Tuesday, June 26, 2012

Might inequality actually be narrowing??

Kevin Hassett and Aparna Mathur of the American Enterprise Institute have just published a study entitled "A New Measure of Consumption Inequality."  The foreword, by Mathur, offers the following overview of the piece:

"In this study, Kevin A. Hassett and I set out to refute the common claim that inequality has grown to the extent suggested in [Thomas] Piketty and [Emmanuel] Saez’s [recent] work [deriving dire conclusions from the fact that U.S. income equality has so vastly grown in recent years and decades.]

"Economists have widely acknowledged that consumption is a better measure of economic welfare than income.  In general, individuals are better able to smooth consumption rather than income over their lifetimes, making consumption a more informative indicator in the study of inequality.  Unlike income, consumption remains relatively steady throughout life since individuals borrow during years with low income and save in high-income years.  Using consumption as the relevant measure of inequality, most studies conclude that, contrary to popular belief, inequality has remained fairly steady over the past thirty years.  Our study retains the focus on consumption inequality and arrives at a similar conclusion .....

"Overall, our analysis reveals a trend toward narrowing of the consumption gap between low-income and other households, contrary to popular perception of the issue.  Public discourse can often become skewed in one direction, therefore it is especially valuable to explore new methodologies in evaluating important issues such as the inequality gap,  Our conclusion in this study debunks the claim of widening inequality."

A couple of preliminary points in response here.  First, Hassett and Mathur don't contradict (nor do they claim to) the Piketty-Saez finding of dramatic increases in income (mainly earnings) inequality.  Second, their empirical finding on consumption inequality is indeed both credible and consistent with other studies.  The question is what to make of it all.  Given what we mean by inequality, and the reasons we might be concerned about it, how should we evaluate the importance of the Hassett-Mathur finding, relative to the Piketty-Saez finding?

On this matter, with all due respect, I must say that I find Mathur's closing claim that the study "debunks the claim of widening inequality" to be extremely wide of the mark - indeed, so much so that it inclines one (not 100% fairly) towards classifying the piece as towards the frivolous advocacy end, rather than the serious scholarship end, of work that AEI publishes.  This is unfortunate, because the study's findings are relevant, and they do help to fill out the overall picture concerning inequality that interested analysts and policymakers should keep in mind.  I'd certainly think things were even worse if consumption inequality had increased as much over the last 20-odd years as income inequality.  But Mathur's argument that consumption is the right standard here begs credulity - even for one, like myself, who has sympathy for a progressive consumption tax, based on arguments that initially seem similar to what she is saying, but in fact are quite different.

OK, let's take a simple comparison, between James the CEO and Joe the factory worker.  Suppose that in 1982 their income ratio was 20 to 1 and their consumption ratio was 10 to 1.  Then we look at their successors in 2012 (James, Jr. and Joe, Jr.) and find that the income ratio is 200 to 1 but the consumption ratio is still 10 to 1.  Does this "debunk[] the claim of widening inequality?"

Mathur's claim that the answer is yes appears to rest on assuming lifetime consumption smoothing.  In other words, she is apparently positing that their relative lifetime positions have not changed!  To make this really simple, suppose everyone knew his or her lifetime earnings (in present value) up front and engaged in perfect smoothing, i.e., spending the same amount each year.  Also, of course, assume zero bequests.  (The paper never mentions either bequests or inter vivos inter-generational wealth transfers.)  Then the fact that the consumption ratio was still 10 to 1 would show that we were somehow being duped by the rising one-year income gap into thinking that relative lifetime earnings had changed - when in fact they hadn't.  (Perhaps, for example, James' earnings had simply become more jammed into his peak earning years.)

As applied to actual people, this cannot possibly be true.  Surely the ratio of lifetime earnings, as between CEOs and people on the shopfloor, has sharply diverged since 1982, even if not quite as much as one might initially think from the Piketty-Saez annual data.  You simply can't make the numbers work in a believable fashion otherwise.  So where is it coming out?  Presumably in deferred consumption that departs from the perfect smoothing model, and above all through transfers to one's heirs.  Do we really believe that, say, Mark Zuckerberg is smoothing perfectly and that his current year consumption is likely to represent the same percentage of his expected lifetime income as in the case of a factory worker?  Leaving out the bequest issue is really not defensible - although I recognize that it raises further issues that it would be hard to discuss adequately in a short and mainly empirical paper.

Another way vastly unequal annual incomes may play out, other than through comparably increased annual consumption, is in terms of retirement and leisure.  CEOs can retire earlier than factory workers, if that's what they want to do. That's an aspect of true consumption and welfare that the studies don't capture.  (Although in fact high earners often like their work enough to be far less eager to retire, even when this is affordable, than are the people with drudgery jobs.)

Suppose that, in 1982, Joe the factory worker was spending 95% of his income on current consumption, while James was spending just under 50% (consistently with the posited relative ratios).  In 2012, we find that Joe, Jr. is likewise consuming 95%, while for James, Jr. it's just under 5%.  Has the claim of rising inequality been "debunked"?  Do we really think that the ratio between James, Jr.'s lifetime earnings and those of Joe, Jr., is likely to be the same as that for James versus Joe?  And can we actually believe that James. Jr. merely has a steeper peak earnings spike than James, rather than hugely increased relative lifetime earnings? That does not strike me as empirically credible.

OK, let's turn to the income versus consumption tax issues that are in the neighborhood, though not expressly invoked by Mathur.  As I discuss in my 2007 article, Beyond the Pro-Consumption Tax Consensus, the actual extent of lifetime consumption smoothing is very important to this debate.  With perfect lifetime smoothing (and holding the bequest issue to one side, as best resolved via the decision whether or not to have an inheritance or estate tax), the case for preferring consumption taxation to income taxation becomes compelling indeed.  But a vulgar version of the lifetime and smoothing-based argument, which mirrors Mathur's statement in the foreword, would clearly be wrong.

Let's look at James, Jr. and Joe, Jr. again, with their 10 to 1 consumption ratio and their 200 to 1 earnings ratio.  Suppose, purely for arithmetical convenience, that we would have a flat-rate tax whether it was levied on income or on consumption.  Thus, the ratio of their current year tax liabilities would be 10 to 1 under a consumption tax and 200 to 1 under an income tax.  Does favoring a consumption tax imply that one really thinks 10 to 1 is the right ratio to keep in mind when comparing these two individuals' relative wellbeing?

The answer is a flat-out No.  Consider here two points about the consumption tax.  First, if James, Jr. did indeed spend his entire salary this year, the ratio of their curent year tax liabilities would be 200 to 1, not merely 10 to 1.  (Again for arithmetical convenience, I ignore the point that Joe, Jr. is only spending 95%, not 100%.)  Second, if there's one thing we know about a well-designed consumption tax, it's that it is neutral as between current and future consumption.  Thus, James, Jr. lowers his current year consumption tax bill by consuming under 5% - but he does not reduce the present value of his (and his heirs') long-term expected tax liability with respect to these earnings.  Thus, in present value-equivalent terms, under the consumption tax he actually is paying 200 times as much tax as Joe, Jr. - not just 10 times as much.

In short, well-thought-out support for consumption taxation rests NOT on the point that current year consumption is the best measure of relative wellbeing - how could it, when extra savings can pay for future consumption? - but rather on the notion that the seeming current-year under-taxation of the bigger saver is actually corrected (at least in theory) over the long run.

It's a well-known fact, by the way, that higher-income people generally spend smaller percentages of their current year income on consumption than do lower-income people.  And this reflects departures from uniform smoothing, not just different relationships between peak and average earnings.  Ignoring this whole point, which pretty much guarantees that annual consumption differences will lag behind rising annual earnings differences, is not, in this setting, intellectual fair play.

It's a shame that the Hassett-Mathur piece, which provides information that is relevant to the inequality debate, is being oversold to support a clearly false proposition, which is that nothing relevant has changed.  But it's still worth reading, and the issue of what significance to ascribe to relatively unchanged consumption ratios does indeed merit further thought.

No comments: