My Ph.D. Micro teacher, David Card, won the Nobel Prize last week.  My best-known piece on Card examines the tension between his research on the minimum wage and his research on immigration.  My most extensive discussion of his work and intellectual influence, however, appears in Chapter 3 of The Case Against Education.  Here’s the excerpt.  Enjoy!

Labor Economists Versus Ability Bias

Labor economists aren’t merely attuned to the possibility of ability bias.  They’ve long felt a professional responsibility to measure it.  But over the last quarter-century, labor economists have surprisingly moved to the view that there’s not much bias to measure.  A famous review of the evidence by eminent economist David Card concludes ability bias is small, non-existent, or even negative.[i]  I call this verdict the Card Consensus.  Many, perhaps most, elite labor economists not only embrace it, but rely on it for practical guidance.  We see the Card Consensus in top scholarly venues like the Journal of Economic Literature.

[T]he return to an additional year of education obtained for reasons like compulsory schooling or school-building projects is more likely to be greater, than lower, than the conventionally estimated return to schooling.[ii]

We see the Card Consensus in top policy initiatives like the Brookings Institution’s Hamilton Project:

[I]t’s possible (and even likely) that individual college graduates have different aptitudes and ambitions, and might even have access to different levels of family resources. All of these factors can impact earnings. However, the evidence suggests that these factors don’t drive the impressive return to college; instead the increased earning power of college graduates appears to be caused by their educational investments.[iii]

Even analysts who don’t cite the Card Consensus enjoy its protection.  Well-publicized calculations of the “value of college” typically ignore ability bias altogether.[iv]  The Card Consensus neuters criticism of this omission.  How can you attack a tacit “0% ability bias” assumption as a fatal flaw when plenty of experts stand ready to defend it as a harmless simplification?

This is a disorienting intellectual situation.  Statistically naïve laymen blithely infer causation from correlation: Since college grads earn 73% more than high school grads, college causes a 73% raise.   Economists who don’t specialize in labor smirk at the laymen’s naiveté; they take sizable ability bias for granted.  But economists who do specialize in labor now largely stand with laymen.  While ability bias is intuitively plausible, the Card Consensus tells us, “Move along, nothing to see here.”

What about abundant research from last section that detects hefty ability bias?  The Card Consensus barely acknowledges it.[v]  Why not?  Labor economists’ most common rationale is that no one can measure all the abilities that cause both academic and career success.  True enough; but that just means ability bias is worse than it looks.  Supporters of the Card Consensus also occasionally muse that high-ability students might leave school sooner:

[S]ome people cut their schooling short so as to pursue more immediately lucrative activities.  Sir Mick Jagger abandoned his pursuit of a degree at the London School of Economics in 1963 to play with an outfit known as the Rolling Stones… No less impressive, Swedish épée fencer Johan Harmenberg left MIT after 2 years of study in 1979, winning a gold medal in the 1980 Moscow Olympics, instead of earning an MIT diploma.  Harmenberg went on to become a biotech executive and successful researcher.  These examples illustrate how people with high ability – musical, athletic, entrepreneurial, or otherwise – may be economically successful without the benefit of an education.  This suggests that… ability bias, can be negative as easily as positive.[vi]

Straightforward rebuttal: Name any ability the well-educated tend to lack.  Outliers have ye always.  But the well-educated are, on average, abler across the board.  No one hears about a kid quitting high school or college and says, “Wow, he must be talented.”

At best, then, the Card Consensus casually throws away a large body of contrary evidence to get off the ground.  But it’s worse than that.  The Card Consensus casually throws away the best evidence.  Worried you’re improperly giving school credit for pre-existing ability?  There’s a clear statistical cure: Measure pre-existing ability to allow an apples-to-apples comparison of people with equal ability but unequal schooling.  The cures the Card Consensus prizes, in contrast, are anything but clear.  Instead of sending researchers in search of better ability measures, it sends them in search of “quasi-experiments” – naturally-occurring situations that mimic experiments.

As a result, labor economists have collected a zoo of alleged educational quasi-experiments.  Some study twins.  As long as identical twins have equal ability but unequal educations, education’s true payoff equals their earnings gap divided by their education gap.[vii]  Other scholars study the effect of season of birth, on the theory that kids who are young for their grade are less legally eligible to drop out of high school.[viii]  Since 2000, researchers have been most transfixed by changes in compulsory attendance laws.  If government forces students who would have dropped out to stay in school, what happens to their income after graduation?[ix]  While technically impressive, all these papers raise more questions than they answer.  To treat changes in compulsory attendance laws as a quasi-experiment, for example, we must assume states change these laws at random – or at least for reasons unrelated to the labor market.

Once a quasi-experimental approach picks up steam, moreover, critics usually uncover deep flaws.  Identical twins with different educations don’t have identical ability; the more educated twin is usually the smarter twin.[x]  Season of birth is not random; it correlates with health, region, and possibly income.[xi]  On closer look, the supposed fruits of U.S. compulsory attendance laws mask unrelated regional trends, especially in the South.[xii]   None of this means quasi-experimental studies of the education premium are worthless, or their critics invariably on target.[xiii]  But compared to directly measuring pre-existing ability, such studies are speculative and unconvincing.  Since the cleanest approach reveals hefty ability bias, and the messy alternatives yield mixed results, we should reject the Card Consensus in favor of the common-sense view that ability bias is all too real.

[i] For a summary, see Card 1999, p.1855.  Card’s article currently has over 3,500 citations.  See also Card 2001.  For approachable reviews, see Angrist and Pischke 2015, pp.209-239, and Oreopoulos and Petronijevic 2013.

[ii] Lindahl and Krueger 2001, p.1106.  Alan Krueger and David Card have repeatedly collaborated, but most of their education research is not co-authored.

[iii] Greenstone and Looney 2011, p.5.

[iv] Perhaps most notably, Georgetown’s Center on Education and the Workforce has published a series of policy analyses implicitly setting ability bias at 0%; see especially Carnevale and Rose 2011 and Carnevale et al. 2011.  The United States Census does the same; see e.g. Julian and Kominski 2011, 2012.

[v] Card’s 1999, p.1834 otherwise exhaustive literature review explicitly makes this choice: “One strand of literature that I do not consider are studies of the return to schooling that attempt to control for ability using observed test scores.”

[vi] Angrist and Pischke 2015, p.213.

[vii] See Card 1999, pp.1846-1852, and Angrist and Pischke 2015, pp.219-222.

[viii] See Card 1999, pp.1837-1838, and Angrist and Pischke 2015, pp.228-234.

[ix] See Angrist and Pischke 2015, pp.223-227, and Oreopoulos and Salvanes 2011.

[x] Sandewall et al. 2014, Bound and Solon 1999, and Neumark 1999.


[xi] Bound, Jaeger, and Baker 1995, pp.446-447.

[xii] Stephens and Yang 2014, esp. pp.1784-1788.  On p.1789, the authors note that quasi-experimental studies of compulsory attendance laws outside the United States detect little or no payoff.


[xiii] Ashenfelter et al. 1999 also discovers signs that quasi-experimental studies reporting larger benefits of education are more likely to be published.