Wes lgg
POLICY RESEARCH WORKING PAPER    1881
The Asian Miracle                                                        The policy differences
between 'accumulation" and
and Modern Growth Theory                                                 "assimilation" growth theories
may be much smaller than
the conceptual or analytic
Richard R. Nelson
differences. Can the Asian
Howard Pack
miracle be explained in terms
of capital investments? Or
were entrepreneurship,
innovation, and learning
significant factors in the rapid
growth of the Asian tigers?
The World Bank
Development Research Group
February 1998



l POLICY RESEARCH WORKING PAPER 1881
Summary findings
In the past 35 years, China, Hong Kong, Korea,               new technologies they were adopting from more
Singapore., and Taiwan (China) have transformed              advanced industrial nations. They see investment in
themselves from technologically backwards and poor           human and physical capital as an essential but far from
economies to relatively modern, affluent economies.          sufficient part of assimilation. In addition, people must
Each has experienced more than a fourfold increase in        learn about, take the risk of operating, and come to
per capita income. In each, a significant number of firms    master technologies and other practices new to the
are producing technologically complex products               country, if not the world. The emphasis for assimilation
competitive with firms in Europe, Japan, and the United      theorists is on innovation and learning, rather than on
States. Their growth performance has exceeded that of        marshalling. If one marshals but does not innovate and
virtually all comparable economies. How they did it is a     learn, development does not follow. These are complex
question of great importance.                                theories that raise as many questions as they answer.
Virtually all theories about how they did it place           Nelson and Pack discuss differences in the way the two
investments in capital stock at the center of the            groups of theorists treat four matters:
explanation. Nelson and Pack divide most growth                 * Entrepreneurial decisionmaking.
theories about the Asian miracle into two groups:               * The nature of technology.
* The "accumulation" theories stress the role of             * The economic capabilities possible with a well-
capital investments in moving these economies "along         educated work force.
their production functions." What lies behind rapid             - The role exports play in a country's rapid
development, according to this type of theory, is very       development.
high investment rates. If a nation makes the investments,      The differences between the theories matter because
marshals the resources, development will follow.             they affect our understanding of why the Asian miracle
- The "assimilation" theories stress the                   happened and because they imply different things about
entrepreneurship, innovation, and learning these             appropriate economic development policy.
economies went through before they could master the
This paper - a product of the Development Research Group - is part of a larger effort in the group to study the impact
of public policy on growth. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington,
DC 20433. Please contact Susheela Jonnakuty, room MC2-615, telephione 202-473-7902, fax 202-522-1154, Internet
address sjonnakuty@worldbank.org. February 1998. (44 pages)
The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about
development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The
papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this
paper are entirely those of the authors. They do not necessarily represenrt the view of the World Bank, its Executive Directors, or the
countries they represent.
Produced by the Policy Research Dissemination Center



The Asian Miracle and Modern Growth Theory
Richard R. Nelson
Columbia University
Howard Pack
University of Pennsylvania
and
The World Bank
We would like to thank M. Abramovitz, G. Akerlof, M. Gersovitz, K. Pavitt, D.
Rodrik, V. Ruttan, L. Westphal, S. Winter, and participants in seminars at the
Johns Hopkins Univeristy,  the Canadian Institute for Advanced Studies,  and
IASSA for comments on earlier drafts. Mu-Yen Hsu provided excellent research
assistance.



/
I



I.    Accumulation and Assimilation Theories of the Asian Miracle
Over the past thirty-five years Korea, Taiwan,China, Singapore, and Hong
Kong, have transformed themselves from technologically backwards and poor, to
relatively modern and affluent economies.   Each has experienced more than a
four  fold  increase  of  per  capita  incomes.    Each  now  has  a  significant
collection of firms producing technologically complex products competing
effectively against rival firms based in the United States, Japan, and Europe.
The  growth  performance  of  these  countries  has  vastly  exceeded  those  of
virtually all other economies that had comparable productivities and income
levels in 1960.   On these grounds alone the question of "how they did it"
obviously is of enormous scientific and policy importance.
It  has  been  less  well  noted  that  their  growth  has  been  historically
unprecedented. The development of Japan in the half century after the Meiji
restoration is widely regarded as comparable.   However,  Japan's per capita
growth rate over this period was less than half that of the Asian nics since
1960. Of course, growth rates in general were slower during this earlier
period.    But  the rate of catch up by the nics  still  is  remarkable.    It
certainly would seem that there is an "Asian Miracle" that cries out for
explanation.
Of  course,  economists  have not  been blind to or unattracted by the
challenge.   Over the last decade a number of different theories have been put
forth purporting to explain the "Asian Miracle."   (Amsden 1989,  Kim and Lau
1994,  Krugman  1994,  Pack and Westphal  1986,  Rodrik  1994,  Westphal  Kim and
Dahlman  1985,  World Bank  1993,  Young  1993).  There  is unanimity  among  the
different theories regarding the identity of some of the key causal factors.
All of the Asian nics have experienced rapid growth of their physical capital
stock. All have been marked by very high rates of investment in human capital.
Virtually all theories about "how they did it" place these investments center
stage in the explanation.
However, there are significant differences in the causal mechanisms
stressed. At the risk of doing some violence to the actual diversity, for our
purposes we find it useful to divide up theories of the Asian Miracle into two
groups.   One group, which we will call "accumulation" theories,  stresses the
role of these investments in moving these economies "along their production
functions."   The other group,  which we will  call  "assimilation"  theories,
stresses the entrepreneurship, innovation and learning that these economies had
to go through before they could master the new technologies they were adopting
from the more advanced industrial nations; it sees investment in human and
1



physical capital as a necessary, but far from sufficient, part of the
assimilation process.
The "accumulation" theory has been pushed hard over the past few years by
several economists, in a way clearly designeci to strip away most of the
"miraculous" from the Asian miracle.   What lies behind rapid development is,
simply, very high investment rates.  Economists who take this point of view do
not deny that adoption and mastering new technology and other modern practices
was an important part of the story.   Rather, the position is that one should
try to explain as much as one can in terms of investments that enable movements
along  a  production  function,  and  see  if  anything  much  is  left  over,  thus
requiring explanation on other grounds.   Several economists who have followed
this path find that, according to their calculations, the lion's share of
increased output per worker can be explained simply by increases in physical
and human capital per worker.  Thus there is little need to assign much of the
credit for the growth "miracle" to entrepreneurship, innovation, or learning,
except insofar as these are terms given to the shift to more capital and
education intensive ways of production. (See e.g. Young 1993, Kim and Lau 1994,
Krugman 1994).
To  assimilation   theorists,  this  point  of  view  seems  odd.               The
technologies that the nics came progressively to master during the 1970's and
1980's were ones with which, in 1960, they had no experience at all... To learn
to use them effectively required the development of new sets of skills, new
ways of organizing economic activity, and becoming familiar with and competent
in new markets.  To do this was far from a routine matter, but involved risk
taking  entrepreneurship  as well as good management.    ( See  e.g.  Pack  and
Westphal  1986,  Dahlman  1994, Amsden  1989.)   What makes  the  Asian  miracle
miraculous  is  that these  countries  did  these  things  so well, while  other
countries were much less successful.  To be sure, adopting the technologies of
the advanced countries required, among other things,  high rates of investment
in physical and human capital, and the nics achieved these high rates.  But to
say that these investments simply enabled these economies to "move along their
production functions" seems a strange use of language.  At the least, it poses
the question of just what is meant by "moving along a production function."
Are we drawing a distinction without a real difference? We do not think
so.   The accumulation account stresses, simply, investments.   The message is
that other countries could have done as well as the successful nics if they had
made the same investment effort. If the nation makes the investments, marshals
the resources, development will follow.  In contrast, the assimilation account
stresses learning about, risking to operate, and coming to master, technologies
2



and other practices that are new to the country,  if not to the world.   The
"marshalling of inputs" is part of the story, but the emphasis is on innovation
and learning,  rather  than  on marshalling.    Under  this  view,  if when  one
marshals but does not innovate and learn, development does not follow.
A convinced accumulationist might respond by saying that, if one educates
the people, and provides them with modern equipment to work with, they will
learn. An assimilationist might respond that the Soviet Union, and the Eastern
European communist economies, took exactly that point of view, made the
investments, and didn't learn.1 There is nothing automatic about the learning
business.  The response of the accumulationist might be that the old communist
countries provided an economic environment where there was no incentive to
learn to be efficient,  either in a technological or an economic sense, much
less to innovate. The assimilation theorist might agree, but then propose that
it is important to understand, therefore, just how the successful nics did it.
The accumulationist would reply that they got the prices right and made the
necessary  public  investments.        Economists  who  stress  entrepreneurship,
innovation, and learning would reply that it is not all that simple, and point
to countries like Spain that have had high investment rates, and have got most
of the prices right, but which are growing at far lower rates than the Asian
nics.
The difference between the theories shows up strikingly in the way they
treat the following four matters: what is involved in entrepreneurial decision
making, the nature of technology, the economic capabilities lent by a well
educated work force, and the role that exporting played in these countries'
rapid development.
Accumulationists pay little explicit attention to firms, seeing their
behavior as being basically determined by the environment - the incentives and
constraints - they face, which determines the actions that are most profitable.
Assimilation theorists,  on the other hand,   see entrepreneurial firms,  and
their ability to learn rapidly, as a critical factor behind the success of
Korea and Taiwan,China, with their behavior supported by their environments,
but only partially determined by external forces.   (See e.g. Hobday 1995, and
Kim forthcoming.) For an assimilation theorist, at least our brand, when firms
contemplate venturing on to ground that is new for them, the profitability of
Easterly  and  Fischer,   1995,  stress  the  low  ex  post  elasticity  of
substitution as an explanation of slow Soviet growth. This could also be
interpreted as reflecting insufficient effort to identify and master new
technology.
3



such venturing is highly uncertain, in the sense of Knight. Some firm managers
will dare to venture; others will choose to stick close to the familiar. Thus,
what firms do is determined by the daring of their decision makers, as well as
by their environment.   And whether an entrepreneurial effort will succeed or
fail also is only partly determined by environmental factors.   It depends, as
well, on the zeal and smarts and learning abilities of firm management and
workers.
Part of the difference here resides in how the different theories see
technology.   Accumulationists seem to believe that the state of technological
knowledge at any time is largely codified in blueprints and associated
documents and that, for a firm to adopt a technology that is new to it but not
to the world,  primarily  involves  getting access  to those blueprints.    In
contrast, assimilationists argue that only a smal:! portion of what one needs to
know to employ a technology is codified in the form of blueprints; much of it
is tacit and learning is as much by doing and using as by reading and studying.
(See e.g.  Nelson and Winter  1982,  and Rosenberg  1994.)    More,  while many
economists believe that technology is defined in terms of engineering and
physical  science,  in  fact  the  lines  between  the  engineering  aspects  of
technology, and the organizational aspects, are blurry, and controlling a
technology often involves knowing how to manage a very complex   division of
labor as much as it involves knowing the relevant lphysics and chemistry.
Both of these differences show up in terms of how the two theories go
about explaining the fact that the nics were able to increase greatly and
rapidly their capital-labor ratios (by more than four fold over the thirty five
years in question) without experiencing a significant decline in the rate of
return to capital.  The accumulationist is inclined to try to explain in terms
of the technological alternatives that were available to firms in the nics.
The ability to hold off diminishing returns is a reflection of the fact that
the nics could draw on a very extensive frontier of technologies that  already
were in use in other countries.  The richness of the frontier was manifest in
the fact that the "elasticity of substitution" was high.
The assimlationist, on the other hand, would argue it is misleading and
incomplete analytically, to try to specify a set of technological possibilities
without considering the decision makers' ability to search and see and
effectively take on board new technology.   That is, what the accumulationist
would explain in terms of the nature of the parameters of a conventionally
defined production function, an assimilationist would explain in terms of
skillful entrepreneurship and learning.
4



Along the same lines, the two theories also differ regarding how they see
the effects of the rapidly rising education levels in these countries.  For the
accunmulationist, rising human capital is treated simply as an increase in the
quality or effectiveness of labor.   Assimilationists,  on the other hand, tend
to see the effects of sharply rising educational attainments, in particular the
creation by these countries of a growing cadre of reasonably well trained
engineers and applied scientists, in ways similar to that sketched out many
years ago by Nelson and Phelps, 1966.   Well educated managers, engineers, and
workers have a comparative advantage in seeing new opportunities and
effectively learning new things.   Thus the growing human capital of the nics
was a very important support for successful entrepreneurship.
Note that the difference *here *mirrors the difference regarding how to
interpret the high ex-post elasticity of substitution.   The accumulationist
sees both the ability to hold off diminishing returns to capital and the high
economic returns to education in terms of the parameters of a conventionally
defined production function.   The assimilationist sees both as manifestations
of the effective learning that was going on.
This same difference between the two theories also shows up sharply in
how they treat the strong export performance of the nic manufacturing firms.
The accumulationists tend to see the steep rise in manufacturing exports as
just what one would expect in economies where the stocks of physical and human
capital were rising rapidly, and shifting comparative advantage towards the
sectors that employed these inputs intensively.   From this perspective, there
is nothing noteworthy about the surge of manufacturing exports, save that it is
evidence that the economic policies of these countries let comparative
advantage work its ways.  In contrast, the assimilationists, while not denying
that the nics were building a comparative advantage in various fields of
manufacturing, tend to highlight the active efforts by government to induce,
almost force, firms to try to export, and the entrepreneurship, innovation, and
learning the firms had to do in order to compete effectively in world markets,
even with government support.
Economists of the assimilation school have argued that exporting
stimulated and supported strong learning in two ways (Westphal, Kim, and
Dahlman,  1985,  Pack and Westphal 1986).   First, being forced to compete in
world markets made the managers and engineers in the firms pay close attention
to world standards.   Second, much of the exporting involved contracting with
American or Japanese firms who both demanded high performance and provided
assistance to achieve it.  The story here clearly is different than one which
sees the development of these new competencies as, simply, the more or less
5



automatic result of changing factor availabilities which called them into
being.
We have noted that the assimilationist's position, at least the one we
espouse, sees the high rates of investment by the nics in physical and human
capital as a necessary, if not a sufficient, component of the assimilation
process.   These high rates themselves are remarkable, even if not miraculous.
Under the argument of the assimilationists, the!se investments were at least
partially induced by, and sustained by, the rapid innovation and learning that
was going on.
Successful entrepreneurship in the nics certainly was facilitated by the
growing supply of well trained technical people. On the other hand, it was not
automatic that newly trained engineers would find work in entrepreneurial
firms.  There  had  to  be  entrepreneurial  firms  in  which  to  work,  or  the
opportunity to found new ones.   Thus in the nics aggressive entrepreneurship
supported and encouraged rapidly rising educational attainment, and served to
make these investments economically productive.   In contrast,  in many other
countries initially as poor as Korea and Taiwan,China, the market for college
graduates  was  almost  exclusively the Governmen'l  bureaucracies,  where  their
skills arguably made little contribution to economic development.
Successful production of new products almost always required that firms
acquire new physical  capital.   There's no question that policies  in these
countries  encouraged saving.   But on the other hand,  what made saving and
investment profitable was the strong and effective innovative performance of
the firms that were entering new lines of business.
We think it apparent that the two broad theories differ both in their
causal structures,  and in the hints they give about  "how to do it."   The
emphasis of the accumulationists is on getting investment rates up and the
prices right.   The message of the assimilation theorists is that successful
industrial development requires innovation and learning to master modern
technologies; effective innovation and learning dJepend on investments, and a
market environment that presses for efficient allocations, but it involves much
more.   And,  indeed,  to a considerable  extent,  the  investments  needed  are
induced by successful entrepreneurship.
Section II considers the argument that careful attention to the numbers
and rigorous calculation supports the accumulationist theory, and there is
little evidence that innovation and learning played much of a role.   We argue
that the commonly used calculations do not do whal their proponents claim.   In
Section III, we propose a different way for discriminating between a change in
6



output accompanied by changes in inputs that can be considered simply "a
movement along the production function," and a change that seems to involve
innovation and learning.   In the light of the argument we develop there,  in
section IV we consider the evidence. We propose that that evidence strongly
supports the assimilationist's  case.   Section V considers in what ways the
differences between the two theories matter.
II.   Why the Standard Calculations in Fact Don't Discriminate
The case put forward by its proponents for the accumulationist theory is
based on calculations of two sorts.   One is a growth accounting.   The other
involves  fitting a dynamic production function.   In both methods the strategy
is, basically, to try to calculate the effect of input growth on output growth,
holding the "production function" constant, and see (under growth accounting)
if anything much is left over as a "residual," or (under production function
fitting) whether the passage of time itself seems to contribute to output
growth over and above what is explained by input growth over time.   We argue
here that, contrary to widespread views in economics, neither kind of
calculation can separate out growth that "would have occurred without technical
advance" from growth that involved technical advance.
It often is not recognized adequately that the simple logic of growth
accounting is only applicable to the analysis of small changes in inputs and
outputs.  (See e.g.  Nelson 1973.)       The procedure basically involves making
estimates of the marginal productivities (or output partial elasticities) of
the various inputs that have changed and, in effect, using these to calculate
the contribution of input expansion to output growth by using a first order
Taylor series. However, in the case of the Asian tigers the investments whose
contribution to growth is being estimated have been cumulatively very large.
While repressed by the format of growth accounting, which usually sets up the
calculations in terms of average yearly changes, and thus makes the changes
appear relatively small, in the countries in question capital per worker
increased more than four times over the past three decades, and years of
average educational attainment also increased greatly.
The   calculations   in   standard   growth   accounting   take   marginal
productivities as estimated by factor prices (or output elasticities as
estimated by factor shares) as exogenous.   However,  under the assumptions of
neoclassical production function theory (which lie behind the growth accounting
logic), large finite changes in inputs can lead to large finite changes in
marginal productivities. Table 1 shows, for a CES production function, what
7



happens to the share of capital, initially at .3, for different proportional
increases   in  the  capital-labor  ratio,   for  different   elasticities   of
substitution. Note that for small increases, capital's share does not change
very much,  regardless of the elasticity of substitution.   For large changes,
however, the elasticity of substitution makes a huge difference in terms of
what happens to capital's share. For example, :300% increase in the capital
labor ratio and an elasticity of substitution of .2 implies a decline in the
capital share from .30 to .0017.
We know that, in the countries in question, despite the large changes in
their quantities, the rates of return on physical capital and on education
stayed high.   Indeed, capital's share of output certainly didn't fall much and
may  have  increased.      We  noted  earlier  that  one  explanation  is  that
technologically  determined  elasticities  of  substitution,  in  the  sense  of
standard production function theory, were quite high, and thus significant
increases in these inputs relative to others had only a modest effect on
marginal productivities as the economy moved along its ex-ante production
function.  Under this explanation a good share of output increase indeed would
have occurred without any technical advance.   This seems to be the implicit
argument  of  the  proponents  of  the  accumulation  theory.    However,  another
explanation is that the elasticities of substitution, defined in the standard
way, were quite low, and that only the rapid taking on board of. new
technologies prevented the sharp diminishing returns and falling partial output
elasticities for the factors that increased most that one would have observed
had these economies stayed on the production functions that existed at the
start of the development traverse.
Consider the latter explanation, which we believe is the correct one.
Under it, innovation and rapid learning are driving growth.  However, a growth
accounting of a standard sort might show a very small residual, or even a
negative one.  The factor shares of the more rapidly growing factors - physical
and human capital - would be and would remain high, as a consequence of the
rapid learning that made their continued expanslon productive. Rapid factor
accumulation was not an exogenous phenomenon but a response to high private
profitability. A growth accounting might "attribute" the lion's share of output
growth to input growth.   There would be little left to explain in terms of
innovation and learning, despite the fact that these are the basic factors
driving growth.
The use by some scholars of the Tornqvist index for the weights applied
to input increases represents acknowledgement that, if one is interested in the
impact on output of finite changes of inputs along a production function,
B



output elasticities can change along the way.    But the use of such an index
(as by Young, 1995) does not deal with the problem highlighted here. The index
uses actual shares for each year throughout the period.  But the actual shares
at the end of the period can be, and in the case in question almost surely
were, affected by the technological changes that occurred over the period. In
general they are not what the shares would have been at the new input
quantities had the production function stayed constant over the traverse. As
shown in Table 1, the evolution of shares depend on the value of the elasticity
of substitution and the magnitude of the increase in the capital-labor ratio.
Also, while not shown in Table 1, they also depend on technical change itself.
It is thus inappropriate to use the observed factor shares as weights since
they assume something one is trying to estimate, namely, TFP growth.2
We want to underline this point because many economists seem to believe
that the absence of a large residual in a growth accounting is strong evidence
that the lion's share of growth was due to movements along a prevailing
production function.   This is just not so if the input changes involved are
large.   Growth accounting alone cannot tell whether the relevant elasticities
of substitution were large or small, and thus cannot distinguish between the
two  stories  sketched  above  about  the  sources  of  growth.    There  is  an
"identification" problem.
One might think that the fitting of a dynamic production function can
avoid this logical limitation of growth accounting,  when input changes are
2 The following equations can be derived for the rate of change of factor
shares for a general production function, Q = f(K, mL) in terms of the initial
share of capital, a, the elasticity substitution, a, and the rate of Harrod-
neutral, labor augmenting technical change, m:
SK  =  (1-a) (1- a )/d] (m - k)              (1)
SL = [a) (1- c )/a] (k'- m)                  (2)
Equations (1) and (2) show that the factor shares utilized in Tornqvist
calculations will be affected by both technical change, in this case labor
augmenting, and changes in capital-intensity. The impact will be smaller the
greater a and m. In estimating the Tornqvist index of inputs,
T = Ģi [1/2 (Si,,  -  Si,t-,) (ln xi,t  -  ln xi, t  ),    (3)-
The Si,, are taken to be exogenous and, assuming competitive input markets, (3)
measures   the contribution of inputs xi  to output.   T is then substracted
from the log difference in output to obtain the desired measure of TFP growth.
But  the  Si,t   are not  in  fact  exogenous but  are  themselves  affected by
technical change as shown in (1) and (2).
9



large and finite.   However,  in practice the identification problem cannot be
resolved this way.
The basic issue in question is how much of experienced growth of output
per worker can be ascribed simply to the large experienced increases in
physical and human capital per worker that were achieved over the period
between 1960 and 1995, and without recourse to the argument that the production
function "shifted".   Let us focus on the growth of physical capital.   Table 2
shows, for a CES production function, the increase in output per worker that
would be generated for different increases in the capital-labor ratio, and for
different  values  of the  elasticity of  substitution.    As with  our  earlier
analysis of what happens to capital's share, for large changes in the capital
labor ratio, the elasticity of substitution matters, a lot.
Consider then the two "explanations," depicted in Figure I, for a large
increase in output per worker, between time one and time two, associated with a
large increase in capital per worker.   In the explanation on the left hand
side, in which the elasticity of substitution is assumed large, much of
experienced labor productivity increase would have occurred even had the
economy stayed on its production function of period one (the dotted curve). The
way the production function is drawn depicts only weak diminishing returns to
increasing capital intensity.   The firm or econoray in question is presumed to
know,  at  time  one,  how  to  operate  effectively  at  much  higher  capital
intensities  than  were  employed  then,  but  chooses  not  to  do  so  because
prevailing factor prices made it more profitable to operate at low capital
intensity. Between time one and time two, factor availabilities changed.
In contrast,  in the explanation  on the right  hand side,  experienced
productivity growth is almost totally the result of the establishment of a new
production function (the solid curve) in that very little productivity growth
would have occurred had the economy remained on its old production function.
Under this explanatory story, at time one the firm or economy in question knew
very little about how to operate effectively at significantly higher capital
intensities.   (The elasticity of substitution that would have obtained if the
firm had been limited to operating technologies it knew initially was very
low.) To have increased capital per worker without learning about and learning
to use new techniques would very quickly have led to low or zero marginal
returns.   Thus the economy,  in order to deal productively with the changed
factor price regime of period two, had to do a lot of "learning," or
"innovating," and in fact it did.
10



Both explanations fit the data at time one and two.  The "levels" and the
"slopes" of the old production functions are the same at time one, and the
levels and slopes of the new production functions are the same at time two.
This point was highlighted by Diamond, McFadden, and Rodriguez (1971), and
Nelson (1973), over twenty years ago. It seems to have been forgotten.
When one "fits" a dynamic production function statistically (through many
not just two points and slopes), how does one discriminate between these two
explanations?   Obviously one needs to place some restrictions  on the  form
fitted, for example, that the rate and direction of "technical advance" be
constants over the period, or that the underlying production function always
have a particular "kind of general shape."  Most of the econometric exercises
we are concerned with here have imposed relatively loose restrictions, although
sufficient to enable a best fitting equation to be calculated.   However, even
if an equation that looks like the left hand side explanation wins the "maximum
likelihood" contest (as in Kim and Lau, 1994), if the constraints on functional
form are relatively loose it is a good bet that an equation that looks like the
right  hand side  explanation  is not  very  far behind.    Standard  regression
techniques of the sort that have been employed do not enable confident
acceptance of one explanation and rejection of the other.
The graphs drawn in Figure 1 are in fact regressions estimated from the
actual data for Korea's manufacturing sector for the years 1962-91. The dynamic
production function fitted to the data is a standard CES, with two inputs-
capital and labor- and constant returns to scale.  To keep the analysis simple
and transparent we constrained technological advance to be neutral and constant
over the period in question.   The key parameters to be estimated are r, the
rate of technological progress, and e, the elasticity of substitution.
In the left hand figure we forced e to be large, .9.  Since growth of K/L
then "explains" a lot of the growth of Q/L, the estimated rate of technological
change,  r, came out low,  .016.   For regression runs in which we set e as
greater than one, the estimated rate of technological change was even smaller.
In any case, once the analyst built in a term accounting for the effects of
rising educational attainments, there would be little room for "technological
advance" in the explanation for economic growth.    In the right hand figure we
constrained e to be low,  .2.   Since under this constraint the growth of K/L
cannot "explain" much of the growth of Q/L, the estimated rate of technological
progress, r, came out high, .045. While growing human capital can cut down on
this figure, taking this factor into account is unlikely to make the rate of
estimated technological progress trivial.
11



Both of these regressions, and one in which all parameters were chosen by
least squares, yield values of R2 of around .98, leaving little to choose among
the regressions on a statistical basis. Again, we want to underline the point.
The fact that the best fit of a dynamic function provides an explanation for
growth in which technological advance plays a small role, and input growth
accounts for the lion's share of growth, does not itself provide strong
evidence against the argument that, in fact, growth would have been far less if
there had not been significant technological advance.   Only the imposition of
oarticular constraints on the dynamic production function enables econometric
techniaue to choose between the explanation on the left had side and the right
hand side of Figure I.  These constraints are basically arbitrary.   And the
imposition of somewhat different ones can change radically the estimated
contribution of technical advance in the attribution.
The authors in question certainly have been careful with their data, and
in the use of their methods.   The problem is that the methods employed just
don't do the job they are thought to do.  Nor, at this stage of our argument,
are we introducing "new data", although we agree that the issue is an empirical
one.      Before considering new evidence,  it would seem important to do some
rethinking of the kind of data that would discriminate between growth where
entrepreneurship, innovation, and learning were central, and growth where they
were not.
III. Back to Basics
How is one to decide between two different explanations, each broadly
compatible with the macroeconomic data, when one stresses "movements along a
production function" and the other emphasizes "entrepreneurship, innovation,
and learning"? We propose that to get an empiric:al answer requires that one
first ask some conceptual questions. What might one mean when one says that an
observed change in inputs and outputs simply reflected a move along a
production function? What might one mean if one argued it was not that simple,
but that entrepreneurship and innovation in fact were involved?   If we agreed
on answers to these conceptual questions we might be able to agree on what kind
of empirical evidence would be relevant.
Regarding what we economists seem to mean by "a move along the production
function,"  reflect  on the  simple  treatment  in undergraduate  microeconomics
texts. The production function, there, is said to be the "efficiency frontier"
of the "production set" - the set of all input-output combinations a firm can
choose among.  One way of explaining the set to students is to say that a firm
12



"knows" a certain set of production techniques or activities, and the
production set is generated by different levels and mixes of those activities
that a firm might choose.   In any case, the firm is viewed as both "knowing
about" each of the alternatives, and "knowing how" to do whatever is associated
with achieving the input-output vector associated with each.
The verbal articulation may admit that there might be modest "set-up"
costs associated with marshalling and organizing to shift operations to a point
within the set that is different from what the firm currently is doing, and
that some adjustments (another form of set-up cost) might be required to get
the new choice operating smoothly, although these shift costs are generally
repressed in the formal modeling.   However,  it seems inconsistent with the
"operating within the production set" idea if the set up costs for shifting to
a new point involved doing a lot of exploratory "search and study" to identify
and get a better feel for alternatives that, up to then, had been unfamiliar to
the firm, and the "adjustment" involved a lot of trial-error-try again learning
by doing and using. At least it would seem inconsistent if the results of
searching and learning were highly uncertain, both to the firm ex ante, and to
an economist trying to predict what the results would be.
Of course, a plausible interpretation of the production set idea might
admit a certain amount of statistical uncertainty regarding inputs and outputs,
particularly if there were unknowable outside forces, like the amount of
rainfall, that bear on the process. But if the decision maker in question has
only very rough ideas about the consequences of trying to do something,  and
initially about how to do it, that something does not seem to be an activity
that  can  be  regarded  as  within  the  unit's  production  choice  set.    The
production set of a firm would appear to be limited at any time to those things
the firm knows about and knows how to do, with good confidence and skill.   Or
at least that is how economists implicitly define the concept.
On the other hand, a move that involves a lot of study of initially hazy
alternatives, or research and development where even the nature of the outcome
is not clear in advance, would, according to these criteria, be regarded as a
"technological" change or "innovation" for the firm in question.  We do not see
how such a move possibly can be regarded as one "along a prevailing production
function," if economists adhere to what they teach about the meaning of choice
sets.
We call attention to the fact that, under the way we are proposing the
distinction be drawn,  a firm's production  set  in principle  could be  very
extensive.   Indeed, much of what some versions of the new neoclassical growth
13



_heory treat as "technological advance" would, under the principles suggested
nere, be regarded as moves along a firm's prevai:Ling production function.   In
_hese models (see e.g. Romer, 1990) investments in R & D are strictly  up-front
costs required to make a product or technique operational.  But in these models
(if not in fact) R & D is strictly a set up cost to make an activity the firm
always  knew about available for use.   There are no Knightian uncertainties
involved.
However, once we get away from particular abstract models, most
economists who have studied the processes empirically understand   that the
_.troduction to the economy  of products or processes significantly different
-rom any employed before   does  not  look like  a move  "along  a prevailing
oroduction  function."       It  is  well  documented  empirically  that,   while
zheoretical engineering calculations at any time encompass a wide range of
techniques not yet brought to practice, the bringing to practice of new
_echnology invariably involves "up front" research and development costs, with
Knightian uncertainties at least early in the process.   (See e.g. Nelson and
Winter, 1982, and Rosenberg,  1994).   While R and D can resolve some of these
uncertainties, there are uncertainties in the R and D process itself. Further,
even after R and D, there almost always are "bugs" at the start of operation,
and it usually takes some time before the operation is really got under
control.  In many cases the attempts at innovating prove unprofitable, and need
to be abandoned, or radically revised.
Of course, in this paper we are dealing with the adoption of technologies
_hat, while new to the firm or country, are not new to the world.   The issue,
then, is whether such changes in the behavior anci performance of firms in the
nics can meaningfully be explained as changed choices within largely unchanging
choice sets.
The accumulationists clearly have in mind that, if a technology is in
effective use in one country, there are ways that firms in other countries can
use to take aboard that technology at relatively low cost, and without
significant uncertainties regarding the outcome of their efforts.   Quite often
detailed descriptions are available.  One can hire consultants who are familiar
with the practices involved.   In many cases one can get assistance from the
firms who are operating the technology, although some license fees may be
required.
The assimilationist, in contrast, is skeptical about easy "technology
.ransfer". To be sure, for many of the technologies that the firms in the nics
adopted there were available engineering texts and articles and the like. Blue
14



prints  and specific handbooks often could be obtained.   There are lots of
consultants for hire.
However, the assimilationist would stress that such a move invariably
involves not only "up front" costs of identifying, learning about, and learning
to master the technique in question, but also significant uncertainties.   The
range  of  options  is  hazy.    Things  often  do not  work  out  as  expected.
Consultants  seldom  can  guarantee  success.    Inevitably  there  is  a  lot  of
learning by doing and using. The costs, and the uncertainties, are greater the
farther the technique being adopted is from those the firm actually has
employed.   in many cases major changes in firm organization may be required.
The firm may need to learn to sense new   markets.   Firms attempting these
changes  can  and  often  do  fail.    Those  that  succeed  do  so  because  they
successfully learn to do things they simply could not do before. That is, they
succeed by expanding their production sets.
IV What Does the Evidence Indicate ?
We can return now to the question of what kind of evidence one would need
to determine whether an observed change was within a prevailing capability or
choice set, or required an expansion of the set of things the organizations in
question knew how to do.  The prior section argued that the standard data and
techniques for deciding simply do not do the job.   We propose here that the
kind of evidence that can be most persuasive involves examination of process,
not simply time paths of inputs and outputs, and that such data are to be found
at a quite low level of aggregation.
However, we believe that, if one has the issues that divide the theories
firmly in mind, aat least some relevant evidence can be gleaned from more
aggregate   statistical  analysis.   Thus  we  have  proposed  that,   for  an
accumulationist, the relationship between a country's growth of output per
worker and its investments per worker is determined by the set of technological
alternatives "out there", which define its available production function. There
would seem to be a presumption that this production function is the same for
all countries, or at least some argument is needed if one is to propose that it
is  not.    On the  other hand,  for an assimilationist,  the  relationship  is
determined, to a considerable degree, by the skill of the firms in a nation in
searching and learning,  capabilities that certainly can differ widely across
nations. For an accumulationist, any significant variation among countries from
15



a fitted cross country "production function" is something of. a bother,  and a
ouzzle  to  be  explained.    But  an  assimilationist  would  expect  to  find
considerable variation. Further, the assimilationist's argument regarding Korea
and Taiwan,China would imply that these countries achieved a significantly
greater increase in output per worker than most other countries with comparable
initial conditions, and comparably high investment rates.
Much of the analysis of the performance of the Asian countries has
emphasized the absolute performance of the countries themselves, particularly
as measured by total factor productivity growth and factor accumulation. To
derive  measures  of  performance  or  to  interpret  such  measures,   strong
assumptions, which we have questioned, are made about the production function.
Another measure  of  performance  which  is  eclectic  rather  than  based  on  a
specific production theoretic base is the estimation of cross country
regressions of the type used by Barro, 1991, and Mankiw, Romer,and Weil, 1992,3
which permit comparison of a given country's performance relative to other
nations. To see whether the performance of the two countries of greatest
interest, here Korea and Taiwan,China, is unusual, we employ the following
estimated  cross  country  regression  equation  to  explain  differences  in
international rates of growth of GDP per capita,4
GDPG = -.0046 -.0308RGDP60 +.0296P60 -.0526GPOP + .0639i   (1)
where GDPG is the growth rate of per capita GDP, RGDP60 is GDP relative to the
U.S. in 1960, P60 is primary school enrollment in 1960 as a percentage of the
relevant age group, GPOP is the growth rate of the population from 1960-85, and
i is the average investent/GDP ratio in 1960-85. The variable i is a proxy
measure for the rate of growth of the capital stock, K = AK/K = iGDP/K. Even
if there is substantial variation in initial capital-output ratios, GDP/K,
differences in the value of i over 25 years will outweigh such dispersion and
yield a good approximation to KW.
These regressions  were developed to test whether the standard Solow-Swan
neoclassical model can explain cross country performance better than endogenous
growth models. However, these models do not invoke strong assumptions about
technical change and factor market pricing that is necessary in estimating TFP
growth within a country over time. For a useful evaluation of this literature
see Crofts, 1991.
16



Figure 2 graphs actual minus predicted GDP growth per capita against i
for countries in our sample that had high investment rates, i > 20 percent. As
can be seen there is an insignificant negative relation between i and actual
minus predicted growth. But among the high investment countries, the Asian
nics, and Korea and Taiwan,China in particular stand out as unusual performers,
even after adjusting for the other variables on the right hand side of (1)
including the potential benefits of being laggards. Table 3 shows the actual
minus predicted growth rates of a number of countries with very high values of
i. Compared with nations such as Greece, Portugal, and Spain, Korea and Taiwan,
China have unusual performance. It is clear that high physical investment
ratios and initial conditions that are thought to be conducive to growth are
not sufficient, alone, to explain the Korean or Taiwanese cases.
There is a large literature which adds additional variables to those
included in equation 1 including some measuring macroeconomic management,
export orientation, and so on. While such variables are of interest, they do
not provide information about the nature of the production performance or the
basis of success of economies in absorbing large quantitities of factor inputs
while others obtained low returns. Our contention is that a critical element
was the technological efforts of firms in Korea and Taiwan,China which allowed
them to successfully initiate new industries and absorb new equipment. While
other countries with high I/GDP ratios could purchase machinery which gave them
the potential to improve their productivity, this could only be successful when
it was combined with domestic effort to absorb the new technology. Moreover,
much of the successful absorption effort is not attributable to formal and
measureable R & D but efforts of firms to learn about new opportunities,
improve  organization  and  inventory  management,  and  undertake  minor  but
cumulatively significant changes in the production process. While proxies for
such  activity  could  be  introduced  in  cross  country  estimates,   their
construction is tenuous and would lead to false concreteness. 50n the other
hand, case studies which are considered later, despite the difficulty of
generalization, are suggestive and provide important insights into the origins
of the exceptional growth shown in Figure 2 and Table 3.
4 This equation is found in Pack and Page (1994). For a review of much of the
literature see Levine and Renelt, 1992.
S For cautions, usually disregarded, about data reliability, see Heston and
Summers, 1991.
17



A major problem with highly aggregated economic data is that it masks the
magnitude and even the nature of the allocational changes going on.   Thus,
earlier we noted that in the 1990's Korean and Taiwanese manufacturing firms
were heavily engaged in producing products that in the 1960's they were not
producing at all. This is strikingly illustrated by Table 4 for Taiwan,China.
In particular, note Taiwan's production of electronic goods, which by the late
1980s were accounting for roughly 21 percent of Taiwanese manufacturing
exports.  In 1960 virtually no electronics goods were produced in Taiwan,China.
In both Korea and Taiwan,China, a considerable amount of the knowledge came
from OECD purchasers of exports (Westphal, Kim, Dahlman 1985, Hou and San,
1993). The transfer of  such information to local firms by importers desiring
lower cost, higher quality products is an important feature of Korean and
Taiwanese  experience.  But  as  Hobday  (1995),  Kim  (forthcoming),    Pack  and
Westphal, (1997) and others report in their detailed firm histories, summarized
below, this information was only the initial foundation upon which firms then
built their technological capacity, first learning rudimentary processing, then
improving their productivity in small ways, then engaging in innovations in
process engineering and product design. While some of the foundations were
acquired from importers, the structure was mainly constructed through intense
efforts on the part of firms.
At  a  slightly  higher  level  of  aggregation  the  rapid  sectoral
transformation in both Korea and Taiwan,China is shown in Table 5. Labor
intensive, technologically simple sectors such as food processing, textiles,
and  clothing  experienced  a  relative  decline  while  capital  and  technology
intensive sectors such as chemicals, metal products, machinery, and electronics
expanded. It is difficult to articulate what it would mean to say that capital
and labor were allocated to these sectors and were routinely incorporated in an
existing well understood production function.   To the contrary,  there was a
widespread perception in both countries that the technological competence of
firms initially was insufficient to undertake efficient expansion in the newer
sectors. In a more formal way, the appendix sets out a model which describes a
growth pattern in which "craft" firms with lower profitability decline in
relative importance while "advanced" firms expand. The two sets of firms have
different production parameters and the aggregate process of growth which
mimics that of a neoclassical process cannot be described as a movement along
an international production function.
18



Within such a world the allocation of activity within the manufacturing
sector almost certainly would be associated with considerable turnover of
fi-ms, with firms going out of business in the declining sectors, and new firms
entering the expanding fields. Such a change in the allocation of activity
within the manufacturing sector almost certainly would be associated with
considerable turnover of firms, with firms going out of business in the
declining sectors, and new firms entering the expanding fields. And within the
expanding areas one would expect to see a certain amount of turnover as some
firms try and fail while others succeed. Unfortunately, we do not have the firm
turnover data that is directly relevant to the phenomena we are characterizing.
However, there are data on the number of firms of different sizes in
Korea and Taiwan,China for several years, and a summary of these data is
presented in Table 6.  The pattern is roughly what one would expect under the
assimilationist's story. There has been a striking decline in the number of
very small firms, most of which very likely were locked into old technologies
and producing traditional products, and a sharp rise in the number of middle
size or larger firms; we conjecture that a large share of these were new firms
entering the new product fields or older firms that succeeded in taking on
board modern technology.   In the early 1970s the productivity of these larger
firms was strikingly higher than that of the small firms that, according to the
story we are proposing, they were replacing.
However, to get at the details of what was going on would seem to require
stuudying individual firms.   Only by studying firms can one see just what was
involved when they came to master new technologies and learn what was needed to
operate in new product fields. As noted above there have been a large number of
detailed studies of Taiwanese and Korean manufacturing firms,  tracing   the
sources of the firms' rapidly growing range of manufacturing competencies.
While a skeptic may argue these are anecdotal, the evidence from several
hundred firm interviews can be regarded as no less compelling than the
imperfect aggregate data which are employed to argue that there was limited
technical change.
For example, a typical description of the production processes in the
Korean engineering industries in the mid to late 1970s is enlightening.   As
shown in Table 5, this sector grew enormously during the last three decades
and became Korea's largest single sector. Yet as late as 1977, its production
processes were described as exceptionally backward not primarily in the type
of equipment utilized but in the organization of production. The following
observations were given by a group of engineers and economists describing
19



Korean machine producing plants as of 1977, fifteen years after Korea's major
effort at industrialization had begun.
The common  pattern was one  of machine  placement  that  is
haphazard rather than allowing for an orderly flow of work.  Floor
space is very crowded and the operation of machining, fabrication
of components, assembly of parts, are scattered in any place that
happens to have available room.   Too much time is spent finding
work, or the next job, or material. In some cases the men have to
find their own area in which to work, perhaps make up some form of
fixtures of their own, or find the means to obtain levels or
measurements to work from. The almost universal characteristic is
one of congestion and a mixing of operations that frequently leads
to deterioration of quality because of improper floor planning.
There is no adequate provision for working space around the main
machines and the aisleways that are normally used to carry the flow
of work are completely congested with work-in process.7
Yet fifteen years later these plants were producing high quality machine
tools for export. The deficiencies described were amenable to improvements
through learning better practice and significant reorganization While it is
possible  to  make  such  learning  tautologically  equal  to  moving  along  an
international production function,  it was costly, the results uncertain,   and
it took place over many years, suggesting a much mTore complex phenomenon, not
replicated in many other countries in which capital accumulation was rapid.
The firm histories provide details of a conmplex interactive process in
which OECD importers furnished some knowledge of production engineering to
facilitate production in low cost firms. As these firms improved their cost and
quality structure, the importers provided specifications for new products which
the local firms manufactured as original equipment manufacturers (OEM). To
maintain their contracts they were forced to constantly reduce cost through
improving productivity. The OEM process thus provided a strong learning
environment in which firms not able to continue to meet quality and cost
specifications in short term opportunistic relationships could easily lose
their markets. To quote from a case study of an internationally known Taiwanese
computer peripherals supplier:
Foreign buyers are an important source of technological
enhancement. Their rigorous specifications are seen as a challenge
for the firm to meet. Equipped with different viewpoints and
accumulated experience, they criticize a lot and suggest other ways
of doing things. Although they do not provide exact blue prints,
'   The World Bank,  1979.  While the production processs described could be
interpreted as a cost minimizing response to the relative cost of labor and
space, the engineers observed that the same amount of space could have been
reorganized in order to achieve a much better flow.
20



their suggestions are invaluable in upgrading the technology level
of the firm. Still, our own R & D is the most important source of
technology. Without this capability, the firm would not be able to
evaluaze research proposals, technology contracts, licenses, or
buyers' suggestions. (Pack, Wang, and Westphal, 1996)
Thus,  while some knowledge was readily obtained from the rest of the
world, as the accumulationists correctly argue, this view is a partial one.
Rather the -ransfers provided the skeleton upon which the flesh of major
industrial prowess was built. Firm histories suggest that when transfers did
occur, they were followed by internal learning and innovative efforts. The R &
D figures and patent statistics shown in Table 7 for Taiwan,China provide one
measure of the large formal domestic research effort which built upon imported
knowledge. Be:ween 1981 and 1991 the number of patents granted to Taiwanese
nationals  quadrupled,   being  roughly  equal  to  foreign  patents  in  1991.
Similarly, formal R & D spending increased from .5 to 2 percent of GDP.
Moreover, fornal R & D is likely to constitute a minor part of domestic
technological effort.
Amsden (1989) provides a history of a Korean textile firm's learning. It
commenced production in 1963 and most of the additions to the plant occurred
before 1977. The machinery was purchased from Belgium, England, Japan, and
Germany and the firm received technical assistance from its suppliers. Labor
productivity shown in Table 8 improved substantially between 1977 and 1986 with
basically unc:anged quantities of machinery in both spinning and weaving.
Output per unit of equipment improved as weJl in spinning and was roughly
constant in weaving. The firm employed a large group of textile engineers to
achieve this improvement in productivity. Two observations based on the data in
Table 8 are relevant for our purposes. First, there was a steady increase in
output per unit of labor after the equipment was installed and this is
explicitly attributable to the firm's effort to improve its performance.
Second, the last row in the table shows the output per unit of input relative
to international best practice in British plants. Twenty five years after its
establishment and after a considerable period of learning, the firm was still
not operating at international best practice, a result hardly consistent with
the view that firms in the NICs moved along an international production
function.
The second example, also provided by Amsden, concerns the Pohang Iron and
Steel Company, generally known as POSCO. Although it initiated production with
imported equipment and with considerable technical assistance from abroad, it
21



too engaged in intensive internal effort to augment its productivity. It did
not simply move instantaneously to international productivity levels.
To improve the performance of each piece of equipment, POSCO
orovided training to its workers. ... Between 1968 and 1979
training courses of one form or another involved roughly 61,4000
workers. Approximately 4,200 people were trained outside the
company, 1,513 overseas. In 1984 alone, 9,900 workers had received
training, some 1,000 of them in computer applications. POSCO also
runs technical training schools in the town of Pohang, and in 1984
established an engineering college that it hopes will evolve along
the lines of MIT. (Amsden, 1989, pp. 305-306.
While the improved general level of education for the labor force was a
precondition for such training to be productive, it is no guarantor of such
effort by firms. Case studies in other countries of firms with relatively high
levels of educated labor do not uncover such evidence of systematic training. 8
For our purposes, one of the most interesting set of firm studies are
those undertaken by Michael Hobday (1995) of Korean and Taiwanese electronics
companies.   Hobday describes in detail how these firms started out,  usually
producing quite simple products, and then progressively moved on to more
complex ones.   In most of the cases he studied,  these new complex products
first were made to order for their foreign customers who, in the early stages,
provided detailed engineering instructions and assistance. Gradually, however,
many of these companies came to be able to do their own design work.   In a
number of cases, recently they have moved on to sell under their own brand
label.    Throughout  the history of these  firms,  one  can  see them actively
working to learn to do the things they were doing better, and to be able to do
more sophisticated and profitable things.   In the early stages, this learning
involved reverse engineering.   As the companies began to do their own design
work, this engineering effort began to be counted as research and development.
He    summarizes  55  firm  case  studies  in  Hong  Kong,  Korea,  Singapore  and
Taiwan,China as follows:
East Asian latecomers did not leapfrog from one vintage of
technology to another.  On the contrary,  the evidence shows that
firms  engaged  in  a  painstaking  and  cumulative  process   of
technological learning: a hard slog rather than a leapfrog. The
route to advanced electronics and information technology was
through  a  long  difficult *  learning  process,  driven  by  the
manufacture of electronics goods for export. (p. 1188).
See, for example, the detailed firm studies in India in Lall,  1987.
22



Linsu Kim (forthcoming) provides a set of analyses of Korean firms, in
several different industries, that show much the same phenomena as does
Hcbday's  study.    The  firms  started  out  using  relatively  unsophisticated
technologies and learned, over the years, progressively to master more
soohisticated ones.   By the 1990s many of these firms were approaching the
technological frontier.   But the paths they took were not simple, and success
never was guaranteed.
The story about the development of Korean and Taiwanese firms told by
:nsden, Hobday, and Kim, is strikingly similar to that told by Odagiri and Goto
(forthcoming) in their study of how Japanese industry learned about and learned
to master the technologies of the West in the years between the Meiji
restoration and the advent of World War II.  They find that a major amount of
searching, exploring, trying, failing, and learning was required before
Japanese firms acquired proficiency in the western technologies they were
adopting and adapting. The decisions of firm managers to get into the new ways
involved major uncertainties.   Odagiri and Goto stress their "entrepreneurial"
nature,  and the innovation and learning that were involved.   Our argument is
that Korean and Taiwanese firms went through much the same process, half a
century later.
To return to our basic analytic argument, we do not think that the
industrial development of Korea and Taiwan,China since the 1950's, or of Japan
a half century earlier (see Saxonhouse, 1974) can be interpreted as "moving
along production functions," at least if that term connotes changing choices
within a largely unchanging choice set.    On the other hand, if the kind of
entrepreneurship, innovation, and learning on the part of firms revealed in the
case studies is considered as perfectly consistent with  the notion of "moving
along a production function," we do not know what that concept would exclude,
and hence it becomes meaningless.
V.    Do the Differences Matter, and If So, How?
The differences between the two theories would appear to matter for two
different reasons.   The first is, simply, regarding how one understands what
happened.   What lies behind the Asian miracle?   The second is that the two
theories might imply somewhat different things regarding appropriate economic
development policy. What kinds of government policies are helpful, and what are
the lessons for other countries?
It  is  apparent  that,  for  many  economists,   one  of  the  strongest
attractions of the accumulation theory is that it is clean and simple, and its
23



basic outlines con_orm with the general theory about economic activity that
ones  finds  in  modern  economic  text  books.    It  is  at  once  delightfully
iconoclastic, and comfortably conservative, to take the miraculous out of the
Asian miracle by proposing that it all was a simple matter of moving along a
production function.   No appeal is needed to the idea of entrepreneurship or
innovation, the sources of which might very well lie outside the effective
province of neoclassical economics.
It also is clear that a major source of resistance to the assimilation
theory is that it seems a complex theory which raises as many questions as it
answers. This raises suspicions that the assimilation theory cannot be cleanly
formulated.   It is a comfort, therefore, that a simpler, more familiar theory
seems capable of providing all the explanation that is needed.
And yet, what is at odds intellectually may be only a small part of the
corpus  of  traditional  economic theory.   More,  that particular part,  which
proposes that production sets can be sharply defined, and that there is a clear
distinction between moving along the production function and having the
production function shift, came into economics only a relatively short time
ago. Perhaps these particular conceptions are not needed for most standard
economic arguments, and maybe they have been accepted too uncritically in any
case.
A strong argument can be made that the assimilationists' perspective is
auite  consistent with an older set of ideas  in economics.   The idea that
economic growth can be explained by increases in the factors of production, and
also by improvements .in their productivity, goes back at least as far as John
Stuart Mill.   However, a striking feature of the earlier analyses of economic
growth, as contrasted with the more contemporary treatments, is that there was
no compulsion to separate sharply between the contributions of different
sources of growth.   For Adam Smith,  increases  in the size of the market,
invention of better ways of performing a task, growing mechanization, and
changing organization of work, all go together.  They would seem to also do so
in Mill.   The early post World War II growth accountants, in particular Moses
Abramowitz, also stressed the interaction of technological advance, growing
physical capital intensity of production, increasing exploitation of scale
economies, rising educational attainments, and changes in the organization of
production,  as factors behind experienced economic growth.   The question of
which of these factors should be interpreted as moving the economy along a
production function and which should be regarded as shifting it seems not to
have been of major concern to these authors.
24



In Section II we argued that standard techniques do not permit one to
separate sharply between movements along and shifts in the production function.
Now we would like to argue  that the very notion that one can make such sharp
splits, even in principle, may not be a useful theoretical premise.
In particular, we would like to argue that "innovation" in practice is a
matter of degree, not kind, and that our growth theory ought to recognize this
explicitly.    For  any  firm  or  organization  at  any  time,  there  are  some
activities that are under practiced control, some that are not at present but
seem easy to learn, others harder, others presently impossible but perhaps with
research and experimentation achievable over the long run.   The problem with
now standard production theory is that it does not recognize these
continuities, but rather presumes a sharp cliff between the known and the
unknown.
The case studies of firms, briefly discussed in Section IV, show them
moving from the known, to the unknown, but cautiously, and drawing from the
known as much as they can.  Yesterday's unknown becomes today's known, and the
firms  venture  further.    An  effective  theory  of  what  has  been  happening
requires, we believe, abandonment of the notion that production sets at any
time are sharply defined, and thus that there is a clear distinction between
moving to another perceived point and innovation. Rather, there is a continuum.
If one explicitly recognizes that that distinction is in fact fuzzy, does
not that mean one has a fuzzy theory?   Not at all.   One of the striking
features of the various "evolutionary models" of economic growth that have been
built over the last decade is that, within them, innovation is treated as a
matter of degree, firms move step by step into the unknown, and in so doing
seldom move very far from the known.
Abandoning the sharp distinction between moving along a production
function and innovation clearly is a big step analytically. Such a step would
involve placing learning and adaptation at center stage of the behavioral
analysis, and letting go of analytic techniques and arguments that presume that
"profit maximization" is something that managers actually are able to achieve,
rather that something they strive for intelligently.   Yet it is arguable that
most of the important and useful propositions about the role of markets and
competition depend on the latter not the former.
The notion that competition tends to force price down towards costs, and
to stimulate reform or elimination of high cost producers, goes far back in
economics.   The argument does not depend on the existence of sharply defined
production sets, or the achievement by firms of policies that actually maximize
25



profits,  given the full set of theoretical alternatives.   It is intelligent
striving that does the job.   Similarly,  the argument that a change in factor
prices will induce behavior that economizes on the factor whose cost has risen
does not require either sharply defined production functions or actual
maximization, but only intelligent striving.
What are the policy implications of taking an assimilationist, or
evolutionary,  view on what happened in the Asian miracle?   Are the policy
prescriptions fundamentally different under an assimilationist theory than
under an accumulationist theory?  In many ways the policy prescriptions in fact
are quite similar, although the reasons behind the arguments differ somewhat.
Both neoclassical and assimilationist theories put considerable weight on
investments in human capital.   By stressing the importance of innovation and
learning, and the role of an educated work force in these processes, the
assimilationist might push even harder on the education front than would a
modern neoclassical economist.
No disagreement either on the importance of investment in physical
capital.  However  the  assimilationist  would  highlight  the  role  of  such
investments as a vehicle for taking aboard more modern technologies, and stress
that if capital formation is not linked to effective entrepreneurship, the
returns to investment almost surely will diminish greatly after a point.   On
the other hand the assimilationist would point to effective entrepreneurship as
a key vehicle for keeping investment rates of return high, and would put less
emphasis on simply trying to lift up the savings rate.
Both theories stress the importance of exporting. However, here too the
reasons  for the emphasis  are somewhat different.    The assimilationist  sees
exporting as an extremely important vehicle for learning, as well as a way of
exploiting evolving comparative advantage.   Thus,  the assimilation  theorist
might be even stronger on the importance of exporting, and willing to bias the
incentive system to induce firms to try to export.
Both theories stress the essential role of private enterprise, profit
incentives, and an environment that stimulates managers to make decisions that
enhance economic development.   A neoclassicalist would focus on getting the
prices  right  and making  necessary public  infrastructure  investments.    The
assimilationist would take a somewhat more complex view on both of these
matters. In particular, an assimilationist might stress the role of government
funding and organization in building up national scientific and technological
infrastructure from which firms can draw assistance.  But under both theories,
26



i  is the energy of private enterprise that is key, and under both there is
deep skepticism about the value of detailed government planning.
Both neoclassical and evolutionary theorists stress the great importance
of  competition.    However  here  too  the  reasons  differ  somewhat,  with  the
proponent of evolutionary theory pushing competition especially in contexts
wnere  innovation  is both  important  and risky.    From  this  point  of  view,
competition  is valuable largely because   choice sets are not clear or not
clearly defined and it is highly valuable, therefore, to get a lot of things
zried.
So, the policy differences between the theories may be significantly
smaller than the conceptual or analytic differences.   This should not be a
surprise.   Economists were stressing the importance of profit incentives, and
competition, and the dangers of government planning, long before the idea of a
sharply defined production set  came into vogue.   Indeed, one can find these
basic arguments in Adam Smith's Wealth of Nations.
27



Appendix: A Simple Evolutionary Model
The model we offer here is totally devoid of substitution possibilities
within a given technology. Rather, all development takes place through the
shifting of resources from one technology, which we will call craft, to
another, which we will call modern. That is, "assimilation" is what is driving
development here. Yet the growth pattern it generates could be interpreted by
a growth accountant or a fitter of dynamic production functions as indicating
that "accumulation" was the basic story. Expansion of physical and educational
capital per worker is an essential part of the process by which the economy
incorporates modern technology into its productive structure. But, on the other
hand, accumulation without assimilation yields no returns.
Within this model a basic constraint on the rate of assimilation is the
effectiveness of entrepreneurship. There always are profits to be made by
expanding the modern sector. The strength of entrepreneurship in responding to
profit opportunities determines the rate at which this happens. This response
can be encouraged by a favorable policy climate. Moroever, a strong
entrepreneurial response may, if successful, generate still more latitude for
the    government  to pursue  additional  desirable  policies.  We  believe  this
interaction accurately depicts an essential ingredient of the "Asian Miracle".
The rapid expansion of human capital, another essen,tial ingredient in our view,
also plays a central role in this model, being necessary for the rapid
expansion of the modern sector.
The model does not contain a third ingredient that we consider central;
-he rapid learning that took place in a firm after modern technology was first
adopted.   The model assumes in effect that such learning took place instantly
and was once and for all, while in fact the firms moved progressively into more
and more complex technologies. Here we choose to keep the model simple and
abstract from the cumulative natUre of learning.
Assume that there are two different kinds of fixed proportions constant
returns to scale technologies, which we will denote c for craft and m for
modern.   Capital per unit of output is the same iLn the two technologies but
output per unit of labor is higher in the modern sector than the craft.   So
also, then, is capital per unit of labor. If factor prices in the two sectors
were the same, unit costs using modern techniques would be lower than costs
using craft technology. However, the modern sector requires "educated" labor
while education is not necessary or productive in craft technology.
28



At the start of the development traverse almost all of capital and labor
is in the craft sector.  However we assume that there is a tiny amount in the
modern sector which serves, in effect, to "seed" .he development process.   At
any time output per input per unit of labor input in the economy or industry as
a whole will be the weighted average of labor productivity in the two
technologies, the weights being the proportion of labor employed by each of the
technologies.. Let a, be output per unit of labor in craft technology and am =
output per unit of labor in modern technology, with ac<am. Then:
Q/L = amLmIL + actL/L       (la)
Q/L = a, + (am -ac)L:/L     (lb)
As Lm/L grows over the development process,  so does Q/L.   Since capital per
unit  of output  is  the  same  in the  two  sectors,  an  increase  in  L /L  is
accompanied by a rise in K/L. Indeed, within this model Q/L and K/L  grow at
the same rate.
Within our model a shift in the proportions of capital in the two sectors
drives  development.   We assume that the price of the product  is the same
whether it is produced by modern or craft technology, and is constant over
time. The latter can be rationalized by presuming that the product is sold on
world markets and hence is insensitive to the quantity produced within the
particular economy in question. We also assume that the cost of capital is the
same in the two sectors. This means that differences in labor cost is the only
factor that affects the relative profitabilities of the two technologies.   We
could modify these assumptions, but making them enables us to tell a cleaner
story.
Let w be the price of labor in the craft sector, and gw its price in the
modern sector,  with g>l.   Thus   g  (for graduation)  reflects  an education
premium.  We assume, however, that g never is so large as to completely offset
the productivity advantages of modern technology.
If one uses  a prime  over a symbol  to denote  an inverse,  then the
difference between the two sectors in cost, and profit, per unit of output, and
capital, can be written:
AC = w(a', -ga'm)         (2)
The   higher   profitability   of  modern   technology   than   craft   provides
29



incentive to shift resources from the latter to the former.  Within this model
the  strength  of  the  response  is  determined  by  the  effectiveness  of
entrepreneurship, denoted by e.
d/dt (log Km/Kc) =ew(a', -ga'm)        (3)
d/dt(log Km/K) =ew(a', -ga'm) (1 -Km/K) (4)
If w and g are constants, the time path of K<m/K (and Qm/Q) will trace out
a logistic function. L,/L will be increasing as these variables grow, but
lagging behind them.   Of course in the limit they all approach one.  Tf w
increases as development proceeds but not g, the rate of expansion of the
modern sector relative to the craft will be accelerated reflecting that, since
modern technology saves on labor, an increased w increases its cost advantage.
An increase in the education premiun, g, over the development trajectory will
diminish the cost advantage of modern technology.  On the other hand a decline
in g, say as educated labor becomes more plentiful, will enhance it.
We know from equations la and lb that, as capital and labor shift to the
modern sector, K/L and Q/L will increase.   If the amount of educated labor is
responsive to demand, human capital also will be increasing. An economist
looking at aggregate data likely would conclude that growth of Q/L was caused
by the growth of physical and human capital per worker (and indeed such growth
of capital was required for growth) and would argue that growth basically was
due  to  "movements  along  the  (economy-wide)  production  function".          This
"exolanation" would repress two things.   First, the force driving growth was
the progressive adoption of modern technology, a technology virtually absent in
the economy before development began. And second, while the profitability of
employing modern technology was motivating the shift, the rate at which the
modern sector replaces the craft was being determined   by the strength of
entrepreneurship.   On the other hand, the traditional analysis would be right
about the rate of growth of human capital being an enabling factor.
Thus consider two economies with exactly the same initial conditions,
facing exactly the same opportunities to adopt modern technology, and having
the  same  input  supply  elasticities.        In  one  the  response  to  profit
opportunities, e, is high, and in the other low.  The expansion of the modern
sector, the growth of physical capital intensity, increases in human capital,
and the advance of labor productivity, all would be faster in the former than
the latter. An economist, thinking in terms of production functions, would try
to explain the differences in terms of different rates of "accumulation", but
30



he key factor behind the scenes would be differences in the entrepreneurlal
resonse to profit opportunities.9
Behind the scenes in the model growth of human capitaL is a enabling
element.  Other things being equal, a high e (resulting in rapid growth of the
..odern sector) will cause a rapid increase in the demand for educated labor. If
_ncreased supply is not forthcoming at the prevailing premium for educated
Labor, under various ways of modeling the dynamics g will rise. This will slow
town the rate of growth of the modern sector associated with a given e.  On the
other hand, a rapid expansion of the educated work force can be absorbed
roductively only if e is high.
Just as, within this model, a high e tends to draw forth expansion of
..uiman capital,  a high e tends to generate high profits in the industry as a
;hole, and hence a source of the savings to finance the investment in the
-odern  sector.    Both    effects  are  of  course  moderated  by  "supply  side'
-ariables. To keep this presentation simple we have not introduced these
explicitly into this model.
Within this model, development is a process driven by a disequilibrium.
-he disequilibrium, and the rate at which it is eliminated, shows up in this
.-odel in the behavior of capital's share over the development traverse. Set the
constant product price as the mumeraire.   Then the share of capital in total
:ncome is:
Sk = (1 - wa'c)Qc/Q  + (1 -gwa'm)Q,\Q (5a)
Sk = (1 -waec) + w(a', -ga>)QQ,/Q      (5b)
he first term of 5b is capital's share in the craft sector. The second term is
...e amount by which capital's share in the modern sector exceeds it's share in
-he craft, times the relative size of the modern sector.
Let b be the common capital output ratio in the two sectors, and r the
eauilibrium  rate  of return on capital.   Assume  that  at  the  start  of thM
development traverse the craft sector is in equilibrium.   Then while capital'e
share in the modern sector is greater than br, since Qm/Q is very small the
share of total capital in the total industry is close to br at the start of tho
zraverse.   We also assume that as development proceeds and the modern sector
- A considerable literature attests to these differences among developing
countries. Contrast, for example, Lall's (1987) description of the behavior of
ndian firms with those of Hobday (1996) and Kim (forthcoming) of the efforts
of Korean and Taiwanese firms. Some of the observed differences may be
attributable to differing policy environments.
31



grows relative to the craft, wg grows to squeeze out excess profits in the
modern sector.  At the end of the development traverse, then, capital's share
again is rb'. However in between, during the course of development, capital's
share will exceed rb'.   While the details depend on the exact specification,
under plausible assumptions capital's share will take an inverted U shaped path
over the development trajectory. As development proceeds and-the modern sector
expands, capital's share first will rise since quasi rents per unit of capital
are higher in the modern sector than the craft, and a growing share of capital
in that sector will more than offset the fact that rising wages will press down
on rents per unit of capital in both sectors.   Later,  as the modern sector
comes to be most of the economy, rising wages will diminish capital's share.
If one notes equation 2, one can see that the expression before QA\Q in
equation 5b is proportional to the rate at which caoital is being shifted from
the craft to the modern sector, and hence the rate at which output per worker
and capital per worker are growing.   Thus capital's share will be high when
capital and output are growing most rapidly. A -growth accountant would
naturally assign a good share of the credit for growth of output to growth of
capital.   If the supply of educated labor just keeps pace with the growth of
employment in the modern sector, human capital also will be growing most
rapidly when output is growing fast.
The foregoing captures the spirit of our argument in the text that, in
the Asian Miracle, both large investments in human capital and forceful
entrepreneurship which resulted in a growing modern sector and diminishing
craft sector (Table 6) were key ingredients, and that they complemented each
other strongly.   Absent the ability and inclination to greatly expand human
capital, aggressive entrepreneurship would have been stymied. Absent aggressive
entrepreneurship, the returns to investment in human capital would have been
low. And when both of these elements were present,, together they made for high
and rising profits in the modern sector which provided the finance for the
large investments in physical capital that were necessary for rapid
assimilation.
32



Figure 1
Alternate Interpretations of Growth
Q/L                                  Q/L
A  1981                                1981
1971
~~~~~~~~~~~~1961 /_17
5jA                                    'A
KIL           I1                      KIL
33



Actual minus Predicted  GDP  Growth  Rate
I       1       1
O       0       0        O       0        C1      0
C1  0       0l       O        1       0       (3
+ g                                     a-
*               ~~~~~~~~~~CD
CD~~~~~~~~~~~.C'



Table 1
Effect of Changes in Capital per Unit of Labor on Share of
Capital in Output With Different Elasticities of Substitution
elasticity  of  10%           100%         200%         300%
substitution    Increase    Increase    Increase    increase
in Capital  in Capital  in Capital  in Capital
per unit    per unit    per uinit    per unit
of labor    of labor    of labor    of labor
Share of Capital in Output (initial share
.30)
.2               .2264        .0261        .0053        .0017
.9               .2978        .2841        .2750        .2687
1.0              .3000        .3000        .3000        .3000
2.0              .3101        .3774        .4260         .4615
Note:  Tables  1 and  2 are derived in the  following manner, Assume  a tCS
production function Q = A(8K-°  + (1-)L-P)L-"P. From this we ean Obtain   the
following relationship between Q/L,  K/L,  and SK,  the share of capital,  in
periods 1 and 2 respectively:
(Q/L), =  8L/     +  (1-6)
(Q/L)2    8(L/K)P +  (1-6)
as well as the share of capital, SK in the two periods as:
K1=   8 +  (1-8),(K0/L0)
SK2_-= 8 + (1-6) (K1/Ll)>
The constant elasticity of substitution production function is undefined for
p=O. In the calculations, the Cobb-Douglas is used when the elasticity of
subsitution is unity. For the calculations in Tables 1 and 2, we define uhits
so that (K/L)0 = 1. Then SKo =.3 implies that S = .3.
35



Table 2
Effect of Changes in Capital per Unit of Labor on Output per
Unit of Labor With Different Elasticities of Substitution
elasticity       10%          100%          20C%         300%
of               Increase    Increase       Increase    Increase
substitution    in Capital  in Capital  in Capital  in Capital
per unit    per unit    per unit    per unit
of labor    of labor    of labor    of labor
Percentage change in output per unit of labor
.2            2.5          8.6           9.2          9.3
.9            2.9          22.4         37.1          48.3
1.0            2.9          23.1         23.1         51.6
2.0            2.9          26.4         26.4          69.0
36



Table 3
Investment Ratios and Predicted Minus
Actual Growth Rates
Country           Investment/GDP         Actual minus predicted
1960-85          Growth Rate of GDP per
Capita
Korea                      24.3                       .024
Taiwan, China              26.5                       .042
Algeria                    25.7                       .008
Spain                      26.5                      -.001
Greece                     26.3                       .008
Ireland                    26.4                       .007
Panama                     25.0                       .002
Portugal                   23.7                      -.002
Poland                     36.8                      -.019
37



Table 4
Changes in Physical Production Levels
Selected Industrial Products
Taiwan (China)  1960-1990
Product                             1960                       1990
Man   Made   Fibers   -             1,762                   1,785,731
millons of tons
Polyvinyl  Chloride  -              3,418                    920,954
millons of tons
Steel  Bars  - millions            200,528                 11,071,9991
of tons
Machine Tools                         0                      755,597
Sewing Machines                    61,817                   2,514,727
Electric Fans                      203,843                  15,217,438
Television Sets                       0                     3,703,000
Motorcycles                           0                     1,055,297
Telephones                            0                     1,055,297
Radiqs                                0                     5,892,881
Tape Recorders                        0                     8,124,253
Flectronic Calculators                0                     44,843,192
Iptegrated     Circuits               0                     2,676,865
(1000)
Electronic Watches                    0                     5,115,695
Shipbuilding (tons)                27,051                   1,211,607
Source; Taiwan Statistical Data Book, 1992, Council for Economic Planning and
Development, Republic of China, Taipei, Table 5-6c.
38



Table i
Share of Current Price Value Added Within Manufacturing
Sector                     Korea     Korea    Korea    Taiwan   Taiwajt
(China)  (China)
1963     1973      1988      1966      1986
Food, bev.                   .34       .18       .11       .29       .11
textiles, cloth.             .22       .22       *15       .15       .16
wood, furn.                  .04       .05       .02       .04       .03
paper, printing              .06       .04       .04       .05       .04
chemicals, petro.            .11       .20       .17       .21       .23
non. met. min.               .04       .05       .04       .07       .03
steel, iron                  .04       .08       .07       ,03       .06
metal prod, machinery,    .12          .16       .36       .10       .18
electronics
other.                       .02       .02       .02       .06       .13
Sources: Korea, United Nations Industrial Development Organization,Handbo6k of
Industrial Statistics, various years, Taiwan (China), birectorate terleral of
Accounting, Budget, and Statistics, The Report on IndUstrial, and Commercial
Census of Taiwan -Fukien Area, The Republic of China) various years.
39



Table 6
Percentage Distribution of Employment by Firm Size
Number of Employees
4-9      10-19      20-49      50-99    100-499      500+
Taiwan, China
1954                      18         13        14          9         16         31
1961                      18         10        14          8         17         34
1971                      8          7         11          9         29         37
Index    of   Value       NA        100         91        100        117       259
Added  Per  Worker,
1971
Korea
1958                      17         16        21         13         21         12
1963                      15         14         16        12         21         22
1975                      4          5          8          9         30         44
Index    of   Value       NA        100        133        193       256        304
Added  Per  Worker,
1971
Source: Ho, 1980, Tables 3.1, D2, D3.
Note: NA, not available
40



Table 7
R & D and Patenting Activity in Taiwan (China)
Year          R&D/GDP          Total          Taiwan         Foreign
Patents         (China)        Patents
Nationals'
Patents
1981            .95a           6,265          2,897           3,368
1986            .98           10,526          5,800           4,726
1991           1.65b          27,281          13,555          13,726
Notes. a, 1984; b, 1990.
Source: Taiwan Statistical Data Book, 1992. Table 6.7, 6.8
41



Table 8
Learning in a Korean Textile Factory
1986
Year                          1977            1986         international.
best practice
Labor Productivity
kilogr4ms per manhour,        52.4            78.5             156.25
r4ng ppinning
kilograms per manhour,        137.1          210.3             324.30
open end spinninga
mneters   per  manhour,       216.2           224.1            360.36
weaving
Machine Productivity
kilograms per spindle,         .20             .23               .21
ring spinning
kilograms  per  rotor,         .91            1.26              1.11
open end spinning'
meters    per    lopms         36.1           35.4              39.8
weaving
Noze; a , initial year is 1979.
SQorce:  Columns  1 and  2 adapted  from Amsden,  1989,  Table  10.4  Column  3
c4lcquateO from coluMn 2 plus coefficients from Pack, 1987, Tables 3.1 and 3.2
4nc4 oalculaions underlying those tables.
42



References
Abramovitz, Moses. "Catching Up, Forging Ahead, and Falling Behind," Journal of
Economic History, June 1986, 46:2, op. 385-406.
Abramovitz, Moses, Thinking About Growth, Cambridge, Cambridge University
Press, 1989.
Amsden, Alice, 1989, Asia's Next: South Korea and Late Industrialization,
Oxford University Press, New York, 1989.
Barro,  Robert,  1991,  "Economic  Growth  in  a Cross  Section  of  Countries",
Quarterly Journal of Economics, 1991, 105: 407-443.
Crafts, Nicholas, 1996, "Post-Neoclassical Endogenous Growth Theory: What are
Its Policy Implications?," Oxford Review of Economic Policy 12:30-47.
Dahlman, Carl. "Technology Strazegy in East Asian Developing Economies,"
Manuscript, July 1994.
Diamond,  Peter; MacFadden,  Daniel; and Rodriguez, Michel.   "Identification of
the Elasticity of Substitution and the Bias of Technical Change: in MacFadden,
Daniel (ed.), An Econometric Approach to Production Theory, North Holland,
Amsterdam, 1972.
Ho, Samuel 1980, "Small-Scale Enterprises in Korea and Taiwan," World Bank
Staff Working Paper No. 384, The World Bank, Washington D.C.
Hou, Chi-Ming and Gee San, 1993, "National Systems Supporting Technical
Advance in Industry: the Case of Taiwan," in Richard R. Nelson, ed., National
Innovation Systems, New York, Oxford University Press.
Hobday, Michael, Innovation in East Asia: The Challenge to Japan, London,
Edward Elgar, 1995.
Kim, J.I.,  and Lau, L.J.  "The Sources-of Economic Growth in   the East Asian
Newly  Industrialized  Countries,"  Journal  of  Japanese  and  International
Economics, 1994, pp. 235-271.
Kim, Linsu, forthcoming, From Imitation to Innovation: Dynamics of Korea's
Technological Learning, Boston, Harvard Business School Press.
Krugman, Paul. "The Myth of Asia's Miracle," Foreign Affairs, December 1994,
pp. 62-78.
Lall, Sanjaya, 1987, Learning to Industrialize, London, Macmillan.
Mankiw, G., D. Romer, and D. Weil, 1991, "A Contribution to the Empirics of
Economic Growth, Quarterly Journal of Economics, CVI.
Nelson,   Richard  R..   "A  Diffusion  Model  of  International   Productivity
Differences in Manufacturing Industry," 1968, American Economic Review 58:1219-
48.
------"Recent Evolutionary Theories About Economic Change," Journal of Economic
Literature, March 1995, pp. 48-90.
------- and Winter, Sidney, An Evolutionary Theory of Economic Change, Harvard
University Press, Cambridge, 1982.
--------.  "Recent Exercises in Growth Accounting:  New Understanding or Dead
End?", American Economic Review, 1973, 73:462-468.
43



--      -, and Phelps, Edmund. "Investment in Humans, Technological Diffusion,
and Economic Growth," American Economic Review, Mav, 1966, 56, pp. 69-75.
Odagiri, Hiroyuki, and Goto, Akira. Technology and Economic Development in
japan, Oxford University Press, forthcoming.
Pack, Howard, 1987, Productivity, Technology, and Industrial Development,
Oxford University Press, New York.
------,  1988,  "Trade  and  Industrialization,"  in  H.B.  Chenery  and  T.N.
Srinivasan, eds., Handbook of Development Economics, North Holland, Amsterdam.
----------, and Larry E. Westphal, 1986, "Industrial Strategy and Technological
Change: theory vs reality," Journal of Development Economics, 22:87-128.
...... ., and John M. Page,  "Accumulation,  Exports  and Growth  in the High
9erforming Asian Countries," Carnegie-Rochester Papers on Public Policy, 1994.
...... , Fang Ye Wang, and Larry E. Westphal, 1997, "Exports and Learning in
Taiwan," ms., processed.
Rodrik, Dani, 1995, "Getting Interventions Right: How South Korea and Taiwan
Grew Rich," Economic Policy
Romer, Paul. "Endogenous Technological Change," Journal of Political Economy,
1990, pp. 71-192.
Rosenberg, Nathan, "Uncertainty and Technological Advance," Manuscript, 1994.
Saxonhouse, Gary, "A Tale of Japanese Technological Diffusion in the Meiji
Period," Journal of Economic History, March 1974, pp. 149-165.
Westphal, Larry E. ; Kim, Linsu; and Carl Dahlman. 1985, "Reflections on
Korea's Acquisition of Technological Capability," in Nathan Rosenberg and
Claudio Frischtak, eds., International Technology Transfer, Praeger, New York.
World Bank. The East Asian Miracle: Economic Growth and Public Policy, Oxford
University Press, Oxford, 1993.
Young,  Alwyn,  1995,  "The  Tyranny  of Numbers:  Confronting  the  Statistical
Realities of the East Asian Growth Experience," Quarterly Journal of Economics,
110:641-80
44






Policy Research Working Paper Series
Contact
Title                            Author                  Date              for paper
WPS1856 Surviving Success: Policy Reform    Susmita Dasgupta      November 1997      S. Dasgupta
and the Future of Industrial     Hua Wang                                   32679
Pollution in China               David Wheeler
WPS1857 Leasing to Support Small Businesses Joselito Gallardo     December 1997      R. Garner
and Microenterprises                                                        37664
WPS1858 Banking on the Poor? Branch       Martin Ravallion        December 1997      P. Sader
Placement and Nonfarm Rural      Quentin Wodon                              33902
Development in Bangladesh
WPS1859 Lessons from Sio Paulo's          Jorge Rebelo            December 1997      A. Turner
Metropolitan Busway Concessions   Pedro Benvenuto                           30933
Program
WPS1860 The Health Effects of Air Pollution    Maureen L. Cropper  December 1997     A. Maranon
in Delhi, India                  Nathalie B. Simon                          39074
Anna Alberini
P. K. Sharma
WPS1861 Infrastructure Project Finance and    Mansoor Dailami     December 1997      M. Dailami
Capital Flows: A New Perspective    Danny Leipziger                         32130
WPS1862 Spatial Poverty Traps?            Jyotsna Jalan           December 1997      P. Sader
Martin Ravallion                          33902
WPS1863 Are the Poor Less Well-insured?   Jyotsna Jalan           December 1997      P. Sader
Evidence on Vulnerability to Income  Martin Ravallion                       33902
Risk in Rural China
WPS1 864 Child Mortality and Public Spending  Deon Filmer         December 1997      S. Fallon
on Health: How Much Does Money   Lant Pritchett                             38009
Matter?
WPS1 865 Pension Reform in Latin America:   Sri-Ram Aiyer         December 1997      P. Lee
Quick Fixes or Sustainable Reform?                                          37805
WPS1866 Circumstance and Choice: The Role  Martha de Melo         December 1997      C. Bernardo
of Initial Conditions and Policies in   Cevdet Denizer                      31148
Transition Economies             Alan Gelb
Stoyan Tenev
WPS1867 Gender Disparity in South Asia:   Deon Filmer             January 1998       S. Fallon
Comparisons Between and Within   Elizabeth M. King                          38009
Countries                        Lant Pritchett
WPS1 868 Government Support to Private    Mansoor Dailami         January 1998       M. Dailami
Infrastructure projects in Emerging   Michael Klein                         32130
Markets



Policy Research Working Paper Series
Contact
Title                            Author                   Date              for paper
WPS1869 Risk Reducation and Public Spending Shantayanan Devarajan   January 1998       C. Bernardo
Jeffrey S. Hammer                          31148
WPS1870 The Evolution of Poverty and       Raji Jayaraman          January 1998        P. Lanjouw
Inequality in Indian Villages    Peter Lanjouw                               34529
WPS1871 Just How Big Is Global Production    Alexander J. Yeats    January 1998        L. Tabada
Sharing?                                                                     36896
WPS1 872 How Integration into the Central  Ferdinand Bakoup        January 1998        L. Tabada
African Economic and Monetary    David Tarr                                  36896
Community Affects Cameroon's
Economy: General Equilibrium
Estimates
WPS1 873 Wage Misalignment in CFA Countries: Martin Rama           January 1998        S. Fallon
Are Labor Market Policies to Blame?                                          38009
WPS1874 Health Policy in Poor Countries:   Deon Filmer             January 1998       S. Fallon
Weak Links in the Chain          Jeffrey Hammer                              38009
Lant Pritchett
WPS1875 How Deposit Insurance Affects      Robert Cull             January 1998        P. Sintim-Aboagye
Financial Depth (A Cross-Country                                             37644
Analysis)
WPS1876 Industrial Pollution in Economic   Hemamala Hettige        January 1998        D. Wheeler
Development (Kuznets Revisited)    Muthukumara Mani                          33401
David Wheeler
WPS1877 What Improves Environmental        Susmita Dasgupta        January 1998        D. Wheeler
Performance? Evidence from       Hemamala Hettige                            33401
Mexican Industry                 David Wheeler
WPS1 878 Searching for Sustainable         R. Marisol Ravicz       February 1998       M. Ravicz
Microfinance: a Review of Five                                               85582
Indonesian Initiatives
WPS1879 Relative prices and Inflation in   Przemyslaw Wozniak      February 1998       L. Barbone
Poland, 1989-97: The Special Role                                            32556
of Administered Price Increases
WPS1880 Foreign Aid and Rent-Seeking       Jakob Svensson          February 1998       R. Martin
39065