A large portion of the rise in the education premium can be explained
by a signaling theory of education which predicts that in the future,
increases in the education level of the workforce will actually cause
the education premium to rise, simply because different workers are being
labeled as “highly educated”. This prediction is supported by past
behavior of the high school education premium. It runs counter to the
view that increases in the relative supply of high education workers will
always lower education’s relative price. Suppose education does not affect
an individual’s productivity, but acts only as a signal of it because
individuals select education based on their productivity, and wages are
determined by productivity. It is shown that this implies additional education
in the economy would not change the wage distribution. The
education premium, or relative price of highly educated workers, is the
ratio of mean high education wages to mean low education wages. If all
workers gained more education, it would mean the “bar” (or productivity
minimum) for a given level of education was being lowered. For example,
suppose “highly educated” referred to a college education. If there were
few college grads, lowering the bar (the most productive non-college grads
becoming college grads) would reduce mean college wages significantly by
adding lower productivity workers. Because there would be many noncollege
grads vs. college grads, a drop in the bar would cause a smaller fall
in the mean non-college graduate wage by removing the most productive
workers. It is shown that this implies the education premium would fall.
However, if the bar was low enough so that there were many college grads
and few non-college grads, the reverse would happen and further declines
in the bar would cause education’s relative price to rise. This effect would
not be due to real changes, but to changes in labeling. To measure how
large this effect could have been, simulations were done to create counterfactual
education premiums for three definitions of “highly educated”:
(1) those with a college degree; (2) those with some college education;
(3) those with a high school education. Premiums were created for the
Census years 1950-2000 that hold the wage distribution the same as the
previous decade, but allow the distribution of education across wage ranks
to be the from the present year. These show what the premiums would
have been if wages didn’t change but education levels changed as in the
data. The simulations for (1) and (2) perform as expected: the simulated
premiums fall when there are more high education individuals, and this
can explain some or all of the observed changes in the education premium
between the past six decades of census data. However, (3) also acts as the
model predicts: because this definition has many more highly educated
individuals, further increases in the supply of highly educated individuals
lower the counterfactual premium. Thus, this model predicts that as the
number of college graduates rises, additional grads will eventually cause
the premium to increase.