Forgive student loans? [updated]

Justin Wolfers lays out the reasons why forgiving student loans is about the stupidest possible stimulus package. My condensed paraphrase:

• It gives money to relatively rich people or those more likely to become relatively rich
• You get awful bang-for-the-buck in giving money where it won't generate spending. 
• It's a lump-sum transfer to those who already have an education so doesn't affect current decisions to invest in human capital except to the extent folks bet on debt forgiveness happening again
• It induces lobbying for such forgiveness in any downturn
• The tradeoff isn't giving free money to people who have debt versus "the corporations"; it's against the next best alternative - giving money to poor people.

Wolfers' final line? "Conclusion: Worst. Idea. Ever. And I bet that the proponents can't find a single economist* to support this idiotic idea."

I hope then that Green MP Gareth Hughes' tweets (1, 2) on the policy proposal don't constitute endorsement. Gareth, you're a smart guy, your work on copyright has been top notch, and you're great on civil liberties. Why keep pushing for utterly absurd economic policies? Why does pro-environment and pro-helping poor people have to be bundled with utterly absurd economics?

Justin Wolfers is hardly a raving free market screw the poor kinda guy. His objective function - what he wants to achieve in the world - is probably very very close to yours, Gareth. There are better and worse ways of achieving economic objectives. Why do the Greens keep going for the sillier ones? [I blame David Suzuki]

As for New Zealand, I haven't recent data on loan take-up by household decile. A 2005 report has the following:

Students coming out of the highest decile schools also go on to carry the biggest student loan debt.

The 2010 report [HT: Greg Dwyer for both] shows those achieving a Bachelors have higher than median earnings a year after study (and even higher earnings three years out), with higher median earnings for the higher level degrees that also correlate with higher debt levels. And, debt repayment through the income tax system depends on income - at zero percent interest on those loans, the real present value of loan balances is lower for poorer people! Forgiving the debt of student loan holders disproportionately gives free money to high-income surgeons. Is that really what the Greens stand for?

Previously:

• Regressive free tuition 
• Obvious points on student loans
• Interest on student loans

* It would be awfully fun to have an iPredict market on which of the usual suspect economists in New Zealand would come out endorsing "Worst. Idea. Ever." I know where I'd put my money and at what prices. 

The impossible trifecta

Kevin Milligan says a Guaranteed Annual Income cannot simultaneously satisfy three goals. Instead, you have to pick two among the following:

1. low tax rate
2. high benefit
3. balanced budget

Treasury here in New Zealand modelled a GAI for New Zealand on the request of the Welfare Working Group (HT: Lindsay Mitchell). What did they find? A GMI paying $300 per week - the mean benefit income among those on benefits - would cost $44.5 billion, or $52.6 billion if we extended it to superannuitants as a replacement for NZ Super. The former could be covered by a flat personal income tax rate of 45.4%; the latter, 48.6%. But full fiscal neutrality would require tax rates of 50.6% and 54.4% - the lower tax rates would be just enough to cover the transfers, but income tax revenues are currently also used to fund more than just transfers.

If we recognize that most parents are beneficiaries via Working for Families and compensate them for the loss of our version of EITC with a $86 per child per week payment, we get a $57.1 billion fiscal cost and a personal tax rate of 50% (or 55.7% for fiscal neutrality).

And, even this rather expensive system leaves the worst off worse off, as it kinda has to. Treasury notes:

Although the Gini coefficient improves under all models, many beneficiaries (including the disabled, carers and sole parents) currently receive more than $300 per week and would be made financially worse off under a GMI scheme. Therefore the GMIs considered could distribute money away from those most in need of government assistance and toward those who have choices and opportunities but choose not to work.

Treasury also warned about potential adverse labour supply responses to the necessary personal tax rates. And, the induced gap between company and personal tax rates would increase IRD's enforcement costs.

Treasury concludes by reiterating Milligan's impossibility:

From the international examples it is apparent that the more equal a society is in the beginning, the lower the returns to a GMI scheme. That is, a New Zealand specific GMI would either be at a level of income too low to reduce poverty, or a level of income that is high enough to reduce poverty but is therefore expensive and hence distortionary through higher tax rates. 

I don't think Gareth Morgan's Big Kahuna scheme is able to escape the trifecta by imposing new taxes on capital or land. Why? Because those aren't free lunches either. If it's worth having a land tax, it's worth doing it regardless of whether we have a GAI. So in the first stage we set the optimal tax structure - and I'm completely unconvinced that a capital gains tax is all that hot an idea anyway (see Seamus's posts here here and here.) But whatever the optimal tax structure, we implement it in stage one. Then, we still have to increase all the tax rates by enough to pay for a GAI if we're going to have a GAI. And the impossibility reasserts itself.

Morgan squares things with a cheaper GAI paying $11k instead of Treasury's $15k. But I have a hard time believing that's a stable political equilibrium. Could NZ politicos really avoid the temptation of adding targeted benefits for the many folks currently on benefits totalling well over $15k? If not, how quickly do we wind up having a GAI on top of a targeted benefit system?

Public Choice as User Guide

Machiavelli's analysis was both positive and prescriptive: how ought the prince maintain and extend his power? Public Choice has sometimes been accused of being a handbook for would-be despots.

Now, Bueno de Mesquita (HT: @Isegoria) has produced a book that leans more closely to this kind of users-guide: The Dictator's Handbook: Why Bad Behaviour is Almost Always Good Politics. This one's going to the top of my Amazon queue. In the meantime, here's Moynihan's review.

There would probably be a market for an intelligent-layman treatment of all the public choice results on voting and elections written as a guide for candidates and junior politicians.

A worthwhile Canadian initiative

(with apologies to the always-excellent Worthwhile Canadian Initiative for appropriating their name).

 Naturally, I watched the Canada-New Zealand game in the Rugby World Cup on Sunday. Such was the mismatch that the outcome was known before the match started, so the biggest point of interest for me was which version of the Canadian national anthem would they sing—the version that starts in French and finishes in English, or the other way round. As it turned out, I was shocked, shocked to not to mention appalled to see that while the New Zealand national anthem was, as is now the norm, sung twice, once in each official language, O Canada was sung only in English.

O.K. so I’m not really worked up. If I had my druthers, the New Zealand national anthem would only ever be sung in Maori and the Canadian only in French, but that is because the words to one’s countries’ national anthems are far less embarrassing when sung in a language that is not your mother tongue. And I have to say that the arrangement of the O Canada that was performed at Sunday's game was the best I have ever herad.

But I did find the contrast interesting. Both countries are officially bilingual, reflecting a bilingual reality in Canada while serving a more symbolic (albeit important) role in New Zealand. But even in sporting events in Canada, the Canadian anthem is only ever sung once. Typically, in matches played in Québec they start off in French and switch to English halfway through, and do the reverse in the rest of Canada.

Let’s face it: to the extent that these pre-game rituals matter for stirring nationalistic passions, most New Zealanders would rather we skipped the anthem and went straight to the haka, so can’t we follow the Canadian example and do the anthem only once, starting off in Maori, and switching to English halfway through. It would make no linguistic sense, since the Maori version is not a translation of the English, but that is true of the two versions of O Canada as well. (“Your forehead is wreathed with glorious flowers”, just doesn’t work in English!)

As an alternative, we could adopt a different Canadian initiative and do what was done in the inaugural performance of O Canada as the official anthem of Canada in 1980—sing both versions simultaneously, so as not to appear to be favouring one language over the other.

Ignatieff on evil

Michael Ignatieff's essay on political evil reminds me of a great article by Lskvayan. Here's Ignatieff:

So, Wolfe’s first lesson is a very old one, but worth repeating. There is method in apparent madness. The world is not divided between a sane world of deliberative politics and an insane world of apocalyptic violence. It is all politics, all the way down. To call a terrorist attack “senseless” is merely to admit that you have not understood its purpose.

Lskvayan provided a nice rational choice story for Stalin's Great Terror. Stalin couldn't tell which of his workers were shirking and which were working. But instituting random draw exile to Siberian death camps with reduced probability of being sent to Siberia for turning in a shirker turns the workers into the monitors and helps solve the principal-agent problem that otherwise obtains. From Lskvayan's abstract:

There is no agreement about the reasons for Stalin's Great Terror of 1937–1939. This paper argues that the problem faced by Stalin was similar to the standard principal–agent problem: the country was run as one enormous firm with Stalin as the only residual claimant. The monetary incentive structure was inadequate and the threat of mass shirking by the agents was real. A simple model of a principal with two agents is developed to address the problem. Assuming that the agents can observe and can reveal each other's shirking, it is shown that, under some assumptions, an equilibrium exists with the following strategy profiles: unless someone's shirking is revealed, the principal is committed to randomly punishing one of the agents with positive probability; an individual agent never shirks and always reveals a co-worker's shirking. A case study of the period is used to check the plausibility of this hypothesis.

The very real output costs of exterminating a lot of workers in Siberia was a second-best solution to the very real output costs of Stalin's slaves' shirking. Other solutions aren't great either; it's the existence of such principal-agent problems and imperfections in Stalin's monitoring ability that provided the space within which small amounts of freedom could exist.

Meanwhile, some neuroscientists say evil doesn't exist.

An iTune

I've avoided iTunes. But there's a new Nirvana box set release with Butch Vig's original mix of Nevermind. And, StatChat gave me $40 in iTunes vouchers. So, off I went. After being unable to install iTunes for a couple of weeks because of some unfinished Apple install locked into my machine, I finally cleared things up. Then spent way too long setting up an iTunes account and re-logging in every time I turned around. But I first logged in as American. There, I saw the $1.29/song and $45 box set.

So I was surprised when my NZ account gave me a $2.39/song, $72 box set.

$1.29 US per song. Take a $0.75 exchange rate: NZ price then $1.72. Add GST*: $1.98. That's still a $0.41 price gap on an individual song. On a 70 song box set, that's $28.70 if you buy the tunes a-la carte.

What about the box set? $45 US = $60 NZ. Add GST*: $69. That's only $3 less! After you put in anything non-trivial for licencing fee differences or just the hassle of dealing with small market issues, the price is basically identical. [I only want the eleven Devonshire Mix tracks, so the box set doesn't much help me.]

If Bruce Hoult is right, things are worse: GST doesn't apply to iTunes because they count as an imported service rather than a good, and imported services are zero-rated. IRD suggests Hoult is right. The price gap is then $0.67 per song ($46.90 for a 70-song set) or $12 for the box set.

We'd always expect per-song pricing to be a worse deal than per album pricing. But I don't know why we'd expect the gap to be so much bigger here than in the States.

When I'd first seen the price gap per song, I'd figured price discrimination: maybe iTunes hits a richer part of the NZ demographic and so optimal price is higher. But surely it would be the high demanders who'd go for the box set over individual songs. It would be surprising for the bundle to sell for more than the components, but I wouldn't have figured a hefty box-set discount consistent with trying to hit the high demand high income cohort. Maybe those looking to buy a box set are more likely to engage in price search (setting up multiple iTunes accounts, comparison shopping with other online outlets) while those wanting individual tunes don't bother with those fixed costs.

Other candidate explanations:

• Less competition from other streaming sources in NZ than in the US.
• But we then require individual tunes are better substitutes than the box set for other streaming services.
• GST and licensing
• Definitely! But are there relatively more costly licensing arrangements for single songs than for box sets?
• Substitutes in form of physical media are way pricier here
• Sure. But why does that get us the box set for roughly the same price as in the States but not individual tunes?
• iTunes prices negotiated in $NZ, not $US, so currency fluctuations carry it.
• This one's pretty reasonable. If iTunes and the label have a set fixed per song price for everything, they'd just stick with that, but they'd need a separate price for a big compilation box set, and that could be negotiated in $NZ at more current exchange rates.
• High demanders have other options, like multiple iTunes accounts in different countries.
• That makes sense too - fixed costs in price search mean few would bother with the hassle for individual songs but might for a big purchase.
• Maybe the high demand, high income cohort, like me, only wants the Devonshire mix and doesn't need another set of mp3s of stuff they already otherwise have.

Other explanations?

Peer review and error rates

Inside Higher Ed points to a 30-year old study I'd not seen:

Although peer review is often portrayed as an institution that arose with the scientific method, Fitzpatrick suggests the roots of peer review were “more related to censorship than to quality control,” serving mainly to concentrate academic authority in the hands of journal editors and, later, their expert reviewers.
While this system created an effective supply-side filter, it was also susceptible to bias, as Douglas Peters and Stephen Ceci demonstrated in a 1982 experiment. Peters, of the University of North Dakota, and Ceci, of Cornell University, took already-published articles in 12 esteemed psychology journals and resubmitted them, changing only the authors’ names and affiliations and the phrasing of the opening paragraphs. Three of the 12 articles were caught by journal editors as duplicates. Of the nine that were not, one was published. The other eight were rejected, most on methodological grounds.

You'd think that's the kind of study that wouldn't now work - it would be too easy for a referee to catch plagiarism on a quick Googling. But maybe not. Nevertheless, I'd expect human ethics committees to object to the deceit involved on the part of the researchers, to deem the deceit and the cost imposed on external reviewers to be too perilous, and to kill the study; replication nowadays seems unlikely.

I wonder if the result is that surprising. Filters are set such that it's better to reject a whole lot of good articles rather than to let an unsound one through; if the signal is noisy, we'd expect that few articles would make it through twice. The experiment shows that articles by prestigious folks get rejected when re-submitted with a less prestigious affiliation, but without a control group of articles originally written by folks from less prestigious place, we can't tell whether the result is just what we'd expect from running things through the filter twice or if it demonstrates bias favouring those affiliated with prestigious institutions.

The rest of the Inside Higher Ed piece features interview with Kathleen Fitzpatrick on her recent book advocating changes in the academic publishing model; it's worth reading.