Cooper Union’s response

Here is the response I received from Jolene Travis, Assistant Director of Public Affairs, Media Relations at Cooper Union, after she read my post.

Dear Felix,
After reading “Occupy Cooper Union” (12/5), I writing to ask for a correction to factual errors cited in your article. I have enumerated the points below and, if you have any questions or would like clarification, I would be available to speak with you directly, as I often do with other members of the press such as Brian Boucher.

1)      He has released little more in the way of financial information than any of his predecessor
President Bharucha and Vice President of Finance T. C. Westcott have met with students, faculty, alumni and staff over the past year to conduct approximately 80 informational and financial meetings. The meetings are ongoing and open to members of the Cooper Union community mentioned in the previous sentence. Additionally, details about the planning process and financials are also available here: (letters from the President), (audited financial statements) and (financial FAQs)

2) A group calling itself Students for A Free Cooper Union has occupied the top floor of the iconic Foundation Building
11 students of the group “Students for A Free Cooper Union” have locked themselves in the 8th floor of the Foundation Building.

3)The Revenue Task Force, they were tasked with finding{revenues which} could not come from other obvious places, like the Board, alumni, or the sale of assets
The relevant paragraph of the Revenue Task Force report reads as follows:
“It should be noted that such areas as minimum trustee contributions, alumni and donor development, and sale of assets were outside the Revenue Task Force’s charge.

4) And so the report comes to the conclusion …that new graduates programs, some of which would charge more than $30,000 per year.
The Deans and faculty for the degree granting three schools will be presenting plans for graduate and other programs (some of the proposed graduate programs could charge in access of $30K) to the Board of Trustees, which will then be reviewed and a final decision is slated for early 2013.

5) Cooper might move to a system where a large part of the budget would come from grad students
The evaluation of the plans by the three schools is an ongoing process and the budget is comprised of many facets including, development efforts at the City, state and federal levels along with alumni, foundations and individual members of the philanthropic community

6) Charging for some students would violate Cooper’s mission, which says that it “awards full scholarships to all enrolled students.”
With the financial realities Cooper Union faces, the Board of Trustees has the authority to create programs that generate revenues and, therefore, can amend the mission statement.

7) A document leaked on Monday, written by a mysterious “Undergraduate Tuition Committee” somewhere within the engineering school, lays out all the reasons why Cooper should start charging tuition to everyone.
The acting dean of the Albert Nerken School of Engineering has organized committees to identify possible revenue streams through the creation of new programs. An analysis, which included 500 variables, was conducted to see at what level of revenue would be needed in order to maintain the full tuition scholarship for undergraduates. This is an ongoing process and no decision has been made.

8)  A year and a half into Bharucha’s tenure, there’s very little reason to believe that he’s the right man for the job –while the current occupation, which was vocally supported at a press conference Tuesday afternoon, seems to provide a pretty strong prima facie case that his university has no faith in him.
Tuesday’s press conference did not reflect the views of the entire student body and faculty. At today’s Board of Trustees meeting, the Board unanimously reaffirmed their confidence in President Jamshed Bharucha’s leadership.

*Also in paragraph 1, where the WSJ was cited as JB’s favorite media outlet it is important to note that The Cooper Union on a regular basis reaches out to a very broad range of local, national and professional media outlets as appropriate to the subject at hand.

Posted in Not economics | 3 Comments

Chart of the day, Apple vs Microsoft edition


Many thanks to Ben Walsh and the Thomson Reuters data team for helping put this chart together. There’s a lot of data here, which isn’t easy to obtain: some 15,000 datapoints in all, going back as far as 1983.

The chart was inspired by this post by Nick Wingfield — the point at which Apple’s market capitalization, in nominal terms, finally exceeded the previous all-time high, which was set by Microsoft in 1999. So we went back and took the closing share price of AAPL and MSFT for every day since they went public, and plotted the market cap against the p/e ratio every day. Apple’s closing prices are in blue; Microsoft’s are in red.

The first thing to notice is the huge difference in p/e ratios at the maximum market cap. The highest AAPL point is a market cap of $627 billion, with a p/e ratio of 15.7. The highest MSFT point, by contrast, is a market cap of $615 billion, with a p/e ratio of 73.

Apple too has traded at whopping great multiples in the past — just look at that long tail of blue dots hugging the x-axis. (In fact, for the sake of clarity I truncated that axis a bit, since it goes on a lot further out than that.) But if you bought AAPL stock at a p/e of 373 you’d look pretty smart right now: the shares have gone up almost 100 times since then. Whereas, if you bought MSFT stock at a p/e of 73, you would just have ended up losing a lot of money.

If you stop looking at individual points and start looking just at the general shapes of the charts, there’s a strong positive correlation on the Microsoft datapoints, and a strong negative correlation on the Apple ones. Generally, with Microsoft, the higher the market cap the higher the p/e; whereas with Apple, the higher the market cap the lower the p/e. (Mathematically, the correlation on the Microsoft data +0.22, while the the correlation on the Apple data is -0.10.)

What this says to me is that the market in Apple shares looks a lot more rational than the market in Microsoft shares. Investors will pay huge multiples for Apple shares when the company looks cheap, but not when the company looks expensive. When Apple breaks the half-trillion barrier, that’s despite the fact that its p/e ratio is low; when Microsoft breaks that barrier, it’s because its p/e ratio is high.

So when Wingfield talks about “investor euphoria” surrounding Apple, bear in mind that the euphoria he’s talking about is very different from the kind of euphoria we saw at the height of the dot-com bubble. Apple is indeed worth a truly enormous amount of money. But investors don’t see the kind of scalability in Apple, circa 2012, that they saw in Microsoft circa 1999. Despite the fact that the global market for smartphones today is vastly greater, and growing much faster, than the global market for Windows software was in 1999.

Update: I’ve put the data here for those who want to play around with it themselves.

Posted in economics | Tagged | 27 Comments

When private-school tuition is tax-deductible

Scott Asen, a former trustee and head of the development committee at Groton, a posh private school, has a revealing op-ed in the NYT. He explains that at such schools, the tuition fees, high as they are, fall well short of covering the annual costs that the schools incur; the difference is made up by donations and, where it exists, by using the school’s endowment. What’s more, that fact is explicitly communicated to parents. “Virtually every private-school parent has heard about ‘the gap’”, he says, and, as a parent, once you know about it, it’s clear what you should do:

To the extent that any family with the wherewithal is paying less than the full cost of the product it is buying through combined tuition payments and donations, that family is effectively being subsidized by other current and past donors. Not only is this ethically unsupportable, but ultimately, it is also financially unworkable.

Asen proposes that all parents who can afford to do so, or who fail to provide financial disclosures proving that they can’t afford to do so, should be “expected” to “fill the gap” with a donation. “Given the strength of the educational product offered by these prestigious schools,” he writes, “I think that for every affluent family scared off by the new policy, there would be another of equivalent means — with an equally desirable child in tow — willing to pay full cost.”

Many parents are following these guidelines already, thanks to pressure from people like Asen. “It’s sort of understood that if your children attend these schools and you can afford it, you will pay all or some of that shortfall,” one father told Jenny Anderson.

The operative word here is “pay”. These payments are just that: money spent to cover the costs of educating Junior. As Asen’s headline puts it, right now private schools are “not expensive enough” — they should cost more. And while the extra money might be couched as a donation for tax purposes, both the school and the parents understand that if you’re merely “filling the gap”, rather than, say, donating a couple of million dollars towards a new gym, then you’re really just covering the annual cost of your kid’s tuition.

Such expenses are not, and should not be, tax-deductible. What Asen is proposing comes very close to tax fraud: he’s clearly saying that if you’re the kid of parents of means, and your parents refuse to fill the gap, then you would no longer be welcome at school. For rich parents, this is to all intents and purposes a tuition hike. But it’s sweetened, a little bit, with the fact that the hike is fully tax-deductible.

Tuition hikes at private schools have never been tax-deductible in the past, and they shouldn’t be tax-deductible now. If Asen wants to suggest that private schools should just raise their rack rate to an amount that covers the cost of tuition, and then use some of the extra revenue to help defray the cost for children of parents outside the 1%, then that I think would be perfectly reasonable.

There’s a curious tension among the kinds of people who send their kids to private school: on the one hand, they do so in large part precisely because these are the schools attended by the kids of the money-is-no-object types. But on the other hand, they like to think that the schools are attended by a diverse group of the best and the brightest, rather than just the richest. And so they support the idea that some carefully hand-picked kids from less flush families should also be able to attend. If Asen’s proposal was a simple tuition hike with some kind of concomitant increase in financial aid, then, it would just represent an institutionalization, within the tuition structure, of diversity principles which are already espoused by most private schools.

But Asen doesn’t go that far. Instead, he takes to its logical conclusion the woeful trend of the transactionalization of philanthropy: the idea that it is entirely normal and expected for every tax-deductible philanthropic donation to be tied to the charity in question providing something of value in return. It starts with the tote bags given out during public-radio pledge drives, and ends with the very institution changing its own name to yours: the Peter G Peterson Institute for International Economics, the Booth School of Business.

In general, where something of real value is given back, only the difference is tax-deductible. If I buy $350 tickets to a charity dinner, and the dinner costs $150, then only $200 can be written off against taxes. And if I get a discounted ticket to the dinner, paying only $100, then none of my payment is tax-deductible. That’s essentially the situation that private-school parents are in right now: they’re buying discounted tickets to an extremely expensive education. If they end up forking over a bit more, but still less than the education costs, then nothing changes, and the extra amount they pay should still come out of post-tax income.

All of which helps to explain why private schools charge less for tuition then they spend on providing it. Within days of their kid being accepted at a private school, parents are informed about the famous gap, and pressured to fill that gap with a donation. In the school’s ideal world, all parents would then fill that gap to the limit of their ability to pay — and rich parents would give even more. The parents, of course, will always be happier paying low tuition and topping it up with a tax-deductible donation than they would be simply paying higher tuition.

But donations have to be voluntary, and no matter how good a school is, there are always worthier charities out there. After all, the US already has universal education: the charitable purpose of the private school is just a marginal increase in the quality of some children’s education — complete with deleterious effects on everybody else. (Just imagine how much better Manhattan’s public schools would be, if all of the island’s super-rich had to send their kids to those public schools.) If I have $10,000 to give to charity, it’s really hard to believe that the first best recipient of that money would be some private school, rather than, say, Doctors Without Borders.

So if schools want parents to donate money, they have to offer something extra in return: they have to offer to educate the parents’ kids. That’s why tuition is set below the cost of education: it forces the parents, in aggregate, to donate the difference, for fear that otherwise the kids’ education might suffer. And it also helps to lay the groundwork for deals like the notorious one between Sandy Weill and the 92nd Street Y: Sandy donated $1 million of Citigroup’s money to the school, and his star analyst managed to get his twin daughters into the pre-school there.

That deal was particularly obnoxious, but it’s entirely commonplace for very rich parents to drop very large hints, along with their private-school applications, that if their kids get in, substantial donations will accompany them. Naturally, the schools are predisposed to accept precisely those kids. Given that those donations appear if and only if the kids attend the school in question, there’s a case to be made that all such donations, even if they’re much larger than the gap, are in effect payment for tuition, and should not be tax-deductible.

I’m no fan of the tax-deductibility of charitable donations in the first place — it’s an enormous tax expenditure which results in a relatively modest amount of extra charitable giving. (Ed Dolan has a good two-part overview of why the deduction is a bad idea.) But let’s put that debate to one side for the time being, and make the reasonable assumption that the deduction is here to stay. Let’s also make the equally reasonable assumption that Asen’s proposal won’t get adopted by any private schools, for the very good reason that it would never pass muster with the IRS.

Then what we’re left with is an admirably clever op-ed by Asen, which sends one message while purporting to say something else entirely. His op-ed is nominally directed at schools, saying that they should change their tuition structure. But the real target is not schools but parents. He’s telling them that private-school cost inflation is enormous, and that they have an ethical obligation to fill the gap, every year, if they can possibly afford to do so. Stripped down, the message is this: “give more, or you won’t like the consequences”.

Development officers sharing Asen’s op-ed with parents will surely be very quick to point out that you can’t make tuition a tax-deductible expense, and that if schools were going to raise their tuition rates, none of that raise would be tax-deductible. As a result, they’ll say, it’s much better for all concerned if the present situation stays in place, and parents top up their tuition fees with tax-deductible donations.

Asen’s op-ed, then, is at heart a fundraising drive aimed at the status quo, rather than a real attempt to change that status quo. As you would expect from someone who spent many years raising funds for a private school. I wonder if the NYT realized that, when they printed it.

Posted in economics | Tagged | 14 Comments

When bloggers go offline

Gone Fishin'.jpeg

Posted in economics | Tagged | 17 Comments

Man U’s weird share price


I have absolutely no idea what on earth might be going on with the MANU stock price. But it’s definitely another datapoint supporting the thesis that IPOs are weird things best avoided. (See also Facebook, of course, and Groupon, and just about any of the other buzzy companies to have gone public of late.)

The chart above — there’s a bigger version here — shows the Man U’s share price since the IPO last week. And the weirdest bit of all is actually the first day, when seemingly nothing happened: the stock never moved more than 2% from its IPO price.

But as Matt Levine noted that afternoon, after the market closed, the underwriters were clearly supporting the stock at the IPO price of $14, and the volumes being transacted at that level were truly enormous: much bigger than the official greenshoe. Either someone was accumulating an absolutely massive chunk of stock in the open market, or else the underwriters sold an insane quantity of shares at the IPO — almost double the final free float of 16.67 million shares. Neither theory was particularly plausible on its face.

Clearly without support, the Man U share price would have dropped below $14. But then something very weird happened when the shares reopened on Monday August 13: after half an hour at $14 per share, they suddenly spiked; at one point that morning they were more than $15 per share. Was Man U stock going up, rather than down? Were the people who bought on the opening day being vindicated? Or was this some weird kind of short squeeze?

Subsequent price action didn’t help clear anything up. On Day 3, Tuesday, someone was clearly providing support at $14.16 per share. And on Day 4, Wednesday, we had a mini-repeat of what happened on Monday: the stock suddenly spiked upwards for no good reason from its former support level.

By the end of the day, however, and on Thursday afternoon, we were back down to the $14 IPO price — and again someone was supporting at that level — until, at 11:10am, they weren’t. At long last, the drop that everybody had expected for Monday morning had finally arrived, and for the time being the stock is floating around somewhere near the $13.40 level.

I could try to construct a narrative which might explain all this, but I’d frankly just be making things up. The one thing that I can be pretty sure of is that we’re not seeing price discovery here: this isn’t the natural forces of supply and demand determining how much Man U shares are worth. What it is, however, I have no idea at all.

Update: This might explain a lot.

Posted in economics | Tagged | 5 Comments

$5,000-an-hour lawyers

David Lat has the numbers today on how much the top partners at Dewey & Leboeuf were making. At the top of the list are Berge Setrakian and Ralph Ferrara, both of whom made around $12.5 million in 2011.

As far as I know, the top recorded rate that any lawyer bills out at is Ted Olson’s $1,800. And obviously the amount of money that law-firm partners make is not the same as the amount they bill out at: they’re not just workers but they’re also part owners of their law firms, and share in the whole firm’s profits. But it’s still an interesting exercise to take annual income and divide it by billable hours to see what top law firm partners can make per billable hour.

And here’s the glorious thing: even if you assume that Setrakian and Ferrara bill 2,500 hours per year, that $12.5 million works out at an eye-popping $5,000 per billable hour.

Of course, at these levels, you’re not (just) being paid for the direct work you do for clients: what you’re really being paid for is bringing new clients into the firm and getting the firm revenue streams which can reach hundreds of millions of dollars. But there’s still a reason why those clients will follow you to the firm, and that reason is that the clients will expect you to do real work for them.

Which helps to clarify exactly where the value lies, in law firms. Junior associates get paid less money than they bill out at; partners get paid more money than they bill out at. At the same time, it’s very unlikely that the clients really think they’re getting $400 an hour in value from a relatively junior lawyer poring over boilerplate at 2am.

In other words, differences in billable rates are basically an accounting fiction, which is used to come up with a calculable final figure to be presented as the bill, but which do not actually reflect the difference in value between various strata of lawyers. In order to do that, you’d be better off dividing annual income by, say, 2,500.

If you’re earning $250,000 a year, then that means you’re earning $100 per billable hour, if you work really hard. And if you earn $12.5 million, that works out at $5,000 per hour.

All of which is to say that Ted Olson’s $1,800 is low. And I suspect that if you multiply the number of hours he bills per year by $1,800, you’ll end up at a fraction of the amount he’s actually paid.

Posted in economics | 8 Comments

More data on 401(k) loans

In my post about Bob Litan and his estimates of 401(k) loan defaults, one of the key bits of weirdness was the way in which he decided that the total number of 401(k) loans outstanding had doubled since 2009. The official Private Pension Plan Bulletin says that 401(k) loans totaled $51.7 million in 2009, but Litan, in his paper on the subject, puts total loans at $104 billion, saying that the 2009 numbers are “outdated”.

Now, Guan Yang points me to some new data from the Labor Department, disclosed in response to a FOIA request. The data is raw, of course, and hard to parse; the loans appear in both Schedule H and Schedule I, for instance. But my colleague Jessica Toonkel asked the chaps at Brightscope to take a look, and this is what they found:

Year Schedule H loans, $ Schedule I loans, $ Total loans, $
2006 40,920,193,973 1,845,546,770 42,765,740,743
2007 43,333,714,360 1,815,617,332 45,149,331,692
2008 48,736,802,777 1,369,538,610 50,106,341,387
2009 48,332,673,154 1,552,443,970 49,885,117,124
2010 52,654,869,898 1,428,015,075 54,082,884,973
2011 709,094 22,082,885 22,791,979

This is raw data from Form 5500, the form that pension plans have to file with the government. Most 5500s for 2011 haven’t been filed yet, so the numbers for 2011 are only extremely partial and are pretty much useless. But what we can see is that there was no big rise in total loans between 2009 and 2010, as Litan and his co-author, Hal Singer, imply. If total loans did go up, they only went up by a single-digit percentage: there’s no way they more than doubled.

To double-check with the data that we do have for 2011, the Brightscope people also looked at total 401(k) loans as a percentage of total defined-contribution assets. That was 1.76% in 2008, 1.73% in 2009, and 1.75% in 2010: pretty constant. In 2011, as I say, we only have very partial data so far. But from the data we do have, covering just under $2 trillion in assets, that ratio is just 1.17%: it’s low, not high. As more data come in, that ratio will surely rise. But again, there’s no reason at all to believe that the number of 401(k) loans rose sharply in 2011.

In any case, these numbers alone should be enough to persuade anybody that the Litan-Singer estimate of as much as $37 billion in annual 401(k) loan defaults is just silly: $37 billion is more than 60% of the total number of 401(k) loans outstanding in 2010. And most people, of course, do pay back those loans.

Posted in economics | 4 Comments

The secret to success in the arts

Nassim Taleb has a short paper out entitled “Why It is No Longer a Good Idea to Be in The Investment Industry”. The basic idea is simple, and well understood:

As a population of operators in a profession marked by a high degrees of randomness increases, the number of stellar results, and stellar for completely random reasons, gets larger.

Taleb’s point in this paper is not that if you have a large enough number of operators, then by sheer statistics some of them are going to get lucky. Rather, his point is that as the number of operators rises, it becomes more and more likely that any given outperformer was simply lucky. If you live in a world of fat tails rather than thin tails, and if you have, say, 1 million operators, then the lucky few get very lucky indeed, even if they don’t have any skill.

In numerical terms: in a thin-tailed (Gaussian) world with 1 million operators, the lucky few will be about 5 standard deviations away from the mean. In a fat-tailed world, by contrast, the lucky few will be somewhere between 70 standard deviations and 170 standard deviations away from the mean. Since you can’t get to that kind of place by skill, the result is a world where all the most successful operators are simply lucky.

Taleb concludes with advice: “if you are starting a career, move away from investment management and performance related lotteries as you will be competing with a swelling future spurious tail”. But I’d take issue with this on two fronts.

For one thing, the spurious tail is not swelling. As I examined back in June, the hedge-fund world, just like the mutual-fund world, is shrinking: maybe not in terms of assets under management, but certainly in terms of the number of active fund managers. And as the number of operators shrinks, the spurious tail shrinks as well.

But on top of that, there’s really limited downside to becoming an investment manager. If you get lucky, you can make a fortune; while even if you do badly you still get paid extremely well. Indeed, if you take Taleb to heart, you need to do very little work at all, since your fortunes will be largely down to luck in any case. Lots of money for little work and a small possibility of massive riches? Is not such a bad deal.

The professions you really want to avoid, after reading Taleb’s paper, are not financial but rather creative. Where do you find millions of people all trying to succeed against the odds? Just look at how many bands there are, how many aspiring novelists, how many struggling artists. Nearly all of them think that if they create something great, that will improve their chances of success in their field. But given the sheer number of people they’re competing against, and given the fact that the number of breakout stars in each field is shrinking rather than growing, the fact is that just about everybody with massive success will have got there by sheer luck.

Sometimes, the luck is obvious: EL James, by all accounts, is an absolutely dreadful writer, but has still somehow managed to become a multimillionaire best-selling author. Carly Rae Jepsen has a catchy pop tune, but is only really successful because she happened to be in the right place at the right time. Dan Colen might be a fantastic self-publicist, but not particularly more so than many other, much less successful artists. And so on.

The fact is that pretty much every successful novelist, every successful pop star, every successful artist is successful mainly because of luck. Oftentimes there’s skill involved too, but if you look hard enough, you’ll find just as much skill in the millions of unsuccessful strivers as you do in the tiny set of people who make it huge. And when you’re trying to make it as a novelist, your downside is the kind of penury that no money manager ever needs to worry about.

So if you’re entering the arts world, broadly speaking, make sure that creating something wonderful is its own reward, because you’re almost certainly not going to get much in the way of recognition from anybody else. And if you do get recognition, you’ll only get it because you’re lucky, rather than through any particular skill.

Posted in economics | Tagged | 27 Comments

If the Reuters blogsite stays down…

Posted in Not economics | 15 Comments

Annals of dubious research, 401(k) loan-default edition

Bob Litan, formerly of the Kauffman Foundation and the Brookings Institution, has recently taken up a new job as director of research for Bloomberg Government, where he’s going to have to be transparent and impartial. But one of his last gigs before moving to Bloomberg — a paper on the subject of people borrowing money from their 401(k) accounts — was neither of those things.

To understand what’s going on here, first check out Jessica Toonkel’s article from Friday about Tod Ruble and his company, Custodia.

Tod Ruble is trying to sell retirement plan insurance that employers say they do not want and their employees may not need.

But the Dallas-based veteran commercial real estate investor is not letting that stop him. Since late 2010, he has started up a company, Custodia Financial, and spent more than $1 million pushing for legislation that would allow companies to automatically enroll employees who borrow from their 401(k) plans in insurance that could cost hundreds of dollars a year.

Once you’ve read that, go back and check out a spate of stories that hit a series of major news outlets in July. Alan Farnham of ABC News, for instance, ran a story under the headline “401(k) Loan Defaults Skyrocket”:

A new study estimates that such defaults might total $37 billion a year, a sharp increase from 2007, when defaults totaled only $665 million.

Similarly, check out Walter Hamilton, in the Chicago Tribune (and LA Times): the headline there is “Defaults on 401(k) loans reach $37 billion a year”. At Time, Dan Kadlec also ran with the $37 billion number, saying that “the default rate on these loans has skyrocketed since the recession”. Similar stories came from Blake Ellis at CNN Money (“Loan defaults drain $37 billion from 401(k)s each year”), Mitch Tuchman at MarketRiders (“401k Loan Default Time Bomb Is Ticking”), and many others.

The only hint of skepticism came from Barbara Whelehan at BankRate. She noted that the study cited Kevin Smart, CFO of Custodia Financial, as a source — and she also noted that “it would be a boon for the insurance industry to get the rules changed, and it is working behind the scenes to do just that. In April, Custodia Financial submitted a statement to the House & Ways Committee arguing for automatic enrollment into insurance coverage for 401(k) loans.”

Whelehan also smelled something fishy in the way the paper was paid for:

This paper by Navigant Economics, which made a big splash in the press, was financially supported by Americans for Retirement Protection. That organization has a website,, but no “about us” link. It does give you the opportunity to sign a petition demanding protection of retirement funds through insurance. Take a look at it, and see if you think the website was created by average Americans or by the insurance industry.

Whelehan was actually breaking news here: there’s no public linkage between Americans for Retirement Protection, the organization which paid for the paper, and the astroturf website. In fact, Americans for Retirement Protection seems to have no public existence at all, beyond a footnote in the paper, which was co-authored by Bob Litan and Hal Singer.

Enter Toonkel, writing her story about Custodia. In the course of her reporting, she discovered — and Custodia confirmed — that Americans for Retirement Protection, and, are basically alter egos of Custodia itself. Custodia would welcome other organizations joining in, but that’s unlikely to happen, because Custodia owns the patents on the big idea that the paper and the website are pushing — the idea that 401(k) loans should come bundled with opt-out insurance policies.

Once you’re armed with this information, it’s impossible not to look at the Litan-Singer paper in a very different way. Its abstract concludes: “We demonstrate that the social benefits of steering (but not compelling) plan participants towards insurance when they borrow are likely positive and economically significant.” And yet nowhere in the paper is there any indication that it was bought and paid for by the very company which has a patent on doing exactly that.

And what about that $37 billion number? Are defaults on 401(k) loans really as big a problem as the paper says that they are? After all, the smaller the problem, the less important it is to introduce an expensive fix for it.

The simple answer is no: 401(k) loan defaults are not $37 billion per year. But the fact is that nobody knows for sure exactly where they are, which makes it much easier to come up with exaggerated estimates. As the paper itself admits, “the sum total of 401(k) defaults ought to be an easily accessible statistic, but it is not”. And the $37 billion, far from being a good-faith estimate, in fact looks very much like an attempt to get the largest and scariest number possible.

So how did Litan and Singer arrive at their $37 billion figure? Let’s start with the only concrete numbers we have — the ones from the Department of Labor, whose most recent Private Pension Plan Bulletin gives a wealth of information about all private pension plans in the country. Every pension plan has to file something called a Form 5500, and the bulletin aggregates all the numbers from all the 5500s which are filed; the most recent bulletin gives data from 2009.

This bulletin has two datapoints which are germane to this discussion. First of all, there’s Table A3, on page 7 of the bulletin (page 11 of the PDF). That shows that loans from defined-contribution pension plans to their own participants totaled $51.7 billion in 2009. Secondly, there’s Table C9, the aggregated income statement for the year. If you look at page 32 of the bulletin (page 35 of the PDF), you’ll see a line item called “deemed distribution of participant loans”, which came to $670 million for the year. If you borrow money from your 401(k) and you don’t pay it back, then that money is deemed to have been distributed to you, and counts as a default. So we know that the official size of 401(k) defaults in 2009 was $670 million — a far cry from Litan and Singer’s $37 billion.

Now the $670 million figure does not account for all 401(k) defaults. Most importantly, in some situations, if you default on a 401(k) loan after having been fired from your job, then the money is counted as an “actual distribution” rather than as a “deemed distribution”.

The Litan-Singer paper goes into some detail about this. “According to a recent study by Smart (2012),” they write, “although Form 5500 reflects actual distributions, there is no way to determine the amount of actual defaults.” They then look in detail at Smart’s figures, footnoting him five consecutive times, and treating him as an undisputed authority on such matters. Their citation is merely “Kevin Smart, The Hidden Problem of Defined Contribution Loan Defaults, May 2012.”

Where might someone find this paper? Here, since you ask: it’s helpfully hosted at And on the front page of the paper, Kevin Smart is identified as the “Chief Financial Officer, Custodia Financial”.

There’s no indication whatsoever in the Litan-Singer paper that the “Smart” they cite so often is the CFO of Custodia Financial, the company which has the most to gain should their recommendation be accepted. And there’s certainly no indication that he’s essentially their employer: that Custodia paid them to write this paper. In fact, the name Custodia appears nowhere in the Litan-Singer paper at all.

It’s instructive to look at the Smart paper’s attempt to estimate the magnitude of the 401(k) default problem. I’ll simplify a little here, but to a first approximation, Smart assumes that 12% of people with 401(k) loans lose their jobs. He also assumes that if you lose your job when you have a 401(k) loan, there’s an 80% chance you’ll default on that loan. As a result, he comes up with a 9.6% default rate on 401(k) loans. He then multiplies that 9.6% default rate by total 401(k) loans of $51.7 billion, adds in some extra defaults due to death and disability, and comes up with a grand total of $6.2 billion in loan defaults per year, excluding the “deemed distributions” of $670 million. Call it $7 billion in total, of which $6 billion could be protected by insuring loans against unemployment, death, and disability.

Now remember that this is a paper written by the CFO of Custodia Financial — someone who clearly has a dog in this race. It’s in Smart’s interest to make the loan-default total look as big as possible, since the bigger the problem, the more likely it is that Congress will agree to implement Custodia’s preferred solution.

But when Litan and Singer looked at Smart’s paper, they weren’t happy going with his $6 billion figure. Indeed, as we’ve seen, their paper comes up with a number six times larger. So if even the CFO of Custodia only managed to estimate defaults at $6 billion, how on earth do Litan and Singer get to $37 billion?

It’s not easy. First, they double the total amount of 401(k) loans outstanding, from $52 billion to $104 billion. Then, they massively hike the default rate on those loans, from 9.6% to 17.9%. And finally, they add in another $12 billion or so to account for the taxes and penalties that borrowers have to pay when they default on their loans.

It’s possible to quibble with each of those changes — and I’ll do just that, in a minute. But it’s impossible to see Litan and Singer compounding all of them, in this manner, without coming to the conclusion that they were systematically trying to come up with the biggest and scariest number they could possibly find. It’s true that whenever they mention their $37 billion figure, it’s generally qualified with a “could be as high as” or similar. But they knew what they were doing, and they did it very well.

When the Chicago Tribune and the LA Times say in their headlines that 401(k) loan defaults have reached $37 billion a year, they’re printing exactly what Custodia and Litan and Singer wanted them to print. The Litan-Singer paper doesn’t exactly say that defaults have reached that figure. But if you put out a press release saying that “the leakage of funds in 401(k) plans due to involuntary loan defaults may be as high as $37 billion per year”, and you don’t explain anywhere in the press release that “leakage of funds” means something significantly different to — and significantly higher than — the total amount defaulted on, then it’s entirely predictable that journalists will misunderstand what you’re saying and apply the $37 billion number to total defaults, even though they shouldn’t.

But let’s backtrack a bit here. First of all, how on earth did Litan and Singer decide that the total amount of 401(k) loans outstanding was $104 billion, when the Department of Labor’s own statistics show the total to be just half that figure? Here’s how.

First, they decide that they need the total number of active participants in defined-contribution pension plans. They could get that number — 72 million — from the Labor Department bulletin: it’s right there in the very first table, A1. But the bulletin isn’t helpful to them, as we’ve seen, so instead they find the same number in a different document from the same source.

That’s as much Labor Department data as Litan and Singer want to use. Next, they go to the Investment Company Institute, which has its own survey, covering some 23 million of those 72 million 401(k) participants. According to that survey, in 2011, 18.5% of active participants had taken out a loan; Litan and Singer extrapolate that figure across the 401(k) universe as a whole.

Finally, Litan and Singer move on to Leakage of Participants’ DC Assets: How Loans, Withdrawals, and Cashouts Are Eroding Retirement Income, a 2011 report from Aon Hewitt which is based on less than 2 million accounts, of the 72 million total. According to the Aon Hewitt report, which doesn’t go into any detail about methodology, when participants took out loans, “the average balance of the outstanding amount was $7,860″. Needless to say, that number was never designed to be multiplied by 72 million, as Litan and Singer do, to generate an estimate for the total number of loans outstanding.

If you want an indication of just how unreliable and unrepresentative the $7,860 number is, you just need to stay on the very same page of the Aon Hewitt report, which says that 27.6% of participants have a loan. If Litan and Singer think that the $7,860 figure is reliable, why not use the 27.6% number as well? If they did that, then the total number of 401(k) loans outstanding would be $7,860 per loan, times 72 million participants, times 27.6% of participants with a loan. Which comes to $156 billion.

But of course we know that there were just $51.7 billion of loans outstanding in 2009; evidently Litan and Singer reckoned that it just wouldn’t pass the smell test if they tried to get away with saying that number might have trebled in a single year. So they confined themselves to merely doubling the number, instead.

Litan and Singer give no reason to mistrust the official $51.7 billion number, except to say that it’s “outdated”. But if it’s outdated, it’s only outdated by one year: it’s based on 2009 data, while the much narrower surveys that Litan and Singer cite are generally based on 2010 data. At one point, they cite the ICI survey to declare that there is “an estimated $4.5 trillion in defined contribution plans”, despite the fact that the much more reliable Labor Department report shows that there was just $3.3 trillion in those plans as of 2009. This, I think, quite neatly puts the lie to the Litan-Singer implication that the problem with the Labor Department numbers is merely that they are out of date, and that when we get numbers for 2010 or 2011, they might well turn out to be in line with the Litan-Singer estimates. There’s simply no way that total DC assets rose from $3.3 trillion to $4.5 trillion in the space of a year or two.

In other words, whatever advantage the ICI and Aon Hewitt surveys have in terms of timeliness, they more than lose in terms of simply being based on a vastly smaller sample base. Litan and Singer adduce no reason whatsoever to believe that the ICI and Aon-Hewitt surveys are in any way representative or particularly accurate, despite the fact that the discrepancies between their figures and the Labor Department figures are prima facie evidence that they’re not representative or particularly accurate. If the ICI and Aon Hewitt surveys were all we had to go on, then I could understand Litan-Singer’s decision to use them. But given that the Labor Department already has the number they’re looking for, it just doesn’t make any sense that they would laboriously try to recreate it using less-reliable figures.

It’s true that the Labor Department’s figures do undercount in one respect: they cover only plans with 100 or more participants — and therefore cover “only” 61 million of the 72 million active participants in DC plans. If Litan and Singer had taken the Labor Department’s numbers and multiplied them by 72/61, or 1.18, that I could understand. But disappearing into a rabbit-warren of private-sector surveys of dubious accuracy, and emerging up with a number which is double the size of the official one? That’s hard to justify. So hard to justify, indeed, that Litan and Singer don’t even attempt to do so.

That, indeed, is the strongest indication that the Litan-Singer paper can’t really be taken seriously. For all their concave borrower utility functions and other such economic legerdemain, they simply assert, rather than argue, that they “believe” it is “more appropriate” to use private-sector surveys rather than hard public-sector data. Such decisions cannot be based on blind faith: there have to be reasons for them. And Litan-Singer never explain what those reasons might be.

Now the move from public-sector to private-sector data merely doubles the total size of the purported problem, while Litan-Singer are much more ambitious than that. So their next move is to bump up the default rate on loans substantially.

There’s no official data on default rates at all, so Litan and Singer, following Smart’s lead, decide to base their sums on a Wharton paper from 2010. Once again, they have to extrapolate from a very small sample: the Wharton researchers had at their disposal a dataset covering 1.5 million plan participants (just 2% of the total). Looking at what happened over a period of three years, from July 2005 to June 2008, the researchers found that the number of terminations, and the number of defaults, remained pretty steady:

These are the numbers that Smart used in his paper: roughly 12% of loan holders being terminated each year, and roughly 80% of those defaulting on their loans.

But these are not the numbers that Litan-Singer use. Instead, they notice that the overall default rate, as a percentage of overall loans outstanding, was roughly double the national unemployment rate at the time. And so since the unemployment rate doubled after June 2008, they conclude that the default rate on outstanding 401(k) loans probably doubled as well.

Do they have any evidence that the default rate on 401(k) loans might have doubled after 2008? No. Well, they have a tiny bit of evidence: they look at the small variations in default rates in each of the three years covered in the Wharton study, and see that those variations move roughly in line with the national unemployment rate. Never mind that the default rate fell, from 9.9% to 9.7%, between 2006 and 2008, even as the unemployment rate rose, from 4.8% to 5.0%. They’ve still somehow managed to convince themselves that it’s reasonable to assume that the default rate today is nowhere near the 9.6% seen in the Wharton survey, and in fact is probably closer to — get this — 17.9%.

This doesn’t pass the smell test. The primary determinant of the default rate, in the Wharton study, was the percentage of loan holders who wound up having their employment terminated, for whatever reason. And so what Litan-Singer should be looking at is the increase in the probability that any given employee will end up being terminated in any given year.

Remember that in any given month, or year, the number of people fired is roughly equally to the number of people hired. When the former is a bit larger than the latter for an extended period, then the unemployment rate tends to go up; when it’s smaller, the rate goes down. But the churning in the employment economy is a constant, even when the unemployment rate is very low.

When the unemployment rate rose after 2008, that was a function of the fact that the number of people being fired was a bit higher than normal, while the number of people being hired was a bit lower than normal. But looked at from a distance, neither of them changed that much. In terms of the Wharton study, what we saw happening to the unemployment rate is entirely consistent with the percentage of loan-holders being terminated, per year, staying pretty close to 12%. Of course it’s possible that number rose sharply, but it’s really not possible that number rose as sharply as the unemployment rate did. And so I find it literally incredible that Litan and Singer should decide to use the national unemployment rate as a proxy for the number of people whose employment is terminated each year.

Well, maybe not literally incredible — the fact is there’s one very good reason why they might do that. Which is that they were being paid by Custodia to use any means possible to exaggerate the number of annual 401(k) loan defaults.

Litan and Singer do actually provide a mini smell test of their own: they say that their hypothesized rise in 401(k) loan defaults is more or less in line with the rise in, say, student-loan defaults or in mortgage defaults over the same period. But those statistics aren’t comparable at all, because Litan and Singer are already assuming that the default rate on 401(k) loans, among people who lose their job, was a whopping 80% before the financial crisis. There’s a 100% upper bound here: you can’t have a default rate of more than 100%. Remember that the whole point of this paper is to provide the case that people taking out 401(k) loans should insure themselves against unemployment: any rise in the default rate from people who don’t lose their job (or die, or become disabled) is more or less irrelevant here. And when your starting point is a default rate of 80%, there really is a limit to how much that default rate can rise; it’s certainly going to be difficult to see it rise by more than 85%, even if you allow a simultaneous increase in the number of people being terminated.

All of this massive exaggeration has an impressive effect: if you take $104 billion in loans, and apply a 17.9% default rate, then that comes to a whopping $18.6 billion in 401(k) loan defaults every year. A big number — but still, evidently, not big enough for Litan and Singer. After all, their number is $37 billion: double what we’ve managed to come up with so far. We’ve already doubled the size of the loan base, and almost-doubled the size of the default rate, so how on earth are we going to manage to double the total again?

The answer is that LItan and Singer, at this point, stop measuring defaults altogether, and turn their attention to a much more vaguely-defined term called “leakage”. Once again, they decide to outsource all their methodology to Custodia’s CFO, Kevin Smart. The upshot is that if you borrowed $1,000 from your 401(k) and then defaulted on that loan, the amount of “leakage” from your 401(k) is deemed to be much greater than $1,000. Litan and Singer first add on the 10% early-withdrawal penalty that you get charged for taking money out of your plan before you retire. They also add on the income tax you have to pay on that $1,000, at a total rate of 30%. (They reckon you’ll pay 25% in federal taxes, and another 5% in state taxes.) So now your $1,000 default has become a $1,400 default.

How does that extra $400 count as leakage from the 401(k), rather than just something that gets added to your annual tax bill? Smart explains:

Most participants borrow from their retirement savings because they are illiquid and do not have access to other sources of credit. This clearly demonstrates that participants who default on a participant loan do not have the financial means to pay the taxes and penalty. Unfortunately, their only source of capital is their retirement savings plan so many take the remaining account balance as an additional early distribution to pay the taxes and penalty, further increasing the amount of taxes and penalties due. These taxes and penalties become an additional source of leakage from retirement assets.

Smart’s 16-page paper has no fewer than 24 footnotes, but he fails to provide any source at all for his assertion that “many” people raid their 401(k) plans in order to pay the taxes on the money they’ve already borrowed. In any event, Smart (as well as Litan and Singer, following his lead) makes the utterly unjustifiable assumption that not only many but all 401(k) defaulters end up withdrawing the totality of their penalties and extra taxes from their retirement plan. And then, just for good measure, because that withdrawal also comes with a penalty and taxes, they apply a “gross-up” to that.

By the time all’s said and done, the $1,000 that you lent yourself from your 401(k) plan, and failed to pay back in a timely manner, has become $1,520 in “leakage”. Add in some extra “leakage” for people who default due to death or disability (apparently even dead people raid their 401(k) plans to pay income tax on the money they withdrew), and somehow Litan and Singer contrive to come up with a total of $37 billion.

It’s an unjustifiable piling of the impossible onto the improbable, and the press just lapped it up — not least because it came with the imprimatur of Litan, a genuinely respected economist and researcher. Custodia hired him for precisely that reason: they knew that if his name was on the front page of a report, that would give it automatic credibility. But for exactly the same reason, Litan had a responsibility to be intellectually honest when writing this thing.

Instead, he never even questioned any of the assumptions made by Custodia’s CFO. For instance: if you’re terminated, and you default on your 401(k) loan, what are the chances that the money you received will end up being counted as an “actual distribution” rather than as a “deemed distribution”? Smart and Litan and Singer all implicitly assume that the answer is 100%, but they never spell out their reasoning; my gut feeling is that it’s not nearly as clear-cut as that, and that it all depends on things like when you lost your job, when you defaulted, and who your pension-plan administrator is.

Custodia’s business, and the Litan-Singer paper, are based on the idea that if people who borrowed money from their 401(k) plans had insurance against being terminated from their jobs, then that would have significant societal benefit. In order for the societal benefit to be large, the quantity of annual 401(k) loan defaults due to termination also has to be large. But right now, there’s not a huge amount of evidence that it actually is: in fact, we really have no idea how big it is.

I can say, however, that Custodia has already won this battle where it matters — in the press. “Protecting 401(k) savings from job loss makes a lot of sense,” said Time’s Kadlec in his post — and so long as Custodia can present lawmakers with lots of headlines touting the $37 billion number and supporting their plan, Litan and Singer will have done their job. The truth doesn’t matter: all that matters is the headlines, and the public perception of what the truth is.

Come to think, maybe this makes Litan the absolutely perfect person to run the research department at Bloomberg Government. On the theory that it takes a thief to catch a thief, Bloomberg has hired someone who clearly knows all the tricks when it comes to writing papers which come to a predetermined conclusion. And he also has a deep understanding of the real purpose of most of the white papers floating around DC: it’s not to get closer to the truth, but rather to stamp a superficially plausible institutional imprimatur onto a policy that some lobbyist or pressure group desperately wants enacted. I can only hope that in the wake of using his talents in order to serve Custodia Financial, Litan will now turn around and use them in order to serve rather greater masters. Like, for instance, truth, and transparency, and intellectual honesty.

Posted in economics, statistics | Tagged , , | 3 Comments