Grade Inflation: It Depends How You Define ‘Educational Quality’

On VoxEU, we have Raphael Boleslavsky and Christopher Cotton’s, “The unrecognized benefits of grade inflation.” The authors write:

Our analysis reveals a surprising link between grade inflation and investment in education quality – schools invest more when they are allowed to inflate grades than when grade inflation is banned. …

With grade inflation, student transcripts convey less information, and therefore the employer relies less on transcripts and more on school reputation when evaluating graduates. In this way, grade inflation increases the incentives that schools have to undertake costly investments to improve quality of education, and the average ability of their graduates. To the extent that school investments and a student’s own study efforts are strategic complements in human capital development, students who anticipate greater investments by schools in turn have greater incentives to increase their own efforts. [Emphasis LSTB]

You could replace the emphasized bits about “education quality” with “wasteful spending” or the like and you’d have an accurate description of what goes on at law schools.

For instance, we have expanding faculties:

Law School Faculty Per School (Calendar-Year Average, Index 1999=100, Excl. P.R.)

(Source: Official Guide, author’s calculations)

We have lots of internal grants and scholarships:

Spending on Internal Grants and Scholarships Per Law School (2013 $)

(Source: ABA (pdf), Bureau of Labor Statistics, author’s calculations)

We have no (net) positive impact on job outcomes:

Percent Employed by Status (NALP)

(Source: NALP)

We have a decline in legal sector labor productivity:

Legal Sector Labor Productivity (2005=100)

(Source: Bureau of Labor Statistics)

…And all this is covered with tuition hikes on the poor souls who are paying full tuition (if that):

Median Full-Time Law School Tuition (2013 $)

(Source: ABA (pdf), FinAid.org, Bureau of Labor Statistics, author’s calculations)

None of this is necessarily the result of grade inflation, which the authors’ model takes to be endogenous when I happen to think it’s exogenous (Hell, even the law school deans say so). If anything grade inflation is a symptom of the same pressures the schools are under to signal their degrees’ prestige to employers. But job outcomes is most of what this all comes down to. If there weren’t such a wide dispersion of jobs and salaries, then there’d be less motivation to engage in these kinds of wasteful behaviors. The free student loans are the accelerant.

However, there’s no reason to believe that, in the face of grade inflation, colleges and universities would improve their reputations by carefully investing in better student outcomes; rather, they would invest the bare minimum of what the employers want to see—not what actually makes the graduates more productive. That’s why the employers complain about how law students take frivolous courses but keep hiring from elite law schools nevertheless.

Revealed preferences, people. Revealed preferences.

The Future of the College Premium Debate

A few months back I picked up a copy of Divided: The Perils of Our Growing Inequality, edited by David Cay Johnston, who signed it when I hobnobbed with the (liberal) one percent a few months back. Its essays make for quick reads on planes, trains, and other forms of transit. One contribution by Jared Bernstein titled, “Inequality Across Generations,” struck me. At the end he advocates “‘college for all who are able’ … as an ambitious investment in building human capital assets for the disadvantaged.” (The essay was originally published in 2007.)

Curious, I looked on Bernstein’s blog and found a 2012 post in which he discusses whether college graduates in low-skill jobs still get a wage premium. The answer’s yes, but I’ll scrutinize why he says so in a moment.

But first, readers keeping score at home will recall that we’ve been sternly warned that regression analyses of higher education “premia” that control for occupations are invalid because “occupation is an outcome variable and not a pretreatment covariate.” And scorekeepers will also recall that I think that argument throws out standard economic theory on occupational wages. When your only tool is regression analysis, every omitted variable looks like a bad control.

(To defend Divided, one of its essays, “Why Do So Many Jobs Pay So Badly?” by Christopher Jencks, states, “The logic of a market economy is that we should all be paid the smallest amount that will ensure that our work gets done, and that is what low-wage workers generally receive.” (68))

I’m not alone, however. Bernstein wrote his blog post in response to findings by two researchers, Paul E. Harrington and Andrew M. Sum, who estimated that in 2012 half (!) of all college graduates under 25 were unemployed or employed in jobs that don’t really require much college education according to the Labor Department. Harrington and Sum should be better remembered for taking on Georgetown University’s Anthony Carnevale, who in 2010 estimated a shortage of college grads by 2018.

Please stop laughing. This is serious.

Harrington and Sum irresponsibly threw all caution into the wind and controlled for occupation in their analysis of college outcomes. They found, unsurprisingly, that college graduates in low-skill (“non-college”) jobs earn significantly less than college graduates in higher skill (“college”) jobs. The literature they cite calls this phenomenon “malemployment.” Harrington’s and Sum’s findings tend to show that occupations matter quite a bit for earnings, leading them to conclude, “If malemployment among college graduates simply does not exist, as the Georgetown forecasters [Carnevale] argue, then there should be little difference in the earnings among college graduates regardless of whether they were employed in college labor market occupations or not.” You can see their results charted in their response to Carnevale in all their brutality.

 

H&S Malemployment

(Yes, you read that right: Although it’s obviously a typo, advanced-degree holders in New England in 2009 earned just 6.6 percent more than high-school graduates if they were in non-college jobs. The next question is how well advanced-degree holders did if they found college occupations—rather than non-college occupations—that didn’t require any advanced training.)

To clarify, it doesn’t matter for Harrington and Sum if you think college mainly signals preexisting abilities or creates Very Important Human Capital. Their point is that there just aren’t enough college jobs to go around.

Bernstein, as well as David Neumark, another college-for-all academic, disagrees with Harrington’s and Sum’s methodology. Both argue that researchers should measure the college wage premium within occupations, and when they do so it’s huge. I’ll stick to Bernstein, since he kicked off today’s adventure:

Bernstein Premium

(Ironic that for all the drinking that supposedly goes on in college the intra-occupational premium for bartenders is scant. Guess they’d learn more about serving drinks in class?)

You might be tempted to ask how, exactly, college makes people 50 percent better at childcare, for example. I think I’m good with children, but it’s not because in my Plato seminar I read the Timaeus, where Socrates recommends educating children the rulers deem worthy and dumping the inferior ones onto the ranks of the proles. (This is also the dialogue where Plato talks about the (metaphorical!) island of Atlantis, which, sadly, is probably the thing he’s best known for.) Neumark, Bernstein, and those who agree with them are invited to satisfy your temptations. For my money, there’s almost no human capital effect on low-skill occupations.

I think it might be useful to go through the occupations Bernstein lists and show their wage dispersions. It turns out that even for the best case scenarios, Bernstein’s big premiums don’t account for a lot in annual earnings.

Occupation Intra-Occupational Premium 10th Percentile Annual Earnings 50th Percentile Annual Earnings 90th Percentile Annual Earnings
Retail Salespersons 49% $16,830 $21,140 $38,820
First-Line Supervisor of Retail Salespeople 42% $23,490 $32,700 $62,830
Waiters/Waitresses 20% $16,300 $18,590 $29,810
Customer Service Reps 45% $19,640 $30,870 $50,570
Cashiers 34% $16,420 $18,960 $27,710
Receptionists and Information Clerks 31% $18,330 $26,410 $38,170
Office Clerks 20% $18,040 $28,050 $45,340
Childcare Workers 49% $16,430 $19,600 $29,770
Home Health Aides 26% $16,690 $21,020 $29,480
Bartenders 9% $16,400 $18,920 $32,780
High School Grad, FT (25+) N/A $15,000-$17,499 $35,636 $70,000-$72,499
Bachelor’s-Degree Holder, FT (25+) N/A $25,000-27,499 $56,929 $100,000+

(Source Bureau of Labor Statistics Occupational Employment Statistics (OES) (2013), Census Bureau Personal Income tables. There’s some apples-to-oranging going on here as Bernstein’s premiums are for 18-30-year-olds and the OES wage ranges are for all ages. I’m also unclear on what Bernstein means by “less than college” in his table, which might include college dropouts or just be high-school graduates who never go to college.)

My point is even if a college education vaults someone into the upper earnings percentiles of a given non-college occupation, there’s little hope that he or she will earn as much as the median full-time college graduate. In many cases such individuals won’t even earn as much as the median high-school graduate. Some premium!

Although Bernstein wrote his post two years ago, the intra-occupational premium is really the endpoint of the debate on the college premium—once you’re willing to recklessly contaminate your regression results with bad controls based on the standard theory heresy that people will be paid “the smallest amount that will ensure that the work gets done” irrespective of educational attainment or student debts.

Except I can’t imagine someone at The New York Times deploying results like Bernstein’s with a straight face. Being in the 90th percentile of waiters/waitresses still means being a waiter/waitress. No, the media will just stick to the misunderstood Average College Graduate, who’s bound to be named Time‘s Person of the Year at some point.

In the meantime I hope we’ll see more work like Harrington’s and Sum’s, which I recommend reading.

NYT Says People Who Use BLS Inflation Data Are Conspiracy Theorists

Do not adjust your TV set. You did not leap into a parallel universe, and I am not suddenly sporting an evil Van Dyke. After enlightening us on how the Labor Department tracks inflation in higher education costs, the Timesresident champion of the elusive average college graduate, David Leonhardt, smugly compared anyone who uses government inflation data in good faith to conspiracy theorists like ShadowStats who think the government is cynically manipulating cost data.

(I’m using an image here instead of a block quote to show you I’m not kidding.)

(I’m using an image here instead of a block quote to prove I’m not making this up.)

One hastens to point out to Leonhardt that using the government’s published indexes, notwithstanding their flaws, to make an argument is not the same as (a) claiming the government is lying, or (b) cooking one’s own measurements based on repudiated methodologies to sell Web site subscriptions at fixed nominal prices. (For the record, I use Education Department data on college costs, though that’s probably problematic as well.)

So what prompted Leonhardt’s self-satisfied editorialization?

Answer: He discovered that until 2003, the Bureau of Labor Statistics tracked college tuition inflation by their stated prices and not their “net tuition” costs, which the bureau now largely tracks. It’s an interesting finding, but it deserved to be raised more professionally—and with better reasoning.

Leonhardt’s argument is that reporting on college tuition costs based on BLS data is “exaggerated” and “deeply misleading” because those data exclude financial aid discounts. He adds that only rich families pay sticker price and then compares college costs to retailers (e.g. Joseph A. Banks) that continuously discount their prices, making their sticker prices meaningless. Hence, he boasts that he’s knocked away yet another (sic) pillar people use to criticize the value of college education.

So what’s the problem here? One, the Joseph A. Banks comparison doesn’t work at all. Retailers that engage in psychological discounting in fact discount their prices—for everyone. They do not ask customers what their annual household incomes are and then charge them (in)appropriately. Colleges really do charge some people sticker price, so the sticker prices aren’t fictitious. Leonhardt hedges this fact by telling us this doesn’t matter because, allegedly, only rich families are charged full price.

The response, my second point, is if this is true so what? What are they paying more money for? Has the quality of education increased for people who pay sticker price? If you’re going to say that these students are paying for the privilege of learning with subsidized, smart people, then you’re going to have to prove that they actually learn more as a result. I’m also going to ask you to estimate when the marginal benefit of adding one more smart, subsidized student to an incoming class outweighs the additional cost to a given student paying full tuition. Note also that “graduating from college while well off” does not in fact ensure that a graduate will be well off going forward.

Finally, Joseph A. Banks’ prices are transparent once customers walk in, see the sticker, and do the math. Colleges, by contrast, advertise a *cough* *garble* *cough* percent-off sale. If this isn’t “real” inflation, why can’t families know up front what they will be charged?

I agree that the BLS should track inflation based on what people pay for goods and services holding the quality of those goods and services constant. However, the best Leonhardt can say about his discovery is that the composition of household spending on college has changed such that some students are asked to pay much more than they would have in the past so that other students can pay about what they would have in the past. If anything, this is a reason to track sticker price inflation, not ignore it.

Leave it to The New York Times to publish a blogger who trolls critics for using government data and then defines inflation away. Maybe Yale should give him an honorary B.S. in applied sophistry.

The Law Apprenticeship Scam

“It’s a cruel hoax. It’s such a waste of time for someone to spend three years in this program but not have anything at the end.”

So says Robert E. Glenn, president of the Virginia Board of Bar Examiners. No, Glenn wasn’t talking about Liberty University’s 34.4 percent employment rate in full-time, long-term, bar-passage-required jobs for its class of 2013, which had an average debt level of $81,045 (only!). Rather he was referring to the low bar-passage rates of Virginia’s law readers, who along with their peers in other states are the subject of a New York Times article, “The Lawyer’s Apprentice.”

Citing data from the National Conference of Bar Examiners, which the Times deserves credit for researching, we learn that only 28 percent of apprentices passed the bar versus 73 percent of ABA law school graduates. This fact prompts the ABA’s Barry Courrier to declare:

“The A.B.A. takes the position that the most appropriate process for becoming a lawyer should include obtaining a J.D. degree from a law school approved by the A.B.A. and passing a bar examination,”

I find this response disappointing for a few reasons: One, even if these statistics are for first-time test-takers only, a 73 percent pass rate is lousy. Law schools should be held to higher standards for what they charge students.

Two, the article appears to tacitly accept the ABA’s position that we can’t have good lawyers without many years of law school (and probably college too). The elephant in this room is selection bias. The reason people go to law school rather than these apprenticeship programs is that law schools broker jobs to people who already do well on standardized tests, to wit, the LSAT. Certainly in the age of PAYE, someone who can crush the LSAT has much better odds of finding a good law job by going to law school than trying to find a lawyer who will train him or her. If anything, law school is a more reliable path to qualifying for the bar exam. Indeed, the article acknowledges that “the lack of class rankings put clerkships with judges and plum gigs at big firms out of reach” for law readers.

If you’re wondering why people who don’t do well on the LSAT go to law school instead of these programs, I give three responses. One, they aren’t widely known and have no advertising. Two, many law students still buy into the versatile JD myth. Three, the largest proportion of people opting out of law school are people who don’t do amazingly on the LSAT anyway. So there. (The Times says these programs are “underpopulated,” but given the effort the would-be apprentices must go through to get established, one might think the problem is that there really isn’t much demand for new lawyers.)

I acknowledge that many of the apprentices interviewed in the article are sincere in their desire to avoid debt and only want to do small practice work. If anything, bar authorities should make it easier for people to choose that route. Instead they offer a post hoc rationalization for credentialism in legal education.

Good News: The Student Loan Crisis Has Been Canceled

…According to the Brookings Institution’s Beth Akers’ and Matthew M. Chingos’ paper, “Is a Student Loan Crisis on the Horizon?” The authors find that not only is there no crisis today, but there won’t be one in the future. (“Crisis” here, I gather, means debtors being unable to make their payments and taxpayers being forced to write-down some billions in student loans.)

Before picking through their paper, though, I have to give special credit to The New York Times‘ David Leonhardt, who crows:

The deeply indebted college graduate has become a stock character in the national conversation: the art history major with $50,000 in debt, the underemployed barista with $75,000, the struggling poet with $100,000. … Such graduates make for good stories (and they tend to involve the peer group of journalists).

This comes mere days after The New York Times Magazine ran an article officially declaring that millennial college graduates who were living with their parents weren’t leaving. Many of the subjects had significant student loan debts and low-paying jobs. I’m not saying the participants were typical of their age group, but I’m impressed that Leonhardt can undercut his own publication. I admire gall.

As for Akers’ and Chingos’ paper, take a look at John Haskell’s response. He argues that the authors commit a composition fallacy by comparing student debt repayment during the more recent economic disaster with the halcyon days of the 1990s.

It’s an excellent point, and I have some of my own to add.

One, on page 4 of the report, the authors aggressively lean on the college premium as evidence that “the growth in debt is not [obviously] problematic.” The idea is that if the gap between college graduates’ earnings and high school graduates’ widens, then college is a good bet. The flaw, and there are many with this kind of thinking, is that both sets of earnings can be falling simultaneously but so long as high school graduates’ earnings are falling faster, then student debt can still be a problem even as the premium is growing.

Two, the authors make an implied structural unemployment argument when they write, “In 2011, college graduates between the ages of 23 and 25 … had employment rates 20 percentage points higher [than high school graduates].” However, not going to college isn’t the cause of lower employment rates among high school graduates. It’s because there’s slack demand for labor in the economy. It’s not too much of a stretch to hypothesize that employers prefer college graduates even for menial jobs.

Three, as always with college premium discussions, not everyone gets the average college degree, and not everyone has the average debt level. The authors only bring this up in their conclusion, which I think is unfair to readers.

Four, Akers and Chingos challenge the rhetoric of a student debt crisis by analyzing data on households with householders aged 20-40 from the Federal Reserve’s Survey of Consumer finances. It’s a minor point, but people who have higher debt levels are probably more likely to be living with their 40+-year-old parents than on their own. I doubt the effect is that large, but it’s something Akers and Chingos should have noted.

Five, it’s one type of composition fallacy to compare past trends with current outcomes, but it’s another to omit prospective factors from one’s predictions. The authors assume today’s college graduates won’t suffer from “cohort risk” due to the persistent output gap. It’s a pretty big if, and Akers and Chingos won’t pay anyone’s student loans if they’re wrong.

Having said that, when the authors find fairly low monthly payment-to-income ratios (excluding debtors making less than $1,000 per year) it may appear too good to be true, but we should acknowledge it.

Monthly Student Loan Payment-to-Income Ratios, 1992-2010

I’m not sure what this means given the simple calculation I did above. It’s pretty surprising that student loans are such a small amount of monthly incomes. It might be that they’re excluding the billions of dollars in student loans that are in default, forbearance, deferment.

Finally, since the conversation on student loan debt is creeping towards amputating graduate school debtors from undergrads, gaze upon Akers’ and Chingos’ Figure 4:

Akers and Chingos Figure 4

The Survey of Consumer Finances is given only once every three years, but even between 2007 and 2010, the spike in graduates’ debt is evident. Who wants to bet that these aren’t Grad PLUS loans? Seriously, that program is not long for this world.

The authors conclude that their results should encourage Washington to not tweak the student loan system based on a perception of widespread financial hardship. They do not, frustratingly, discuss any of the existing indicators of a present student debt crisis. 11 percent of student loan balances are delinquent, 11 percent of debtors (minimum) with federal loans are in default, and $322 billion out of $1.043 trillion in federal loans are in deferment, forbearance, or default. (Calculated from here) Since we know the economy isn’t roaring forward and won’t without systemic reform, it’s hard to believe that all these loans will be repaid in full. If this doesn’t count as a crisis, what does?

 

Someone Needs to Teach the NYT How to Use the Internet

…Okay, not the Internet (I hope), just FRED, but it’s an unfortunate performance by David Leonhardt, who was the star of an outing earlier this week.

In “The Jobless Rate for Community-College Graduates Is Also Low,” he teaches us:

The federal government’s main educational-attainment categories are fairly blunt. In particular, the “some college” category includes a wide array of people: Those who have dropped out of college without earning any degree, those who have earned a two-year degree meant to lead directly to a job (such as in nursing) and those who have earned an academic two-year degree that is often a first step toward a bachelor’s degree.

[T]he Bureau of Labor Statistics does distinguish among the different versions of “some college.” (It would be even better if the bureau made this data easily available to the public.)

I get the feeling Leonhardt does not use government data much. For one, I’ve never seen “some college” include associate’s degrees in any dataset. For another, the data series Leonhardt refers to are in fact easily available to the public and can be found on FRED (Federal Reserve Economic Data).

Here’s Leonhardt’s chart:

Unemployment Rates by Educational Attainment

Here’s my version from FRED (just “some college” and “associate degrees” types):

 

(Click to view source data)

See? Now we can replace him with a robot?

Okay, I’m just being a meanie. Leonhardt’s ignorance is an honest mistake and I’m sure if he spent a couple weekends learning how to use FRED he wouldn’t be wasting his readers’ time telling them information is unavailable when it is. However, his ignorance is a thin end of a wedge. It may seem trivial, but FRED represents one of the great advances of this decade: easy access to government information. Instead of trying to find Brazil’s GDP in an encyclopedia, one can look on Fred or the IMF’s or World Bank’s Web site. FRED doesn’t have all datasets yet, but journalists should be turning to it first when researching government data. I’m sure they will be within a decade, but ones like Leonhardt really need to familiarize themselves with it.

But since we’re on the topic of the unemployment of the educated, I thought I’d take another crack at it. Leonhardt compares unemployment rates by education to conclude that, as the title of his article suggests, two-year colleges are a good bet too—so long as students get the degrees.

Comparing these unemployment rates doesn’t sit well with me because they all use different denominators. The high unemployment rate for those who have less than a high school diploma might be less significant if there aren’t that many non-high-schoolers in the labor force. Likewise, a low unemployment rate for college graduates might be problematic if there’s a large number of them. So, using FRED, I gathered the unemployment level (not the rate) by education for the 25 to 34 age bracket (yes, FRED now has that), and I decided to do a cross-sectional analysis of the unemployed by their educational attainment.

The data only go back to 2000, but here you are:

Cross-Section of Unemployment Level by Education

The results are slight but the trends are clear: The educational attainment of the young unemployed is rising. As of last April, for the first time, half of them had at least some college education. In May 2007, about 22 percent of the young unemployed had an associate’s degree or bachelor’s degree; now it’s 28 percent. Meanwhile the percent of the unemployed who haven’t finished high school, whose 11.5 percent unemployment rate typically alarms Leonhardt and his peers, has fallen from 24 percent to 15 percent of the total.

You might be tempted to say that the unemployment level doesn’t include people who drop out of the labor force. True, but (a) that’s not Leonhardt’s argument, and (b) the labor-force dropouts might be more educated too. Go ahead, prove me wrong.

The point is, if Leonhardt et al. are right about how crucial education is, then we would expect the proportion of educated unemployed to be declining. On the contrary, we see that as we educate more people, it just substitutes the uneducated unemployed with the educated. It calls into question the mantra that “college graduates are less likely to be unemployed.”

**********

UPDATE: Allow me to clarify my thoughts here. Nothing I’ve written here should be interpreted as saying that, for example, people shouldn’t finish high school because they’re a smaller proportion of the unemployed. The unemployment rate by education does matter for the obvious reasons we read about, like employers not wanting to hire uneducated people. However, my point is to draw attention to the evidence that if the more education people attain, the better-educated the unemployed become. Irrespective of the unemployment rate by education, this phenomenon supports the signaling theory because we’d expect the unemployed to be less educated over time.

When Will Robots Replace Journalists Urging Everyone to Go to College?

…Is the question that crossed my mind reading David Leonhardt’s, “Is College Worth It? Clearly, New Data Say,” for The New York Times.

The title alone tells you exactly how this article will play out. The “new data” will specify the returns to higher education for the “average college graduate,” who is, apparently, everyone who goes to college. There will be some quotes from notable college-for-all economists who haven’t left campus for the real world in ages. There will then be liberal-esque dismissal of student loan debt and how tough it is for college grads to find good jobs. The article will close with a bunch of hail-Mary pseudo-arguments about the consequences of not sending everyone to college. There might be a line about naughty for-profits.

Leonhardt’s article fits just about all the points.

Americans with four-year college degrees made 98 percent more an hour on average in 2013 than people without a degree.

There’s your “average” earnings premium.

“We have too few college graduates,” says David Autor, an M.I.T. economist, who was not involved in the Economic Policy Institute’s analysis.

There’s your cloistered economist.

But what about all those alarming stories you hear about indebted, jobless college graduates?

The anecdotes may be real, yet the conventional wisdom often exaggerates the problem. Among four-year college graduates who took out loans, average debt is about $25,000, a sum that is a tiny fraction of the economic benefits of college. (My own student debt, as it happens, was almost identical to this figure, in inflation-adjusted terms.)

Student debt excuses, check. They aren’t a problem because someday grads will have the college job that’ll enable them to pay them off. Debtors need to have faith, but don’t worry about David Leonhardt, he’ll do just fine because he’s just as average as you are with his job at the NYT.

Those who question the value of college tend to be those with the luxury of knowing their own children will be able to attend it.

As the economy becomes more technologically complex, the amount of education that people need will rise. At some point, 15 years or 17 years of education will make more sense as a universal goal.

And there’re your hail-Marys. (I’d love to know Leonhardt’s source for the first one.)

Nothing on the perfidious for-profits, but Leonhardt’s defensiveness is entertaining.

It’s important to emphasize these shortfalls because public discussion today — for which we in the news media deserve some responsibility — often focuses on the undeniable fact that a bachelor’s degree does not guarantee success. But of course it doesn’t. Nothing guarantees success, especially after 15 years of disappointing economic growth and rising inequality.

In other words, if college doesn’t pay off, don’t expect David Leonhardt to solemnly assess the situation. He’ll just blame “inequality” instead.

If NYT articles on higher education are going to be so one-sided and filled with unsubstantiated claims about its critics, why can’t we just replace their authors with robots?

Follow

Get every new post delivered to your Inbox.

Join 138 other followers