The following is the text of some brief remarks that I offered at a conference in Berkeley last Friday honoring Alan Auerbach upon his retirement. Guidelines for these remarks suggested that one discuss recent work of one's own that reflected Alan's influence and intellectual presence in any of his multiple fields.
My most recent, though hardly my only, Alan-influenced, piece is called Time Is, Time Was: Evaluating the Use of the Life Cycle Model as a Fiscal Policy Tool, which recently appeared in an Elgar Research Volume on Law & Time. It responds to Alan’s important recent work with Larry Kotlikoff & Darryl Koehler, using intra-generational accounting to measure US economic inequality & fiscal progressivity.
Alan’s work with Larry and Koehler (which I’ll call AKK to save time) does this by using lifetime spending power, in lieu of such snapshot measures as income or wealth. It finds less economic inequality, and more progressivity, than you’d find using the snapshot metrics. I would guess that not all members of the Berkeley Economics faculty, even limiting it to those in this room, agree 100% with the paper’s analysis. But the analysis would simply be right, leaving nothing further to discuss, if one fully granted the premises that, over the full lifecycle, people exercise consistent rational choice, in the presence of complete markets, leaving aside liquidity constraints.
In a standard analysis of consumer choice between, say, movies and pizza, people seek to equalize the marginal utility of the last unit they consume of each commodity. AKK applies the same approach to consumption in different periods, on the view – surely correct – that these are in effect separate “commodities,” each subject to its own declining marginal utility as one consumes more of them. But my piece argues that equalizing marginal utility across periods is considerably more challenging than doing so for movies and pizza, and also is subject to various heuristics & decisional metrics that would be irrational in the absence of real world decision costs. Plus, changes in information that aren’t fully insurable may have an impact.
I conclude that the underlying model, under which it basically doesn’t matter when one earns a dollar, in determining when one spends it, is not sufficiently descriptively accurate to be treated as more than an important orienting benchmark. Like such other “it doesn’t matter” theories as the Coase Theorem, the Efficient Markets Hypothesis, and the Modigliani-Miller Theorem, its value lies more in its showing us where to look for falsifying conditions, than in its full empirical validity.
Does this mean that we should keep on using traditional snapshot metrics such as income and consumption after all? Not at all. They still have all the flaws that AKK rightly attribute to them.
I conclude that there is no simple answer to the question of how lifetime, as opposed to shorter periods (themselves requiring further definition) should be used in measuring economic inequality, fiscal progressivity, or the question of why (and how much) inequality matters. Indeed, perhaps more important than any particular conclusion is the need for continuing methodological humility and agnosticism in how we think about these issues.