WSJ’s Student Loan Coverage Improves: More Facts, Fewer Deadbeats

And not just facts, neutral facts, which is how reporting is supposed to be. I’ve criticized The Wall Street Journal‘s student loan coverage, but its most recent article on the topic, “U.S. to Forgive at Least $108 Billion in Student Debt in Coming Years,” is a start in the right direction.

Okay, the title could use some work. More accurately, it should be something like: “GAO Projects U.S. Will Forgive $108 Billion in Student Loans in Coming Years.” It’s 76 characters, which is too long for most SEO-obsessed editors, but it doesn’t characterize a possibility as a certainty.

Conversely, the WSJ neglects to cite another GAO study on the subject of student debtors’ earnings. Its data are nearly two years old, but they show that 72 percent of people on income-sensitive repayment plans were earning $20,000 annually or less. Not even 10 percent of IBR and PAYE participants (157,000) made more than $40,000 per year.

Thus, the WSJ’s reasoning still follows a shaky line of reasoning:

(1) IBR participants’ debts are high,

(2) High debts are only feasible for grad students taking out Grad PLUS loans,

(3) Graduates tend to find jobs with high incomes and have low unemployment rates,

(4) So the benefits of IBR go to high-income people.

The prior GAO study pokes holes in (3) and (4). Income is the independent variable, not debt, and incomes are low. Still, the WSJ’s reporting this time inserts enough adverbs to qualify these claims that I’m going to give this an earned “C.” There is no grade inflation on this blog.

Oddly, in its haste to cover the GAO’s attacks on the government’s accounting for student loans, the WSJ neglects to include immanent compensating factors that will raise student debtors’ incomes: tax cuts, stimulus, job growth, a harried Fed, and 3-4 percent growth in the near future. Things will rapidly get better for America’s student debtors.

Gini Coefficients: How to Calculate Them in Excel and What The F They Mean

Social scientists and wonk readers are familiar with Gini coefficients, and while I’ve used them a few times myself, I have no idea what they really mean.

Okay, fine, I know what a Lorenz curve is, and I can explain it mathematically. But ultimately any understanding of the Gini coefficient is inherently relative, e.g. “They say .3 is good for income distribution, but anything above .5 is bad. The U.S. is .45, so that’s bad … ish.” It doesn’t say anything about the actual distribution.

Indeed, the first two Wonkblog posts I found in a brief Internet search describe the Gini coefficient essentially as a (well-known) measure of inequality. No details.

Vox is slightly better: “[I]t’s an incredibly abstract idea that’s difficult to verbally describe. [Thanks, Vox.] The advantage to using a gini coefficent is that in principle it summarizes all the information about the distribution of income and thus facilitates easy comparisons.”

Easy comparisons? To what? Other Gini coefficients? Whatever.

Still, aside from eyeballing a bunch of Lorenz curves—fun fact: I’m going to make you eyeball a bunch of Lorenz curves today—a Gini coefficient is totally undescriptive because it’s a decimal without any units.

This post tries to remedy that, but first, per the title here’s a big ol’ array formula for calculating Gini coefficients in MS Excel. The source is Excel and UDF Performance Stuff. I chose the Angus Deaton version and improved on it by replacing the ROW formulae with COUNTIFs because those account for blank cells in the data. It also avoids the pitfall of negative data numbers, which mess up Lorenz curves. However, as an array formula, it must be entered with CTRL-SHIFT-ENTER otherwise it won’t work. Behold:


*Where “Datarange” is something like, “$B$1:$B$1000”.

The Web site contains a few other methods, but this one is the most comprehensive and isn’t limited to 4,000 data points. I’ve tested this version with hand-cranked Ginis, that is, sorting the data by size, creating a cumulative sum of them (the Lorenz curve), calculating a bunch of trapezoids among them to find the area under the Lorenz curve, adding those up, subtracting them by equivalent triangle of equality, and then dividing them by that triangle.

So what does the Gini coefficient mean? In other words, what kind of distribution can one imagine when provided with a given Gini number?

It’s a damn good question that I figured out an answer to. I created a bunch of columns in Excel with a bunch of RAND() formulae, gathered some statistics, and then averaged the results. The RAND() formula produces a random number from 0 to 1. The more of them you use, the bigger, more robust the distribution you get. A .1 is as likely as a .9, etc. The mean and median will be .5. But what’s the Gini coefficient of a large number of random numbers within a set range?

Answer: One third.

This is huge for me because now I know when someone says that the Gini coefficient for the income distribution in some country is .333333, I can visualize that it’s equivalent to a random income distribution. Because one third is such a round number, I had a hunch that taking each RAND() formula to different powers would yield other round Gini coefficients, so, here’s a table of Gini coefficients as a power of each data point from a random distribution. (And yes, this blog endorses the divisible properties of the number 12.) I’m adding similar information for the resulting Lorenz curves. It’s quite intriguing—and quite tedious because it takes a few minutes for my computer to spit out the numbers. But this blog wouldn’t be anything if it didn’t hog some clock cycles:

1/12 .040 48.9% 82.5% 89.1% 94.4% 97.6% 99.1% 100.0% 92.3%
2/12 .077 24.2% 68.0% 79.4% 89.1% 95.3% 98.3% 100.0% 85.7%
3/12 .111 12.1% 56.1% 70.7% 84.1% 93.1% 97.4% 100.0% 80.0%
4/12 .143 06.1% 46.3% 63.0% 79.4% 90.9% 96.6% 100.0% 75.0%
5/12 .173 03.1% 38.2% 56.1% 74.9% 88.8% 95.7% 100.0% 70.6%
6/12 .200 01.6% 31.5% 50.0% 70.7% 86.7% 94.9% 100.0% 66.7%
7/12 .226 00.8% 26.0% 44.5% 66.8% 84.6% 94.0% 100.0% 63.2%
8/12 .250 00.4% 21.4% 39.7% 63.0% 82.6% 93.2% 100.0% 60.0%
9/12 .273 00.2% 17.7% 35.3% 59.5% 80.7% 92.4% 100.0% 57.2%
10/12 .294 00.1% 14.6% 31.5% 56.2% 78.8% 91.6% 100.0% 54.6%
11/12 .315 00.1% 12.0% 28.1% 53.0% 76.9% 90.8% 100.0% 52.2%
12/12 .334 00.0% 09.9% 25.0% 50.0% 75.1% 90.0% 100.0% 50.0%
16/12 .400 00.0% 04.6% 15.7% 39.7% 68.3% 86.9% 100.0% 42.9%
20/12 .454 00.0% 02.1% 09.9% 31.5% 62.1% 83.9% 99.9% 37.6%
24/12 .500 00.0% 01.0% 06.3% 25.0% 56.4% 81.0% 99.9% 33.4%
30/12 .555 00.0% 00.3% 03.1% 17.7% 48.9% 76.9% 99.9% 28.6%
36/12 .600 00.0% 00.1% 01.6% 12.5% 42.4% 72.9% 99.9% 25.1%
42/12 .636 00.0% 00.0% 00.8% 08.9% 36.8% 69.2% 99.9% 22.3%
48/12 .666 00.0% 00.0% 00.4% 06.3% 31.9% 65.6% 99.9% 20.1%
54/12 .692 00.0% 00.0% 00.2% 04.4% 27.6% 62.3% 99.9% 18.2%
60/12 .714 00.0% 00.0% 00.1% 03.1% 24.0% 59.1% 99.8% 16.7%
66/12 .733 00.0% 00.0% 00.0% 02.2% 20.8% 56.1% 99.8% 15.4%
72/12 .749 00.0% 00.0% 00.0% 01.6% 18.0% 53.2% 99.8% 14.3%
84/12 .777 00.0% 00.0% 00.0% 00.8% 13.5% 47.9% 99.8% 12.6%
96/12 .799 00.0% 00.0% 00.0% 00.4% 10.2% 43.1% 99.7% 11.2%
120/12 .833 00.0% 00.0% 00.0% 00.1% 05.8% 35.0% 99.7% 09.1%
144/12 .856 00.0% 00.0% 00.0% 00.0% 03.3% 28.3% 99.6% 07.7%
240/12 .909 00.0% 00.0% 00.0% 00.0% 00.3% 12.3% 99.4% 04.8%

And here is the same information for the Lorenz curves.

1/12 .040 00.0% 08.2% 22.3% 47.2% 73.3% 89.2% 100.0% 48.0%
2/12 .077 00.0% 06.8% 19.8% 44.5% 71.5% 88.5% 100.0% 46.2%
3/12 .111 00.0% 05.6% 17.7% 42.0% 69.8% 87.7% 100.0% 44.4%
4/12 .143 00.0% 04.6% 15.7% 39.7% 68.2% 86.9% 100.0% 42.9%
5/12 .173 00.0% 03.8% 14.0% 37.4% 66.6% 86.2% 100.0% 41.4%
6/12 .200 00.0% 03.1% 12.5% 35.3% 65.0% 85.4% 100.0% 40.0%
7/12 .226 00.0% 02.6% 11.1% 33.4% 63.5% 84.7% 100.0% 38.7%
8/12 .250 00.0% 02.1% 09.9% 31.5% 62.0% 83.9% 100.0% 37.5%
9/12 .273 00.0% 01.8% 08.8% 29.7% 60.5% 83.2% 100.0% 36.4%
10/12 .294 00.0% 01.5% 07.8% 28.1% 59.1% 82.5% 100.0% 35.3%
11/12 .315 00.0% 01.2% 07.0% 26.5% 57.7% 81.8% 100.0% 34.3%
12/12 .334 00.0% 01.0% 06.2% 25.0% 56.3% 81.1% 100.0% 33.3%
16/12 .400 00.0% 00.5% 03.9% 19.8% 51.2% 78.3% 100.0% 30.0%
20/12 .454 00.0% 00.2% 02.5% 15.8% 46.5% 75.6% 100.0% 27.3%
24/12 .500 00.0% 00.1% 01.6% 12.5% 42.3% 73.0% 100.0% 25.0%
30/12 .555 00.0% 00.0% 00.8% 08.9% 36.7% 69.3% 100.0% 22.3%
36/12 .600 00.0% 00.0% 00.4% 06.3% 31.8% 65.8% 100.0% 20.0%
42/12 .636 00.0% 00.0% 00.2% 04.4% 27.5% 62.4% 100.0% 18.2%
48/12 .666 00.0% 00.0% 00.1% 03.1% 23.9% 59.3% 100.0% 16.7%
54/12 .692 00.0% 00.0% 00.0% 02.2% 20.7% 56.3% 100.0% 15.4%
60/12 .714 00.0% 00.0% 00.0% 01.6% 18.0% 53.4% 100.0% 14.3%
66/12 .733 00.0% 00.0% 00.0% 01.1% 15.6% 50.7% 100.0% 13.4%
72/12 .749 00.0% 00.0% 00.0% 00.8% 13.5% 48.1% 100.0% 12.6%
84/12 .777 00.0% 00.0% 00.0% 00.4% 10.2% 43.3% 100.0% 11.2%
96/12 .799 00.0% 00.0% 00.0% 00.2% 07.6% 39.1% 100.0% 10.1%
120/12 .833 00.0% 00.0% 00.0% 00.1% 04.3% 31.7% 100.0% 08.4%
144/12 .856 00.0% 00.0% 00.0% 00.0% 02.4% 25.7% 100.0% 07.2%
240/12 .909 00.0% 00.0% 00.0% 00.0% 00.3% 11.2% 100.0% 04.6%

And as promised, some Lorenz curves:


Ooooh, the colors.

And for a fun digression, here’s a Lorenz curve of U.S. earnings.


(Source: Social Security Administration)

The data aren’t presented in a way that can be properly sorted, so I can’t calculate the area under the curve, but at the median, the Gini coefficient is probably about .5. At the 90th percentile, it’s more like .8. I wouldn’t be surprised if U.S. income polarization is worse than .45.

So, now you know how to estimate a Gini coefficient in MS Excel, and if someone throws Gini numbers at you, you can look at the tables to get a sense of what they mean in terms of the maximum data point. It’s still a number without any units, but at least you can see the relationships among the data points more clearly.

Indiana Tech Accused of ‘Bait and Switch’

By students of the soon-to-be-closed law school? NO! By a lawyer representing an aggrieved faculty member, according to Fort Wayne’s News-Sentinel:

[Indiana Tech’s board of director’s] decision “throws into chaos the lives and academic plans of the student body. The law school’s tuition is just under $20,000. You don’t have to be a lawyer to be repulsed by this outrageous bait and switch.”

I predict very few lawyers are repulsed by Indiana Tech’s decision. I’m not the first to opine on it, but Indiana Tech School of Law’s demise benefits humanity. It was never fully enrolled, only one of its twelve graduates passed the bar, and at last the board saw the writing in the blue book. Whether it will herald more law school closures is debatable. I think many of its peers will see Indiana Tech as an Icarus rather than a bellwether.

If I were cynical, I’d suspect Indiana Tech knew it was going to fail and used its provisional accreditation as a sword to rescue its students from the ignominy of starting their legal educations over at a more sound ABA school than a shield against total failure.

Otherwise, I have very little to say on this subject, except to remind everyone of those bygone days five years ago when Indiana Tech School of Law was a glint in its board’s eyes—and its Feasibility Committee was warning that there would be an attorney shortage so unbearable that we’d have to import foreign lawyers. Seriously, it was that dishonest.

Now Indiana Tech’s president, Arthur Snyder, concedes, “Over the course of time it has become apparent that the significant decline in law school applicants nationwide represents a long term shift in the legal education field, not a short-term one.”

Many voices warned Indiana Tech not to open a law school. It ignored them and made a $20 million mistake. But don’t expect it to apologize to its students for its hubris—they’re the ones who really paid.

WSJ Has No Idea Who Benefits From IBR/PAYE/REPAYE/ETC

A hypothetical: Jill and Jack live in the same town. Jill has many healthy habits but is a nurse who spends time around infected people, Jack less so. The town is hit with a case of spectrox toxaemia, a dangerous disease. The government offers to immunize people. Jill decides to be immunized; Jack does not. Jill does not get sick; Jack does. So, epidemiologists, did Jill not contract spectrox toxaemia because she was immunized or because of her healthy habits (or luck)?

If you’re The Wall Street Journal, the answer is her habits. Most of us would believe otherwise, given how dangerous spectrox toxaemia is and Jill’s contact with its victims.

Likewise, this line of reasoning animates the WSJ’s opinion of the government’s income-sensitive repayment programs for student debtors, which it claims benefit higher-debt people with better credit scores than lower-debt people who don’t. It’s unintuitive, if you’re the WSJ apparently, but it makes more sense to those of us familiar with the student debt system.

Here’s how it works: People who take out lots of debt might not in fact have the incomes to repay them, so they choose an income-sensitive repayment because the alternative is … Default! Thus, looking at how much they borrow is less important than looking at how much they’re paid.

Last year, in fact, the Government Accountability Office explored this topic and found that most people in income-sensitive repayment programs were earning less than $20,000 annually. So the Jills aren’t so different from the Jacks after all.

Sure, if there were no IBRs/PAYEs/REPAYEs/ETCs, then these Jills with good borrowing habits would be more likely to take deferments and forbearances, but their debts would still not be repaid. That’s because debts that can’t be repaid will not be repaid, no matter what someone’s credit score or how much they borrowed. What matters is what they earn, and college graduates don’t earn much these days.

And if you think the Jills have too much debt, then the problem isn’t IBR/ICR/REPAYE/ETC, it’s that the government lends too much money to people for degrees they don’t need.

‘Law Deans Are Running Bait and Switch Operation’

…In Australia.

Yes folks, y’all can add Down Under to the list of countries with too many law students chasing too few legal jobs, according to a surprisingly scathing opinion piece in The Australian Financial Review by a law instructor at Macquarie University.

A few quotes:

“Law student numbers are out of hand. Nearly 15,000 finish their degree each year, and enter a market where there are only 66,000 solicitors.” Yikes.

“Law deans are running a bait and switch operation. They hold out the promise of a legal career, while adding to the unemployment queue.”


“[The deans’] claim that the legal problems undertaken in law tutorials are a platform for a generalist degree – that will see students who miss out on a job as a lawyer well placed to enter other high-paid spheres of the economy – is a self-serving myth.”

All this is just more evidence that American legal education does a better job of training law deans to advocate their positions than other countries do. The best they can offer here is the versatile-law-degree argument, but if they want to avoid a government crackdown, they’ll have to lean on increasing diversity in the profession or at least gussy up what they have with some kind of human-capital analysis. Otherwise, these Southern Hemispherians will end up like their Japanese counterparts.

Council of Economic Advisors: College Pays. Grad School? Sh!

Or, “The Reality Behind AEI’s Reality Behind the Student Debt ‘Crisis'”

Speaking of student loans, I am directed to the American Enterprise Institute’s response to the Council of Economic Advisor’s (CEA’s), “Investing in Higher Education: Benefits, Challenges, and the State of Student Debt” (pdf).

Because I try to deliver early on my post titles rather than bury them, here’s the report’s chart on the crucial but under-emphasized dispersion of earnings by educational attainment for 35-44 year-olds with payroll incomes. (This cohort doesn’t seem so representative to me of recent student borrowers—and not in a good way, but that’s a different issue.)

CEA--State of Student Loans--Figure 5

Eyeballing the chart, more than a quarter of graduate-degree holders earn less than the median 4-year-degree holder in the same age bracket, and the bottom 25 percent of grads earn about $45,000 or less. The B.A.s earn between about $20,000 and $130,000 while the grads make roughly between $30,000 and $170,000. Graduates in between the 75th and 90th percentiles haul in nearly half the total difference. This wide dispersion cries for more analysis because graduate borrowing amplifies student debt loads. High debts and low incomes, even for this small group of debtors, tend to discredit the human capital hypothesis and the purpose of student lending.

But back to the reality behind the reality behind the- etc.

Critical readers should always be on their guards whenever someone characterizes the “student debt crisis.” Frequently it’s a strawman of the crisis writers want to discuss rather than how much of the unpayable debt will be written down in the future. In the AEI’s case, the crisis is, “[T]he macroeconomic impact of high debt levels.” Here, AEI takes this to mean the stock of $1.3 trillion of debt.

The AEI post turns to its education scholars, startlingly Jason Delisle, who perhaps has moved on from the New America Foundation. Delisle focuses first on the claim that “student debt is holding back the economy.” The CEA report attempts to discredit this position in six ways. One, student debt is not as big as the mortgage bubble (which I don’t think I’ve seen anyone argue for a few years now). Two, hardship today will be offset by the future productivity unleashed by education. Three, everyone borrowed student loans when the opportunity costs were lowest, so high debt levels are in step with the economy and not undermining it. Four, student debt is only slightly reducing homeownership among young people. Five, student loans only reduce auto debt for high-balance debtors. Six, student-loan debts reduce small-business formation and their incomes somewhat, but other factors are involved.

The study concludes, “Had the same students received an education without as many loans, the recovery would likely have been stronger, but not substantially so. Most individuals, and the economy as a whole, will benefit from the education made possible by student loans” (56).

In other words, the Obama administration is asking everyone to double down on its hope that all this education will pay off someday and the government won’t have to write down hundreds of billions of dollars in unpayable education debt, whether by forgiveness promises in repayment plans or new legislation. It’s a theme that crops up elsewhere in the report, and it suffers from two problems. One, higher education doesn’t correspond to higher aggregate incomes; rather it seems to be swapping high-school grads with college grads while keeping incomes flat. If college boosts incomes like video-game power-ups, then we’d expect exponential growth in aggregate incomes, but we’re not. And anyone who thinks the payoff will come later must explain why intervening variables aren’t involved, e.g. occupational differences, which would explain the wider earnings dispersions for the credentialed. The CEA gives us no confidence in its education bet.

Problem number two is that the report tends to side against studies produced by the Federal Reserve Bank of New York (especially those by Meta Brown, et al.) in favor of research producing more satisfying results. The impacts might be trivial, but the NY Fed found that youngish student debtors weren’t getting mortgages or were more likely to live with their parents than the unindebted (links buried here). Meanwhile, the report shoos away the Bennett hypothesis by claiming a lack of consensus, with the caveat that there may be some “administrative bloat” in colleges and universities. Consensuses are tough rhetorical animals to wrestle with and should require significant evidence to prove. A few studies here and there will not do it.

So back to AEI. When Delisle writes, “[Advocacy groups] say student debt is forcing people to delay things like buying a house, starting a family, all productive things. This report is pretty clear that isn’t the case,” he’s wrong. The report clearly concedes that student debt is negatively affecting the economy, albeit to a small degree, and thanks in part to wishing away contrary NY Fed studies and insisting that all the education will pay off someday.

To clarify, student debt is a notable if not primary contributor to a generational disaster dominated by the trade deficit or slack aggregate demand—and new student borrowing is declining—but the CEA report isn’t the source to show it. So that’s a strike against Delisle.

He asks:

Why are millions of borrowers flocking to enroll in a program [IBR, PAYE, REPAYE, etc.] that allows them to cap their student loan payments at a small share of their income if the return on an educational investment are large? Something seems amiss there. I’ve done a lot of work showing that the income-based repayment program is too generous as a result of Obama administration changes, which may explain this disconnect.

I’ve answered the first question already: There is no large, aggregate return to higher education. As to Delisle’s work on the changes to IBR, it’s never demonstrated that the programs are too generous because it’s based on lopsided, self-verifying hypotheticals. In fact, according to a GAO study, in 2014 only 2 percent of debtors in IBR or PAYE plans earned more than $80,000, so Delisle’s mythical IBR deadbeat is not a serious policy concern. Amusingly, Delisle’s reaction to the GAO study at the time was to blame debtors for not making enough money, gasping that they’d use IBR plans for long-term rather than short-term debt relief.

AEI then turns to resident scholar Andrew Kelly, who writes, “Lower interest rates [proposed by Democrats] won’t help folks with small balances who aren’t repaying nearly as much as they’ll help those with average or large balances, most of whom have no trouble repaying because they have the highest educational attainment!”

I have problems with the Warrenian interest-rate proposals too, but Kelly makes the frequent mistake of flipping the income and debt variables to conclude that high-balance debtors are deadbeats, even though the CEA report shows a wide income dispersion for graduate-degree holders.

I admit I didn’t give the CEA report a thorough read, but it looks like the AEI scholars didn’t either.

No Libertarians, the ABA Does Not Control The Supply of Lawyers

Writing for Forbes, University of Chicago law professor Todd Henderson explains to us “Why Lawyer Salaries Are Skyrocketing.” Although he attributes most of the cause of the big-law salary hike to the libertarian red-tape boogeyman, Henderson opens the article with long-falsified supply-side reasoning.

On the supply side, the American Bar Association operates a state-approved cartel, which uses a licensing regime to artificially limit the supply of legal services. In a recent white paper, the White House came out against occupational licensing in general, and breaking the ABA cartel would be a good first step in addressing the staggering growth in lawyer pay.

The last time I recall encountering the “ABA attorney shortage” claim in any depth was two years ago when Michael Lind on Salon told us that that the ABA controls the supply of lawyers. Henderson’s argument though more predictably libertarian is nevertheless surprising because only a month ago The New York Times explored law-graduate underemployment in depth. The natural question is, how can Henderson discuss an attorney shortage while graduates a state away from him struggle to find work at far less pay?

In recent years bar-passage rates have played a role in graduate underemployment to some extent, but not all of the 5,004 unemployed or unsurveyed class of 2015 graduates failed the bar. Another 5,400 graduates were in JD-advantage jobs, which frequently includes positions that could be filled with people with less education. These graduates should be pushing lawyer pay down, and this is prior to any discussion of whether big law salaries should track inflation.

Then of course, there’s the fact that payroll lawyers’ incomes have been flat for quite a while.

10th to 90th Percentile Dispersion of Annualized OES Lawyer Incomes

From a business perspective, law firms could also take the same amount of money and substitute more new associates for the same (or less) pay to cover demand for their services. That is, if demand for their services is really an issue.

Then of course, there’s the ABA’s accrediting power, which a Department of Education panel threatened with a one-year suspension not because it’s refusing to accredit more law schools but because it’s accrediting law schools with insufficient regard to graduates’ employment outcomes.

Cleary other forces are responsible for the ~$20,000 big-law pay raise. I insist I’m not a biglawologist and other voices such as Steven Harper are vastly more credible than I am on the subject, but anyone who thinks ABA rules are choking lawyer supply doesn’t have much credibility when it comes to regulatory boogeymen either.