Professional Documents
Culture Documents
Daron Acemoglu
Journal of Economic Literature, Vol. 40, No. 1. (Mar., 2002), pp. 7-72.
Stable URL:
http://links.jstor.org/sici?sici=0022-0515%28200203%2940%3A1%3C7%3ATCIATL%3E2.0.CO%3B2-P
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained
prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in
the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/journals/aea.html.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic
journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers,
and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take
advantage of advances in technology. For more information regarding JSTOR, please contact support@jstor.org.
http://www.jstor.org
Tue Oct 30 14:10:14 2007
Journal of Economic Literature
Val. XL (Marcla 2002), pp. 7-72
episodes. For example, there were already the idea that technological advances fa-
signs of significant technology-skill vor more skilled workers is a twentieth-
complementarity in the 1910s. Claudia century phenomenon. In nineteenth-
Goldin and Lawrence Katz (1998) argue century Britain, skilled artisans
that the spread of batch and continuous- destroyed weaving, spinning, and
process methods of production in- threshing machines during the Luddite
creased the demand for skills. They and Captain Swing riots, in the belief
add, ". . . the switch to electricity from that the new machines would make
steam and water-power energy sources their skills redundant. They were right:
was reinforcing because it reduced the the artisan shop was replaced by the
demand for unskilled manual workers factory and later by interchangeable
in many hauling, conveying, and assem- parts and the assembly line (e.g., John
bly tasks" (p. 695). Over this period, James and Jonathan Skinner 1985;
capital-intensive industries increased Goldin and Katz 1998). Products pre-
the demand for skills considerably (see viously manufactured by skilled artisans
Goldin and Katz 1998, table 3), and the started to be produced in factories by
scope of these industries expanded with workers with relatively few skills, and
the sharp fall in the price of electricity many previously complex tasks were
(see, for example, Arthur Woolf 1984, simplified, reducing the demand for
p. 178). The rapid increase in the im- skilled workers.2 Joel Mokyr (1990, p.
portance of white-collar and clerical oc- 137) describes this process vividly:
cupations gave another boost to the de- First in firearms, then in clocks, pumps,
mand for skills. Generalizing from the locks, mechanical reapers, typewriters, sew-
experience of the 1920s, Harry Jerome ing machines, and eventually in engines and
(1934, p. 402) argued that ". . . in the bicycles, interchangeable parts technology
proved superior and replaced the skilled
future . . . there is considerable reason
artisans working with chisel and file.
to believe that the effect of further
[mechanization] will be to raise the
"n the absence of detailed econometric evi-
average skill required." dence, it is difficult to eneralize from the histori-
The early twentieth-century evidence cal examples and concfude that technical change
was so powerful that Zvi Griliches was definitely skill-replacing during the nine-
teenth century. Still, the only econometric evi-
(1969) suggested capital and skills are dence that I am aware of supports this view. Usin
intrinsically complementary. Richard the 1850 Census of Manufacturers, James an8
Nelson and Edmund Phelps (1967), Skinner (1985) find that there was more ra id sub-
stitution of capital for skilled workers t&n un-
Finis Welch (1970), Theodore Schultz skilled workers. I t can also be ar ued that techni-
(1975), and Jan Tinbergen (1975) also cal change always increases t%e demand for
argued that technological developments "skills," and the artisans who were hurt as a result
of new technolo y were not "skilled" since they
increase the demand for skills. Events lacked the flexi%ility to adapt to the required
since then support this notion. Personal changes. This argument is not totally convincing,
computers, computer-assisted produc- since the artisans earned considerably more than
other laborers (for example, James and Skinner
tion techniques, and robotics appear to 1985 report over 60 percent wage differentials for
complement skilled workers, replacing building and printing workers relative to laborers
many labor-intensive tasks. In this light, in the 1850s). So the artisans possessed skills that
were being rewarded by the market, and the stan-
it is perhaps natural to view the in- dardization of the production process destroyed
crease in inequality over the past sev- these rewards. On the other hand, it has to be
eral decades as a direct consequence of noted that many of the skill-replacing technologies
of the nineteenth century ma have also increased
technical change. the demand for engineers andmanagers (see, e.g.,
Although the consensus is now broad, Goldin and Katz 1998).
Acernoglu: Technical Change, Inequality, and the Labor Market 9
Interchangeable parts were in fact very an acceleration in skill bias during the
much designed to be skill-replacing past few decades appears to be the main
(unskill-biased). Eli Whitney, a pioneer cause of the increase in inequality.
of interchangeable parts, described the We can understand the behavior of
objective of this technology as: technical change by recognizing that
the development and use of technol-
. . . to substitute correct and effective op-
erations of machinery for the skill of the art-
ogy is, at least in part, a response to
ist which is acquired only by long practice profit incentives.3 When developing
and experience; a species of skill which is not skill-biased techniques is more profit-
possessed in this country to any considerable able, new technology will tend to be
extent. . . . (quoted in H . J. Habakkuk 1962, skill-biased. I suggest that the early
p. 22)
nineteenth century was characterized
The experience of the nineteenth and by skill-replacing developments be-
early twentieth centuries led Harry cause the increased supply of unskilled
Braverman (1974) and Stephen Marglin workers in the English cities (resulting
(1974) to argue that technical change from migration from rural areas and
was "deski1ling"-a major purpose of from Ireland) made the introduction
technical change was to expand the divi- of these technologies profitable. In
sion of labor and simplify tasks pre- contrast, the twentieth century has
viously performed by artisans by break- been characterized by skill-biased
ing them into smaller, less skill-requiring technical change because the rapid in-
pieces. Braverman (1974, p. 113), for ex- crease in the supply of skilled workers
ample, suggested that the first principle has induced the development of skill-
of management and production tech- conlplementary tech~lologies.The re-
niques of the period was "dissociation of cent acceleration in skill-biased tech-
the labor process from skills of the work- nical change is in turn likely to have
ers. The labor process is to be rendered been a response to the rapid increase
independent of craft, tradition, and the in the supply of skills during the past
workers' knowIedge." several decades. However, I also argue
A longer view therefore suggests that that despite the acceleration in skill
technological advances may not have bias, we are most likely not in the
always increased the demand for skills. midst of a "technological revolution";
In fact, most early nineteenth-century what has changed is not necessarily
innovations appear to have replaced the overall rate of progress, but the
skilled workers and expanded tasks per- types of technologies that are being
formed by the unskilled. But then, why developed.
have technological advances been skill-
biased in the twentieth century? And, 3 Precedents of this approach include Jacob
are technological changes the major cause Schmookler (1966), who emphasized demand pidl
and the extent of the market as key determinants
of the recent increase in inequality? of innovations; the endogenous growth theory,
This essay attempts to answer these e.g., Paul Romer (1990), Gene Grosslnan and El-
questions. It has two main theses: hanan Helpman (1991), and Philippe Aghion and
Peter Howitt (1992); the induced innovation the-
ory, including Syed Ahlnad (1965), Charles Ken-
The behavior of wages and returns to nedy (1964), Paul Samuelson (1970), Yujiro
schooling indicates that technical Hayami and Vernon Ruttan (1970), and Paul
change has been skill-biased during the David (1975); and recent work including my own,
Acernoglu (1998, 1999b, 2000), Acelnoglu and
past sixty years and probably for most Fabrizio Zilibotti (1999), and Michael Kiley
of the twentieth century. Furthermore, (1999).
10 Journal of Economic Literature, Vol. XL (March 2002)
was relatively stable before the 1970s, ior of entrepreneurs driven by a variety
because the rate of skill accumulation of nonprofit motives. Demand for skills
in the U.S. economy was as rapid increased faster during the past thirty
as the constant pace of skill-biased years, this approach would maintain,
technical change (e.g., Katz and Kevin because of a teclanological revolution led
Murphy 1992; Murphy, Craig Riddell, by the microchip, personal computers,
and Paul Romer 1998; David Card and and the internet.6 New technologies of
Thomas Lemieux 2000). The recent in- the early nineteenth century were skill-
crease in inequality is then explained, replacing (unskill-biased) because the
not by a major technological change, technological frontier then only en-
but by a decline in the growth rate of abled the invention of skill-replacing
the supply of skills. The second possible techniques.
answer comes from the acceleration hy- Nevertheless, there are several prob-
potlaesis, which maintains that there has lems with this approach. First, although
been an acceleration in skill bias begin- a number of papers, including Green-
ning in the 1970s or the 1980s. Accord- wood and Yorukoglu (1997), Andres
ing to this hypothesis, there is a dis- Hornstein and Krusell (1997), and Ga-
continuity in the growth rate of the lor and Moav (2000), show that rapid
demand for skills. The most popular technical change may lead to slower to-
version of this hypothesis claims that tal factor productivity (TFP) growth,
there has been a notable acceleration in the slow rates of TFP and output
the skill bias of technology, driven by growth of the past several decades are
advances in information technology, or difficult to reconcile with a technologi-
perhaps a "Third Industrial Revolution." cal revolution during this time period.
So was there an acceleration in skill Second, demand for skills appears to
bias? This question is difficult to an- have accelerated starting in the late
swer, as we lack direct measures of the 1970s, precisely when the supply of
degree of skill bias. To tackle this ques- skills increased very rapidly. Exogenous
tion, one therefore needs to look at a technology theories do not explain the
variety of evidence often pointing in timing of this acceleration.7
different directions. I conclude below 6See, among others, Krueger (1993), Eli Ber-
that skill-biased technical change is man, John Bound, and Griliches (1994), and David
likely to have accelerated over the past Autor, Katz, and Krueger (1998) for evidence that
several decades. This conclusion is the rapid spread of computers has increased the
demand for skills. See Per Krusell, Lee Ohanian,
based on the sharp increase in overall Victor Rios-Rull, and Giovanni Violante (2000),
inequality starting in the 1970s and on Oded Galor and Daniel Tsiddon (1997), Green-
the fact that returns to schooling rose wood and Yorukoglu (1997), Aghion and Howitt
(1998, ch. 9), Caselli (1999), Galor and Olner
over the past thirty years despite the Moav (2000), Violante (2000), Yonna Rubinstein
unusually rapid increase in the supply and Tsiddon (1999), Aghion, Howitt, and Violante
of educated workers. (2000), and Eric Gould, Moav, and Bruce Wein-
berg (2000) for models in which ra id technical
Why did the demand for skills accel- f
change increases the demand for ski1 s and causes
erate over this ~ e r i o d ?And why has a rise in inequality.
new technology favored more skilled 7Naturally, supply and demand ma 7 have
moved together because supply responded to de-
workers throughout the twentieth cen- mand. I argue below that the large increase in the
tury, but not during the nineteenth cen- supply of educated workers was not in anticipation
tury? One approach would view tech- of, or in response to, high returns, but driven by a
variety of other factors. More generally, I often
nology as exogenous, stemming from focus on the effect of the supply of skills on tech-
advances in science or from the behav- nology not because I view supply as exogenous,
12 Journal of Economic Literature, Vok. XL (March 2002)
that they do not provide a natural expla- zational change, labor-market institu-
nation for the fall in the wages of tions, and international trade have in-
low-skill workers. Although a number of teracted with technical change in a fun-
papers, including Caselli (1999), Green- damental way, amplifying the direct
wood and Yorukoglu (1997), and Galor effect of technical change on inequality,
and Moav (2000), show that technologi- and likely causing the decline in the
cal revolutions may be associated with a wages of less-skilled workers.
fall in the wages of low-skill workers, it Therefore, the overall picture that
is difficult to see how sustained techno- emerges is not necessarily one in which
logical change can be associated with technology is the only factor affecting
an extended period of falling wages of the distribution of income. On the con-
low-skill workers and stagnant average trary, the underlying thesis of this essay
wages. This leads to the next question is that technology itself is no more than
for this essay. an endogenous actor. To explain the
Why did the real wages of low-skill changes in the distribution of income,
workers fall over the past several de- and to forecast what other changes may
cades? There are a number of possible happen in the future, we need to under-
answers. First, labor-market institu- stand the forces that shape technologi-
tions, for example labor unions, under- cal progress, and how technology inter-
went important changes over the past acts with the overall organization of the
thirty years, and these changes may labor market.
have reduced the wages of many manu- There is considerable uncertainty on
facturing workers, causing an increase many issues, and both more theoretical
in inequality and a decline in the real and more empirical work are needed.
wages of low-skill workers (e.g., Rich- Two areas deserve special attention.
ard Freeman 1991; John DiNardo, Ni- The first is the differential behavior of
cole Fortin, and Lemieux 1995; David residual inequality and returns to
Lee 1999). Second, international trade schooling during the 1970s. Most
between skill-scarce less-developed coun- economists view changes in residual in-
tries and skill-abundant rich economies equality as related to changes in labor-
increased over this period, and this may market prices. It is therefore puzzling
have put downward pressure on the that during the 1970s, while returns to
wages of low-skill workers in the United schooling fell, residual and overall in-
States (e.g., Adrian Wood 1994; Edward equality increased. I argue below that
Leamer 1995). Third, there has been a models based on a single skill index
transformation of the way in which (one type of skill or many types of skills
firms are organized, or perhaps in the that are perfect substitutes) are unable
way that firms and workers match (see, to explain this pattern. Instead, we need
for example, Acemoglu 1999a; Michael models with multidimensional skills.
Kremer and Eric Maskin 1999; Timothy Moreover, for this type of model to ex-
Bresnahan 1997; Bresnahan, Erik Bryn- plain the behavior of residual inequality
jolfsson, and Lorin Hitt 1999; and during the 1970s and the 1980s, techno-
Autor, Frank Levy, and Richard Mur- logical progress needs to have changed
nane 2000). Although each of these fac- the demand for different types of skills
tors could have been the cause of the differentially. The endogenous technol-
recent changes in the wage structure, I ogy models discussed above provide a
argue that their direct effect has been possible explanation for why different
limited. Instead, I suggest that organi- dimensions of skills may have been
14 Journal of Economic Literature, V o l . X L ( M a r c h 2002)
1 I I I I 1 I
39 49 59 69 79 89 96
year
Relative Supply of College Skills and College Premium
Figure 1.The Behavior of the (log) College Premium and Relative Supply of College Skills (weeks worked by
college equivalents divided by weeks worked of noncollege equivalents) between 1939 and 1996. Data from
March CPSs and 1940, 1950, and 1960 censuses.
very sharply during the 1980s. This in- illustrates two more important patterns:
crease in the returns to schooling has 1. Overall wage inequality started to
been one of the major motivating facts increase sharply in the early 1970s after
for the empirical inequality literature a period of relative stability-prior to
(e.g., Bound and Johnson 1992; Katz the 1970s, the 90th, SOth, and 10th per-
and Murphy 1992). centiles of the wage distribution fol-
There have also been important lowed each other closely, but came
changes in the overall distribution of apart sharply in the 1970s.
wages. Figure 2 plots the 90th, 50th 2. Median wages stagnated from 1975
and 10th percentiles of the overall onwards, while workers at the 10th
wage (weekly earnings) distribution percentile of the wage distribution (i.e.,
for white male workers between 1963
and 1997 (with the 1963 values for this would likely contribute to the composition ef-
all series indexed to lOO).9 This figure fects. Moreover, male-female wage difference nar-
rowed substantially over the same time period as
"ample constructed as described in the appen- well. School quality for black men also under-
dix. I focus here on wage inequality for white men went significant transformation (e.g., Welch 1973;
since labor market partici ation of women in- Card and Krueger 1992), and this could create
!'
creased substantially over t l e sample period, and significant composition effects.
16 Jozcrnal of Economic Literature, V ol. XL ( M a r c h 2002)
150 1
year
Indexed Wages for White Males 1963-97
Figure 2. Changes in the Indexed Value of the 90th, 50th, and 10th Percentiles of the Wage Distribution for
White Males (1963 values normalized to 100).Data from March CPSs.
I I I I I I I
65 70 75 80 85 90 95
year
Residual Inequality Measures for White Males 1963-97
Figure 3. 90-50, 50-10 and 0.5 x 90-10 Differentials from Log \lJeekly Wage Regressions for White Males
Aged 18-65.
This is where I begin as well. I intro- where p 5 1, and Al(t) and Ar,(t) are factor-
duce a simple framework that links augmenting technology terms. Although
wages to supply of skills and to demand all the results of interest hold for a general
generated by the technology possibili- constant returns to scale F ( . ) function, I
ties frontier of the economy. There are focus on the CES production function to
two types of workers, skilled and un- simplify the discussion. I drop the time
skilled (high and low-education work- argument when this causes no confusion.
ers), who are imperfect substitutes. Im- T h e elasticity of substitution between
perfect substitution between the two skilled and unskilled workers in this
types of workers is important in under- production function is o = 1/(1- p). I
standing how changes in relative sup- refer to skilled and unskilled workers as
plies affect skill premia. For now I gross substitutes when the elasticity of
think of the unskilled workers as those substitution o > 1 (or p > O), and gross
with a high school diploma, and the complements when o < 1 (or p < 0).
skilled workers as those with a college Three noteworthy special cases are: (i)
degree.l"o the focus in this section is o + 0 (or p + -DO) when skilled and un-
on returns to schooling (or between- skilled workers will be Leontieff, and
group inequality), and I use the terms output can be produced only by using
"skill" and education interchangeably. skilled and unskilled workers in fixed
In practice, education and skills are proportions; (ii) o + when skilled and
only iinperfectly correlated, so it is use- unskilled workers are perfect substi-
ful to bear in mind that since tliere are tutes, and (iii) o + 1, when the pro-
skilled and unskilled workers within the duction function tends to the Cobb
same education group, an increase in Douglas case. The value of the elasticity
of substitution will play a crucial role in
12This classification of workers into skilled and the interpretation of the results that fol-
unskilled groups is obviously appropriate only for
the twentieth-century United States. In many low. I n particular, in this framework,
other countries or eriods, hi h school graduates teclinologies either increase the pro-
could b e considere%to be "skiyled" workers. More ductivity of skilled or unskilled workers,
generally, the two-skill model is a convenient
simplification, less realistic than a world wit11 a that is, tliere are no explicitly skill-
continuum of imperfectly substitutable skills. replacing or unskilled-labor-replacing
Acemoglu: Technical Change, Inequality, and the Labor Market 19
technologies.13 But, as we will see be- This equation implies au;~/aH/L > 0: as
low, depending on tlie value of the elas- the fraction of skilled workers in the
ticity of substitution, an increase in A17 labor force increases, the unskilled wage
can act either to complement or to should increase. Similarly, the skilled
"replace" skilled workers. wage is
Tlie production function (2) admits
three different interpretations:
1. There is only one good, and
skilled and unskilled workers are iinper- which yields a t u ~ / a H / L< 0; everything
fect substitutes in tlie production of tliis else equal, as skilled workers become
good. more abundant, their wages should fall.
2. The production function (2) is also Combining these two equations, the skill
equivalent to an economy where con- premium-the wage of skilled work-
sumers have utility function [Ye + Y;]~/P ers divided by the wage of unskilled
defined over two goods. Good YI, is pro- workers-is14
duced using only skilled workers, and Y1
is produced using only unskilled work-
ers, with production functions YI, = AhH,
and Yl =AlL. In tliis interpretation, it is
important that the economy is closed.
We will see in section 6.3 that different
results obtain when international trade Equation (4) can be rewritten in a more
in tliese two goods is allowed. convenient form by taking logs,
3. A mixture of the above two
whereby different sectors produce
goods that are imperfect substitutes,
and liigh- and low-education workers are
employed in both sectors. Naturally, the skill premium in-
Although the third interpretation is creases wlien skilled workers become
more realistic, I generally use one of the more scarce, i.e.,
first two, as they are easier to discuss.
Since labor markets are competitive,
the unskilled wage is
This is the usual substitution effect, and
shows tliat for given skill bias of technol-
ogy, as captured by Ah/&, the relative
demand curve for skill is downward slop-
ing with elasticity l/o = (1- p). Intui-
13A more general formulation would replace
equation (2) with the production function tively, an increase in HIL creates two
different types of substitution. First, if
14For some parameter values, skilled workers
nlay have lower wages than the unskilled, i.e.,
where B , and Bi, would b e directly unskilled-labor- o 5 1. One may want to impose
replacing and skill-replacing technologies, and an H
increase in b, would correspond to some of the
tasks previously performed by the unskilled being
taken over by the skilled (see, e.g., Johnson and
Stafford 1999 on this). For most of the analysis
here, there is little to b e gained from this more
r
to avoid this. Alternative1 , one could assume that
skilled workers can use t l e technologies nornlally
used by the unskilled, A,, and be more (or equally)
general production function (but see section 5.3). productive at this than the unskilled.
. . . .
20
.
f
Skill premium
. \ Journal of Economic Literature, Vol. X L (March 2002)
..........................
Relative demand
for skills
college graduate workers (relative to
high school graduates) tliat were as high
as 40 percent quickly fell to less tlian 20
percent. The extent of substitution was
also clear. First, many college graduate
workers could not find employment in
skilled jobs: Angrist (1995) shows a
I I
sharp increase in the unemployment
rn' ........................... .......------
I rate of college graduates, and Ze'ev
I A l l - b i a s e d tech. Schiff and E h u d Ya'ari (1989) report
I
change
I
that only one in eiglit Palestinian gradu-
H/L Hf/L'
ates could find work in liis profession,
Figure 4. The Relative Demand for Skills witli the rest working as unskilled labor-
ers, mainly in the construction industry.
Second, preinia for tasks usually per-
skilled and unskilled workers are produc- formed by more educated workers fell
ing the same good, but performing dif- sharply. Between 1984 and 1987, the
ferent functions, an increase in the num- preiniuin for administrative and inana-
ber of skilled workers will necessitate a gerial jobs (relative to manual laborers)
substitution of skilled workers for tasks fell from .32 to .12, while the preiniuin
previously performed by tlie unskilled. for clerical workers fell from .02 to -.08
Second, if skilled and unskilled work- (see Angrist 1995 for details).
ers are ~ r o d u c i n gdifferent goods, tlie As equation (6) sliows, the elasticity
greater number of skilled workers will of substitution, o, is important for the
lead to a substitution of the consumption behavior of tlie skill premium when
of tlie unskilled good by the skilled good. supply changes. The elasticity of substi-
In botli cases, this substitution hurts the tution is also crucial for the response of
relative earnings of skilled workers. the skill premium to changes in tech-
Figure 4 draws tlie relative demand nology. Unfortunately, this parameter is
for skills, equation (5), against the rela- rather difficult to estimate, since it re-
tive supply of skills, H I L , which is taken fers to an elasticity of substitution that
to b e given for the purposes of this ex- combines substitution both within and
ercise. An increase in the relative sup- across industries. Nevertlieless, there
ply, from H I L to H'IL', moves tlie equi- are a number of estimates using aggre-
librium point along the downward gate data that give a range of plausible
sloping relative demand curve, and values. The majority of these estimates
reduces tlie skill premium from OI to of. are between o = 1 and 2 (see for exain-
An interesting case study of the re- ple Freeman 1986). The response of
sponse of the returns to schooling to an college preiniuin for Palestinian labor
increase in the supply of skills is pro- reported in Angrist (1995), for example,
vided by the experience in the West implies an elasticity of substitution be-
Bank and Gaza Strip during the 1980s. tween workers with sixteen years of
As Josliua Angrist (1995) illustrates, scliooling and tliose witli less than
there was a very large increase in tlie twelve years of schooling of approxi-
supply of skilled Palestinian labor as mately o = 2.
Palestinian institutions of liigher educa- Given the focus of tliis essay, it is also
tion, totally absent before 1972, began important to know how the skill premium
to open. Angrist shows that premia to responds to technology. Differentiation
Acemoglu: Technical Change, Inequality, and the Labor Market 21
of (5) shows that tlie result depends on elasticity of substitution between skilled
tlie elasticity of substitution: and unskilled workers greater than 1.
I t is also useful to compute the aver-
age wage in tliis economy. Without con-
trolling for changes in tlie educational
composition of the labor force, tlie
Therefore, if o > 1 (i.e., p E (0, I]), then average wage is
improvements in the skill-complementary
technology increase the skill premium.
This can be seen in figure 4 as a sliift out
of the relative demand curve, moving the
wliich is also increasing in H I L as long as
skill premium from o to o". The con-
the skill premium is positive (i.e., o > 1
verse is obtained when o < 1: tliat is,
or A;(H/L)P - A: > 0). Intuitively, as the
wlien o < 1, an improvement in the pro-
skill composition of the labor force
ductivity of skilled workers, Ah, relative
improves, wages will increase.
to tlie productivity of unskilled workers,
Therefore, tlie results so far imply
Al, sliifts the relative demand curve in
that in response to an increase in H I L :
and reduces tlie skill premium. This case
appears paradoxical at first, but is, in 1. The relative wages of skilled workers,
fact, quite intuitive. Consider, for ex- the skill premium o = Z U ~ / Z decreases.
O~,
ample, a Leontieff (fixed proportions) 2. Wages of unskilled workers increase.
production function. In this case, wlien 3. Wages of skilled workers decrease.
A!, increases and skilled workers become 4. The average wage (without control-
more productive, the deinand for un- ling for education) rises.
skilled workers increases by more than
the deinand for skilled workers. In some It is also useful to liighlight tlie iinpli-
sense, in this case, the increase in Ah is cations of an increase in Ah on wage
creating an "excess supply" of skilled levels. First, an increase in A!,, with A1
workers given the number of unskilled constant, corresponds to an increase in
workers. This excess supply increases the Ah/Al; tlie implications of this change
unskilled wage relative to the skilled on the skill premium were discussed
wage. Tliis observation raises an iinpor- above. Moreover if Ah increases, every-
tant caveat. I t is tempting to interpret tliing else being equal, we expect both
improvements in teclinologies used by tlie wages of unskilled and skilled work-
skilled workers, A!,, as "skill-biased." ers (and therefore the average wage) to
However, when the elasticity of substitu- increase: in tliis framework, technological
tion is less tlian 1, it will be advances in improvements always increase all wages.
technologies used with unskilled work- T h e most central result for our pur-
ers, Al, that increase the relative produc- poses is tliat as H I L increases, the skill
tivity and wages of skilled workers, and premium, a, should fall. In terms of
an increase in Ah relative to Al will be figure 4, tlie increase in supply cor-
"skill-replacing." responds to a rightward shift in tlie ver-
Nevertlieless, the conventional wis- tical line from H I L to H'/L', which
dom is that the skill premium increases would move the econoiny along tlie
wlien skilled workers become relatively downward sloping demand curve for
more-not relatively less-productive, skills. But this tendency of tlie skill pre-
which is consistent with o > 1. I n fact, mium to fall could be counteracted by
as noted above, most estimates sliow an changes in technology, as captured by
22 Journal of Economic Literature, Vol. XL (March 2002)
ln (Al,/Al). Therefore, this simple for- of wages and employment. Autor, Katz,
i$ulation encapsulates the essence of and Krueger (1998) report employment
tlie two forces that Tinbergen (1975) and wage-bill shares for different
emphasized: groups of workers in their appendix ta-
The two preponderant forces at work are ble A l . If we assume a specific value for
technological development, which made for o , we can translate tliese numbers into
a relative increase in demand and hence in clianges in Al,/Al. In particular, notice
the income ratio . . . and inci-eased access that the relative wage bill of skilled
to schooling, \vhich made for a relative workers is given by
decrease. ( p . 35, italics in original)
TABLE 1
EMPLOYMENT
SHARES
AND SKILL-BIASED
TECHNICAL 1940-1990
CHANGE,
1940 0.004 0.21 0.016 0.31 0.035 0.38 0.140 0.37 0.303 ,055 0.392 0.63
1950 0.006 0.24 0.011 0.28 0.030 0.37 0.146 0.38 0.219 0.47 0.313 0.56
1960 0.013 0.29 0.030 0.37 0.080 0.48 0.189 0.43 0.343 0.59 0.476 0.69
1970 0.017 0.32 0.069 0.47 0.179 0.61 0.199 0.45 0.485 0.70 0.652 0.81
1980 0.042 0.40 0.157 0.59 0.486 0.81 0.270 0.52 0.643 0.80 0.933 0.97
1990 0.090 0.50 0.470 0.81 1.777 1.18 0.357 0.60 1.064 1.03 1.673 1.29
Note: The first panel gives the ratio of the employment of skilled relative to unskilled workers, and the wage bill of
skilled to unskilled workers for the corresponding skill categories. These data are taken from Autor, Katz, and
Krueger (1998). Some college refers to those with more than high school (hence the measure is those with more
than high school divided by those with high school or less). College graduate refers to all those with a college
degree, and college equivalent is defined, as in Autor et al., as those with a college degree + 0.5 x those with some
college (correspondingly, the unskilled are defined as those with high school and less + 0.5 x those wit11 some
college). The bottom panel gives the implied technology shifts usi;glequation (9) above for different values of the
elasticity of substitution. The demand index D is defined as ( A j r / A l ) ~
employed in all sectors-and the sec- complementarity during tlie 1910s and
toral composition should adjust in order 1920s. In light of this evidence, one
to clear the market. The evidence is very might consider the bulk of tlie twenti-
much the opposite. Berman, Bound, eth century to have been characterized
and Griliches (1994) and Murphy and by skill-biased technical change, though
Welch (1993) show a steady increase in whether technical cliange during the
the share of college labor in all sectors. early twentieth century was or was not
This discussion leads to my first con- skill-biased is not central for tlie focus
clusion, which I highlight for future of this paper.
reference.
4. Steady- Demand and Acceleration
Conclusion 1. The past sixty years must Hypotheses
have been characterized by skill-biased
technical change. The previous section liighlighted the
importance of skill-biased technical
Furthermore, Goldin and Katz (1998) cliange over the past several decades. A
provide evidence of teclinology-skill first liypothesis is that skill-biased
24 Journal of Economic Literature, Vol. X L ( M a r c h 2002)
the college wage-bill share is related to the demand for skills, since industries
the demand for skills as shown by equa- that are highly computerized may de-
tion (9). The results indicate that in an mand more skilled workers for other
industry where computer use increased reasons as we11.18 In fact, when Autor,
by 10 percent, the college wage-bill Katz, and Krueger (1998) run the above
share grew by about 0.15 percent a year regressions for the 1960-70 college
faster between 1980 and 1990, and 0.3 wage-bill shares, they obtain
percent faster a year between 1990 and
1996.
Although this evidence is suggestive,
it does not establish that there has been Therefore, industries investing more in
a change in the trend growth of skill- computers during the 1980s were al-
biased technology. As pointed out in ready experiencing more skill upgrading
conclusion 1 above, the only way to during the 1960s, before the spread of
make sense of the post-war trends is to computers (though perhaps slower than
incorporate skill-biased technical change after the 1980s, since the coefficient
over the whole period. Moreover, here is about half of that between 1980
Goldin, and Katz (1998) present evi- and 1990). This suggests that at least
dence suggesting that capital-skill com- part of the increase in the demand for
plementarity may have been as high skills coming from highly computerized
during the 1910s as during the recent industries may not be the direct effect of
period because of increased demand for computers, but reflect a secular long-run
skills coming from the introduction of shift toward more skilled workers. So
electricity in most manufacturing pro- faster skill upgrading by highly comput-
cesses. Similarly, even though there erized industries is not inconsistent with
were few computers in workplaces the steady-demand hypothesis.
before the 1970s, other technological The third, and probably most power-
developments may have increased the ful, piece of evidence in favor of an ac-
demand for skills as rapidly as-or more celeration in skill bias also comes from
rapidly than-computers. Therefore, the Autor, Katz, and Krueger (1998). They
question is whether computers and the document that the supply of skills grew
associated information technology ad- faster between 1970 and 1995 than be-
vances increased the demand for skills tween 1940 and 1970-by 3.06 percent
more than other technologies did dur- a year during the latter period com-
ing the 1950s and the 1960s, or even pared to 2.36 percent a year during the
earlier. This question cannot be an- earlier thirty years. In contrast, the col-
swered by documenting that computer- lege premium increased between 1970
ized industries demand more skilled and 1995 by about 0.39 percent a year,
workers.17 while it fell by about 0.11 percent a
Cross-industry studies also may not year during the earlier period. If de-
reveal the true impact of computers on mand for skills had increased at a steady
pace, the college premium should have
17This argument is related to a point first raised
by Lawrence Mishel and Jared Bernstein (1994). 18 Mark Dorns, Timothy Dunne, and Kenneth
They ointed out that rnuch of the evidence pre- Troske (1997) show that new technolo ies (but
s e n t e l i n favor of the impact of technolo on in-
equality shows that there has been skZK-biased
1
not con~puters)are adopted by plants t at have
rnore skilled and inore highly aid workers, and
technical change during the 1980s, not that skill- ants do not increase tgeir wages or de-
biased technical change has accelerated relative to
earlier periods.
btP
:id ~ d skills
r
technologies.
after the ilnplenlentation of these
Acemoglu: Technical Change, Inequality, and the Labor Market 27
also fallen since 1970.1Qoreover, more flexible time trends show an accel-
Autor, Katz, and Krueger (1998) docu- eration in the relative demand for skills
ment greater within-industry skill up- during the 1970s or the 1980s (the cubic
grading in the 1970s, 1980s, and 1990s and quartic time trends are almost iden-
than in t h e 1960s, which is also con- tical, hence practically indistinguishable
sistent with more rapid skill-biased in the figure).
technical change during these later A fourth piece of evidence comes
decades. from Greenwood and Yorukoglu (1997)
A simple regression analysis also con- and Per Krusell e t al. (2000). These
firms this point. I combined t h e data authors argue, based on t h e work of
from the March CPSs and decennial Griliches (1969), that equipment capital
censuses used in figure 1 above. Using is more complementary to skilled work-
these data, a regression similar to that ers than unskilled workers. This prem-
of Katz and Murphy for t h e period ise may b e reasonable since advances in
1939-96 yields similar results: equipment often appear to substitute
machines for tasks previously per-
formed by unskilled workers. Following
t h e work by Robert Gordon (1990) and
with an R h f 0.63 and an implied elastic-
ree en wood, Zvi Hercowitz, and Krusell
(1997), these papers document that t h e
ity of substitution of 1.8, which is some-
post-war period has witnessed a secular
what larger than the estimate of Katz
decline in the relative price of equip-
and Murphy. However, adding higher
ment capital, and argue that t h e associ-
order terms in time (i.e., time squared,
ated increase in the stock of equipment
time cubed, etc.) improves the fit of the
capital led to skill-biased technical
model considerably, and these higher-
change. Moreover, they argue that this
order terms are significant. I n figure 5, I
relative decline accelerated in t h e early
plot the implied time trends from regres-
1970s, and the associated acceleration
sions with higher-order terms as well as the
in t h e stock of equipment capital
linear trend (all numbers were rescaled
increased the demand for skills.
to fit in one graph). All three of these
Krusell e t al. (2000) formalize their
1Qhe college premium fell between 1940 and approach by assuming the following
1970 in part because it is estimated to be very high production function:
in the 1940 census. There may b e reasons to be
suspicious of data quality from this census, be-
cause (i) the education variable was different, (ii)
r = K : [ ~ ~ L+P( 1 - bl)(bz~,h
there may have been an overstatement of years of
+ ( 1 - bz)Hh)b/h](l- a)/P
schooling, possibly by as much as a factor of 1.5 or where K, is structures capital (such as
2 for sorne cohorts, and (iii) there was no self-
employment income in this census. But it is not buildings), and K, is equipment capital
clear whether any of these rneasurernent problems (such as machines). The parameter
will cause an upward bias in the college premiurn. ol = 1 / ( 1 - A) is the elasticity of substitu-
a
In any case, the level of the colle e premiurn from
this census is not out of line wit other historical
evidence (see, e.g., Goldin and Katz 2000; and
tion between equipment and skilled
workers, and o z = 1 / ( 1 - y ) is the elastic-
Piketty and Saez 2001). Moreover Autor, Katz, and ity of substitution between unskilled
Krueger show that even ignoring data from the
1940 census, there is evidence for an acceleration workers and the equipment-skilled
in the skill bias of technical change. For example, worker aggregate. If o l > o z (i.e., y > A),
for the range of the values for the elasticity of sub- equipment capital is more complemen-
stitution between o = 1 and 2, skill-biased techni-
cal change appears more rapid during the 1980s tary to skilled workers than unskilled
than in the 1970s and the 1960s. workers, and an increase in K, will
28 Journal of Economic Literature, Vol. XL (March 2002)
o linear time trend o quadratic time trend
+ cubic time trend A quartic time trend
I I I I I - ----- 7I
39 49 59 69 79 89 96
year
Alternative Time Trends for the Relative Demand for Skills
Figure 5. Estimates of Time Trends froin Regressions of In on In (HIL),year, yea+, year3 and year4
between 1939 and 1996 (with observations in 1939, 1949, 1959 froin the decennial censuses and
observations for 1963-96 froin the March CPSs)
lo
year
Figure 6. Behavior. of the Log Relative Price of Equipment Capital, 1963-92
TABLE 2
THE EFFECTOF THE RELATIVE
PRICE OF EQUIPMENT
ON SKILLED
PREMIA
Relative price
Note: This table reports the regression of the log college prerniuin on a linear time trend,
the log relative supply of skilled workers and various measures of the relative price of
equipment capital. For comparability, all data taken from Krusell et al. (2000).
section 2, weighs in favor of a marked Pierce view, instead, suggests that the
change in labor market prices and de- increase in residual inequality and t h e
mand for skills during recent decades. bulk of t h e rise in overall inequality
This argument is based on the view that are related to changes in labor market
changes in residual inequality reflect prices, and therefore to t h e supply and
changes in labor market prices, a thesis demand for skills. Second, according to
put forth by Chinhui Juhn, Murphy, this view, the rise in residual inequality
and Brooks Pierce (1993)." This view is is an important piece of evidence in the
important for an interpretation for the debate on acceleration: residual in-
recent changes in wage structure for equality, which was stable during the
two reasons. First, as the evidence in 1960s, began to increase rapidly during
section 2 indicates, much of the recent the early 1970s (figures 2 and 3), indi-
increase in overall inequality is d u e to cating a discontinuity in labor market
the rise in residual inequality. If resid- prices and, most likely, in the rate of
ual inequality were unrelated to the de- increase of the demand for skills.
mand and supply of skills, the frame- A concrete example is useful for clari-
work here could only account for a fying why residual inequality is linked
relatively small fraction of the increase to labor market prices. Suppose that
in overall inequality. T h e Juhn-Murphy- two otherwise identical individuals dif-
fer in terms of their unobserved skills
" Of course, an alternative-and inore cynical-
(for example, in terms of interpersonal
view would b e to interpret residual inequality as
"a measure of our ignorance" (as Moses Abraino- skills, motivation, specific skills for their
vitz 1957 did for T F P ) . When a standard wa e
regression such as (1) provides a good fit, t e a
residuals will be less disperse. Nevertheless, given
job, o r IQ)." Denote t h e unobserved
the variety of skills that we are unable to measure fl By unobserved skills, I inean skills that are
in standard data sets, much of the residual will not observed by the econometrician. These skills
plausibly reflect rewards to some unobserved could be-and are likely to be-observed by em-
skills. ~ l o y e r sT
. hese types of skills are often referred to
Acemoglu: Technical Change, Inequality, a n d the L a b o r Market 31
skill of individual 1 by a1 and that of behavior of overall and residual in-
individual 2 by a2 > al, and assume that equality since the 1970s suggest a
wages are given by marked shift in the demand for skills
In w,t = 28ta, + yth,, (13) over the past several decades. I therefore
tentatively conclude:23
where yt is the price of h skills at time t,
while 8t is the price of a skills. Since
these individuals are identical in all re- Conclusion 2. T h e behavior of returns
to schooling and residual inequality
spects other than their unobserved skills,
a , we have the variance of log wages over the past three decades suggests an
acceleration in the demand for skills
(or of residual wages) among these two
individuals as: beginning in the 1970s or 1980s.
Var(1n to) = - a1)2.
5. Acceleration in Skill Bias
Now if at a later date, t', this variance
What explains the more rapid in-
increases to Var(1n w)', and we know that
these two individuals are still identical crease in the demand for skills over
t h e past several decades? T h e first pos-
in all other respects and that a2 - a1 has
sibility is a change in labor market insti-
not changed, we can interpret the in-
tutions. T h e second is t h e role of in-
crease in Var(1nto) as reflecting an in-
creased international trade. T h e third is
crease in the price of unobserved skills,
more rapid skill-biased technical change.
8t. T h e discussion in the appendix
shows that t h e bulk of t h e increase in
I argue in section 6 that changes in
labor market institutions and t h e in-
overall and residual inequality cannot
creased importance of international
b e explained by composition effects,
trade cannot explain the change in labor
i.e., by changes in a z - a l , so this in-
crease is most likely d u e to a rise in t h e market prices by themselves. Moreover,
price of and demand for unobserved the evidence discussed in section 4 is
consistent with new technologies play-
skills during the 1 9 7 0 ~ . 2 V h e r e f o r e ,
ing an important role in changing t h e
with this interpretation, t h e rapid rise
wage structure. So h e r e I begin with
in t h e residual inequality during the
changes in technologies, and in particu-
1970s and t h e 1980s indicates a more
lar, I discuss "pure technological" ap-
rapid increase in demand for skills
proaches where technology is t h e only
during these decades than earlier.
factor determining t h e demand for
Overall, there is a variety of evidence
skills.
suggesting an acceleration in skill bias
over the past 25-30 years. Although not 5.1 "Techtzological Re2jolutiotzs" atzd
all evidence is equally convincing, the Acceleratiotz in Skill Bias
rise in the returns to schooling over t h e
T h e first group of technological theo-
past thirty years, despite the very rapid
ries links the acceleration in skill bias to
increase in the supply of skills, and the
exogenous technological developments,
and argues that a "technological revolu-
as "unobserved ability." This does not iinply that
these unobserved skills are necessarily synony- tion" led to more rapid skill-biased
mous with IQ or other single dirneilsional skill technical change beginning in t h e 1970s
indices.
'"ut differently, the absence of large coin osi "The most important caveat is that the avail-
tion effects indicates that the rise in r e s i d u 3 in: able data suggest relatively slow skill-biased tech-
equality likely reflects changes in the demand nical change during the second half of the 1990s
for unobserved skills rather than changes in the (see Mishel, Bernstein, and John Schinitt 1998;
"supply of unobserved skills." and Murphy and Welch 2000).
32 Journal of Economic Literature, Vol. XL (March 2002)
or the 1980s. I n terms of the model de- acceleration is that new technologies
veloped above, this corresponds to a are more complementary to skilled
more rapid increase in Ah/Az during this workers than to unskilled workers-for
period, translating into greater skill pre- example, there are more rapid advances
mia. Many of the proponents of this in the technologies used by skilled
view argue that the acceleration in skill workers, as captured by Ai, above. Rapid
bias is, at least in part, related to infor- technological progress then corre-
mation technology and computers (for sponds to an acceleration in slcill bias.
example, Krueger 1993; Berman, Bound, An alternative perspective, building on
and Griliches 1994; Autor, Katz, and an idea originally suggested by Nelson
Krueger 1998; Berman, Bound, and and Phelps (1966), also focuses on the
Machin 1998). effect of rapid technical change on in-
An interesting version of this story is equality, but puts the emphasis on the
the one developed by Krusell e t al. ability of skilled workers to deal with
(2000) discussed above. They argue that the introduction of new technologies.
the demand for skills accelerated as a According to this view, demand for
result of the more rapid decline in skills will automatically increase during
the relative price of capital equipment periods of rapid technological change.
beginning in the early 1 9 7 0 ~ The .~~ Welch (1970) gave an early succinct
Krusell e t al. theorv is attractive since it
i
summary of these two views. T h e first
provides a unified framework in which view, which we may call the accelera-
we can identify both the cause of the tion hypothesis, sees human capital as a
steady increase in the demand for skills, factor of production, and the more rapid
and the source of the more r aL ~ i dskill- increase in the demand for skills results
biased technical change, though the evi- from an acceleration in technical change
dence provided in the previous section that favors this factor of production. In
casts some doubt on the link between other words:
the relative price of equipment and the
demand for skills. [Tlechnical change may not be neutral
between skill classes. It may b e that incre-
T h e main idea of all these theories of
ments in technology result in increments in
the relative productivity of labor that are
241 classify this approach as one of exogenous positively related to skill level. (Welch 1970,
technology, since the driving force, the decline in p . 38)
the relative price of equipment capital, is assumed
exogenous.
An alternative interpretation of the approach by In contrast, the second vie\v-the
Krusell e t al. (2000) is that the main determinant Nelson-Phelps hypothesis-argues that25
of the demand for skills is not technology-skill
corn lementarit but capital skill complementarity. 25A recent paper by Samuel Bowles, Herbert
, . . -
I berieve that t e dist~nctionbetween technology- Gintis, and Melissa Osborne (2001) has a similar
skill versus capital-skill is not very useful. Capital- classification of different approaches. They call
skill complementarity could play an im ortant role the first view Walrasian, and the second Schumpe-
only in a model as in Greenwood e t a? (1997) or terian. I n the first, skills that enable workers to
Krusell e t al. (2000), where new capital embeds produce more are valued, while in the second, it is
superior technologies. In this sense, it is a combi- skills that enable workers to deal with changes in
nation of new capital and new technologies that is economic and social environments. They also pro-
increasing the demand for skills. Moreover, Autor, pose an alternative, Coasian, view, where "skills"
Katz, and Krueger (1998) show that demand for are workers' ca acity to work within organizations
more educated workers across industries is af- and follow autIority From the point of view of
fected by high-tech capital (e.g., computers), but the analysis here, this view is not fundamentally
not by equi rnent capital, suggesting further that different from the Walrasian view, since workers
it is new tec\nologies, not simply capital intensity, are paid more for skills that increase their marginal
that matters for inequality. contribution to the profits of their employers.
Acemoglu: Technical Change, Inequality, and t h e Labor Market 33
ITlhe ~
L i I
r o d u c t i v i t vof education would be approaches, it is useful to consider a
positively related to the rate of change in use- simple model based on Galor and Moav
ful technology (the ability to change) and to
the size of the tecl~nological gap (room for (2000). Suppose that
innovation). In this case, if the rate of utiliza-
tion of technology is accelerating, or if the A1 = $l(g)a and All = $lla (14)
technology gap i s growing, t h e return to
education will rise relative to other inputs. where a is a measure of aggregate tech-
(p. 38) nology, and g is the growth rate of a , i.e.,
g = b / n . The presumption that skilled
Studies building on the Nelson- workers are better equipped to deal with
Phelps hypothesis include Galor and technological progress can be captured
Tsiddon (1997), Greenwood and Yoru- by assuming that @'I < 0. Galor and Moav
koglu (1997), Caselli (1999), Aghion (2000) refer to this assumption as the
and Howitt (1998, ch. 9 ) , Galor and "erosion effect," since it implies that
Moav (2000), Violante (2000), Rubin- technical change erodes some of the es-
stein and Tsiddon (1999), Aghion, tablished expertise of unskilled workers,
Howitt and Violante (2000), and Gould, and causes them to benefit less from
Moav and Weinberg (2000).2'3These pa- technological advances than skilled work-
pers argue that there has been a tech- ers do. Substituting from (14) into (4),
nological revolution in the U.S. econ- the skill premium is
omy starting in the 1970s and relate the
rise in inequality to the increased
demand for skills resulting from the
technological revolution. For example,
Greenwood and Yorulzoglu (1997) draw
a parallel between the First and the
Second Industrial Revolutions and what
has been happening in the U.S. econ-
Therefore, as long as 4'1 < 0, more rapid
omy since 1974. Caselli (1999) develops
technological progress, as captured by a
a similar theory where a technological
higher level of g, will increase the skill
revolution increases the demand for
premium.
workers who can switch to the sectors
Theories that explain the increase in
that benefit from the introduction of
inequality as a result of rapid tech-
new technologies.27
nological progress have a number of
To get a basic understanding of these
attractive features.28 First, many econo-
" See Aghion (2001) for an approach that com- mists and commentators view the ad-
bines the Nelson-Phelps insi hts with the Schum- vances in computer and information
%;
peterian notion of creative- e struction to discuss
technology a
the impact of the diffusion of computers on -, as break with the tech-
inequality. nologies of the past, and so are open to
27The explanation offered by Rubinstein and
Tsiddon (1999), and Aghion, Howitt, and Violante 28These theories also predict that inequality
(2000) is somewhat different. They argue that should increase when new technologies are being
there is increased uncertainty at times of rapid introduced, but should decline when these new
technological change, and more skilled workers technologies are standardized and being used rou-
are better able to cope with uncertainty. This idea tinely by many firms (see, for example, Galor and
is also related to a thesis first put forth by Michael Tsiddon 1997, or Aghion, Howitt, and Violante
Piore and Charles Sabel (1984) that the oil price 2000). So far, there seems to be no evidence of a
shocks increased the uncertainty faced by produc- decline in inequality in the United States, but per-
ers, and induced them to change the organization haps the years to come will see a return to the
of production toward organizational forms that levels of inequality experienced during the 1960s,
require adaptable workers. vindicating this approach.
34 Journal of Economic Literature, Vol. XL (March 2002)
the idea that we might be in the midst I t is difficult to imagine how a new
of a technological revolution. Second, a and radically more profitable technol-
variety of evidence supports the notion ogy will first lead to 25 years of substan-
that skilled workers have a comparative tially slower growth. Although, in an
advantage in coping with rapid tech- influential paper, Paul David (1990)
nical change. Ann Bartel and Frank argues that the spread of electricity to
Lichtenberg (1987) show that firms in- American manufacturing was also slow
troducing new technologies hire more and productivity gains from electrifica-
skilled workers. Bartel and Nachum tion were limited until the 1920s, the
Sicherman (1998) document that re- parallel with the recent productivity
turns to unobserved ability appear to be slowdown should not be overstated.
higher in industries with more rapid First, though productivity growth from
technical change. Andrew Foster and electrification was sluggish during the
Mark Rosenzweig (1996) provide evi- early 1900s, the U. S, economy overall
dence from developing countries that had a much higher level of output
more educated workers are better growth than the growth levels experi-
placed to take advantage of advances in enced over the past three decades. Data
agricultural technology. from table A-XVIII of John Kendriclz
T h e main difficulty with both the (1961) imply that output growth be-
theories based on the acceleration and tween 1899 and 1909 in the U.S. econ-
the Nelson-Phelps hypotheses is that omy was 4.2 percent a year, while be-
they rely explicitly on rapid technical tween 1909 and 1919, it was 3 percent,
change in recent decades." There is and between 1920 and 1929, output
little direct evidence that the decades grew by 3.6 percent a year. Second, as
between 1970 and 1995 have been a pe- noted by Stephen Oliner and Daniel
riod of rapid technical change, how- Sichel (1994), computers and other ad-
ever. First, this period has experienced vanced office equipment have only been
sluggish T F P and output growth rela- a trivial part of the aggregate capital
tive to earlier periods. Greenwood and stock of the U.S. economy until the mid-
YorulzogIu (1997) and Hornstein and 1990s. I t is therefore unlikely that the
Krusell (1996) argue that the slow T F P whole of the U.S. economy has been
growth itself may be an outcome of the adapting to the changes in this rela-
more rapid technical change. According tively small part of the capital stock. Fi-
to this argument, new revolutionary nally, as shown by Ernie Brendt, Cath-
technologies first reduce productivity erine Morrison, and Larry Rosenblum
growth as firms and workers spend time (1994), more computerized sectors did
learning to use these technologies. not perform any better in terms of labor
productivity growth over this period;
this pattern is also difficult to recon-
29Another argument that applies only against cile with a computer-led technological
approaches based on the Nelson-Phelps hypothe-
sis is that historical evidence is not necessarily in revolution.
line with a view that demand for skills always in- I t is also useful to note that although
creases during times of rapid technical change. As computers have no doubt increased our
discussed in the introduction, the major techno-
logical changes of the early nineteenth century a standards of living and quality of life
ear to have been large1 skill-replacing (unski!: over the past thirty years, they may be
[iased), even though t i e g seem as radical as much less radical innovations than cer-
computer technolo y This suggests that it is
the skill bias of tec%*ology, not merely its ra id tain previous new technologies (see
arrival, that is important for the demand for skilys. Gordon 1998). To gain perspective,
Acemoglu: Technical Change, Inequality, and t h e Labor Market 35
consider the difference that the tele- increase in the supply of skills. There is
graph made to a world in which the no a priori reason to expect the accel-
fastest medium of communication was eration in skill bias to coincide with the
pigeons. Mokyr (1990, p. 124) describes rapid increase in the supply of skills.
this as follows: Those who want to subscribe to the
exogenous technological progress view
The telegraph had an enormous impact o n
19th-century society-possibly as great as
have to explain this as a chance event.31
that of the railroad. Its community and politi- 5.2 Endogenous Skill-Biased Technical
cal value was vast, as was its effect in coordi-
nating international financial and commodity
Change
markets. Unlike the railroad, it had no close T h e theories discussed so far a r e -
substitutes, the closest being homing pigeons
and semaphore.
sume technical change to be skill-
biased b y nature (or, at the very least,
O r consider the difference that the auto- recent technologies to have increased
mobile and air conditioning made to the the demand for ;kills d u e to exogenous
"
quality of life, and electricity and in- reasons). A different perspective is to link
terchangeable parts made to the manu- the types of technologies that are devel-
facturing sector. As also pointed out by oped and adopted to (profit) incentives,
Gordon (1998), compared to these improve- or to demand pull as emphasized by
ments, the switch from mainframes to Schmookler (1966).s2
PCs, or from telephone to e-mail, or from Historical evidence is consistent with
the typewriter to the word processor may the notion that profit incentives and op-
be modest.30 aortunities are imnortant
L L
for the de-
A final problem for all of the ap- velopment and introduction of new
proaches based on exogenous techno- 3 l O n e could argue that the supply of skilled
logical developments is the coincidence workers increased because, during the 1960s,
in the timing of this change, and the workers anticipated that there was going to be a
technological discontinuity in the decades to
rapid increase in the supply of skilled come, and responded to this by increasing their
\vorlzers. Recall that there was a very education. This story appears quite unreasonable,
large increase in the supply of college however. There is no evidence that anyone, let
alone teenagers, foresaw the technological devel-
graduate \vorlzers during the late 1960s opments of the 1970s and the 1980s as early as the
and the early 1970s (figure 1 and table 1960s. Moreover, the increase in the supply of
1 show the large increase in the em- skills can be largely explained by two factors.
First, the Vietnam era draft laws encouraged
ployment share of college workers be- young males to stay in college lon e r (and indi-
tween 1970 and 1980). So the accelera- rectly also influenced female enroflments) Sec-
tion in skill bias is either concurrent ond, college enrollments were on an upward trend
since the early 1950s, and much of the increase in
with, or immediately follows, this large the suppl of college graduate workers is ac-
counted &r by the interaction of this upward
30Another ar ument is that our ability to mea- trend with the very large relative size of the baby
sure T F P grow$ may have deteriorated following boom cohorts.
a change in technological regime, However, as 32This is also the approach taken by the endog-
Martin Bailey and Gordon (1988) document, pro- enous rowth theory, which determines the overall
ductivity slowdown has been concurrent in many rate o f technical change-but not the degree of
sectors, some of them with little problems in skill bias-from profit incentives (e.g., Romer
measuring output or output uality. It is interest- 1990; Gene Grossrnan and Elhanan Helpman
ing to note, however, that evi3ence in favor of this
hypothesis may yet emerge, especial1 since pro-
ductivity growth has been quite rapidduring the
2
1991; A hion and Howitt 1992). An advantage of
the e n ogenous-technolog perspective in this
context is that it provides cLar reasons for the de-
past three years (but see Gordon 1998, and more gree of skill bias to increase even without a change
recently, Dale Jorgensen and Kevin Stiroh 2000, in the overall pace of technological improvement
on this). -see below.
36 Journal of Economic Literature, V o l . X L ( M a r c h 2002)
technologies. Fernand Braudel (1984, cotton gin on the South was nothing
p. 566) took a strong position on this: short of spectacular. In the court case
[Tlhe efficient application of technology over the patent rights, Judge Johnson
lags, by definition, behind the general move- wrote:
ment of the economy; it has to be called on,
sometimes several times, to meet a precise "[. . . as a result of the cotton gin] . . .
and persistent demand. individuals who were depressed with poverty,
and sunk with idleness, have suddenly risen to
An interesting example of the timing wealth and respectability. Our debts have been
of technological development respond- paid off, our capital increased; and our lands
are treble in value." (Quoted in Constance
ing to profit incentives is given by the Green 1956, p . 92)
introduction of the electric street car in
U.S. cities during the late nineteenth Within a short time, Eli Whitney's gin
century. In his history of electricity in turned the United States from a cotton
the United States, David Nye (1990) importer into the largest cotton exporter
describes this as follows: in the world.
Cities grew larger, better transportation Schmookler (1966) provides a famous
was needed, so t h e [electric] trolley was in- argument for the importance of demand
vented, called into being by the crowded late pull in the development of many tech-
nineteenth century cities . . . . By the 1870s nologies. H e documents rapid innova-
large cities had ceased to b e accessible by
foot, or built to the scale of pedestrians, and
tions in railroads follo\ving increased
traffic congestion was terrible. ( p . 85) purchases of railroad equipments, and
more generally argues that industries
This created the profit opportunities to with greater investments experience
develop and introduce the electric trol- faster technological progress because
ley. The technological requirements had the returns to such progress are greater.
been met long before, and awaited these A natural next step is then to argue that
profit opportunities. Nye writes, "How- the degree of skill bias in technical change
ever great the need for the electric trol- is also determined by profit opportuni-
ley after 1870, it was hardly a new idea; ties and by the demand for different
it had been the object of experiment types of technologies. Here, by endoge-
during four decades" (p. 86). nous (skill bias) technology approach I
Another example of the type of inno- mean the view that the degree of skill
vation responding to profit incentives is bias in technical change is influenced
provided by the cotton gin. In the late by profit incentive^.^^
eighteenth and early nineteenth centu- A key determinant of profitability is
ries, only green seed cotton, which was market size. As Schmookler (1966)
difficult to clean, could be grown in stated in the title of two of his chapters:
most of the American South, and Brit- "The amount of invention is gooerned by
ain imported most of its cotton from the t h e extent of t h e market." T h e most suc-
West Indies, Brazil, and India. A ma- cessful businessmen have always been
chine to remove the seeds was essential aware of this. For example, Matthew
for the success of American cotton. In
contrast to almost all other textile inno- 33An alternative explanation for the increase in
wage inequality based on endogenous technical
vations that were taking place in Eng- change is offered by Huw Lloyd-Ellis (1999). In
land and Europe, such a machine, the his model, it is the interaction between endoge-
cotton gin, was developed in the United nous technical change and the slowdown in the
growth of labor quality that leads to increased in-
States in 1793 by Eli Whitney in re- e uality. Skill bias is not endogenous in Lloyd-
sponse to this need. T h e impact of the Eylis' model.
Acemogku: Technical Clzange, Inequality, a n d t h e L a b o r Market
Relative Wage
1 Long-run Relative
uemand for Slalls
Long-run Rel. Wage
Short-run Response
I I I
r
Exogenous Shift in Relative Supply
Figure 7. The Dynamics of the Relative Wage of Skilled Workers in Response to an Increase in the Supply of
Skills with Elldogellous Skill-Biased Technical Change
Relative Wage
I I
I I
I I
I I
I
I
I
I
I
Long-run Relative
Demand for Skills
Short-run Response
Short-run Response ..............
I
I
I
I I
Cohen and Chang-Tai Hsieh 2000). of rural industry came to the cities.
This response suggests a change in Fourth, "there was a large influx of
the production structure toward more labor from Ireland" (p. 137). Finally,
skilled workers, consistent with the "technical changes in agriculture in-
theories outlined in this section.39 creased the supply of labor available to
Despite this evidence showing simul- industry" (p. 137).
taneous increases in the supply of, and According to the endogenous tech-
demand for, skills in a number of episodes, nology hypothesis, this increase in the
it is difficult to distinguish exogenous supply of unskilled labor should encour-
and endogenous technical change. The age unskill-biased technical change. And
exogenous technical change theory as predicted by this approach, there
maintains that technical change is often were inajor skill-replacing (unskill-
slzill-biased. Endogenous technical change biased) technologies introduced during
theory instead suggests that new tech- this period, most notably the factory
nologies should be skill-biased when system replacing tasks previously per-
the supply of skills increases. Since the formed by slzilled artisans. Moreover, in
supply of skills has increased most of his classic study, Habakkuk argues that
the time over the past one hundred or the increase in the supply of labor was
so years, the implications of the two an important inducement to the devel-
theories are quite similar. The increase opment of factory methods. He also
in the supply of unskilled labor in the quotes an American historian, Handlin,
English cities during the early nine- to explain why the adoption of factory
teenth century provides an interesting methods in the United States was
contrasting event for the two ap- solnewhat delayed. Handlin wrote:
proaches. Bairoch (1988, p. 245) de-
[N]o matter what degree of standard-
scribes this rapid expansion as follows: ization technical process of manufacturing
". . . between 1740 and 1840 the popu- reached, the absence of a cheap labor supply
lation of England . . . went up froin 6 precluded conversion to factory methods.
inillion to 15.7 million. . . . while the (p. 146)
agricultural labor force represented 60-
Habakkuk placed much more impor-
70% of the total work force in 1740, by
tance on the role of wages in determin-
1840 it represented only 22%." Habak-
ing innovation decisions, a view that
kuk (pp. 136-37) also emphasizes the
later became known as the "Habakkuk
increase in the supply of unskilled labor
hypothesis." But he also emphasized the
in English cities, and attributes it to
different availabilities of skilled labor in
five sources. First, enclosures released
Britain and the United States. He wrote:
substantial labor from agriculture.
- Sec-
ond, "population was increasing very [ I ] n both countries this provided manufac-
rapidly" (p. 136). Third, labor reserves turers with an incentive to adopt and devise
methods which replaced skilled by non-
skilled . . . [but ] . . . the English had a
39 Since Israel can be approximated by a small stronger incentive than the Americans to
open economy, another possibility is a change in replace skilled by unskilled labor. ( p . 152)
the output mix and trade patterns. Gandal, Han-
son, and Slaughter (2000) and Cohen and Hsieh
(2000) find no evidence for this, and show that With a similar reasoning to Habalzlzuk
demand for skills increased in all Israeli sectors. and Schmookler, the endogenous tech-
Cohen and Hsieh (2000) also argue that because nology view suggests that nineteenth-
many Russian immigrants initially worked in low-
skill occupations, the supply of skills to the Israeli century businessmen took advantage of
economy may not have increased by much. the rapid increase in the supply of labor
Acemoglzc: Technical Change, Inequality, and the Labor Marlcet
in the cities by developing the factory As discussed above, a major problem
system. According to this view, there is for the "technological revolution" mod-
a close link between the skill-replacing els is the slowdown in TFP. Since the
technologies of the nineteenth century endogenous technology approach places
and the change in the factor supplies the emphasis on which type of tech-
faced by employers-"the reserve army nologies are developed, it is not incon-
of unskilled labor." Although these his- sistent with the slow growth in TFP
torical examples are informative, they during the past thirty years: an accel-
do not reveal whether endogenous tech- eration in the skill bias of new technolo-
nology choices are iinportant in under- gies does not require faster technical
standing more recent skill-biased tech- progress. The evidence presented in
nical change. A systematic study of how Richard Newell, Adain Jaffee, and
technologies respond to large changes Robert Stavins (1999) is consistent with
in the relative supply of skills is clearly the notion that changes in the direction
a worthwhile future research project.40 of technical change can happen without
An iinportant aspect of the endoge- an increase in the overall rate of pro-
nous technology theory is that it inakes ductivity growth. They show that inno-
relatively tight predictions regarding vation in air conditioners responded to
the future path of technical change. changes in energy prices by becoming
While, with exogenous skill bias theo- inore energy-efficient, but find no in-
ries, there is no clear reason to expect a crease in the rate of productivity
further acceleration or deceleration in growth. In fact, Aceinoglu (1998) and
the skill bias of new technologies, ac- Kiley (1999) show that the increased
cording to this endogenous skill bias effort of firms to develop more skill-
theory, the future path of technical biased technologies could run into de-
progress should be closely tied to the creasing returns, and hence inay cause a
path of the supply of skills. If the relative slowdown in TFP growth. Intuitively,
supply of skills continues to increase, the overall productivity growth in the
we should expect further skill-biased economy is maximized with a balanced
technical change.41 distribution of resources toward devel-
oping skill-biased and unskill-biased
40It is also useful to note that the skill bias of technologies (due to decreasing returns
technology most likely res onds not only to to each activity). During periods of
changes in the relative soppfy of skills, but to a rapid skill-biased technical change, all
variety of other factors. Recent work b Marcus
Mobius (2000) and David Thesmar a n J Mathias resources go into developing skill-
Thoenig (2000) shows how the size of the product biased machines, and cause a decline in
market, the degree of competitive pressure and advances in unskill-biased technologies.
instability facing firms may affect the way firms
choose to organize, and therefore the demand for Because of the decreasing returns to
skills. Moreover, Mobius suggests that these
changes reduced the demand for skills during the
nineteenth century as there was greater stan- the 1980s has come to an end. I t also has to be
dardization of products, but increased this de- noted that the increase in the relative supply of
mand during the past several decades as the need college educated workers in the 1990s may have
for flexibility increased. How the organization of been less than the revious trend (see figure I ) ,
and this may have aPfected the technolo y choices
roduct markets and the extent of competition af-
fect technology choices and the demand for skills
is a very promising area for future research.
B
of firms. In particular, with the increased abor force
participation of less skilled workers, there may
41Although evidence from the 1990s suggests now be a sufficient number of unskilled workers
that skill-biased technical change is now slower, supplying labor at low wages to make the further
there is not yet sufficient evidence to decide development of unskilled-labor-cornpleinentary
whether the rapid skill-biased technical change of technologies quite profitable.
44 Journal of Economic Literature, Vol. XL (March 2002)
scale, improveinents in the other sector ers.43 In the basic framework of section
will not fully offset this decline, and 3, average wages always increase when
overall TFP growth will fall.42 the supply of educated workers in-
Finally, it is useful to discuss briefly creases, and all wages should rise in re-
the response of skills to technology. sponse to an increase in the productiv-
The analysis so far treated the supply of ity of skilled workers, Ah. Yet, over the
skills as exogenous, and investigated the past thirty years the wages of low-slzill
implications of the supply on the de- workers have fallen in real value, which
mand for skills. Naturally, the supply contrasts with their steady increase in
of skills will also respond to econornic the previous thirty years.
incentives: more workers are likely to Models of faster technological prog-
acquire skills when skill premia are ress would naturally predict that un-
higher. Such supply choices can be eas- skilled workers should benefit froin this
ily incorporated into this framework. faster progress. The endogenous tech-
Suppose, for example, that the relative nology approach discussed in the pre-
supply of skills, HIL, is an increasing vious subsection, on the other hand,
function of the skill prerniurn, a.In this predicts that there inay be no improve-
case, if the long-run deinand curve for ments in the technologies for unskilled
skills is upward-sloping, we can have workers for an extended period of time
an equilibrium path in which both the because skill-biased innovations are
relative supply of skills and the skill more profitable than unslzill-biased in-
premium increase together over tiine novations. Yet in that case, their wages
(see Acemoglu 1998). This equilibrium should be stagnant or increase slowly,
configuration gives us an attractive in- but not fall.
terpretation for the joint behavior of Some of the studies mentioned above
college skills and the college pre- have suggested explanations for the fall
mium in the U.S. economy over the past in the wages of low-skill workers. For
sixty years (see figure 1): high returns example, recall that Galor and Moav
to schooling encourage education, (2000) argue that faster technological
which in turn induce inore skill-biased change creates an "erosion effect," re-
technical change, increasing returns to ducing the productivity of unskilled work-
schooling again. ers. Using equation (3) from above, in
the simplified version of their model
discussed in section 5.1, the unslzilled
5.3 A Puzzle: The Decline in Wages
wage is ZUL = $z(g)a[l + $;(H/L)P](~ - P)/P,
of Low- Skill Workers
where the $1 function captures the ero-
A coinmon shortcoming of all the sion effect. The rate of growth of un-
pure technology approaches discussed skilled wages will be ~ ~ L / Z U=Lg(1 + E$),
in this section is that they do not natu- where E$ is the elasticity of the $1 func-
rally predict stagnant average wages tion which is negative by the assump-
and/or falling wages for unskilled work- tion that $'I < 0. If this elasticity is less
than -1, an acceleration in econoinic
42The view that too much effort toward inl- growth can reduce the wages of
roving the skill-biased technologies may be re-
fated to the T F P slowdown is consistent with the
pattern of sectoral T F P growth observed recently. 43 However, recall that if the increase in non-
As Gordon (1998) and orgensen and Stiroh (2000) wage benefits is taken into account, avera e wa es
I,
document, there has e e n rapid T F P growth in % fp
increased over this period. So the more ro ust act
computer-producing sectors, but mediocre, or might be the fall in the real wages of low-skill
even disappointing, T F P growth in other sectors. workers.
Acemoglu: Technical Change, Inequality, and the Labor Market 45
market both amplifies the effect of cause of the increase in inequality. In-
technology on wage inequality, and stead they have become powerful actors
provides a possible explanation for this only by interacting with technical
decline. change, amplified the direct effect of
technical change on inequality, and
6. Ramifications of Tech7zical Change contributed to the fall in the wages of
low-skill workers.
This section discusses how technical 6.1 Organizatio7zal Change a7zcl
change can affect labor market prices I7zequality
by transforming the organization of the
labor market. T h e idea that technology A variety of evidence suggests that
affects the organization of production, important changes in the structure of
and the institutions around it, is an old firms have been taking place in the U.S.
one. Marx put it in a dramatic fashion: economy over the past 25 years. More-
"The hand-mill gives you society with over, it seems clear that a major driving
the feudal lords; the steam-mill, society force for this transformation is changes
with the industrial capitalist." T h e argu- in technologies (hence the view that
ment here is not as extreme, but re- technical change is essential for the
lated: recent technological developments changes in the organization of firms).
may have led to important changes in F o r example, team production and
the organization of production. other high-performance production
My focus here is on three sets of methods are now widespread in the
changes that could account for the fall U.S. economy (e.g., Casey Ichniowski,
in the wages of low-skill workers: the Giovanna Prennushi, and Kathryn
transformation of the organization of Shaw 1997, or Eileen Applebaum and
firms; change in labor market "insti- Rosemary Batt 1994). Similarly, Peter
tutions,"46 particularly the decline in Cappelli and Steffanie Wilk (1997) show
unionization; and the interaction between that there has been an increase in the
international trade and technical change. screening of production workers, espe-
Organizational change often destroys cially from establishments that use
the types of jobs that pay high wages to computer technology and pay high
low-skill workers. Deunionization re- wages.
duces the bargaining power of low-skill Murnane and Levy (1996) report case
workers. And international trade with study evidence consistent with this
less developed countries (LDCs) in- view. From their interviews with human
creases the effective supply of unskilled resource personnel at a number of com-
labor and depresses the marginal value panies, they describe the change in the
product of less skilled workers in the hiring practices of U.S. companies. A
U.S. economy. Therefore, a11 three nlanager at Ford Motor colnpany in
changes could be responsible for the 1967 describes their hiring strategy as
changes in the U.S. wage structure and follows: "If we had a vacancy, we would
for the decline in the wages of low-skill look outside in the plant waiting room
worlzers. Nevertheless, I argue that these to see if there were any warm bodies
factors by theinselves are not the major standing there. If someone was there
and they looked physically OK and
4 6 1 am using the term institutions loosely here,
to capture the rules of the game in the labor mar-
weren't an obvious alcoholic, they were
ket, patterns of bargaining, as well as government hired" ( p . 19). In contrast, comparable
labor lnarlcet policy. companies in the late 1980s use a very
Acemoglu: Technical Change, Inequality, and the Labor Marlcet 47
0.35 1 I I I I I
1983 1985 1987 1989 1991 1993
Year
Figure 9. The Evolution of the Percentage of Employment in the Top and Bottom 25 Percentile Industry-
Occupation Cells (weight-at-the-tails of the job quality distribution)
and Gilles Duranton (2001). Kremer mulation implies that if the productivity
and Maskin consider a production func- (ability) of unslzilled workers, A/, is very
tion which distinguishes between man- low relative to A/,, they pull down the
agers and workers. They show that a productivity of skilled workers. In
change in technology or an increase in contrast, when they work in separate
the dispersion of skills may encourage firms, slulled workers are unaffected by the
high-skill workers to match with other productivity of unskilled workers. More-
high-skill workers, rather than work as over, p > l, which implies that improve-
managers in establishments employing ments in the productivity of skilled work-
low-skill workers. ers has more effect on the productivity
Here I outline a simple model, syn- of new-style organizations. The parame-
thesizing Kremer and Maskin (1999) ters Bp and B, capture the relative effi-
and Acemoglu (1999a), that captures ciency of old and new-style production
the effect of the technical change on functions.
the organization of production, and via T h e labor market is competitive, so
this channel, on wage inequality. T h e the equilibrium organization of produc-
basic idea is simple. As the productivity tion will maximize total output, given
of skilled workers increases, it becomes by Bp[(A/L)P+ (A~,lao)p]l/p + B,A!(H - ho),
more profitable for them to work by where 1x0 E [0,HI is the number of
themselves in separate organizations skilled workers employed in the old-style
rather than in the same workplace as organizations. For all cases in which
the unskilled workers. This is because ho > 0, the solution to this problem will
when the skilled and the unskilled work involve
together, their productivities interact,
and unskilled workers may put downward +
W H = BpAEh$- ' [ A ~ L PAEla$](l-P ) / P
(22)
pressure on the productivity of skilled P
= BsA,,
workers (for example, because unskilled i.e., skilled workers need to be paid B,A!
workers have to implement part of the to be convinced to work in the same
production process designed by skilled firms as the unskilled workers. The
workers). unskilled wage is
Specifically, suppose that firms have
access to the following production
functions
the old-style production function: Now consider an increase in Ah. Dif-
Y = Bp[(A/L)P+(Ahlzo)~]~/~, ferentiating (22) yields alao/aA/, < 0, that
the new-organization production function: is, there will be fewer skilled workers
Y =BB!~N, working with the unskilled. Moreover,
differentiation also gives a(Ahho)/dAh< 0,
where ho is the number of skilled work- so the efficiency units of skilled workers
ers employed in the old-style firms, and that the unskilled work with declines.
1 % is~ the number of skilled workers em- From (23), this implies that clw~/dAl,< 0.
ployed in new organizations separately Therefore, skill-biased technical change
from unskilled workers. The fact that encourages skilled workers to work by
when they are employed in the same themselves, and as a result, unskilled
firm these two types of workers affect wages fall. Intuitively, because, in the
each other's productivity is captured by old-style organizations, the productiv-
the CES production function. This for- ity of skilled workers depends on the
Acewzoglu: Technical Change, Inequality, and t h e Labor Market 49
ability of unskilled workers, when the this evidence does not yet enable an as-
skilled become even more productive, sessment of whether changes in organi-
the downward pull exerted on their pro- zational forms have been an important
ductivity by the unskilled workers be- contributor to the changes in labor
comes more costly, and they prefer to market prices, and future research is
work in separate organizations. This re- required to determine the role of
duces the ratio A/,ho/L and depresses organizational change in the rise in
unskilled wages. As a result, improve- inequality.
ments in technology, which normally 6.2 Institutional Change
benefit unskilled workers as in section
3, may hurt unskilled workers because Labor market "institutions," includ-
they transform the organization of ing minimum wage laws, the impor-
production. tance of collective bargaining in wage
An increase in B , / B p , which raises the determination, and perhaps social
relative profitability of the new organi- norms, undoubtedly have an important
zational form, also leads to further seg- effect on the distribution of wages. T h e
regation of skilled and unskilled work- Great Compression in the wage distri-
ers in different organizations. This last bution after the Second World War is
comparative static result is useful since difficult to explain without invoking
Bresnahan (1999) and Autor, Levy and changes institutions and social norms
Murnane (2000) argue that by replacing (see, e.g., Goldin and Robert Margo
workers in the performance of routine 1992). Could changes in labor market
tasks, computers enabled a radical change institutions also account for the in-
in the organization of production.47 crease in wage inequality and the dis-
This is reminiscent of a technological continuity in that demand for skills
change that makes the new-organization during the past decades?
production function more profitable. Two major changes in labor market
These organizational stories are at- institutions over the past 25 years are
tractive since they provide a unified ex- the decline in the real value of state and
planation for the changes in the wage federal minimum wages and the reduced
structure and the apparent changes in importance of trade unions in wage de-
the organization of firms. An interesting termination. Although many economists
recent paper by Eve Caroli and Van suspect that these institutional changes
Reenan (1999) provides evidence sug- may be responsible for the changes in
gesting that changes in wages have been the structure of the U.S. labor market
accompanied by changes in organi- (see Freeman 1991; DiNardo, Fortin,
zational forms. Acemoglu (1999a) and and Lemieux 1995; Lee 1999), I will
Kremer and Maskin (1999) also provide argue that these changes are unlikely
evidence suggesting a number of orga- to be the major cause of the recent
nizational changes in the U.S. economy increase in U.S. wage inequality.
during the past 25 years. Nevertheless, The real value of the minimum wage
was eroded throughout the 1980s as the
"A related perspective is offered by Aghion nominal minimum wage remained con-
(2001), who also argues that computers replace
unskilled tasks. He suggests that colnputers are a stant for much of this period. Since the
"general-purpose technology," so their diffusion minimum wage is likely to increase the
follows an inverted S shaped pattern. H e sug- wages of low-paid workers, this decline
gests that as inore and inore firms adopted com-
puters over the past decades, demand for unslzilled may b e responsible for increased wage
worlcers fell rapidly. dispersion. DiNardo et al. (1995) and
50 Journal of Economic Literature, Vol. XL (March 2002)
Lee (1999) provide evidence in support increase in wage inequality. Unions
of this hypothesis. Although the contri- often coinpress the structure of wages
bution of the decline in the ininirnu~n and reduce skill premia (see for example
wage to increased wage dispersion can- Lloyd Reynolds 1951; or Freeman and
not be denied, most economists believe James Medoff 1984). Throughout the
that this is unlikely to have been a ma- postwar period in the U.S. economy,
jor factor in the changes in the U.S. unions negotiated the wages for many
wage structure. First, only a small frac- occupations, even indirectly influenced
tion of male workers are directly af- managerial salaries (see DiNardo,
fected by the rniniinu~nwage (even in Kevin Hallock, and Pischlze 2000).
1992, after the minimum wage hike of Unions also explicitly tried to compress
1990-91, only 8 percent of all workers wage differentials. This suggests that the
between the ages of 18 and 6 5 were decline of unions may have been a major
paid at or below the minimum wage). cause of the changes in the structure of
Although the minimum wage may in- wages.
crease the earnings of some workers Although deunionization could in
who are not directly affected, it is un- principle be an important factor in the
likely to affect the wages above the me- structure of wages, the extent and tim-
dian of the wage distribution. But as ing of deunionization suggests that it is
figure 3 shows, the difference between not the major driving force of the in-
the 90th percentile and the median crease in inequality. First, wage in-
behaves very similarly to the difference equality increased in many occupations
between the median and the 10th per- in which prices were never affected by
centile." This implies that whatever unions (such as lawyers and doctors).
factors were causing increased wage Moreover, in the United States, de-
dispersion at the top of the distribution unionization started in the 1950s, a period
are likely to have been the major cause of stable wage inequality. During the
of the increase in wage dispersion 1970s, though unionization fell in the
throughout the distribution. Second, the private sector, overall unionization rates
erosion in the real value of the mini- did not decline rnuch because of in-
mum wage started in the 1980s, creased unionization in the public sec-
whereas, as shown above, the explosion tor. Overall union density was approxi-
in overall wage inequality began in the mately constant, around 30 percent of
early 1970s (subject to the caveat raised the work force, between 1960 and 1975.
in footnote 11). I t was the anti-union atmosphere of the
The declining importance of unions 1980s and perhaps the defeat of the air-
may be another important factor in the traffic controllers' strike that led to the
largest declines in unionization, dating
4Qxcept during the early 1980s when there is a much of deunionization after the rapid
Inore rapid increase in inequality at the bottonl of increase in inequality during the early
the \va e distribution, most likely due to the fall-
q,
inf sea value of the minimum wage. There are 1970s.49 Evidence from other countries
su stantlally larger increases in inequality at the also paints a similar picture. For exam-
bottonl of' the distribution aillong female worlzers ple, in the United Kingdom, wage in-
or in a sanlple that combines Inale and fernale
worlzers. This reflects the larger effect of mini- equality started its sharp increase in the
lnuin wage laws on female earnings. Fi uie 3
sliows that there are factors, other than J l e ' d e - 49 An interesting recent paper by Henry Farber
cline in the real value of the minimum wage, hav- and Bruce Western (2000) dates the nlajor decline
ing a major effect on the top and the bottonl of in union activity to the early 1980s, a few months
the male wage distribution. before the air-traffic controllers' strike.
Acemoglu: Technical Change, Inequality, and the Labor Market 51
that the United States starts trading with price of skill-intensive goods in the
a set of LDCs that have access to the LDCs before trade. The saine argument
saine technology as given by Ah and Al, as above implies that fl > p w , i.e., trade
but are relatively scarce in skills. Denote with the skill-abundant United States re-
the total supplies of skilleq and upskilled duces the relative price of skill-intensive
yo5kers in the LDCs by H and L where goods in the LDCs. This implies that
H / L < H / L , which siinply reiterates that a " < & ; after trade wage inequality
the United States is more abundant in should fall in the LDCs.
skilled workers than the LDCs. Although in theory increased trade
After full trade opening, the product with the LDCs can be the cause of the
markets in the United States and the rapid increase in the deinand for skills,
LDCs are joined, so there will b e a most evidence suggests that the direct
unique world relative price. Since the effect of increased international trade
supplies of skill- and unskill-iniensive on the U.S. labor market has been rela-
goods are Ah(H + H) and Al(L + L ) , the tively minor. First, as equation (27)
relative price of the skill-intensive good shows, the effect of international trade
will be works through a unique interzjening
mechanism: inore trade with the LDCs
increases the relative price of skill-
intensive goods, p, and affects the skill
premium via this channel. In fact, in
The fact t h a ~R ~ V > pus follows irninedi-
this simple framework, the percentage
ately from H / L < H / L . Intuitively, once
increase in the skill premium is propor-
the United States starts trading with
tional to the percentage increase in the
skill-scarce LDCs, deinand for skill-
relative price of skill-intensive goods.
intensive goods increases and pushes the
Perhaps the most damaging piece of
prices of these goods up.
evidence for the trade hypothesis is that
Labor demand in this econoiny is de-
most studies suggest the relative price
rived from product demands. The skill
of skill-intensive goods did not increase
premium therefore follows the relative
over the period of increasing inequality.
price of skill-intensive goods. After
Robert Lawrence and Slaughter (1993)
trade opening, the U.S. skill preiniuin
found that during the 1980s the relative
increases to
price of skill-intensive goods actually
fell. Jeffrey Sachs and Howard Shatz
(1994) found no major change or a
where the fact that coIV > w U S is an iinme- slight decline. A more recent paper by
diate consequence of pw > pus. There- Krueger (1997) criticized the methods
fore, trade with less developed countries and data used by these studies, and
increases wage inequality in the United found an increase in the relative price
States. of skill-intensive goods. Nevertheless,
The skill premium in the LDCs will the increase in these prices is relatively
also be equal to wlV after trade opening small, so would not be able to account for
54 Journal of Economic Literature, V o l . X L (Ma,rch 2002)
the large increase in the skill premium the same point by showing that the con-
experienced in the U.S. economy. tent of unskilled labor embedded in the
Second, with trade as the driving U.S. iinports is sinall relative to the
force, increased production of skill- changes in the supply of skills taking
intensive goods should be drawing place during this period.51
workers away froin other sectors. In Although the above arguments sug-
contrast, as documented by Murphy gest that increased international trade
and Welch (1993), Berman, Bound, and with the LDCs is not the major cause of
Griliches (1994) and Autor, Katz, and the changes in the wage structure by it-
Krueger (1998), all sectors, even those self, they d o not rule out a powerful ef-
producing less-skilled goods, increased fect of international trade when it inter-
their demands for more-educated work- acts with technical change. In a world
ers. This pattern is not consistent with with endogenous technical change, in-
trade being the main driving force of creased international trade could affect
the increase in the demand for skilled the types of technologies developed and
workers (though one has to bear in adopted by firms, and have a large ef-
mind the increase in outsourcing in fect through this channel. This possibil-
interpreting this fact; see Robert ity was first raised by Wood (1994), who
Feenstra and Hanson 1999). argued that trade with the LDCs will
Third, a direct implication of the lead to defensive skill-biased innova-
trade view is that, as shown above, tions. Wood, however, did not develop
while demand for skills and inequality the mechanism through which such de-
increased in the United States, the con- fensive innovations could occur. I now
verse should have happened in the illustrate liow trade causes skill-biased
LDCs that have started trading with technical change using the endogenous
the inore skill-abundant U. S, economy. The technology lllodel developed in section
evidence, however, suggests that inore 5.2 (this analysis draws on Aceinoglu
of the LDCs experienced rising in- 1999b).
equality after opening to international Suppose that the United States starts
trade (see Hanson and Ann Harrison trading with the LDCs as discussed
1994; or Donald Robbins 1995). Al- above, and assume that the LDCs al-
though the increase in inequality in a ways use U.S. technologies. Therefore,
nuinber of cases inay have been due to the supply of skilled and un%killed
concurrent political and econoinic re- goodsAin the LDCs is Yh FAJH and
forms, the preponderance of evidence Yl =AlL where as before H / L < H / L .
is not favorable to this basic iinplication T h e immediate effect will be an
of the trade hypothesis. increase in the relative price of
Finally, a nuinber of econon~istshave skill-intensive goods as illustrated by
pointed out that U.S. trade with the
LDCs is not iinportant enough to have a 51 This is probably the wealzest criticisnl against
the trade view, and many studies have pointed out
inajor impact on the U.S. product mar- how international trade could have a larger effect
ket prices and consequently on wages. on U.S. labor market prices in the presence of la-
Paul Kruglnan (1995) illustrates this bor market rents. For example, George Borjas and
Valerie Rainey (1995), Dani Rodrik (1997), and
point by undertaking a calibration of a Dube and Reddy (1999) have argued that the
siinple North-South model. Katz and threat of international trade may reduce wages,
Murphy (1992), Berman, Bound, and especially in sectors with substantial rents, and
this change in bargaining ower may affect the
Griliches (1994) and George Borjas, earnings of unslzilled wor&rs more, increasing
Freeinan, and Katz (1997) emphasize inequality.
Acemoglu: Technical Change, Inequality, and the Labor Market 55
equation (27). Now, recall from the librium price of skill-intensive goods
analysis in section 5.2 that there is a cannot change either. Combining equa-
relative price effect on the direction of tions (18) and (27) gives the post-trade
technical change (because developing skill bias of technology as
technologies to produce the inore ex-
pensive good is more profitable). There-
fore, trade, by making the skill-intensive
goods more expensive, encourages more
skill-biased technical change.
To determine exactly how the direc-
tion of technical change will be affected
by trade, we need to know the market where AfS/AfS is the pre-trade skill bias
sizes for new technologies after trade of technology in the United States.
opening. I t is plausible to assume that The implication is that when the di-
trade opening with the LDCs will not rection of technical change is endoge-
have a major effect on the enforcement nous, trade between the United States
of intellectual property rights in the and the LDCs will induce skill-biased
South. In that case, trade opening will technical progress. T h e result is not
induce skill-biased technical change in only that trade leads to an increase in
the United States. Specifically, as long skill premia, but that this can happen
as after trade opening the United States without the counterfactual iinplications
does not start producing technologies of the standard trade models discussed
for unskilled workers in the LDCs, the above.
relative market sizes for the two types T h e first implication of this induced
of technologies remain at HIL. This im- skill bias is that the impact of trade on
plies that the technology inarket clear- labor markets may b e much larger than
ing condition, equation ( I S ) , no longer predicted by the standard trade models,
holds. In particular, since p j y > ~p j T J ~ which helps against the criticism that
froin equation (27), there will now only the amount of trade the United States
be incentives for skill-biased technical undertakes with the LDCs is not large
change. This process continues until enough. Second, because trade causes
plTJ = p ~ / p j v
= (H/L)-1, so that incentives skill-biased technical change, the fact
to develop the two different types of that all sectors have increased the em-
technologies are balanced. This implies ployinent of skilled workers is consis-
that the skill bias of technical change is tent with trade being the underlying
still determined by equation (18) from cause of the increase in inequality.
section 5.2, i.e., by U.S. domestic rela- Third, for the same reason, there is a
tive supplies alone. Intuitively, the rela- force counteracting the decline in in-
tive price of skill-intensive goods plays equality in the LDCs implied by trade:
two roles in this model. T h e first is to these economies use U. S. technologies,
clear the market for goods (i.e., equa- which are becoming more skill-biased.
tion 27), and the second is to ensure Finally, and quite strikingly, trade
equilibrium in the technology market leaves the relative price of skill-intensive
(cfr. equation 18). Since the technology goods in the United States unchanged
inarket clearing condition relates the in the long-run. Recall that changes in
relative price of skill-intensive goods to relative prices are the usual intervening
the relative supplies in the U.S. market, mechanism in trade models, so a nuin-
which do not change, the long-run equi- ber of studies have concluded that trade
56 Journal o f Econonzic Literature, V ol. X L ( M a r c h 2002)
has not been an important factor in the 7. Changes i n Resiclual Inequality
increase in inequality because the rela-
tive price of skill-intensive goods has T h e previous sections highlighted
not increased much (e.g., Lawrence and that there are major unanswered ques-
Slaughter 1994; Sachs and Shatz 1995). tions regarding the causes of the in-
In this model, the long-run relative crease in inequality, the reasons for the
price of skill-intensive goods in the faster skill-biased technical change in
United States is unaffected by trade. the past few decades, and the determi-
More generally, induced skill-biased nants of the fall in the wages of low-
technical change in the United States skill workers. These questions awaiting
implies that trade will increase the further research notwithstanding, we
price of slzill-intensive goods by only have a reasonably simple and useful
a limited amount, while still having framework, and the beginnings of con-
a major effect on the U.S. labor sistent answers. In contrast, in this and
inarlzet . the next section, I discuss areas where
Notice finally that the interaction be- answers are much inore tentative, and
tween international trade and technical there is a greater need for future re-
change inay help to explain the decline search. I begin with residual inequality.
in the wages of low-skill workers. In- A major issue that most models dis-
creased international trade acts as an cussed so far failed to address is the dif-
increase in the supply of unskilled ferential behavior of returns to school-
workers, and as shown in our basic ing and residual inequality during the
framework in section 3, this would put 1970s. I argue in this section that an
downward pressure on unskilled wages. explanation for this pattern requires
Borjas, Freeman, and Katz (1987), for models with multi-dimensional slzills.
example, provide evidence that in- 7.1 A Single Inclex Model
creased international trade during the of Residual Ineqz~ality
1980s reduced the wages of high
school graduate worlzers (though they The simplest model of residual in-
also suggest that the effect of the immi- equality is a single index model, in
gration of less-skilled workers was which there is only one type of skill,
greater). though this skill is only imperfectly ap-
Overall, international trade could still proximated by education (or experi-
be a major driving force of the changes ence). Expressed alternatively, in a sin-
in the wage structure. However, for gle index model observed and
increased trade to have such a large unobserved skills are perfect substi-
effect on the structure of wages-and to tutes. Consider, for example, the model
avoid the aforementioned counterfac- developed above, but suppose that in-
tual implications-it must cause a stead of skills, we observe education,
change in the path of technological e.g., whether the individual is a college
progress. So the combination of trade graduate, which is imperfectly corre-
opening and endogenous technical lated with true slzills. A fraction 4, of
change gives an alternative theory for college graduates are highly skilled,
acceleration: skill-biased technical change while a fraction $,, < $, of noncollege
accelerates, neither exogenously nor in graduates are highly skilled. Denote
response to the changes in the supply the skill premium by o = w H / w L .The
of skilled workers, but due to trade college premium, i.e., the ratio of
opening. average college to noncollege wages, is
Acemoglu: Technical Change, Inequality, and the Labor Market 57
zero disturbance term. Here yt is the price
of observed skills, while 8t is the price
of unobserved skills. T h e education
-
-
@cw+ ( 1 - 0,) premium can be written as
+ (1 - $n)'
$n(J)
while within-group inequality, i.e., the In at E E(ln w,t I hlt = 1) - E(ln zolt / hZt= 0 )
= yt + @t(A~t - Aot)
ratio of the wage of high-wage college
graduates (or noncollege graduates) to where Ait = E(ln zolt 1 hlt = 1) and Aot is
that of low-wage college graduates (or defined similarly. Residual inequality
noncollege graduates), is ~owzthln= (J) (since can be measured by Vnr(A,t 11%= 0 ) and
high-wage workers in both groups earn Var(AztI h = 1) .
Z U H , while low-wage workers earn W L ) . T o simplify the discussion, assume
As long as @, and @,, remain constant, we that there is perfect sorting into educa-
and w1ulthln will always move together. tion, in the sense that there exists a
Therefore, an increase in the returns to threshold z such that all individuals
observed skills-such as education-will with unobserved ability above n obtain
also be associated with an increase in the education. Then within-group inequal-
returns to unobserved skills. ity among high and low-education work-
This framework provides a natural ers will move in opposite directions as
starting point, linking between- and long as the price of observed skills, 8, is
within-group inequality. I t predicts that constant: as n declines (and average
within- and between-group inequality education increases), Var(Azt1 h = 1) will
should move together. But during the increase, but Vnr(A,t I h = 0 ) will fall. In-
1970s, returns to schooling fell while tuitively, there are more and more
residual (within-group) inequality in- "marginal" workers added to the high-
creased sharply. W e can only account education group, creating more unob-
for this fact by positing a decline in 4, served heterogeneity in that group and
relative to @,, of a large enough magni- increasing within-group inequality. But
tude to offset the increase in a; this in contrast, the low-education group be-
would ensure that during the 1970s the comes more homogeneous. Therefore,
college premium could fall despite the without a change in the prices for unob-
increase in within-group inequality. A served skills, this approach cannot ac-
large decline in 4, relative to @,, would count for the silnultaneous increase
predict a very different behavior of the in inequality among both low- and
college premium within different co- high-education groups.5"
horts. Yet the appendix shows little evi- 7.3 Churning and Residual Inequality
dence in favor of this. I therefore con-
Another approach emphasizes that
clude that the single index model cannot
workers of all levels of education may
explain the changes in residual inequality
face difficulty adapting to changes. This
during the 1970s and the 1980s.
7.2 Sorting ancl Residual Inequality S 2 A natural variation on this theme would be a
situation in which y and 8 move to ether How-
Another approach would conlbine ever, this x,ill run into the same pro%lems as the
educational sorting with an increase in
the demand for skills. Suppose that
a
sin le index model: if y and 8 always move to-
get er, then such a model would predict that
within-group ine uality should have fallen during
wages are given by In w,t = 8tn, + yth, + ~ , t the 197'0s Thcre're, models based on sorting also
require a mechanism for the prices of observed
where h, is a dummy for high education, and unobserved sltills to move differently during
a, is unobserved ability, and elt is a mean the 1970s.
58 Journal of Economic Literature, Vol. XL (March 2002)
has been argued by Piore and Sabel 1970s. Therefore, a mechanism that
(1984), and more recently by Aghion, could lead to differential behavior in
Howitt, and Violante (2000) and Gould, the prices to observed and unobserved
Moav, and Weinberg (2000). Violante skills is still necessary.54
(2000) offers an alternative, but related
theory, where increased churning is 7.4 A Two-Irzdex Model
driven by the greater productivity gap of Residual Inequality
between different firms (because of
Since models based on a single index
faster embodied technical change) and
of skill (or models where different types
increased skill dynamics of ex ante
of skills are perfect substitutes) are in-
identical workers (which he contrasts
consistent with the differential behavior
with the argument that it is innate abil-
of returns to schooling and within-group
ity differences that matter). According
inequality during the 1970s, an obvious
to all these approaches, an increase in
next step is to consider a two-index
inequality results from more rapid tech-
model where observed and unobserved
nical change, not because of skill bias
skills are imperfect substitutes. There
but because of increased "churning" in
are good reasons to expect that in the
the labor market. Aghion, Howitt, and
real world skills are multidimensional.
Violante (2000), for example, suggest
For example, the social psychologist
that only some workers will be able to
Howard Gardner (1988) makes a strong
adapt to the introduction of new tech-
case that we should always think of
nology, and this will increase wage
skills as multidimensional, and that the
inequality.
standard IQ measures fail to capture
An advantage of this approach is that
this multidimensionality. There is a
it is in line with the increased earnings
long tradition in economics building on
instability pointed out by Gottschalk
the Roy model which uses models with
and Robert Moffit (1994). However,
multidimensional skills to analyze wages
there is relatively little evidence, other
and returns to schooling (e.g., Robert
than this increase in earnings instabil-
Willis and Sherwin Rosen 1975). In the
ity, supporting the notion that there is
context of the determinants of earnings,
more churning in the labor market. The
a recent paper by Bowles, Gintis, and
data on job creation and job destruction
reported by Steven ~ a v i s John
, Halti- 54An interestin theory similar to the churning
wanger, and Scott Schuh (1996) shows models that coulf lead to such a differential be-
havior is advanced by Galor and Tsiddon (1997).
no increase in job reallocation during They draw a distinction between abiljty and edu-
the 1980s or the early 1990s, and most cation, and argue that returns to abilit increase
evidence does not indicate much of a faster during periods o f rapid tecXnological
change. I f we view the 1970s as a period o f ra id
decline in job stability over this period technolo ical change, as suggested above, this t i e -
(e.g. Francis Diebold, David Neumark, ory woul% imp1 an increase in the returns to abil-
and Daniel Polsky 1997; or Henry ity (unobserve8 skills) durin this period Never-
theless, this explanation is st81 not consistent with
Farber 1995).53 Also theories based on the facts because high-ability individuals are more
churning do not naturally predict a di- likely to be high education, so rapid technolo ical
vergence between returns to education pro ress should also increase returns to schoofing.
Pergaps a combination o f this mechanism with dif-
and residual inequality during the ferential sorting into education, or with imperfect
substitution between high- and low-education
53More recent evidence indicates that there workers, might be able to account for the diver-
may have been a decrease in job tenure during the gence between returns to schooling and residual
later parts o f the 1990s (see Neumark, Polsky, and ine uality during the 197Os, but such a model has
Daniel Hansen 1999). n o t i e e n developed yet.
60 Journal of Econojmic Literature, Vol. XL (March 2002)
Overall, single index models are not substantially in the United States and
capable of explaining the changes in re- the United Kingdom, but remained
sidual inequality over the past thirty fairly stable in many continental Euro-
years, and we do not yet know how im- pean econon~ies(see for example Davis
portant various factors are. Analysis of 1995; Gottschalk and Timothy Smeeding
the determinants of residual inequality 1999).
and the reasons why there was an explo- T h e standard explanation for this
sion in overall inequality beginning in divergent behavior, succinctly summa-
the 1970s remains a inajor research rized by Kruginan (1994) and O E C D
area. (1994), and sometimes referred to as
the Krugman hypothesis, maintains that
8 . Cross-Country Patterns inequality did not increase as much (or
not at all) in Europe because labor mar-
So far, I have focused on U.S. wage ket institutions there encourage wage
inequality patterns and on the incen- compression, limiting the extent of in-
tives to develop new technologies coin- equality. This can be captured in the
ing from the U.S. supply of skills. The competitive framework of section 3,
cross-country dimension presents a where firms are always along their rela-
number of challenges, in particular be- tive deinand curve, by assuming that
cause it is difficult to explain why in- labor inarket institutions iinpose an ex-
equality increased much inore in soine ogenous skill premium G = W H / W L . This
countries than others. In addition, for implies:
the endogenous technology view, it is
important to know whether it is the
relative supply of skills in each country
or in the world as a whole that deter-
mines the direction of technical change. where the level of enlployinent of un-
I now briefly discuss these issues. skilled workers, 1, will generally be less
than their labor supply L because of
8.1 Dzjrferences in Inequality Patterns
wage compression. A inore conlpressed
Although the tendency toward wage structure-i.e., a lower 6-will in-
greater inequality has been a feature in crease the unemployment of unskilled
Inany developed and less developed workers, given by L - 1.
countries (see Freeman and Katz 1995; T h e view that wages are more com-
and Berinan and Machin 2000), there pressed in Europe clearly has soine
are also inarked differences in the be- merit. Francine Blau and Lawrence
havior of within- and between-group in- Kahn (1995) show that the inajor differ-
equality across these countries. Katz, ence in overall inequality between the
David Blanchflower, and Gary Loveman United States and Inany continental
(1995), Murphy, Riddell, and Roiner European econonlies is not in the 90-
(1998), and Card and Lemieux (2000) SO differential, but in the 50-10 differ-
show that the differential behavior of ential. This suggests that the lninimuin
the supply of skills can go a long way wage, strong unions, and generous
toward explaining the differences in transfer progranls in Europe are in
the returns to schooling, especially be- part responsible for the relative wage
tween the United States, Canada, and coinpression in Europe.
the United Kingdom. Nevertheless, it is Nevertheless, the Krugman hypothe-
puzzling that wage inequality increased sis runs into two difficulties. First, unless
Acemoglu: Technical Change, Inequality, and the Labor Market 61
firms can either produce with some productivity of unskilled workers (see
existing technology, a , or upgrade to a Acemoglu and Pischke 1999 for this
superior technology, a' = u + a , at cost y. argument in the context of training).
The profit to upgrading the technology As long as the cost of upgrading to
used by a skilled worker is (1 - P)aq - y, the new technology, y,is small, i.e., less
both in the United States and Europe. than yu, firms will upgrade both with
The new technology will therefore be skilled and unskilled workers in the
adopted when United States and Europe. I n this case,
cross-country inequality levels will be
stable. This corresponds to the situation
Note that there is a holdup problem, dis- in the 1950s and the 1960s. In contrast,
couraging upgrading: a fraction P of the if y is high, for example, because the
productivity increase accrues to the technological improvements of the 1980s
worker due to rent sharing (Paul Grout are more expensive to implement, there
1984; Aceinoglu 1996). may be a divergence in inequality be-
The incentives to upgrade the tech- tween the two economies. For instance,
nology used by unskilled workers, on if y E (yU,a), then new technology will not
the other hand, differ between the be adopted with unskilled workers in
United States and Europe. In the the United States, but it will be used
United States, this profit is given by with unskilled workers in Europe. As a
(1- p)a - y. So, the new technology will result, while wage inequality increases
be adopted with unskilled workers if in the United States, it will remain
stable in Europe. Therefore, a simple
story for cross-country differences in
inequality trends emerges from this
Clearly, yu < ys, so adopting new tech- model: wage compression encourages
nologies with skilled workers is more the use of more advanced technologies
profitable. The returns to introducing with unskilled workers, and acts to rein-
the new technology are different in force itself in Europe. In contrast, tech-
Europe because the minimum wage is nological developments can harm the
binding for unskilled worlzers. To sim- earnings of low-skill U.S. workers who
plify the discussion, suppose that even are not protected by this type of wage
after the introduction of new technology, compression. Whether the interaction
the i~linimum wage binds, i.e., between wage compression and tech-
g > P(A + a ) . Then, the return to intro- nology choice could be important in ex-
ducing the new technology in Europe plaining European inequality and un-
with unskilled workers is a - y, and firms employment patterns is an area for
will do so as long as y < a. Since a > yU, future study.
firins in Europe have greater incentives
to introduce advanced technologies with 8.2 International Determinants
unskilled workers than in the United
of Teclznology
States. This is because the binding mini- The endogenous technology frame-
mum wage in Europe makes the firm the work developed above links the skill
Jill1 residual claimant of the increase in bias of technology to the relative supply
the productivity of unskilled workers. of skills. There are a number of inter-
This highlights that in an economy with a esting and difficult issues that arise
compressed wage structure, firms inay when we consider the international di-
have a greater incentive to increase the mension. Here I simply mention some
Acemoglu: Technical Change, Inequality, and t h e Labor Market 63
TABLE A1
CITANGES
ANNUALIZED IN OVERALL
AND RESIDUALWAGEINEQUALITY
((FROM KATZAND AUTOR)
Note: The numbers give 10 x annualized changes from Table 4 of Katz and Autor (2000). 90-10 is the difference
between the 90th and 10th percent of the log wage or residual distlibution, and 50-10 is the difference between the
median and 10th percent of the corresponding distribution. The residuals are estimated from log earnings regres-
sions with nine education dummies, a quartic in experience and their interactions. See notes to Tables 3 and 4 in
Katz and Autor (2000).
and the experience controls. T h e sample includes sources: decennial censuses, and March CPSs and
all full-time full-year workers between the ages of May CPSs (and later Outgoing Rotation Group
18 and 65, and except those with the lowest 1 per- files-ORGs). These numbers show no significant
cent earnings. Earnings for top coded observatiolls change in residual or overall inequality during the
are calculated as the value of the to code times 1960s, and consistent increases in inequality from all
1.5. The relative supply of skills is c a t u l a t e d from sources during the 1970s and the 1980s. For example,
a sample that includes all workers between the the data from the Census and the March CPSs indi-
ages of 18 and 65. It is defined as the ratio of cate that the 90-10 differential increased about 0.10 a
college equivalents to non-college equivalents, cal- year between 1970 and 1979, while the 50-10 dif-
culated as in Autor, Katz, and Krueger (1998) ferential increased by about 0.11 a year during the
using weeks worked as weights. I n particular, col- same period. The May CPS data show a smaller
lege e uivalents = college graduates + 0.5 x work- increase in the 90-10 differentials during this
\
ers wit some college, and noncollege equivalents =
high school dropouts + high school graduates + 0.5 x
riod, but a corn arable increase in the 50-10
ferelltial OveraK although there is less uniformity
iz:
workers with some college. among data sources regarding the behavior of re-
Samples used for overall and residual wage in- sidual inequality than returns to schoolin (see
equality include ollly white male full-time full year Katz and Autor 2000), there is considerabfe evi-
workers between the ages of 18 and 65, and ex- dence that residual and overall inequality started
clude those earning less than half the real value to increase during the 1970s.
of the 1982 minimum wage converted from nomi-
A3. Can Cornposition Effects Explain
nal dollars using the personal consumption expen-
diture deflator (see Katz and Autor 2000). Earn- Inequality Changes?
ings for top coded observations are calculated as A possible explanation for the patterns we ob-
the value of the top code times 1.5. serve could be changes in the distribution of un-
observed skills- or more concretely, composition
A2. The Belanolor of Ooerall Inequality
effects. For example, the average ability of workers
During the 1970s with high education may have increased relative to
I n an important paper on the effect of labor that of workers with low education over time.
market institutions on inequality, DiNardo, Here, I document that the increase in the returns
Fortin, and Lemieux (1995) provide evidence sug- to education and residual inequality are not smply
gesting that in the May CPSs, there is no increase due to composition effects. Note first that compo-
in inequality during the 1970s. I n table A l , I dis- sition effects cannot by themselves explain the re-
lay numbers from the survey by Katz and Autor cent changes in inequality: as noted in subsection
b 0 ), who report changes in residual inequality 7.2, compositioll effects suggest that ine uality
for the past four decades from three different among educated and uneducated workers s ould %
Journal of Economic Literature, Vol. XL (March 2002)
TABLE A2
COMPOSITION
EFFECTS
Bornin19- 06-10 11-15 16-20 21-25 26-30 31-35 3640 4145 46-5 51-55
Year J +
Panel A
1950 1.448 1.370 1.175 1.093
1960 1.551 1.564 1.525 1.421 1.303 1.132
1970 1.680 1.656 1.613 1.539 1.392 1.153
1980 1.567 1.560 1.538 1.402 1.222 1.063
1990 1.798 1.761 1.723 1.674
Panel B
Panel C
Note: The top panel gives the college premium from the census indicated at the beginning of the row for cohoits
born in the five-year intervals indicated at the head of the column. For example, the first number is for individuals
born between 1906-10 from the census of 1950. The college premium is defined as the wages of workers from that
cohort with a college degree or more hvided by the wages of workers from that cohoit with twelve years of
schooling. The bottom panel gives the change in the college premium for a given cohort between the two indicated
dates and the difference between the wage growth of two neighboring cohorts as indicated by equations (34) and
(35).All data are from the decennial censuses for white males born in the United States.
move in opposite directions. This suggests that Consider changes in cohort quality first. If, as
changes in the true returns to skills have played at many claim, the U.S. high school system has be-
least some role in the changes in inequality. come worse, we might expect a decline in Ao, with-
More generally, to get a sense of how important out a corresponding decline in Al,. As a result,
composition effects may be, consider a variant of AIt -Aot may increase. Alternatively, as a larger
equation (13) above with two education levels, high fraction of the U.S. population obtains higher edu-
h = 1 and low 1%= 0, and suppose wages are given by cation, it is natural that selection into education
(i.e., the abilities of those obtaining education)
will change. I t is in fact possible that those who
are left without education could have very low un-
where 12, is a dumm for high education, a, is un- observed ability, which would translate into a low level
observed ability, a n X q t is a mean zero disturbance of Aot, and therefore into an increase in Alt -Aot.
term. Define the (log) education premium-the Although these scenarios are plausible, theoret-
difference between the average wages of high and ically the opposite can happen as well. For exam-
low education workers-can be written as ple, many academics who have been involved in
the U.S. education system for a long time com-
In ot = E(ln tu1t1 lxl = 1)- E(ln wit 1 h , = 0) plain about the decline in the quality of universi-
= yt + Alt - Aot ties, while the view that American high schools
where A1, = E(a, 1 lzi = 1) and Aot is defined simi- have become much worse is not shared universally
larly. The increase in the education premium can (e.g., Krueger 1998). The selection argument is
be caused by an increase in yt (a true increase in also more complicated than it first appears. I t is
the returns to skills) or an increase in Alt-Aot. true that, as long as those with high unobserved
There are basically two reasons for an increase in abilities are more likely to obtain higher educa-
A1,-Aot: (1) changes in cohort quality, or (2) tion, an increase in education will depress Aot. But
changes in the pattern of selection into education. it will also depress Alt. T o see why, assume that
Acemoglu: Technical Change, Inequality, and the Labor Market
there is perfect sorting-i.e., if an individual with
ability n obtains education, all individuals with TABLE A3
ability a' > a will do so as well. I11 this case, there CHANGESI N INEQUALITY
BY COIIORT
will exist a threshold level of ability, n, such that (FHOM JUHNET AL. 1993)
only those with a > a obtain education. Next con-
sider a uniform distribution of a , between bo and
bo + b,. Then, Panel A: 90-10 Differentials for Log Weekly Wages
Year of
inarket entry 1964 1970 1976 1982 1988
and
Krueger, Alan. 1993. "How Computers Have logical Creatioity and Economic Pr0gre.s~.NY:
Changed the Wage Structure: Evidence from Oxford U. Press.
Microdata, 1984-1989," Quart. J . Econ. 110, , ed. 1993. The Briti.sh Industrial Reuo1z~-
pp. 33-60. tion: AII Econoinic Perspectioe. CA: Westview
. 1997. "Labor Market Shifts and the Price Press.
Puzzle Revisited," Princeton U. Ind. Rel. work. Murnane, Richard J. and Frank Levy. 1996.
paper 375. Teaching the Ba.sic Neeo Skills. NY: Free Press.
. 1998. "Reassessing the View That Amer. Murphy, Kevin M.; W. Craig Riddell and Paul M.
Schools Are Broken," Princeton U. Ind. Rel. Romer. 1998. "Wages, Skills and Technology in
work. paper 395. the United States and Canada," in General Pur-
. 1999. "Measuring Labor's Share," Ainer. pose Technologies. E . Helpman, ed. MA: MIT
Econ. Reu. 89, pp. 45-51. Press.
Krueger, Alan B. and Jorn-Steffen Pischke. 1997. Murphy, Kevin M. and Finis Welch. 1992. "The
"Observations and Conjectures on the U.S. Em- Structure of Wages," Quart. J . Econ. 107, pp.
ployment Miracle," NBER work. paper 6146. 255-85.
Krugman, Paul. 1994. "Part and Prospective , 1993. "Industrial Change and the Rising
Causes of High Unemployment," in Reducing Ilnportallce of Skill," in Uneuei~ Tide.s. S.
Unenzployment: Current Issues and Policy Op- Danziger and P. Gottschalk, eds. NY: Russell
tions. MO: Fed. Reserve Bank Kansas City. Sage.
, 1995. "Growing World Trade: Causes and , 2000. "Wage Differentials in the 1990s: Is
Conseauences." Broolcines u
Pav. Econ. Act. 1,
8
the Glass Half Full or Half Empty?" unpub. pa-
pp. 327'-77. per.
Krusell, Per; Lee Ohanian, Victor Rios-Rull and Nelson, Richard and Edmund Phelps. 1966. "In-
Giovaillli Violante. 2000. "Capital Skill Comple- vestment in Humans, Technological Diffusion
Journal of Econonaic Literature, Vol. XL (March 2002)
and Economic Growth," A E A Papers Proceed- Sachs, Jeffrey and Howard Shatz. 1994. "Trade
ings 56, pp. 69-75. and Jobs in U.S. Manufacturing," Broolcings
Newell, Richard; Adain Jaffee and Robert Stavins. Pap. Econ. Act. 0:1, pp. 1-69.
1999. "The Induce Innovation Hypothesis and Samuelson, Paul. 1965. "A Theory of Induced In-
Energy-Saving Technological Change," Quart. novations Along Kennedy-Weisacker Lines,"
J . Econ. 114, pp. 907-40. Reu. Econ. Statist. 47, pp. 444-64.
Nickell, Stephen and Brian Bell. 1996. "The Col- Scherer, Michael. 1984. Innountion and Growth.
lapse in Demand for the Unskilled and Unem- MA: MIT Press.
ployment across the OECD," Oxford Reu. Econ. Schiff, Ze'ev and Ehud Ya'ari. 1989. Intifada, The
Pol. 11, pp. 40-62. Palestinian Uprising-ism el'.^ Third Front. NY:
Neumark, David; Daniel Polsky and Daniel Han- Simon and Schuster.
sen. 1999. "Has Job Stability Declined Yet? Evi- Schmookler, Jacob. 1966. Invention and Econoinic
dence for the 1990s," J. Lab. Econ. 17:4, S29- Growth. MA: Harvard U. Press.
64. Schultz, Theodore. 1975. "The Value of the Ability
Nye, David E . 1990. Electrifying America: Social to Deal with Disequalibria," J. Econ. Lit. 13,
Meanings of a Neeo Technology, 1880-1940. p p 827-46.
MA: MIT Press. Thesmar, David and Mathias Thoenig. 2000. "Crea-
OECD. 1994. Jobs Study. Vols. 1 and 2. Paris and tive Destruction and Firm Organization Choice:
Washington, DC: OECD. A New Look into the Growth-Inequality Rela-
Oliner, Stephen and Daniel Sichel. 1994. "Com- tionship," Quart. J . ECOII.115:4, pp. 1209-38.
puters and Output Growth Revisited: How Big Tinbergen, Jan. 1975. Income Difference: Recent
Is the Puzzle?" Brookings Pap. Econ. Act. 0 3 , Research. Amsterdam: North-Holland.
p p 273-317. Violante, Giovanni L. 2000. "Technological Accel-
Pierce, Brooks. 2001. "Compensation Inequality," eration, Skill Transferability and the Rise in Re-
forthcoming, Quart. J . Econ. sidual Inequality," Ulliversitp College London
Piketty, Thomas and Emannuel Saez. 2001. "In- mimeo.
come Inequality in the United States, 1913- Welch, Finis. 1970. "Education in Production," J .
1998," MIT and Harvard mimeo. Polit. Econ. 78, pp. 312-27.
Piore, Michael and Charles Sabel. 1984. The Sec- , 1973. "Black-\Vhite Differences in Re-
ond 1ndu.strial Divide. NY: Basic Books. turns to Schooling," Anzer. Econ. Reu. 63:5, pp.
Reynolds, Lloyd G. 1951. The Structure of Labor 893-907.
Markets; Wages and Labor Atobility in Theory Williamson, Jeffrey. 1990. Coping with City
and Practice. NY: Harper. Groeoth Dzcrin the British Indzcstrial Reuolz~-
Riddell, Craig. 1993. "Unionization in Canada and t i o n MA: Camfridge U Press.
the United States: A Tale of Two Countries," in Williamson, Jeffrey G. and Peter H . Lindert. 1980.
Sinall Differences That Matter. David Card Anzerican Ineqz~ality: A Atacroeconomic His-
and Richard Freeman, eds. IL: U. Chicago tory. NY: Academic Press.
Press. Willis, Robert and Sherwin Rosen. 1975. "Educa-
Robbins, Donald J. 1995. "Trade and Inequality in tion and Self-Selection," J . Polit. Econ. 87, pp.
Seven Developing Countries: Summary and 817-36.
Synthesis of Country Studies," Harvard, HIID Wood, Adrian. 1994. North-South Trade, Employ-
mimeo. nzent and Inequality: Changing Fortunes i i ~a
Rodrik, Dani. 1997. Has Globalization Gone Too Slcill Drioen World. Oxford: Clarendon Press.
Far? Washington DC: Institute Int. Econ. Woolf, Arthur G. 1984. "Electricity, Productivity,
Romer, Paul M. 1990. "Endogenous Technological and Labor Saving: American Manufacturing
Change," J . Polit. Econ. 98:5, S71-S102. 1900-1929," Explor. Econ. Hist. 2 1 2 , pp. 176-
Rubinstein, Yonna and Dani Tsiddon. 1999. "Cop- 91.
ing with Technological Progress: The Role of Zuboff, Shohana. 1988. In the Age of the Sinart
Ability i i ~Atalcing Inequality So Per.sistent," Tel Machine, The Future of Work and Power. NY:
Aviv U. work. paper. Basic Books.
http://www.jstor.org
LINKED CITATIONS
- Page 1 of 10 -
This article references the following linked citations. If you are trying to access articles from an
off-campus location, you may be required to first logon via your library web site to access JSTOR. Please
visit your library's website or contact a librarian to learn about options for remote access to JSTOR.
[Footnotes]
3
Endogenous Technological Change
Paul M. Romer
The Journal of Political Economy, Vol. 98, No. 5, Part 2: The Problem of Development: A
Conference of the Institute for the Study of Free Enterprise Systems. (Oct., 1990), pp. S71-S102.
Stable URL:
http://links.jstor.org/sici?sici=0022-3808%28199010%2998%3A5%3CS71%3AETC%3E2.0.CO%3B2-8
3
Induced Bias in Innovation and the Theory of Distribution
Charles Kennedy
The Economic Journal, Vol. 74, No. 295. (Sep., 1964), pp. 541-547.
Stable URL:
http://links.jstor.org/sici?sici=0013-0133%28196409%2974%3A295%3C541%3AIBIIAT%3E2.0.CO%3B2-6
3
Factor Prices and Technical Change in Agricultural Development: The United States and
Japan, 1880-1960
Yujiro Hayami; V. W. Ruttan
The Journal of Political Economy, Vol. 78, No. 5. (Sep. - Oct., 1970), pp. 1115-1141.
Stable URL:
http://links.jstor.org/sici?sici=0022-3808%28197009%2F10%2978%3A5%3C1115%3AFPATCI%3E2.0.CO%3B2-Q
NOTE: The reference numbering from the original has been maintained in this citation list.
http://www.jstor.org
LINKED CITATIONS
- Page 2 of 10 -
3
The Supply of Skilled Labour and Skill-Biased Technological Progress
Michael T. Kiley
The Economic Journal, Vol. 109, No. 458. (Oct., 1999), pp. 708-724.
Stable URL:
http://links.jstor.org/sici?sici=0013-0133%28199910%29109%3A458%3C708%3ATSOSLA%3E2.0.CO%3B2-5
4
Education in Production
F. Welch
The Journal of Political Economy, Vol. 78, No. 1. (Jan. - Feb., 1970), pp. 35-59.
Stable URL:
http://links.jstor.org/sici?sici=0022-3808%28197001%2F02%2978%3A1%3C35%3AEIP%3E2.0.CO%3B2-7
6
Capital-Skill Complementarity and Inequality: A Macroeconomic Analysis
Per Krusell; Lee E. Ohanian; José-Víctor Ríos-Rull; Giovanni L. Violante
Econometrica, Vol. 68, No. 5. (Sep., 2000), pp. 1029-1053.
Stable URL:
http://links.jstor.org/sici?sici=0012-9682%28200009%2968%3A5%3C1029%3ACCAIAM%3E2.0.CO%3B2-B
6
Technological Revolutions
Francesco Caselli
The American Economic Review, Vol. 89, No. 1. (Mar., 1999), pp. 78-102.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28199903%2989%3A1%3C78%3ATR%3E2.0.CO%3B2-S
9
Black-White Differences in Returns to Schooling
Finis Welch
The American Economic Review, Vol. 63, No. 5. (Dec., 1973), pp. 893-907.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28197312%2963%3A5%3C893%3ABDIRTS%3E2.0.CO%3B2-1
10
Measuring Labor's Share
Alan B. Krueger
The American Economic Review, Vol. 89, No. 2, Papers and Proceedings of the One Hundred
Eleventh Annual Meeting of the American Economic Association. (May, 1999), pp. 45-51.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28199905%2989%3A2%3C45%3AMLS%3E2.0.CO%3B2-Z
NOTE: The reference numbering from the original has been maintained in this citation list.
http://www.jstor.org
LINKED CITATIONS
- Page 3 of 10 -
11
Labor Market Institutions and the Distribution of Wages, 1973-1992: A Semiparametric
Approach
John DiNardo; Nicole M. Fortin; Thomas Lemieux
Econometrica, Vol. 64, No. 5. (Sep., 1996), pp. 1001-1044.
Stable URL:
http://links.jstor.org/sici?sici=0012-9682%28199609%2964%3A5%3C1001%3ALMIATD%3E2.0.CO%3B2-K
24
Capital-Skill Complementarity and Inequality: A Macroeconomic Analysis
Per Krusell; Lee E. Ohanian; José-Víctor Ríos-Rull; Giovanni L. Violante
Econometrica, Vol. 68, No. 5. (Sep., 2000), pp. 1029-1053.
Stable URL:
http://links.jstor.org/sici?sici=0012-9682%28200009%2968%3A5%3C1029%3ACCAIAM%3E2.0.CO%3B2-B
24
Long-Run Implications of Investment-Specific Technological Change
Jeremy Greenwood; Zvi Hercowitz; Per Krusell
The American Economic Review, Vol. 87, No. 3. (Jun., 1997), pp. 342-362.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28199706%2987%3A3%3C342%3ALIOITC%3E2.0.CO%3B2-X
24
Capital-Skill Complementarity and Inequality: A Macroeconomic Analysis
Per Krusell; Lee E. Ohanian; José-Víctor Ríos-Rull; Giovanni L. Violante
Econometrica, Vol. 68, No. 5. (Sep., 2000), pp. 1029-1053.
Stable URL:
http://links.jstor.org/sici?sici=0012-9682%28200009%2968%3A5%3C1029%3ACCAIAM%3E2.0.CO%3B2-B
25
The Determinants of Earnings: A Behavioral Approach
Samuel Bowles; Herbert Gintis; Melissa Osborne
Journal of Economic Literature, Vol. 39, No. 4. (Dec., 2001), pp. 1137-1176.
Stable URL:
http://links.jstor.org/sici?sici=0022-0515%28200112%2939%3A4%3C1137%3ATDOEAB%3E2.0.CO%3B2-X
32
Endogenous Technological Change
Paul M. Romer
The Journal of Political Economy, Vol. 98, No. 5, Part 2: The Problem of Development: A
Conference of the Institute for the Study of Free Enterprise Systems. (Oct., 1990), pp. S71-S102.
Stable URL:
http://links.jstor.org/sici?sici=0022-3808%28199010%2998%3A5%3CS71%3AETC%3E2.0.CO%3B2-8
NOTE: The reference numbering from the original has been maintained in this citation list.
http://www.jstor.org
LINKED CITATIONS
- Page 4 of 10 -
33
Endogenous Technological Change and Wage Inequality
Huw Lloyd-Ellis
The American Economic Review, Vol. 89, No. 1. (Mar., 1999), pp. 47-77.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28199903%2989%3A1%3C47%3AETCAWI%3E2.0.CO%3B2-U
44
Changes in Unemployment and Wage Inequality: An Alternative Theory and Some Evidence
Daron Acemoglu
The American Economic Review, Vol. 89, No. 5. (Dec., 1999), pp. 1259-1278.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28199912%2989%3A5%3C1259%3ACIUAWI%3E2.0.CO%3B2-N
References
Changes in Unemployment and Wage Inequality: An Alternative Theory and Some Evidence
Daron Acemoglu
The American Economic Review, Vol. 89, No. 5. (Dec., 1999), pp. 1259-1278.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28199912%2989%3A5%3C1259%3ACIUAWI%3E2.0.CO%3B2-N
NOTE: The reference numbering from the original has been maintained in this citation list.
http://www.jstor.org
LINKED CITATIONS
- Page 5 of 10 -
The Economic Returns to Schooling in the West Bank and Gaza Strip
Joshua D. Angrist
The American Economic Review, Vol. 85, No. 5. (Dec., 1995), pp. 1065-1087.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28199512%2985%3A5%3C1065%3ATERTSI%3E2.0.CO%3B2-O
NOTE: The reference numbering from the original has been maintained in this citation list.
http://www.jstor.org
LINKED CITATIONS
- Page 6 of 10 -
Technological Revolutions
Francesco Caselli
The American Economic Review, Vol. 89, No. 1. (Mar., 1999), pp. 78-102.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28199903%2989%3A1%3C78%3ATR%3E2.0.CO%3B2-S
The Dynamo and the Computer: An Historical Perspective on the Modern Productivity
Paradox
Paul A. David
The American Economic Review, Vol. 80, No. 2, Papers and Proceedings of the Hundred and
Second Annual Meeting of the American Economic Association. (May, 1990), pp. 355-361.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28199005%2980%3A2%3C355%3ATDATCA%3E2.0.CO%3B2-X
Technical Change and Human-Capital Returns and Investments: Evidence from the Green
Revolution
Andrew D. Foster; Mark R. Rosenzweig
The American Economic Review, Vol. 86, No. 4. (Sep., 1996), pp. 931-953.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28199609%2986%3A4%3C931%3ATCAHRA%3E2.0.CO%3B2-N
NOTE: The reference numbering from the original has been maintained in this citation list.
http://www.jstor.org
LINKED CITATIONS
- Page 7 of 10 -
Investment and Wages in the Absence of Binding Contracts: A Nash Bargaining Approach
Paul A. Grout
Econometrica, Vol. 52, No. 2. (Mar., 1984), pp. 449-460.
Stable URL:
http://links.jstor.org/sici?sici=0012-9682%28198403%2952%3A2%3C449%3AIAWITA%3E2.0.CO%3B2-2
NOTE: The reference numbering from the original has been maintained in this citation list.
http://www.jstor.org
LINKED CITATIONS
- Page 8 of 10 -
Factor Prices and Technical Change in Agricultural Development: The United States and
Japan, 1880-1960
Yujiro Hayami; V. W. Ruttan
The Journal of Political Economy, Vol. 78, No. 5. (Sep. - Oct., 1970), pp. 1115-1141.
Stable URL:
http://links.jstor.org/sici?sici=0022-3808%28197009%2F10%2978%3A5%3C1115%3AFPATCI%3E2.0.CO%3B2-Q
NOTE: The reference numbering from the original has been maintained in this citation list.
http://www.jstor.org
LINKED CITATIONS
- Page 9 of 10 -
Education in Production
F. Welch
The Journal of Political Economy, Vol. 78, No. 1. (Jan. - Feb., 1970), pp. 35-59.
Stable URL:
http://links.jstor.org/sici?sici=0022-3808%28197001%2F02%2978%3A1%3C35%3AEIP%3E2.0.CO%3B2-7
NOTE: The reference numbering from the original has been maintained in this citation list.
http://www.jstor.org
LINKED CITATIONS
- Page 10 of 10 -
NOTE: The reference numbering from the original has been maintained in this citation list.