Prof. Bryan Caplan

Econ 812


Week 12: Labor Economics

I.             Human Capital Theory

A.           I assume you are all familiar with the calculation of present discounted values, or PDVs.  Recall that the lower interest rates are, the more future benefits count.

B.           While PDVs are most-frequently calculated for businesses, the idea is completely general.  You can calculate the PDV of adding insulation to your home.

C.           Similarly, you can calculate the PDV of attending school.

D.           This is the key intuition behind human capital theory.  We can think about labor market decisions like any other investment.

E.           Ex: Should you get another year of school?  Add up the PDV of your foregone earnings during school and the extra income you expect to get after you've completed the schooling.

1.            Note: Since you forego earnings first, and get a raise afterwards, education makes less and less sense as interest rates rise.

F.            What else can you do for your career, and how do you decide if they are good investments?

1.            Co-writing a paper with a faculty member

2.            Putting your cv on fancy paper

3.            A computer projector

II.            The Return to Education

A.           An enormous empirical literature tries to estimate the return to education. 

B.           Underlying motivation: Many economists see credit market imperfections as a serious problem, especially if there is no obvious collateral.  An unusually high rate of return to education would confirm their suspicions.

C.           So how do you calculate the return to education?  Basic estimates start with an assumption that makes analysis highly tractable: Foregone earnings are the ONLY cost of education. 

D.           Then ignoring finite lifespan, a regression of log earnings on a constant and years of education gives you a rate of return estimate.  Just look at the coefficient on education.  A coefficient of .1 indicates that a year of education raises earnings by 10%.  In other words, if you give up one year of income, you earn 10% extra every year thereafter - just like a consol.

E.           Using this approach on NLSY data, you get an estimated 12.6% real rate of return to education (controlling for no other factors).

F.            But this number is surely too high:

1.            You do not reap the benefits of increased earnings forever.  This is a slight effect, since the lost years are far in the future.  Return drops to 12.56%

2.            It costs resources to educate people.  Counting these costs drastically reduces the rate of return.  With annual tuition of $15,000, estimated return falls to 6.5%.

3.            There is also a return to experience; you have the subtract this rate from the return to education to figure out how much extra you get if you go to school instead of work.

4.            This is an estimate of the average, not the marginal rate of return.  (The marginal rate would be lower.  Can you explain why?)

5.            The estimate tacitly assumes school completion probability is 100%, when it’s actually far lower.

6.            It does not control for intelligence, which is highly correlated with education.

III.          Intelligence and Human Capital

A.           We all have an intuitive notion of what is means to be "intelligent."  Empirical research on intelligence is one of the best-developed areas of psychology.

B.           In practical terms, researchers usually measure intelligence with IQ (Intelligence Quotient) or related tests.  These tests have come under angry attack on a number of grounds.  We'll briefly consider each in turn:

1.            Cultural bias

2.            "There is no one thing that constitutes 'intelligence.'"

3.            Imperfection

C.           Complaint #1: "Cultural bias."  There are large group differences in performance on IQ tests.  Jews do about 1 SD better than average, blacks about 1 SD worse.  Critics blame this on cultural bias - supposedly, the tests measure familiarity with middle-class lifestyles rather than ability.  Unfortunately for this argument, it has been carefully tested and shown to be wrong.  If you use IQ tests to predict performance on practical tasks – like ability to drive a tank through an obstacle course – IQ tests actually overstate the performance of members of groups with low average IQs. 

D.           Complaint #2: "There is no one thing that constitutes 'intelligence.'"  Everyone is good at some things and bad at others, or so the claim goes.  Still, the fact is that for a wide range of mental problems, people who are good at some are usually (not always) good at all of them, and vice versa.  Think about the SAT Verbal versus Math scores.  There are some people who are great at Verbal and terrible at Math, but there are a lot more who are great at both or terrible at both.

E.           Complaint #3: Imperfection.  There are several varieties of this complaint.  One is that the same person has received very different test scores at different times.  Another is that world-renowned geniuses (Feynman is a common example) got low IQ scores.  All this may be true, but it's irrelevant.  IQ scores are more reliable than anything else, and if you tested 100 geniuses their average score would be very high.

F.            Intelligence is a lot like "strength."  There is some ambiguity, but at root we know what we mean, we know there are real differences, and we know that people who are strong by one measure are usually strong by other measures, too.

G.           There is a second debate about the extent to which IQ is hereditary or environmental.  There is no time to resolve this here, but evidence from carefully-constructed twin and adoption studies finds that the variance is about 80% genetic.  Unclear where the remaining 20% comes from - it doesn't seem to be family environment.

H.           Why do I bring all this up?  Because controlling for IQ sharply reduces the measured return to education to a mere 7.5%.  (1 extra percentile of IQ bumps you up .7%; a year of education is thus worth about as much as 11 percentiles of IQ).

I.             Estimated return with $15,000 tuition drops to 3%.

IV.          Signaling and the Social Rate of Return

A.           Main idea of credit market imperfections: social return exceeds private return.

B.           The empirical case in the NLSY looks quite weak once you make a few obvious adjustments.  (Of course, some might simply say that the case is weak precisely because governments already so heavily subsidize education).

C.           All of these calculations assume, though, that education actually increases productivity and thereby raises social output.  But recall that there is a competing hypothesis: signaling.

D.           Insofar as education is signaling, when one worker becomes more educated, his wages go up.  But at the same time, all other workers look relatively worse, and their wages go down.  The effect on productivity of additional signaling is zero.

E.           Implication: the previous estimates only show the private rate of return.  The social return will be lower.

F.            If 50% of education's effect is signaling, the estimated rate of return falls to -.3%!  If it is 90% signaling, it falls to -5.5%.

G.           Note that there is a simple policy government could use to improve the market's efficiency: taxing education.  In the signaling model, education wastes real resources.  Taxing education would preserve the relative ranking but use fewer resources.

H.           In reality, of course, governments almost always massively subsidize education.

I.             If education were unsubsidized, you might not be able to afford it; but then you probably wouldn't need it to get a good job either.  Firms would switch to apprenticing and other ways to find out your "type."

V.           Nominal Rigidities

A.           One unusual feature of labor markets that has often been discussed is nominal rigidities.  Even though labor seems to satisfy the assumptions of perfect competition quite well, nominal wages rarely fall even in the face of surplus labor.

B.           Neoclassical theory does not rule this out.  Nominal rigidity could exist simply because of menu costs.

C.           But menu costs seem pretty small relative to the value of the product.  This has led behavioral economists to blame it on money illusion and/or fairness.

D.           Evidence: Numerous psychological studies indicate that most people have money illusion to some degree. 

E.           Even when you make the point explicit, respondents evaluate employers' "fairness" partly in nominal terms.  In one study, people were asked to evaluate two firms' behavior when both are making "small" profits. 



Inf=0, Raise=-7%


Inf=12%, Raise=+5%


F.            Who cares about fairness?  There is also evidence that disgruntled workers' performance worsens.  "Wage cuts hurt morale."  Effort is partly about incentives and partly about trust.

G.           Ask employers: How do workers respond to wage cuts versus layoffs?

1.            The UC Berkeley pay cut.

H.           Note: Nominal rigidities could potentially be corrected by simply inflating them away.  In practice, of course, this is more easily said than done.

VI.          Efficiency Wages

A.           Unpleasant working conditions in an occupation decrease labor supply and raise wages.  This wage premium is generally known as a "compensating differential."

B.           With symmetric information, then, employers can induce workers to work harder by paying them more, and markets still clear.

C.           However, with asymmetric information, matters are more complex.  What happens if workers know more about their effort level than their employer does?

D.           Employers might threaten to fire you if they catch you shirking, but in competitive markets, the fired worker can immediately get a job just as good as his last job.

E.           So what might employers do?  They might raise workers' pay above the market-clearing level in order to make the threat to fire them serious.  That way, if they get fired, it will be hard for them to find a job that is just as good as the one they lost.

F.            What happens if all employers think this way?  Then everyone raises wages above the market-clearing level, and a permanent labor surplus emerges.

G.           If you hire the unemployed workers at a lower wage, then given certain assumptions, their performance falls faster than the wage.  This makes them unemployable, even though they are identical to the employed workers.

H.           Note that this is a real model.  Inflation raises the equilibrium nominal efficiency wage 1:1.

I.             Some economists use the efficiency wage model to argue for industrial policy.  You can increase total output by taxing the employed to subsidize jobs for the unemployed.

J.            However, the efficiency wage problem can also be mitigated by simply making unemployment less pleasant.  So it could just as easily be seen as an argument against unemployment insurance, welfare, etc.