A quick data dump post.
Karl Smith gets this right: “If you look across the demographic survey’s what looks to be the biggest effect of the recession has been to delay adulthood for folks born around 1990. Rather than graduating high school or college and getting their own apartment and their own car, they stayed with or moved back in with their parents. This drove down household formation and drove down the demand for automobiles.” The Census webpage has similar thoughts and numbers. Unemployment is the biggest driver of young people not forming households, and as youth unemployment very slowly comes down we should see household formation pick up.
Or should we? Steve Waldman counters in comments: “One might argue that the funds a young adult born in 1970 would have directed to rent payments and transportation have, for a young adult born in 1990, been redirected towards servicing student debt.”
Meanwhile Businessweek announces: “Only 9 percent of 29- to 34-year-olds got a first-time mortgage from 2009 to 2011, vs. 17 percent 10 years earlier.” (h/t Matt Stoller) Also cited in the Businessweek article, the National Association of Consumer Bankruptcy Attorneys is shooting distress signal flares into the air based on interviews with their frontline members who see young people struggle with student debts.
Can we get any data on this? One initial place to look is the ratio of student debt payments-to-income. How much of income goes to servicing student debts, and how has that changed?
Here’s a current economics journal review, Student Loans: Do College Students Borrow Too Much—Or Not Enough? Spoiler alert: they think not enough. What surprised me is how thin the “not enough” data is on the month-to-month (as opposed to lifetime frameworks). Review:
There is little evidence to suggest that the average burden of loan repayment relative to income has increased in recent years. The most commonly referenced benchmark is that a repayment to gross income ratio of 8 percent, which is derived broadly from mortgage underwriting, is “manageable” while other analysis such as a 2003 GAO study set hte benchmark at 10 percent…the student would need an annual income of about $25,456 [to handle a $212 monthly student loan payment]…Overall the mean ratio of student loan payments to income among borrowers has held stead at between 9 and 11 percent, even as loan levels have increased over time (Baum and Schwarz 2006; Baum and O’Malley 2003).
(I get annoyed with economists’ indifference between a college wage premium driven by increasing college wages versus declining non-college wages, especially as tuition costs are rising. But that’s for another day.)
The 2006 Baum and Schwarz study cited finds that there are two trends – rising debt but steady student debt-to-income ratios – and that the “contrast in these two trends can be attributed to a combination of rising earnings, declining interest rates, and increased use of extended repayment options. Studies of loan repayment in the United States in 1997 and 2002 revealed that while average total undergraduate debt increased by 66 percent, to $18,900, average monthly payments increased by only 13 percent over these five years. The mean ratio of payments to income actually declined from 11 percent to 9 percent because borrower.”
But that study ends in 2002. Has anyone done an analysis on this since? I can’t find one, so we’ll have to do a back of the envelope one.
Heidi Shierholz, in new college grads losing ground on wages, shows the rising earnings of new college graduates in the late 1999s that end up driving the 2006 Baum and Schwarz study above. However, according to Shierholz, since 2003 entry level college graduate wages have been flat. Meanwhile, average undergraduate tuition student loan debt has gone from $18,900 (according to above) to $24,800 (according to Demos). That’s an average loan payment increase of about $60/month, from $228 to $285.
If monthly student debt-to-income was 9% with $18.9K debt, then the average monthly student debt-to-income with no increase in income and $24.8K debt is 11.25%. Notice that is above the 10% GAO “healthy” student debt ratio and impacts many more people as a percentage of the young. Now interest rates are lower and payments can probably be stretched out; however unemployment is higher for recent college graduates (and their parents), so income is also going to be a bit lower. Let’s say they cancel. We also know that graduating into a recession has a long-lasting impact on wages, so let’s assume students don’t expect a rapid rebound in wages once they start working.
So now the practical questions becomes: how much an average increase of 2.25% of student-loan to income payments, that extra $60/month, impacts how twenty-somethings move out of their parents house? I’d be interested in arguments, but I can’t imagine it’s going to be the breaker for a large percentage of the population. And I imagine it is a weaker determinate than, say, credit-card debt incurred during joblessness on the balance-sheet front. I’d like to see more arguments, but at this point I don’t see student debt putting a pickup in household formation in check – it will decline a bit than it would otherwise, but young people might equally just spend less on housing and rentals to balance out their monthly debt ratios.
Rortybomb 
Reblogged this on Bay Area Real Estate Trends and commented:
They will definitely impact first-time homebuyers!
So we’re reducing the potential of future housing bubbles (young people are too hobbled by debt to rush in and push up prices) and transferring what could have been speculative froth there to the education sector? and kids here, like those in Singapore, are more likely to live with mom and dad for their 20s (whom they LIKE a lot, see any study of millennials), and not rush into an early and possibly awful household?
I’m no apologist for megadebt for students. But let’s be really clear on exactly what they’re missing out on when we try to save them from loan payments. And let’s also reduce the pay of the parasitic finance sector first…I am sick to death of “progressives” salivating at cutting professor/doctor/nurse pay first and then thinking “oh some day we’ll get around to taxing the billionaires/hedgies/finance parasites.” Cuz the docs/profs are a lot weaker, right?
you made a mistake that confused the hell out of me. i actually started to write a longer comment questioning your methodology, but you look to have just substituted the word “tuition” when you wanted “debt” in this sentence:
“Meanwhile, average undergraduate tuition has gone from $18,900 (according to above) to $24,800 (according to Demos). “
Ugh. Sorry about that. Fixed above.
Mike — My speculations over at Modeled Behavior were off-the-cuff and data-free, but if we want to think seriously about this, I think we have to be very careful.
Anecdotally, it’s been my experience, as I suspect it has been yours, that student debt is much more prevalent and a much bigger deal for the cohort of students in college today than in the early 1990s, when I had the privilege. It is dangerous to put “affordability” in terms of the cash-flow cost per month. I invited that by phrasing my comments to Karl Smith in terms of redirected cash flows, which was artless on my part, my bad! But putting costs in terms of dollars-per-month financing is the trick of car salesmen trying to sell you on upgrades or real-estate agents trying to persuade you of how much home you can “afford”. It makes burdens seem small that are not small at all.
Contemplating student debt in terms of averages is also dangerous, because the “wage premium” hides a tremendous amount of dispersion in outcomes ex post and uncertainty for students ex ante. It is easy for financial analysts to compare the lifetime expected wage premium in NPV terms to the NPV of tuition + often subsidized interest payments, and find that students are underborrowing. But computations of the NPV of the wage premium, while fine in aggregate, are badly overestimated from the perspective of a rational individual student, who must apply a severe risk premium to her discount rate if she is not overconfident of success. A rational indebted college graduate finds herself with an asset of very uncertain value and a liability she cannot escape in almost any circumstance. Such a student reasonably perceives herself to be in a precarious financial circumstance.
This brings us to the question of “what has changed?” Students borrowed in the early 1990s, students borrow today, do we have any reason to think that recent borrowing behavior will be more discouraging of household formation than 20 years ago? I think that the answer is yes. Here’s why:
1) Balance sheet effects: Since the mid-1990s, the average real debt burden among indebted students by 62%, while the average wage premium, ignoring dispersion effects, has risen by only 27%. And the median wage premium has risen much less than the average premium. Much of the increase in the college wage premium is in the right tail. (Look at Figure 5 of Avery & Turner, comparing 1978 to 2008) Outcome dispersion among college graduates has increased dramatically. The disproportionate increase in indebtedness means that, regardless of the absolute value of the expected wage premium relative to the total cost of education + finance, an average indebted student holds a much more heavily levered asset than a student who graduated in the early 90s. The increase in outcome dispersion and ex ante uncertainty suggests a typical college graduate’s rational estimate of the financial value of a dollar’s expected wage premium should be lower than in the early 1990s. The combination of these two effects, increased outcome uncertainty and increased leverage, creates what any young capitalist would recognize as a dangerous capital structure, more dangerous than it was back in my day. A rational college graduate would avidly try to adjust to a capital structure commensurate with her risk position, by retiring student debt more quickly than a graduate in the early 1990s. At the margin, she will trade off housing consumption to retire debt more than her predecessor would have.
2) Simple prevalence of debt: According the the Avery & Turner review, 19% of undergraduates borrowed in the early 1990s, while 35% did in 2007-2008. For the 16% of students who would not have borrowed, the cash-flow delta, even if objectionable monthly affordability terms, is not $60 on average (your cite from 1999), but the full $280 per month. That’s a substantial fraction of a potential new household’s rent. There are complexities: did the students who did not take out loans in the early 1990s finance their college lifestyle with burdensome credit card debt more than today’s formal student debt carriers do? Perhaps, but over the period (until the financial crisis), credit card borrowing was growing in general and credit standards declining, so that would be a countertrend result. Regardless, it could be true. Perhaps students became more savvy about substituting subsidized student debt for other forms of finance of the period (although the nondischargability of student debt calls that savviness into question). But the null hypothesis has to be that a substantially larger fraction of students in the early 1990s financed their educations out of family support and current income than are able to do so today. These debt-free early 1990s graduates could much more easily afford to set up house than their successors can today.
There is some empirical evidence that college indebtedness has effects on students’ decisions with respect to career that are disproportionate when viewed from a standard, lifetime consumption-smoothing perspective. This is consistent with the notion that graduates consider their debt position to be burdensome, and are willing to make tradeoffs to retire it, even tradeoffs that seem disproportionate, like altering choices of major and vocation. See e.g. Rothstein and Rouse (which is cited by Avery & Turner). The study is not a “slam dunk” — their population is too elite to be broadly representative and their results depend upon model specification more than I would like. But it’s a very hard question to address empirically. If it is true that student loans weighs upon students enough to alter career choices, it is hardly implausible that retiring that debt would matter enough to delay household formation.
Overall, I think a lot of trends may be conspiring to alter the norm of immediate postcollegiate household formation. In a sense, prolonged “doubling up” is just a continuation of the “two income trap”: as fixed household costs (rent or mortgage payments, transportation, education costs as financed over time, health care insurance, etc) have grown, the operating leverage of households have increased. More adults per household is one way to counter this trend. If it became obligatory over time to have two incomes per household, then it also became more desirable to have three or four. The United States has had very strong norms of filial independence, and prior to the financial crisis those norms held. But as JorgeO on Twitter reminds us, those norms are unusual from a global perspective. Economically, the US may have been like a supersaturated solution, out of equilibrium but deceptively stable until the financial crisis catalyzed a sudden crystallization. I am not making a firm prediction that household formation won’t revert to late 20th Century trends. I don’t know. In general I think it is dangerous to suppose changes in behavior induced by the crisis period are nothing more than “blips”. It seems pretty clear that the Great Depression induced a lot of lasting behavioral changes, via formal regulatory channels, but also via altered norms and assumptions among the public who survived it. The traumas of the Great Recession might do so as well.
wow, great comment. i had thought of a few of those issues (minus the quantitative rigor, citations, and general erudition – i’m just a guy!) earlier when mike’s error confused me. may i add one more issue? what about the increase in advanced degrees? i went a little bonkers looking through census data, getting really frustrated that reporting formatting seems to have changed between 2000 and 2009, but then i found a decent handy article:
http://chronicle.com/article/As-Graduate-Student-Population/128402/
which links to this report:
http://nces.ed.gov/pubsearch/pubsinfo.asp?pubid=2009166
The article tells us, from the the report: “Over all, the number of graduate students has increased by 57 percent since 1988.” and “More than three-quarters of professional students borrowed against the cost of their education.” That report indicates the average graduate degree loan is $18,500.
I know we are talking about a relatively small part of the population, but there has definitely been a lot of reporting and information on a law school bubble, and anecdotally speaking, a number of my college-educated cohort friends earned JD’s only to eventually take work that doesn’t require one. When we add in post-college degrees, does that change the debt picture as well?
The Baum paper from 2006 linked above and the 2003 study they cite (Table 2) breaks numbers down between undergraduates and graduates. The numbers are much higher for graduates students. Post-college degrees will definitely change the debt picture, I’m not sure exactly in which ways, and it should be pointed out that many young people are taking the recession to get post-college degrees (one big driver of the increasing rate of student loans).
One of the things I hope the most for from the CFPB is that there is better data on all this stuff available for researchers.
thanks! missed that. i’m not sure why they get such higher levels than the $18,500 i cited above (maybe because they only include those who carry debt? unclear if the other study included those with zero debt). i guess what i am getting at is whether we’ll continue to see significant growth in advanced degrees or if it’s just a symptom of the recession. there certainly has been an explosion in growth of masters and professional programs, but it’s unclear if that can be sustained. also unclear if the graduate degree will continue to increasingly be viewed as a necessary prerequisite for a middle class lifestyle that might have been more attainable with just a bachelors a generation ago.
Hey Steve,
I’m very interested in this beyond the “recovery winter” arguments and more towards the question of how much student debt relief will impact aggregate demand now and in the future (there’s a ton of debate at high-levels on the equivalent question for underwater homes, while student loan relief tends to get filed immediately under worst idea ever by economists), so thanks for the detailed comment. A few points.
I would be shocked if there wasn’t a substantial rise in the number of young people who never leave home as a result of this recession – however I think we’d have to look to hysteresis and labor force participation as the driver. Youth poverty is very high, but that’s looks to be a lot more young people making less than $11K a year. I also think we’ll likely see a decline in homeownership rates among the young, as a lack of sustained labor means a lack of downpayments and cushions for mortgage origination (to speak nothing of tightened standards). I also think high rents in desirable, urban areas may drive added increases in people per household. But I have a hard time seeing student debt for people born in 1990 as a major new problem outside these contributing factors.
The Rothstein study cuts both ways – I remember (it’s been a bit) Rothstein finding increased incomes for those with more debt (or at least those with debt) as a result of job choices focused more on remuneration. Debt, as we expect, generates liquidity problems, which in turn increases effort, job search intensity, a lower reservation wage and more of an emphasis on money-making (I worked on the analogy of student loans as a faculty tax here). More generally, we could see student debt compel more market labor than we would otherwise, which would thus mean more money available for housing. Debt as a labor stick, and the stick works. I find this result unpleasant, but it follows from the analysis.
I’ve heard various numbers saying that debt-servicing as a whole has gone up tremendously for the young over the past 15 years but couldn’t verify it. If so, I’d like to see what has driven that. I wonder if the more critical balance-sheet effect would be young people keeping a semblance of a life off of credit-cards while unemployment, but that is speculation.
(A lot of the bad scenarios here are driven by for-profit schools, which tend to have very high debt loads and very low value-added.)
The cultural shifts on college and affordability strike me as up in the air; a system of indenture where people tithe 10% of their income for a long while to get the ability to fully develop their talents and capabilities, abandonment of public education and increasing relief through tax-rebates and deferment looks to all be in motion, but could also all be resisted.
Pingback: Secondary Sources: Student Loans and Household Formation, Fed Rules, Confidence and Jobs - Real Time Economics - WSJ
With regard to the final question:
“So now the practical questions becomes: how much an average increase of 2.25% of student-loan to income payments, that extra $60/month, impacts how twenty-somethings move out of their parents house? I’d be interested in arguments, but I can’t imagine it’s going to be the breaker for a large percentage of the population.”
I think what this misses is the variance in actual debt loads. Though the data could prove me wrong, my guess is that most people cluster around certain debt loads, so while the mean debt load might have only risen slightly, the actual debt for groups above the mean is substantially higher, perhaps around $40K, and for the other group (commute to college/community for 2 years, etc.) is more like $10k.
This would imply that it is only a certain subset of the 1990s cohort who are actually staying at home. For instance I worked from home at a bookstore for the first year out of college (2010), before moving out the second year to live in an apartment with friends. On a 10 year repayment plan my bill is about $500 a month (I want to a quasi-state school but lived off campus and worked part time). As a result, I will certainly not be buying a house, or a car, any time soon. But I do have other friends who have much less debt, and for whom the recession is but a momentary setback.
I would venture to guess that higher ed tuition pools in a similar way, with schools all aiming to charge about the same as their peers within any single tier, and averaging debt loads/tuition would obscure this.
Mike — I think we are on the same page. Does student debt constitute a “major new problem”? Obviously not since 2007, when the kids were still alright. When there is a continuous change over time, there is never a moment when it constitutes a major new problem. But if you do a comparison over a longer period — I do think that 20 years is about right — you can see a major change that explains part of why “kids today” are behaving as though they are more financially insecure than kids 20 years ago. Ultimately that’s what “doubling up”, at least initially, is about. The question of whether what we are observing is a “blip” or a secular change turns, at least initially, on whether what drives a revealed sense of precarity is temporary or permanent. I’m glad to see you refer to hysteresis as well, because social norms are path dependent and don’t straightforwardly reverse as economic fortunes reverse, so that a sufficiently prolonged “blip” can turn out to have permanent effects.
Overall, the reason why kids are staying home is, I think, because they view their economic situations as poor and precarious. Current unemployment and the forward-looking labor market are a large part of that. For a people who form expectations based on experienced history, the distribution of potential labor markets and macroeconomies is permanently altered. Outcomes a typical graduate in 2006 would not make tradeoffs to insure against a graduate in 2012 will, even if the headline numbers revert to normal.
But objectively, the cost of education and the fact increasingly uncertain outcomes must be financed by increasingly inescapable debt has, drip by drip, made a serious contribution to the precarity of new college grads. (As you say, the burden is more numerically outrageous for graduate and professional students. It is offset there by greater expectations of the value of the asset, but the wide dispersion in outcomes renders that asset also of questionable value for many students relative to the certainty of the debts.) There is no point in time where this suddenly “mattered”, and again hysteresis is a good word. There is no “major new problem”, but if you could replace the distribution of student debt burdens of today with the (inflation-adjusted) distribution that prevailed in 1990, I think that would be a very significant stimulus to all kinds of less precautionary behavior, including household formation. In terms of magnitudes, would an increase in current labor market participation rates matter more or less than a reversion to 1990 debt burdens? I don’t know. The precaution is overdetermined, and estimating magnitudes over larger ranges than we can observe in a clean experiment is almost impossible to do well. I think the largest (im)possible effect would come if you could eliminate the experience of the “great recession”, that the change what labor market participants now believe is possible is the single largest cause of precautionary behavior. But current labor market conditions matter a lot, as do the objectively deteriorating financial position of the individual (rather than aggregate) students once the cost of an education and of financing it are taken into account.
Indeed, if we extrapolate cost of education trends forward, and if we assume (as I do throughout, unapologetically) that individuals cannot accurately predict their lifetime labor market outcomes, the effect of increasingly levered balance sheets would have eventually become sufficient to depress household formation entirely on its own. In extremis, a college education becomes an impossibly expensive lottery ticket that grants a minority of winners a culturally “middle class” life and condemns the rest to a negative but nondischargeable financial position.
Re the Rothstein study and income effects, sure, to the degree that kids choose more lucrative majors, that tradeoff may reduce the degree to which graduates have to trade-off household formation. But of there are kids making the (conventionally irrational!) choice to accept second-best careers out of concern over debt loads, it strikes me as implausible that there are not kids making the (much more justifiable) choice of going for the first-best career and managing the debt burden via the easily reversible choice to double up. Plus, using career choice as a means of managing debt by increasing expected income is one of those rational individually, collectively absurd strategies, like using liquidity to manage risk. Increased supply of graduates in the fields widely considered lucrative will depress median wages in “lucrative” fields. In fields like finance and law, there will always be an extreme upper tail, but those fields will come to look more like the lottery ticket of postcollegiate life anyway. Unless there is labor market demand in “lucrative” fields that is sufficient to offer the most students a secure and reasonably affluent lifestyle, shifting majors by students can be individually useful but cannot solve the problem in aggregate.
Pingback: The “Ohioans use puppets to explain stuff” Daily Quinn « The Daily Quinn
If we define excessive student loan debt as a debt-service-to-income ratio of 10% or more, then 2.4% of Bachelor’s degree recipients had excessive debt in 1992-93 and 12.4% in 2007-08. (These results are based on a synthesis of data from the quadrennial National Postsecondary Student Aid Study and US Census Bureau data on income by educational attainment.)
My analysis of the relationship between default rates and debt ratios suggests that default rates begin climbing rapidly after a 15% debt-service-to-income ratio. I tend to promote a rule of thumb that total education debt at graduation should be less than the student’s expected annual starting salary and ideally a lot less (debt-to-income ratio of 1:1 or less), which corresponds to a debt-service-to-income ratio of 13.8% for the unsubsidized Stafford loan. Borrowers who graduate with total education debt that is less than annual income should be able to repay the loans in about 10 years. Students whose debt exceeds their income will need an alternate repayment plan, like extended repayment or income-based repayment to afford their monthly loan payments. But this means they will be in repayment for more than 10 years.
About a third of Bachelor’s degree recipients in 2011 had enough federal student loan debt to qualify for a repayment term of 20 or more years. This means these students will still be repaying their own student loans when their children enroll in college. (But note that there’s some evidence that borrowers with excessive debt delay life cycle events like getting married and having children.) They will not have saved for their children’s college educations and will be less willing to borrow parent education loans to help their children because they will still be up to their eyebrows in debt. So a key concern is the cascading effects of too much debt.
Mark Kantrowitz
Publisher of Fastweb.com and FinAid.org
Pingback: Zunguzungu on Privatization and Brutalizing Campuses; My Favorite Graph Juke of the Week (OECD Education Edition) | Rortybomb
Pingback: Sunday Reading « zunguzungu