This isn’t anything too new, but Scott Winship has some of the best writing that inequality in the United States is an overblown issue. There’s not much competition because most people – liberal and not – seem to have taken the position that inequality is bad, and getting worse. It’s not that Winship disagrees with that statement, per se, just that it is exaggerated and, presumably, shouldn’t be the pillar of national conversation. Most of the smart conservatives I read don’t deny the dangers of inequality as much as the notion that we can do something about it.
I think that’s a more defensible position. While Winship presents the strongest case that we shouldn’t be worried, I am not convinced.
Winship’s argument is excellent because he presents the case against inequality with almost complete fairness (passes the ideological Turing Test, that is). Still, there is one implicit assumption in the tale that makes the argument just a little too good:
Taken together, their errors severely undermine the liberal case about the consequences of inequality. True, a careful examination of the evidence does not establish that inequality is harmless, or that it has nothing to do with our other economic problems. Economic data cannot prove a negative. But they can fail to prove a positive, and they do fail to prove the claims that underlie the left’s basic economic narrative. They reveal little basis for thinking that inequality is at the root of our economic challenges, and therefore for believing that reducing inequality would meaningfully address our lagging growth, enable greater mobility, avert future financial crises, or secure America’s democratic institutions.
Let’s say an economist wants to show that “an extra hour of education improves outcomes”. He doesn’t show that this research hypothesis is true as much as he rejects his null hypothesis that “an extra hour of education does not improve outcomes”. When Winship says “economic data cannot prove a negative [but] they can fail to prove a positive” he’s asking liberals to reject the null that “inequality does not harm economic mobility”.
But why should he get the null? That is, why isn’t the debate framed such that conservatives (as an unfairly broad term for people who don’t care about inequality) have to reject the null that “equality does not harm economic growth”. Note, that Winship is right that economists can’t prove a negative. It’s entirely possible that with a given set of data we fail to reject both hypotheses. However, the choice of null changes the debate completely, and Winship ignores this remarkable design choice.I will argue – on generous assumption that we do not have data to reject either null – that it is the liberals who are right: our null hypothesis should be that inequality harms social mobility. (Studies that have been published after Winship’s essay, by the way, give statistically significant evidence that there is a connection, but that later.)
Imagine that an economist wants to study the effect of war on general welfare. He has a welfare measurement machine, but is unsure how well it works. The data is murky, and it’s not clear he can produce a statistically significant result. He has two choices of a null “war does not reduce general welfare” and “peace [= the lack of war] does not reduce general welfare”. It is clear that the former would be a ridiculous null. Enter theory.
For all the rap theoretical social science receives, without it empirics cannot mean much. There are good human theoretical reasons to believe war causes all kinds of damage. In fact, if we had the resources we could probably show a statistically significant proof to this effect. But it doesn’t matter that we don’t. Anything that results in the inane waste of economic resources and human life like war does not improve welfare other things equal.
Not even the neocons would disagree with the idea that everything else equal peace is preferable to war. On the other hand, it would be equally dumb for a public economist to choose “100% tax rates do not reduce efficiency” as his null. Chances are, given the randomized experiment, he could achieve a significant result whatever his null was, but even if he couldn’t it’s a dumb null. Carl Sagan once said “extraordinary claims require extra ordinary evidence”. This is kinda like that.
Of course, equality promoting policies might harm growth – make everyone equally poor is a GOP buzzword, right? In this case, we should take it as default that inequality harms mobility and compel conservatives to reject the null that “equality does not harm growth” – at which point we can have an honest discussion about the tradeoffs.
So what is the theory connecting inequality with immobility? Quite obvious. If we define social mobility in relative terms (i.e. the proportion of those born in the bottom fifth that make it to the top fifth) there is, by definition, a scarcity of space at the top. However, if those born at the top don’t move down, everyone else is competing for a smaller number of seats. As inequality increases, those at the bottom are less competitive resulting in immobility. Winship hides behind scientific language on this point:
[Claims like the above] are made by a number of respected academic economists, such as Miles Corak and Timothy Smeeding. As hypotheses, they make intuitive sense. But attempts to rigorously test them have been few and far between. More often than not, analysts fail to make much of a case beyond showing that things have worsened. Since income inequality has risen, they presume it must have caused these other problems.
In economics not everything can or should be tested. It’s an empty argument to claim that we don’t have something “rigorous” to prove something that is blatantly obvious. 50% of students from Exeter end up in the Ivy League. Maybe one kid from Camden does. There are many nuances, ifs, buts, and outs, but nothing changes the huge discrepancy. Things are intuitively sensible for a reason.
Unfortunately economists don’t have a lab to regress the same world with different Gini coefficients. We have to make do with the one we have. And “correlation is not causation” just isn’t what it used to be:
But the arguments in both The Spirit Level and Krueger’s analysis are deeply flawed. The basic issue in both cases is the same problem that has bedeviled other arguments about inequality: Correlation does not always indicate causation. High-inequality countries and low-inequality countries differ in many ways. Kenworthy, for instance, has noted that Nordic education policies could be behind the Great Gatsby Curve. More devastatingly, Manhattan Institute scholar Jim Manzi found that the Great Gatsby Curve was just as evident when countries’ inequality levels were replaced with the sizes of their populations.
Well, yes. And didn’t Jevons say something about Sunspots and Corn? Winship wants to throw the most obvious theory out of the window and give up the quest for policy to a randomized experiment. There’s almost nothing “devastating” about the Manzi study. There’s theory that suggests a relationship between mobility and equality. If there is something as convincing for a relationship between mobility and population, I haven’t heard about it. The relationship between population size and mobility is completely irrelevant to that between mobility and inequality, and it’s a rhetorical trick that obscures the fact that those who want to ignore inequality as a social problem are shifting the burden of proof even when theory is stacked against them.
Now, to be fair, theory also suggests that the most effective way of overcoming inequality – tax and transfer – is constrains growth. I’m entirely open to have a debate about whether there is a practically efficient way to overcome inequality. I’m progressive utilitarian and think this would be worth it but fully understand that a coherent argument can be formed to the contrary. That is a debate worthwhile having. Accusing those of overstating inequality with the empty sigh that there are no “rigorous” studies is not.
Much of Winship’s argument focuses on a lack of cross-national correlation between inequality and any host of factors. And while everyone knows that correlation does not imply causation, very few seem to remember that lack of correlation does not imply lack of causation. It’s entirely possible, indeed probably, that two events are causally connected but obscured by an unobservable variable. (At the bottom of this post is a simple heuristic model to this effect).
I have no problem with Winship’s points against the Stiglitz hypothesis that inequality increases the risk of financial crisis. The theory behind this is weak (and very easily refuted without citing a lack of “rigorous studies”) and, unlike the connection between mobility and equality, fails the “extraordinary claims require extraordinary evidence” test. But this weak point seems to be overblown in Winship’s argument against the inequality-obsessed left. Very few of us seem to think it caused the crisis or dampened the recovery. Even Paul Krugman is on record to the contrary. The dedication of so much word space to a relatively irrelevant part of the opponent’s argument hits a grey line on the ideological Turing Test.
The demand for rigor is suddenly lifted when it comes to inequality’s effect on political cycles. Winship cites some very good studies that conclude “rising inequality probably has not lowered expenditures in support of safety nets.” That may or may not be true. It’s an interesting study, but way to complex a system to achieve any statistically relevant answers. There are theories which may argue either way, and that’s a fair conversation. But I’m curious why Winship extends this freedom to politics but not to macroeconomics, which is only marginally less complex.
Indeed, there is good evidence citing a scary connection between money given to judges and the rulings they receive. Sure, this isn’t inequality per se, but it’s not difficult to see the link between a stratified business lobby which is more influential than ever. (And it’s not like Democrats are all that much better in this respect).
It is telling that Winship’s essay is the among the best essays questioning our obsession with inequality. But implicit through the whole piece is an assumption that it is the left who has something to prove, and the right can take the moral and economic costs of redistribution for granted. And yet, if you remove phrases like:
In other words, the claims of inequality’s political effects are even more difficult to demonstrate, and even more poorly established empirically, than are the claims of its economic effects.
There isn’t much left to the whole thing, except debunking the idea no one believed to begin with. I sound sarcastic not because I have something against Winship or this essay – it’s the best stuff on the subject as I’ve said multiple times – but because when you think about it the demand for statistical rigor is an easy out. (As is, granted, the statement that the demand for rigor is an easy out – but we have theory on our side).
If we didn’t have common sense theory I’d firmly be with Winship on this point. But we do. Common sense is a lot better than people give it credit for. Tax incidence is a frequent example some economists like throwing around explaining why common sense is useless. But most people know that taxing a business can hurt the consumer. That is common sense. In any case, maybe there are theoretical/intuitive arguments that inequality doesn’t hurt social mobility. Taking that to it’s logical conclusion – meaning a Gini coefficient in the 70% plus range – would likely break any such argument because, fact of the matter is, inequality does hurt mobility. We don’t need randomized, significant studies to prove it. And, believe it or not, it turns out there is good, randomized evidence that proves our case.
Finally, there’s another very important case for greater equality. If social mobility was perfect any amount of inequality, excluding Rawlsian concerns for another day, would be okay. But social mobility isn’t perfect. This means that those at the bottom have a shrinking chance at all the growth. They really do have to wait for something to trickle down, when the past twenty years have shown that isn’t happening. So completely irrespective of the fact that inequality reduces social mobility, if there is any immobility at all, then there is a moral imperative to redistribute to correct for this immobility.
Let’s say we’re in a two class society with 50% of the growth accruing to 10% first class of the population and the other 50% to to the 90% second class. If mobility is imperfect, that is those born in the second class have less than a 10% chance of entering the first class, it is the responsibility of the state to equalize the probability each citizen earns the marginal growth in the economy as a whole. Hence if we parametrize social mobility to be perfect at 1, for all values below 1, some level of social redistribution is important. Given a constant level of social immobility, the importance of redistribution increases as inequality increases. Given a constant level of inequality, the importance of redistribution increases as immobility decreases.
Considering inequality increases immobility, one doesn’t have to think hard about why the left is obsessed. I’m a raging neoliberal, but count me among the hysterically worried.