then, may be that the worldview required to overturn the consumer welfare standard in antitrust is undermined by progressives’ own attraction to antitrust as a vehicle for progressive change.
In what was, I think, the only math course I took in college, the professor said: “when you are trying to solve a problem, look for the asymmetries and ask: why? Asymmetries happen for a reason, and if there is no reason, then the asymmetry shouldn’t be there.”
It is a staple of tax theory that the income tax touches both labor and investment income whereas the consumption tax touches only labor income.
Why this asymmetry?
The Conventional Account
The standard explanation sheds no light on the question. It goes like this.
Suppose that you work for a period l at wage w, invest the resulting income of wl at interest rate r, and then spend the proceeds.
Under an income tax of rate t, you will pay twl in tax on your labor income, leaving you with wl-wlt=wl(1-t) remaining to invest. Your investment will grow to wl(1-t)+wl(1-t)r=wl(1-t)(1+r), but the interest earned on the investment, wl(1-t)r, counts as income as well, so it, too, will be taxed at rate t. You will therefore pay twl(1-t)r in additional tax, leaving you with wl(1-t)(1+r)-twl(1-t)r=
to spend after all taxes have been paid.
By contrast, under a consumption tax, you will be able to invest all of your labor earnings, wl, to obtain wl(1+r) after your investment pays out. Assuming that the income and consumption tax rates are the same, you will then pay tax twl(1+r) when you go to spend your income on consumption items, leaving you with a total of wl(1+r)(1-t)=
The expressions for the income tax and the consumption tax both have a tax term, 1-t, applied directly to labor income, wl, suggesting that they both tax labor income in the same way. But the expression for the income tax also applies the same 1-t term to the rate of return on investment, r, which the expression for the consumption tax does not. This suggests that the income tax taxes investment returns whereas the consumption tax taxes only labor income.
Indeed, if we construct an expression for an income tax that applies to labor income but not to investment income, we end up with our expression for a consumption tax.
We start with labor income wl and then apply a tax of twl, as we did in analyzing the original income tax, to obtain wl(1-t). This amount is again invested to obtain wl(1-t)(1+r). But now we are done—we do not apply a tax on investment returns. The result,
is identical to the expression that we obtained for the consumption tax.
We must, then, conclude that a tax on consumption taxes only labor income whereas the income tax taxes both labor income and investment income. (The table at the end summarizes after-tax value under this and a number of other scenarios that I will discuss.)
Why this asymmetry?
It is not intuitive.
A consumption tax is a tax on money going out, whereas an income tax is a tax on money going in. If everything that comes in must eventually go out, should not the income tax and the consumption tax be identical?
If the money the worker spends on consumption was generated in part through financial investments, shouldn’t a tax on everything he spends (i.e., on consumption) end up taxing the return on investment as well as labor income?
If there is no good reason for an asymmetry, then it should not be there.
If it should be there, then we have not really understood tax policy until we have understood why the asymmetry is there.
It turns out that the conventional account of the difference between consumption and income taxation both ignores a broader symmetry between the two approaches and is equally mum regarding a good reason for the asymmetry.
A Latent Symmetry
Let’s start with the broader symmetry that the conventional account omits.
Our method will be to try to find—or construct—symmetry between the income and consumption taxes. If we can figure out what we need to make the treatments symmetrical, we can then ask whether the absence of the things that are needed for symmetry is justified.
Comparing Workers with Rentiers
The first thing to realize is that the conventional account jumbles the tax treatment of labor income together with that of investment income. It tells the story of a laborer who starts with labor income and then invests it, suggesting that all income originates in labor. It starts with wl.
But not all of us are so unfortunate.
Some people don’t work—don’t have to work—because they get all of their income from financial investments.
The money they invest is so great that the return it throws off is large enough to make them prefer not to work for labor income.
There’s an old word for such people: rentiers. As Piketty has shown, they are making a comeback.
To avoid any quirks in tax treatment created by jumbling the experience of the rentier together with that of the worker, let’s consider the worker who gets all of his income from labor and the rentier who gets all of his income from investments—rather than consider a worker who plays rentier with his labor income, as in the conventional account.
Let’s look first at the rentier and the income tax.
The rentier has some financial endowment—call it p for property—which he grows to p(1+r)=p+pr through investment. At this point, the income tax is applied to his interest income, pr, and he is taxed the amount tpr. This leaves him with p+rp-tpr=
We see that the effect of the application of the income tax to investment returns reduces the after-tax growth rate of the investment from r to r(1-t).
Now let us consider the experience of the workers under the income tax when he does not invest his earnings in financial assets.
In that case, the worker earns wl, pays tax on it, and does not invest, ending with
The income tax treatment of the worker clearly differs from that of the rentier. We should be surprised that they do.
The income tax taxes both labor income and investment income at the same rate t, and here we have a rentier who has generated only investment income and a laborer who has generated only labor income. Shouldn’t the two be taxed in the same way?
The answer holds the key to our problem.
We can find it by continuing to try to construct symmetry between the treatment of the rentier and the worker.
One way to do that would be by eliminating the 1 term in the parentheses in our expression for the rentier’s after-tax income. That would give us pr(1-t), which is analogous to the expression for the worker’s after-tax income, lw(1-t).
But doing that would not make any sense. Eliminating the 1 is equivalent to making the claim that when you invest an amount p you obtain interest on your investment, pr (which is then taxed to obtain pr(1-t)), but not the return of the amount you invested, p.
But that’s not how finance works. When you invest an amount, you don’t lose your capital and gain only the interest paid on it. Instead, you get your capital back, plus interest.
We can also create symmetry by adding a 1 to the expression for the worker’s after-tax income, to obtain
Here is where things get interesting.
Labor as Capital
Adding that 1 in makes the claim that labor income, wl, is equivalent to the interest paid on an investment of working hours l. The worker invests l by working, and that investment grows into l(1+w)=l+wl, the worker’s l hours of work plus labor income wl.
But the “labor capital” that the worker has invested, l, is not taxed, only the return on that capital, wl, is taxed, just as, in the case of financial assets, the capital, p, is not taxed, and only the return on capital, pr, is taxed. As a result, the worker pays twl on value of l + wl, and ends up with l+wl-twl=l(1+w(1-t)), our symmetrized expression for labor income.
That is, with this change, the wage, w, becomes the analogue of the interest rate, r—it becomes the rate at which labor is compensated, just as r is the rate at which investment is compensated. And the number of hours worked, l, becomes the analogue of the dollar amount invested, p—it becomes the amount of labor with which the worker is endowed, just as p is the amount of investment capital with which the rentier is endowed.
But what does it mean to say that a worker invests l hours of labor and receives those l hours plus a return equal to labor income in exchange?
Of course, if one works l hours, one never gets those hours back again. Perhaps one receives the satisfaction of having worked l hours of honest labor alongside one’s income on that labor.
Or the l hours represent one’s ability to work over a given period of time, and, as this ability does not diminish from one period to the next, at the end of each period one starts over afresh with the same amount of hours on hand to invest in labor.
We do not need to resolve this question, however, because, for purposes of comparing consumption and income taxation, the nature of a labor hour matters only to the extent that this affect’s the hour’s exchange value. But labor hours have no exchange value.
Labor hours cannot be transferred.
One’s labor time is personal to oneself.
Only I can can expend my hours on work because only I have those hours.
You can pay me for my work—that’s the return that you pay me on my investment of my time in laboring for you—but you yourself cannot work my hours. If I could transfer my hours to you, then you could work 24 or 48 or 72 hours in a single day, for you could work my hours and your own ours and the hours of others.
But that, of course, is impossible.
I am a capital asset that only I can use.
Because labor hours cannot be transferred, they have no price. They cannot be exchanged for cash, cannot be spent on consumption goods or services, and cannot be taxed.
When we speak of financial investment, we speak of p(1+r) and when we speak of labor, we ought to speak, analogously, of l(1+w).
That is, we should think of labor hours in the same way as we think of financial assets, and we should therefore think about the taxation of labor income in the same way as we think about the taxation of investment income.
When we tax investment income, we don’t tax the financial asset that is invested—we don’t tax “savings”—but only the interest on that asset—the return on investment.
Just so, when we tax labor income, we don’t tax the labor asset that is invested—those labor hours—(how can we?) but only the interest on that asset, which is the interest rate—here called the wage—applied to the asset in the form of hours worked. That interest on the labor asset is otherwise known as labor income.
Thus the rentier’s after-tax investment value is p(1+r(1-t)) and the worker’s after-tax investment value is, similarly, l(1+w(1-t)).
But in the case of the worker we don’t write down l(1+w(1-t)) and instead write down wl(1-t) because the fact that l is not transferable, has no cash value, can’t be spent on consumption, and can’t be taxed causes us to forget that it is there.
Financial capital is a transferable thing—it’s dollars and cents, or things that can be exchanged for them—-and we can and do tax it and consume it. The existence of p is therefore constantly before our mind’s eye and we do not omit to write p(1+r(1-t)) rather than, analogously to the worker’s case, pr(1-t).
But what does this hidden symmetry between the income tax as applied to the rentier and the income tax as applied to the worker tell us about the asymmetry that we set ought to conquer, which is the asymmetry in the consumption tax treatment of labor and (financial) investment?
That answer is: a lot.
It shows why the asymmetry exists.
Eliminating the Asymmetry by Treating Labor as Capital
For if we acknowledge that labor income is the return on an investment of labor time, and if it were possible to transfer one’s endowment of labor hours, l, so that it could be taxed or consumed, just one’s endowment of financial assets, p, can be taxed or consumed, then the consumption tax would no longer be equivalent to a tax on labor income; it would differ both from a tax on labor income and from a tax on investment income and would differ from both in the same way.
Thus the asymmetry between the consumption tax treatment of labor and investment income would be eliminated.
If we acknowledge that labor income is the return on labor hours, and if one’s labor endowment were transferable—and so taxable and consumable—then a worker who generates labor income but does not invest in financial assets would generate value equal to l(1+w) from working and then pay tax tl(1+w) on this value, leaving the worker with consumption equal to
But, as we have already seen, under the income tax, the worker’s after-tax value would be l(1+w(1-t))—it would differ from value under a consumption tax in that the factor 1-t would be applied to the wage instead of to the entire investment.
And precisely the same would be true of the income and consumption taxes with respect to financial investments.
We have already seen that under an income tax the rentier’s after-tax value would be p(1+r(1-t)). Under a consumption tax, the rentier would invest p, obtain p(1+r), and then pay the consumption tax on that amount, leaving p(1+r)(1-t). Thus, here again, the difference would lie in the application of the factor 1-t to the interest rate (the analogue of the worker’s wage) in the case of an income tax. The table at the end summarizes these results.
It follows that, as a general matter, we cannot say that the consumption tax taxes labor income but not financial income.
The conventional account is not generally true.
The consumption tax is neither a tax on labor income nor a tax on financial income.
Rather, in principle, it is a tax on wealth—it taxes both the return on capital and the capital itself, whether that capital is labor capital or financial capital.
Because all wealth—the capital and the return on capital—must, ultimately, be consumed.
We also see from this analysis that the income tax and the consumption tax are not the same, whether applied to labor income or financial income. The income tax taxes the return on capital whereas the consumption tax taxes both capital and its return.
The answer to the question how a tax on what goes in can differ from a tax on what goes out is that what goes in is not just income but rather wealth—capital plus returns thereon—and so an income tax will not fully cover it. Thus the tax on what goes out—the consumption tax—which does cover everything that goes out, will differ from the income tax.
The Untransferability of Labor Hours as the Source of the Asymmetry
So much for the general structural symmetry of the income and consumption taxes. We must now ask why the conventional account finds asymmetry.
Is the conventional account simply mistaken, or is there a reason why, in practice, we must depart from the general structure?
It turns out that we have already encountered the answer. There is a good reason why, in practice, there is asymmetry. That reason is the untransferability of labor hours.
We cannot, in fact, say that under a consumption tax the after-tax value enjoyed by a worker is l(1+w)(1-t), because that equals l(1-t)+w(1-t), implying that the tax authority taxes tl labor hours, and the worker consumes l(1-t) labor hours.
But that’s impossible, because those hours can’t be transferred, either to the government or anyone else.
Unlike the rentier, who, at the end of the day, spends both his investment capital and his returns on consumption, the worker cannot spend his labor hours on consumption because he cannot trade them. He generates cash for consumption only through the returns he generates on his labor, which are paid to him in dollars or other tradable commodities.
His value after application of the consumption tax is, therefore, his labor income—his return on his labor hours—wl, less a tax twl on those returns, or wl(1-t), plus his labor capital, l, which he still holds, even if he cannot trade or consume it. So it is l+wl(1-t)=l(1+w(1-t)).
But that makes the consumption tax identical to the income tax—the same identity found by the conventional account. For we have already seen that under the income tax, he is left with l(1+w(1-t)), and as the income tax touches only labor income, wl, the fact that he cannot transfer his labor hours does not change the analysis.
Financial assets are consumable, however, and so the nonconsumability of labor hours counsels no change to our expression for the rentier’s value after application of the consumption tax. It remains p(1+r)(1-t) and so remains different from the expression for the rentier’s value after application of the income tax, p(1+r(1-t)).
So far from failing to tax financial capital, the consumption tax in fact does tax it—p(1+r)(1-t)=p(1-t)+rp(1-t), so financial capital, p, is reduced by 1-t—and it is the fact that the consumption tax taxes financial capital, while not taxing labor capital, that accounts for the difference in the way the consumption tax treats the two forms of endeavor.
Thus the nontransferability of labor introduces the asymmetry between the treatment of labor and investment income by the consumption tax that we see in the conventional account—and this is evident even without telling the story about the reinvestment of labor income in financial assets through which the conventional account makes this point.
The conventional account does not, of course, say that the worker’s after-tax value is l(1+w(1-t)), but rather that it is wl(1-t)(1+r). That is because the conventional account doesn’t recognize that labor hours are labor capital—or doesn’t care whether they are, because they are nontransferrable—and so doesn’t bother to write down the 1 in l(1+w) when it writes down the worker’s starting income. The conventional account omits the 1 and writes down wl instead. According to the conventional account, the worker’s wealth is exclusively his return on labor capital.
And, as already noted, the conventional account goes on to imagine the worker playing rentier and investing his labor income in financial assets. Thus, in the case of the consumption tax, the worker’s income is invested in financial assets and grown to wl(1+r) before being taxed down to wl(1+r)(1-t). So, in the conventional account, wl is in effect substituted for p in our expression for the rentier’s after-tax value of p(1+r)(1-t).
Similarly, in the income tax context, we have labor income of wl that is taxed down to wl(1-t) and then ploughed into financial investments. So wl(1-t) is substituted for p in our expression for the rentier’s after-tax income, p(1+r(1-t)), to obtain wl(1-t)(1+r(1-t)), the familiar result of the conventional account in the case of an income tax.
We can now restate what we have learned in terms of the conventional account.
The fact that after-tax value under a consumption tax in the conventional story has the same form, wl(1+r)(1-t), as a tax on labor income (that is then invested tax free), wl(1-t)(1+r), is not due to any failure on the part of a consumption tax to tax investment returns. After all, wl(1+r)(1-t)=wl(1-t)+wlr(1-t), so investment returns wlr are taxed down to wlr(1-t).
Rather, it is due to the fact that labor time, l, is not taxed under a consumption tax. If it were, then, adding 1+r terms (to reflect financial investment) to the expressions we generated for the worker above, we would find that the after-tax value expression under a consumption tax would be l(1+w)(1+r)(1-t), whereas the expression for a tax on labor income that is then invested tax free would be l(1+w(1-t))(1+r), which is a different quantity. It is only by deleting the first 1 to be found in each expression that they collapse into each other. (Equivalently, we could render labor hours untaxable—pulling the first 1 in the case of the consumption tax out of the parentheses—as we did above, and achieve the same result.)
Taxing Capital Would Restore Symmetry
Let us return to our simpler comparison of the tax treatment of worker and rentier in which the worker does not plow his earnings into financial assets.
Our understanding of the source of the asymmetry in consumption and income taxation helps us see how to eliminate it.
We have seen that the key to the asymmetry is that the consumption tax taxes capital and the income tax does not, but labor capital is untaxable, so the consumption tax and the income tax collapse into each other in the case of labor income—but not in the case of financial income.
The way to make the income tax equivalent to the consumption tax is, therefore, to redefine the income tax to apply to financial capital.
Make the income tax a wealth tax.
That would make the consumption tax as applied to investment income the same as the income tax, and the consumption tax and the income tax would, then, be equivalent both when applied to labor income and when applied to investment income.
Because no amount of redefinition of tax rules can make labor hours transferable, such a wealth tax would not apply to labor capital, and so the income tax as applied to labor income and the consumption tax would continue to be identical for practical purposes.
But now the income tax as applied to financial investments would be identical to the consumption tax, for now the rentier would pay tax both on p and his return rp, and so he would have his investment outcome, p(1+r) less tax tp(1+r), or p(1+r)(1-t), and the consumption tax on the rentier’s investment outcome would be the same as the income tax that he pays—tp(1+r)—leaving the rentier with the same after-tax value of p(1+r)(1-t).
This situation is slightly more complicated when we translate this insight back into the conventional account, because in the conventional account the worker is also the financial investor.
The conventional account’s expression for the income tax applied both to labor and financial income, wl(1-t)(1+r(1-t)), differs from that for the consumption tax, wl(1+r)(1-t), in two ways.
One is that the second 1-t term is applied directly to r rather than to the entire expression. The fix of taxing financial capital solves this problem. The worker’s financial capital—wl(1-t)—would, then, be taxed alongside the return on that capital, rwl(1-t), and so we would have after-tax value of wl(1-t)(1+r)(1-t), bringing the expression closer to that for the consumption tax.
But that still leaves the first 1-t term as an extraneous term. That term is present because the worker uses his labor income for purposes of financial investment and his labor income is taxed when it is earned. In effect, the worker who invests would be taxed twice under a wealth tax. Once when he generates labor income and again when he converts that labor income to financial capital and invests it.
This is the sort of problem we were able to ignore in comparing the experience of the worker with that of a separate rentier.
The income and consumption taxes can be brought into equivalence by recognizing a deduction for labor income that is invested in financial markets, in which case the worker will invest wl instead of wl(1-t), and the worker’s after-tax income will therefore become wl(1+r)(1-t)—identical to that for the consumption tax.
Labor income that is earmarked for investment is properly treated as investment capital, and so it should be taxed after it is invested, not when it is first acquired. If we tax labor income that is earmarked for investment, then we should tax all financial assets at the time that they are acquired—that is, before they are invested—and then again when the investment pays out.
The proposed fix to the asymmetry in consumption taxation—taxing financial capital and deducting labor earnings that are invested in financial markets—has a name: it is called a “cash-flow consumption tax.” The fact that it is called a consumption tax and not a wealth tax, even though it is collected when cash flows in rather than when it flows out, is an acknowledgement that this sort of tax on inflows is equivalent to a tax on outflows.
The Radicalism of the Consumption Tax
Defenders of the income tax (as presently defined) like it because it taxes more. Under the conventional account, after-tax income under the income tax is wl(1-t)(1+r(1-t)), which is less than after-tax value wl(1+r)(1-t) under a consumption tax because an additional 1-t term is applied to the interest rate, r, under the income tax.
But after-tax income is lower under the income tax only because, under the conventional account, labor income is taxed once before it is invested, and that happens only because the conventional account assumes that a financial investor’s source of income is labor.
For the rentier, the situation is reversed. We have already seen that the rentier subject to an income tax ends up with p(1+r(1-t)) in after-tax income whereas the rentier subject to a consumption tax ends up with p(1+r)(1-t). In the latter expression, the 1-t is outside of the parentheses—reducing the entire magnitude contained in the parentheses—and so after-tax income is lower than in the case of the income tax. That is because the rentier pays tax on his financial capital, not just his returns.
Because the consumption tax is a wealth tax; it is more radical than the income tax.
The truly rich, who do not work but live off of their financial assets, do worse under a consumption tax than under an income tax.
It is only the worker, who scrapes together savings to invest in financial markets, who ends up being taxed more heavily under the income tax than under the consumption tax.
That same math teacher who taught me to question asymmetries also liked to call out students who weren’t paying attention in class. “Are you an angry young man?”, he would ask.
But I would like to tax the rentier’s capital.
Worker income invested (conventional account)
Worker value (transferable labor)
Rentier value (transferable labor)
Worker value (untransferable labor)
Rentier value (untransferable labor)
Worker value (labor capital ignored)
Rentier value (labor capital ignored)
Worker income invested (conventional account with cash-flow consumption tax instead of income tax)