Standards of evidence
In a broadly excellent discussion of theories relating inequality and growth, Jared Bernstein writes:
all of this research is relatively new, and while it makes suggestive connections, there is not enough concrete proof to lead objective observers to unequivocally conclude that inequality has held back growth.
Bernstein absolutely right, of course. But really? “concrete proof to lead objective observers to unequivocally conclude” Is that a remotely meaningful standard of evidence for, well, anything?
People on the political right, including many respected economists, make strong claims that ceteris paribus taxation is bad for growth. They certainly have plausible models in which it would be. But as an empirical matter, the fair thing to write would be there is not enough concrete proof to lead objective observers to unequivocally conclude that taxation has held back growth. The evidence is very conflictatory! To steal Bernstein’s apt metaphor, there are a lot of moving parts! It is in fact almost certainly false to claim that taxation is always and everywhere bad for growth, and almost certainly true that there are circumstances under which it has been and would be good. If you are behaving as a scientist, it’s kind of a shitty, stupid question, “how does XXX affect growth?” How does molybdenum affect life? They are related! But if you start running regressions of liveliness against concentrations of molybdenum, you won’t get very far. If you want to study the relationship between molybdenum and life, or XXX and growth, for any XXX, you’ll have to characterize mechanisms and offer detailed, contingent accounts. You won’t find simple, black box relationships.
But we are not always, or even usually, behaving as scientists. The Tax Foundation will tell you right off that taxes are bad for growth, much worse than spending cuts. Studies prove it, and if you disagree you are simply wrong. Steve Roth aptly wonders why so few voices among “respectable progressives” are willing to even give fair consideration to the case that inequality might be an impediment to growth. I think he has a point. This isn’t a general phenomenon. It’s not like “liberals are cautious scientists, while conservatives run roughshod over the truth”. Progressive economists are willing to assert, in the same stentorian, authority-of-science voice as the Tax Foundation people, that fiscal multipliers are real or that evidence against expansionary austerity is incontrovertible. But on connections between inequality and the macroeconomy, it feels like respectable progressives are always looking for an excuse to say there’s no there there. People who are usually very smart make very thin arguments that are frankly beneath them to cast doubt on the relationship.
Now let’s be perfectly clear: there is no reliable quantitative relationship between inequality and growth, just as there is no reliable quantitative relationship between taxation and growth, between government spending and growth, monetary policy and growth, or pretty much anything else and growth. There are studies, which pare and tease their panels in ways they justify on some grounds or other, and those studies yield conclusions. You can buy the assumptions, methodologies, and mechanisms implicit in those parsings or not, it’s your choice. But you won’t find a clear, incontrovertible relationship between any simple thing and developed world, per-capita growth. It’s too complicated a phenomenon. You have to buy someone’s stories, and interpret the numbers through those stories, to claim the evidence is strong.
But that doesn’t mean there is nothing at all that we can say about inequality and the macroeconomy. We can, for example, say that marginal propensity to consume effects are real. The intellectual history of MPC goes something like this:
There is and has always been an obvious, intuitive, and robust stylized fact that people with higher incomes save greater fractions of their incomes than people with lower incomes. In the post-Great-Depression intellectual climate, which was acutely sensitive to the dangers of demand shortfalls, this suggested, uncomfortably to some, that inequality could be a macroeconomic hazard.
Milton Friedman pointed out that differing marginal propensities to consume observed in the data might have nothing at all to do with inequality. If people try to smooth consumption over time, then in a stochastically equal society (one in which everyone’s expected incomes are the same, but each individual is subject to random fluctuations in any period), we would observe that individuals who happen to have unusually high incomes in one period save a lot, to cover the periods where they will have unusually low incomes. Friedman’s elegant Permanent Income Hypothesis suggested people just spend a constant fraction of their lifetime incomes in every period, so (under perfect information about that “permanent” income) each individual’s spending would be constant and apparently different marginal propensities to consume would be due solely to fluctuations in income, with no actual changes in spending.
For reasons that would be baffling, if I weren’t so cynical about the economics profession, the Permanent Income Hypothesis was generally accepted as sufficient explanation of observed MPC effects in cross-sectional data, and the issue was considered closed. You were naive and ill-informed if you thought MPC effects in the data had anything to do with inequality. They had been explained.
Of course, you had to be an idiot to believe that the Permanent Income Hypothesis fully accounted for MPC effects. Undoubtedly consumption smoothing explains a part of cross-sectional variation in marginal propensities to consume, but you don’t need careful empirics to prove that it can’t explain all of them. Why not? Because not consuming leaves a residue, something called savings, which becomes wealth. If across the income spectrum everyone spent and saved in equivalent proportions, we’d expect no cross-sectional variation in terminal wealth as a proportion of lifetime income. But in real life, much of the bottom of the income distribution dies with zero or negative wealth (i.e. they stiff their creditors), while those near the top of the distribution leave large bequests. An intergenerational Permanent Income Hypothesis could only explain this if poorer people expect their kids to be much wealthier then the children of moguls. Which is not so plausible.
If things that are obvious don’t persuade you, if something has to have tables in the back and be peer-reviewed to qualify as “rigorous”, you are a very severely deluded human. Nevertheless, a few courageous researchers have done the work of examining in numerical detail whether the Permanent Income Hypothesis is sufficient to account for variations in spending, and the answer is always no. I’ve cited ‘em before, I’ll cite ‘em again: “Why do the rich save so much?” by Christopher Carroll; “Do the Rich Save More?”, by Karen Dynan, Jonathan Skinner, and Stephen Zeldes. I’m sure if MPC makes a comeback in macroeconomic conversations, someone prestigious will find some way to parse the data differently and explain it all away again. That researcher will surely die rich.
OK. So inequality-related MPC effects are real. But what to they have to do with growth? Nothing at all, in an unconditional sense. I’ll go further than Bernstein. It’s worse than “there is not enough concrete proof to lead objective observers to unequivocally conclude that inequality has held back growth.” There’s little reason at all to think that inequality has held back growth, in the past tense, through an MPC channel. Why not? Because we didn’t observe in the past anything that looked like an intractable insufficiency of aggregate demand! Past-tense, we reconciled inequality with growth. MPC effects suggest that one way to generate more demand would be to broaden the distribution of income. They do not imply that broadening the distribution of income is the only way. The macroeconomic footprint of increasing inequality lies not in growth, but in the interest rates and financial chicanery that were necessary to support that growth. Call it the monetary offset.
Prior to 2008, we found means of supporting aggregate demand despite an almost certain drag imposed by increasing inequality. Those means included a broad mix that included fiscal policy (we ran deficits), unsustainable equity booms, the “democratization of credit” and unsustainable credit booms, and of course straightforward monetary policy. Real interest rates have collapsed since the early 1980s. The reason we might talk about inequality is not because it is mechanically, unconditionally, here-is-the-regression-now-STFU connected with growth. It’s because many of us have decided that other, more “conventional”, demand stimulants have run their course, that repeating them or increasing the dose won’t work, or would have adverse side effects we’d prefer to avoid.
We have a large menu of ways we can try to support demand. We can go the Scott Sumner route, double down on monetary policy. We could do big, old-style fiscal stimulus, have the government give money to those who lobby best without worrying about fairness or income distribution. We could embrace inefficient health care provision and build more university rec centers. We could have the financial sector figure something out again, some means of enabling those who otherwise wouldn’t be able spend to do so. We can find ways of persuading rich people to spend more. We can import demand from elsewhere, like China and Germany. We can try electronic money and negative interest rates. We are, as they say, free to choose.
But the reality of MPC effects means that, along with all those other possibilities, broadening the distribution of income would be expansionary and narrowing that distribution would be contractionary, ceteris paribus. If, like Larry Summers, it pains you that maybe the “natural interest rate” is negative now, the reality of MPC effects means that policy which broadens the distribution of income would help push it positive, and put us back into more comfortable territory. If, like me and Pope Francis, you think that present levels of inequality are horrific for human and communitarian reasons, then among the many macro policies that might support demand, it is rational to tilt towards those more likely to engender a broad distribution. It is quite irrational, as I think some well-meaning economists do, to hold MPC effects to much higher standards of evidence than the mechanisms that justify other interventions, because “economics is not a morality play” and reducing inequality would be the moral thing. Better to err on the side of human welfare rather than reputational purity.
I happen to think that the macroeconomic case for reducing inequality is much stronger than the case I’ve made here. I think the character of growth is badly misshapen when demand is narrowly sourced, that technological stagnation is mostly a distributional problem, that institutional correlates of growth are harmed by increasing inequality. But those are all more speculative claims. You can tell me the “jury is still out” on those. But the jury is not out, it never reasonably has been out, on the reality of distribution-related MPC effects. I’ll disagree, respectfully, if you claim that for supply-side or libertarian reasons we should ignore that reality and prefer other means of supporting demand (or that we should not worry about supporting demand at all). But don’t say “it’s unclear” whether income distribution affects aggregate demand, holding other factors constant. Of course it does.