KAKASHEV: ECONOMIC DEVELOPMENT AS SELF-DISCOVERY

ECONOMIC DEVELOPMENT AS SELF-DISCOVERY*

Ricardo Hausmann and Dani Rodrik

Revised

April 2003

John F. Kennedy School of Government

Harvard University

79 Kennedy Street

Cambridge, MA 02138

(617) 496-3740

ricardo_hausmann@harvard.edu

John F. Kennedy School of Government

Harvard University

79 Kennedy Street

Cambridge, MA 02138

(617) 495-9454

dani_rodrik@harvard.edu

ABSTRACT

In the presence of uncertainty about what a country can be good at producing, there can be great

social value to discovering costs of domestic activities because such discoveries can be easily

imitated. We develop a general-equilibrium framework for a small open economy to clarify the

analytical and normative issues. We highlight two failures of the laissez-faire outcome: there is

too little investment and entrepreneurship ex ante, and too much production diversification ex

post. Optimal policy consists of counteracting these distortions: to encourage investments in the

modern sector ex ante, but to rationalize production ex post. We provide some informal

evidence on the building blocks of our model.

* We thank the Ford and Rockefeller Foundations and the Carnegie Corporation of New York

for financial support, and Zoë McLaren and Anton Dobronogov for research assistance. We also

thank Bill Easterly, Murat Iyigun, Devesh Kapur, Michael Klein, Andres Rodriguez-Clare, Peter

Schott, Sebastian Edwards, and participants at the NBER-IASE meeting in Monterrey, Mexico,

November 15-16, 2002, for helpful comments.

ECONOMIC DEVELOPMENT AS SELF-DISCOVERY

Ricardo Hausmann and Dani Rodrik

Introduction

The theory and practice of economic development have converged in the last two decades

on a remarkably simple view of growth fundamentals. Stated in its starkest form, this view is

that economic growth requires two things: foreign technology and good institutions. This

perspective is well grounded in the neoclassical model of economic growth, which predicts that

poor countries will experience rapid convergence with advanced economies once they have

access to state-of-the-art technologies and their governments respect property rights. From this

perspective, failure to grow can be attributed to one or both of two pathologies. One is the

“closed-economy” pathology, in which governments retard technological progress by reducing

access to foreign trade and investment and imported capital equipment and intermediate goods.

The other is the “corruption” pathology, in which political leaders either willfully fail to respect

property rights and screw things up deliberately in order to enrich themselves and their cronies or

fall into inefficient games in which sound money and fiscal solvency perish. The natural

remedies for these pathologies are economic openness and improved governance. With these

remedies in place, economic growth should follow naturally. In the words of a recent paper on

growth: “Once a developing country government establishes the rules to a fair game and ensures

their enforcement, it would be well advised to stand back and enjoy the self-generating growth”

(Roll and Tallbott 2001). Reforms in the areas of governance and openness have accordingly

become the cornerstones of development strategy in virtually every country during the last

fifteen years.

Actual development experience presents at best an awkward fit with this conception of

growth basics. We point in particular to two important types of evidence that seem to us to run

counter to the consensus view. The first of these relates to the economic performance of Latin

American countries during the 1990s. By the standards of the consensus view, the quality of

policymaking in Latin America has been unmistakably and significantly better in the 1990s than

it was two or three decades before. For example, Lora’s (2001) index of structural reform—

measuring the degree to which government intervention has been reduced in trade, finance,

taxation, and state ownership—shows the average for 16 Latin American countries rising

steadily from around 0.34 in 1985 (out of a maximum of one), to 0.58 in 1999

1. Yet these

economies’ response to the reforms has been extremely disappointing. Per capita income

declined in these countries from an average of 22.9 percent of the US level in 1985 to 17.7 in

1999, a relative decline of 22.7 percent

2. Economic growth in the 1990s has been on average

much lower than in the decades before 1980, even though the region was closed to trade and had

poorer institutions by most benchmarks in the earlier period. In fact, only three Latin American

countries (Chile, Uruguay, Argentina) have outperformed in the 1990s their record during 1950-

1980. Of these three, only Chile remains a clear success. Uruguay’s performance has hardly

been exemplary, as its growth rate looks good only in relation to an even worse performance

prior to 1980, and together with Argentina, it now lies in ruins. Why is growth so low in a

region that has tried so hard to adopt the consensus agenda?

Morley, using Lora’s (1997) methodology extends the index back to the early 1970’s. He finds that the index

moves from 0.34 in the early 1970’s to 0.82 by 1995.

Note that by 1985 Latin America was in the midst of a debt crisis which had led to a major recession. This should

have made the subsequent recovery larger.

The other side of the coin, and the second strand of ill-fitting evidence, is presented by

the experience of countries that have had greater success. Some of the most important among

these countries—South Korea and Taiwan since the early 1960s, China since the late 1970s, and

India since the early 1980s—have done extremely well under quite heterodox arrangements. All

these countries have emphasized exports and none grossly violated property rights. But their

strategies bear only passing similarity to today’s consensus precepts. South Korea and Taiwan

retained high levels of protection for a long time, and made active use of industrial policies.

When Korea was hit with the Asian financial crisis in 1997, its economy had so many

institutional weaknesses by orthodox standards that many observers located the crisis’ roots in

Korea’s “governance” problems. China achieved phenomenal growth rates without formally

enacting private property rights—something that would have seemed impossible to many

economists had the Chinese miracle not taken place. India barely reformed its incredibly

cumbersome trade and industrial regime before its economy took off in the 1980s. And even

after more ambitious reforms were enacted in the early 1990s, the Indian economy remained

among the world’s most protected. Of course heterodoxy did not produce payoffs everywhere.

Most countries with protected economies and poor protection of property rights languished or

retrogressed. But the fact that the world’s

most successful economies during the last four

decades prospered doing things that are more commonly associated with failure is something that

cannot be easily dismissed.

We present here a different perspective on economic development, one that has more

room for the Latin American and Asian “anomalies” noted above. Our focus is on a particular

type of learning that we believe has not received enough attention in the literature on economic

development: learning what one is good at producing.

3 We emphasize that this is a key

challenge in the process of transformation into a modern economy. Neither economic theory nor

management science is of much help in helping entrepreneurs (or the state) choose appropriate

investments among the full range of modern-sector activities, of which there could be tens of

thousands, once one moves beyond broad categories such as “labor-intensive products” or

“natural-resource based products.” Yet making the right investment decisions is key to future

growth, as it determines the pattern of specialization.

4 In these circumstances, there is great

social value to discovering that cut flowers, soccer balls, or computer software can be produced

at low cost, because this knowledge can orient the investments of other entrepreneurs. But the

initial entrepreneur who makes the “discovery” can capture only a small part of the social value

that this knowledge generates. As the cases we shall discuss in the empirical section of the paper

reveal, other entrepreneurs can quickly emulate such discoveries. Consequently,

entrepreneurship of this type—learning what can be produced—will typically be undersupplied,

and economic transformation delayed.

This perspective differs from the standard view in an important way. In the neoclassical

model, it is presumed that the production functions of all extant goods are common knowledge.

Our starting point is that this is not a good assumption for developing countries. Much

technology is “tacit,” meaning that it cannot be easily codified into blueprints that allow easy

For a paper that is focused on learning about the institutions that are appropriate to the local setting, see Mukand

and Rodrik (2002).

This kind of indeterminacy of specialization is different from the indeterminacy that originates from scale

economies or learning-by-doing. The latter are the subject of Gomory and Baumol (2000): “Today there is not one

uniquely determined best economic outcome based on national advantages. Today’s global economy does not

single out a single best outcome, arrived at by international competition, in which each country serves the world’s

best interests by producing just those goods that it can naturally turn out most efficiently. Rather, there are

many

possible outcomes

that depend on what countries actually choose to do, what capabilities, natural or human, they

actually develop.” (Gomory and Baumol 2000, 5, emphasis in the original).

application (Nelson 1990, Evenson and Westphal 1995, Lall 2000). Moreover, even when the

production techniques used in the advanced countries are transparent to outsiders, their transfer

to new economic and institutional environments typically require adaptations with uncertain

degrees of success.

5 This is an aspect of technology transfer that is amply documented in the

literature. In their survey on technology transfer, for example, Evenson and Westphal (1995)

emphasize the role of domestic tinkering in the successful adoption of foreign technology. They

list such adaptations as “technological efforts related to raw material control, product and process

quality control, production scheduling, repair and maintenance, changes in production mix, as

well as others including episodic trouble-shooting to overcome problems encountered in the

course of operations” (1995, 2249).

6 We shall discuss these issues at greater length when we

turn to empirical evidence.

If learning what a country is good at producing requires an investment

and the returns to

that investment cannot be fully appropriated, the problem faced by potential entrepreneurs in

developing countries is identical to the problem faced by innovators in the advanced industrial

countries. However, the policy environments facing the “innovators” in the two settings are

quite different. Typically, the intellectual property regime protects discoverers of

new goods

through the issuance of temporary monopolies, i.e., patents. But the investor in the developing

country who figures out that an

existing good can be produced profitably at home does not

For an important theoretical analysis of “learning to learn,” and how its presence may be responsible both for lack

of convergence across countries and for specialization in production, see Stiglitz (1987). Zeira’s (1987) model of

investment, in which a firm does not know its profit function and can discover it only through investment, come

especially close to capturing the essence of our approach. In our model, unlike in Zeira’s, knowledge acquired in

the process of discovering one’s costs spills over to other potential entrepreneurs.

Another source of uncertainty is about trading possibilities--whether there is adequate demand out there for a

particular product, for example. This type of informational problem is discussed in Rauch and Watson (2001) in

their model of “international trade intermediaries,” and “network intermediation.” Such entrepreneurship may be

under-supplied because intermediaries may not have enough incentive to maintain and expand their informational

networks.

normally get such protection, no matter how high the social return.

7 Indeed, ease of entry by

competitors (i.e., imitators or copycats) is normally judged to be an important indicator of how

well markets function—the lower the barriers to entry, the better. Free entry makes the nonappropriability

problem worse, and undercuts the incentive to invest in discovering what a

country is good at producing. Laissez-faire cannot be the optimal solution under these

circumstances, just as it is not in the case of R&D in new products.

We develop a general-equilibrium framework in the next section to clarify the analytical

and normative issues. We begin with a small open economy that is initially specialized in

traditional activities where there is no uncertainty. There also exist a large number of modern

activities, as yet not exploited, all of which have an uncertain productivity. In the first period,

entrepreneurs have the option of setting up a “firm” of a fixed size and discovering the true costs

of production in the specific activity they have invested. In the second period, they can operate

as a monopoly, and choose to do so if their costs turn out to be lower than the relevant world

price. In the third period, costs of all active firms become common knowledge and there is free

entry, eliminating (through adjustments in wages) all excess profits. We characterize in this

setting both the laissez-faire equilibrium and the social optimum (in which the social planner can

allocate resources at will, but has no more information than is available to the private sector).

We highlight two failures of the laissez-faire outcome: there is

too little investment and

entrepreneurship in the first period (unless the second period is “very long”), and

too much

production diversification in the second. Optimal policy consists of counteracting these

As Evenson and Westphal (1995) note, the kind of local adaptations that are normally required “are seldom

associated with inventions that are patentable abroad. They do not yield improvements that are sufficiently

inventive relative to the known state of the art” (1995, 2249).

distortions: to encourage investments in the modern sector ex ante, but to rationalize production

ex post.

Let us now return to the anomalies that we used to motivate this paper. What does our

framework have to say about the possible causes of the disappointing Latin American

performance? Our model offers three clues as to why the investment response may have been

anemic despite increased openness and improved institutions. First, and most significantly, the

binding constraints to growth may have been elsewhere. Entrepreneurship may have been

constrained by inadequate inducements to discover costs in new activities (given that the private

returns to such investments lie far below social returns), and not mainly by inadequate property

rights or lack of access to imported technologies. Second, to the extent that market-oriented

reforms have increased the mobility of firms, they may have actually reduced incentives to invest

in new activities. Ease of entry speeds up emulation, and reduces the appropriability of cost

discoveries. In fact, if this is the case, reforms have a bigger payoff in the short run than in the

long run: new entrants exploit the benefits of past cost discoveries, but once these are exhausted,

the economy is dragged down by the lower efforts in innovation.

8 Third, to the extent that the

reforms have increased productivity in traditional sectors alongside potential new activities, they

may have increased the resource cost of entrepreneurship in the modern sectors. What we can

say, at the very least, is that the reforms of the 1980s and 1990s have paid scant attention to the

problem of spurring investment in non-traditional activities when returns to entrepreneurship in

such activities are subject to non-appropriability for reasons that we highlight in this paper.

This is consistent with the evidence from Latin America. Initial estimates of the growth effects of reforms

(Easterly, Loayza and Montiel 1997, Lora and Barrera 1997, and Fernandez-Arias and Montiel 1997) found

relatively large and permanent effects. Using more recent data, Lora (2002) finds that the growth payoff to reform is

smaller and more transitory than originally estimated.

With regard to Asia, our framework helps us understand why the provision of rents by

governments (through trade protection, temporary monopolies, subsidized credits, and tax

incentives) has gone hand in hand with industrial growth and diversification. These rents may

have been needed to stimulate the cost discovery process. Detailed accounts documenting these

rents in South Korea and Taiwan (see for example Amsden 1989, Wade 1990, and Evans 1995)

are otherwise impossible to square with the conventional understanding of what constitutes

desirable economic policies. At the same time, our framework highlights how rents can backfire

if governments do not complement them with policies that rationalize industries and discipline

firms that end up with high costs. We would hypothesize that the absence of such discipline was

a hallmark of import substitution policies (ISI) in Latin America. We will come back to this

distinction after we develop our analytical framework.

An important antecedent to our paper is Hoff (1997), which modeled Bayesian learning

in an infant industry setup.

9 The starting point of that paper, as with ours, is uncertainty about

the true costs of production under local conditions. Hoff goes on to show how this can generate

a motive for policy intervention that differs from that in standard learning-by-doing models. In

Hoff’s model, both firms and the social planner are risk averse, which generates some interesting

wrinkles. (For example, the optimal subsidy to “infant industries” need not be increasing in the

informational barrier to entry.) We shall ignore risk aversion in our model, so we can focus

directly on the externality generated by learning the costs of new activities.

The outline of the paper is as follows. We describe and analyze the formal framework in

the next section. We then turn to some suggestive empirical evidence. We provide evidence on

three aspects of our framework in particular. First, we show that there is a large element of

We came across the paper by Hoff (1997) after the working paper version of this paper was circulated, so we failed

to refer to it in earlier versions.

uncertainty at a disaggregated level as to what a country will be good at producing. Second, we

discuss evidence relating to the difficulties entailed in importing technology off-the-shelf. Third,

we summarize some case studies of domestic technology diffusion from successful incumbents

to emulators. We then turn to a discussion of the policy implications of this approach.

Conceptual framework

We embed our policy analysis in a model of a small open economy with two sectors,

modern and traditional. We distinguish between these two sectors according to whether costs of

production are known. The modern sector is made up of

n goods that are not currently produced,

and the costs of which are discovered only after production is attempted. Letting

ci denote unit

costs of production of good

i, we assume that costs depend on unobserved productivity

parameter

in the following manner:

(1) ,

i

i

c bw

where

w is the wage rate and b/ θ

is the number of workers needed to produce a single unit of the

good

i. We have in mind many different types of uncertainty: producing a good that has not

been locally produced previously requires learning about how to combine different inputs in the

right way, figuring out whether local conditions are conducive to efficient production, and

discovering the true costs of production. We use uncertainty over

θi as a proxy for all this. We

assume that the ex-ante distribution of

is uniform over the interval [0, 1] and that the θi are

i.i.d. across the

n sectors. Note that modern-sector production uses only labor and has constantreturns

to scale technology once productivity is known.

The unobserved productivity parameters

θi is a property of the individual goods, and not

of entrepreneurs: all entrepreneurs who run firms producing good

i will operate with productivity

10 We shall assume that each modern-sector firm is of a given size, fixed (by appropriate

choice of units) to one unit of good

i’s output. Each entrepreneur can run one, and no more than

one, modern-sector firm.

Discovering

requires setting up the firm, which in turn entails the hiring of b units of

labor. In this initial stage (period 1), the firm produces nothing. Let

m denote the number of

entrepreneurs who choose to establish firms in period 1. The total amount of (sunk) investment

in the first period is correspondingly

mbw. We will refer to m interchangeably as the quantity of

“investment” or “entrepreneurship.”

In period 2,

become known for those m goods in which investments have been made.

Entrepreneurs have the option of producing a unit of the good and earning

p (an exogenous price

fixed on world markets

11), or, if unit costs bw/ θ

turn out to be larger than p, to close the firm at

no extra cost. In this period, there is no entry into the modern sector so that entrepreneurs who

produce will earn excess profits. (Even though

p is fixed, so is output thanks to the assumptions

that firm size is fixed and an entrepreneur cannot run more than a single firm.) This transitional

period of monopoly profits can be motivated in one of two ways. It could be that it takes time

Hence uncertainty is associated with costs of production rather than with entrepreneurial talents. Once an

entrepreneur discovers costs in a given sector, there is a large number of entrepreneurs who can emulate the

incumbent. Some other models of industrialization emphasize instead the selection of talented entrepreneurs who

can best undertake the innovations needed for modern production. A recent paper by Acemoglu, Aghion, and

Zilibotti (2002), for example, distinguishes between investment-based and innovation-based strategies, arguing that

the latter become critical as the economy catches up with the frontier. In that second stage, what is key is the

selection of good entrepreneurs rather than the selection of low-cost activities. To the extent that this is a good

representation of the development process, our model can be thought of as applying largely to the early stages of

industrialization.

Note that these are goods that are already being produced in other, more advanced countries. So saying that there

are well-known, fixed prices is not at odds with the assumption that none of them are being produced at home

currently.

for the

to become common knowledge. Alternatively, θ

can be immediately known, but it

could take time for an “imitator” to set up a firm. We denote the length of period 2 as

Finally, in period 3 (which lasts until infinity), there is free entry into the modern sector,

and excess profits are eliminated. The latter happens in our case via upward adjustments in the

wage rate

The model is closed by describing production in the traditional sector. We assume that

the traditional sector operates under constant returns to scale, and employs labor and a fixed

factor. It will be convenient to use a specific functional form, so we write the production

function in the traditional sector as

y = ( l − s )α , where l is the total labor force of the economy,

is employment in the modern sector, and α is the factor share of labor in the traditional sector.

The diminishing marginal returns to labor in the traditional sector implies that the modern sector

faces a positively-sloped labor supply curve. Adjustments in wages will therefore play an

important equilibrating role for our economy. We assume that

n is small relative to l , so that we

do not have to keep track of the implications for the labor force as the number of entrepreneurs,

varies. The price of the traditional sector is fixed at 1 as the numeraire.

Social optimality: The full-information case. Consider first the case of an omniscient

social planner, who not only can allocate production at will, but also has full knowledge of the

for all the sectors without any prior investment. The full-information social optimum is easy to

describe. Rank modern-sector goods by their

, and let θmax denote the highest θ

. Let h(s)

denote the value marginal product of labor in the traditional sector (

h(s) = α( l − s )α-1). Then the

optimum is characterized by the following three conditions:

(2a)

> ( ) ⇒

max

h l

b

p

θ s = l (full specialization in good with θ

= θmax);

(2b)

h s l

b

h

(l ) ≥ p ≥ (0) ⇒ 0 ≤ ≤

max

;

(2c)

< (0) ⇒

max

h

b

p

θ s = 0 (full specialization in the traditional sector).

For modern-sector production to be viable, labor productivity in the good with

θmax must be high

enough that condition (2c)—value marginal product of labor in the traditional sector exceeds that

in the modern sector even when the economy is completely specialized in the former sector—can

be ruled out. We assume this is the case. This leaves the other two possibilities, where the

economy either fully specializes in the

θmax sector (2a), or is diversified between this and the

traditional sector (2b). Note that due to constant costs in the modern sector, it never pays for the

social planner to operate any modern-sector good with

< θmax. If the modern-sector is active,

the optimal thing to do is to concentrate all resources in the

θmax sector until either wages rise

sufficiently to equate the marginal value product of labor in the traditional sector or all labor is

exhausted.

Social optimality: Incomplete information. Now turn to a more realistic social planning

problem, where the planner can still allocate resources at will, but has to incur the same learning

costs as private entrepreneurs do. The planner has to decide how many investments,

m, to

undertake in period 1. Each modern-sector good is ex-ante identical with an expected

θ = ½.

However, once the productivity parameters of

m of the goods are revealed, only the one with the

highest

will matter: as before, the social planner will never want to produce a modern good

with a lower

than the maximum that is already known. So what matters to the planner is not

the expected

(which is independent of m), but the maximum of m draws (which does depend

m). Let the latter be denoted by ) ( max m θ . Since the ex-ante distribution of θ

is uniform over

[0, 1] and the draws are independent, the expected value of the rank statistic

θ max (m) has the

simple form

E( ) ( max m θ ) ≡ ) (m θ )

m/(1+m). We note that ) (m θ )

is increasing in

m, but at a

decreasing rate (

θ '(m)> 0,θ ''(m)< 0 ). So the social planner gains by increasing investment in

first period learning, but there exist diminishing returns.

In period 1,

mb units of labor are allocated to “investment” (s1

mb, where a subscript

denotes the time period). First-period output is therefore:

(3) ( )

α ( )α 1 1 y = l − s = l − mb

Productivity in the modern sector on periods 2 and 3 is not known for sure ex ante, so a riskneutral

social planner will use the expected productivity ) (

m θ )

to anticipate the future labor

market equilibrium. Labor allocation will be determined by setting the value marginal product

of labor equal in the two sectors in expected terms:

(4) ( ) ( )

1 ˆ

α − α −

l s

b

p m

Since the social planner will allow free entry starting from period 2 (and not period 3, as in the

decentralized case), there is no distinction between periods 2 and 3 in the social planning

problem, so we use the subscript “2” denote outcomes in both periods. Note also that we are

assuming a diversified equilibrium (corresponding to case 2(b) above). Equation (3) implicitly

defines

s2 as an increasing function of m. Using (3), the (expected) outputs of the modern (x) and

traditional (

y) sectors can be expressed in turn by

(5)

 





 





 





 





= = −

−

ˆ( ) ˆ( ) ˆ( )

θ θ θ

b

l p m

b

m

b

x s m

(6)

ˆ( )

( )

2 2

−

 





 





= − =

b

y l s p m

Since the social planner faces an intertemporal trade-off, we need to specify a discount

rate. For reasons that will become clear later, it is convenient for our purposes to specify the

discounting in the following way. Let the discount factor between period 1 and 2 be given by

δ,

while the instantaneous discount rate is

ρ. We shall express the present discounted value from

the vantage point of period 1 of a stream of continuous payments

K that are received during

period 2 and 3 as

δK/ρ . (Obviously, δ is a function of both ρ and the “length” of period 1.

Since we treat both of the latter as constants, there is no harm in treating

δ as a constant as well.)

The social planner’s maximand can now be written as the present discounted value of

production evaluated at world prices:

(7)

1 ( 2 2 ) max W y y px

= + +

with

y1, y2, and x2 as defined in (3), (5), and (6). The associated first-order condition is given by:

(8) ˆ'( ) ( )

1 0,

θ −α − α − =

s m b l mb

b

p

where

s2 is as defined implicitly by (4). The first term in this equation expresses the marginal

benefit of first-period investment: an increase in

m increases expected future productivity in the

modern sector, a gain that is spread over the equilibrium level of future employment in the

modern sector (

s2). The latter feature introduces an element of increasing returns to scale to

investment, since

s2 is itself an increasing function of m. The countervailing forces are the

diminishing return to ) (

m θ )

and the rising marginal cost of drawing labor away from the

traditional sector (captured by the second term in (8)). We assume these countervailing forces

dominate, so that the second-order condition is satisfied and we have an interior solution. It is

easy to verify, under these conditions, that the social planner’s choice of

m, as defined implicitly

by (8), is increasing in

p and decreasing in b.

Decentralized equilibrium. We now characterize the market equilibrium under the

conditions discussed previously, namely that the entrepreneurship decision is decentralized and

there is no entry (“imitation”) during period 2. That means that any entrepreneur who decides to

produce in period 2 earns a monopoly profit during the length

T of the second period. Let 1/θ

denote the expected value of 1/

θ . Then expected profits equal π =

− bw2 , where w2 is the

(expected) equilibrium wage that prevails in period 2. The present discounted value of these

profits, since they accrue only for a time of length

T, equals ( ) π

ρ ) 1− e− T ≡ R(T , with

(0)=0, 0<R(T)<1, and R’(T)>0. Under free entry, expected profits are just offset by the

investment cost that each entrepreneur incurs in setting up a firm in period 1. The condition for

zero-profits ex ante is therefore given by:

(9)

( )( ) 0 1

R T p

− bw2 − bw ≤

ρ θ

with the equation holding as an equality whenever

m>0. Even though m does not appear

explicitly here, this equation determines the quantity of aggregate investment since

m affects the

level of wages in periods 1 and 2 through the labor-market constraint.

The labor-market clearing equation for period 1 is:

(10)

w =α (l − mb)α − .

To determine the corresponding equation for period 2, we need to know the proportion of

investments that actually lead to production. Once

bw is sunk, an entrepreneur will choose to

remain in the modern sector as long revenues cover variable costs. Denote by

θ ~ the limit

productivity such that this is true, i.e.

p

bw

2 θ~ = . All entrepreneurs who draw θ [θ max (m),θ~] i ∈

will choose to produce in the second period. Therefore the expected level of employment in the

modern sector in period 2 is

p

s

( max (m) ~)mb ( max (m) bw2 )

= θ −θ = θ − . The labormarket

equilibrium in this period can then be written as

(11)

max 2

[ ( ) ]

−

 



 



= − −

α θ

p

w l m bw

Equations (9), (10), and (11) determine the three endogenous variables in the system,

m, w1, and

w
2

Outcomes in period 3 can be determined separately, but are not needed for our purposes.

We just note that with free entry in the third period, all but the highest productivity firm close

down, and imitation drives profits in that activity down to zero. The mechanism that achieves

rationalization of production is this: lower productivity activities are driven out as wages get bid

up by the entry of more firms into the highest-productivity activity. Since entrepreneurs make

zero profits in period 3, the investment level is determined only by outcomes in periods 1 and 2.

We now discuss the salient features of the decentralized equilibrium. First, note from (9)

that

m=0 for T sufficiently close to zero. The interpretation is straightforward: entrepreneurs

earn monopoly profits only during the duration of the second period. If the length of this period

shrinks to zero, there is no incentive to invest in discovering costs. If imitation is immediate, the

equilibrium level of investment and entrepreneurship is zero. Second, over the range of

T for

which

m>0, m, w1, and w2 are all increasing in T. That is, investment increases as the monopoly

period is extended. Third, as long as monopoly prevails, modern-sector production is diversified

and does not specialize in the good with the highest (known) productivity. Once imitation kicks

in with free entry, production in the modern sector is rationalized and only highest-productivity

activity survives.

Compared to the social optimum discussed previously, therefore, the decentralized

equilibrium suffers from two distinct inefficiencies. First, the level of investment and

entrepreneurship in the market equilibrium (call it

mc) does not coincide with that required for

social optimum (call it

m*). This can be verified easily by comparing the first-order condition of

the social planner (eq. 8) with the free entry condition of market equilibrium (eq. 9). Note that

the second term is the same for the two equations (substitute (10) into (9)), but that the first,

capturing the benefits of investment, differ. The planner cares about the economy-wide benefits,

while the entrepreneur cares about monopoly profits. Where

mc stands in relation to m* is

ambiguous in general. If monopoly profits are short-lived (

T is small), then mc < m*, and there is

too little investment and too few entrepreneurs in the market equilibrium. This is the case that

motivates the analysis of this paper. On the other hand, if free entry is delayed for very long (

is large), then

mc > m*, and there is over-investment motivated by the pursuit of monopoly

profits.

The second source of inefficiency is the existence of monopoly power in period 2. In our

framework, the costs of monopoly arise not from price-setting behavior (which is constrained by

the small-open economy assumption), but from the fact that restricted entry yields too little

specialization within the modern sector. Monopoly prevents the flow of resources into the

highest-productivity activity, and allows activities that will not eventually survive to do so.

The bottom line of the framework is that as long as

T is non-zero but not too large, the

laissez-faire equilibrium delivers (a)

too little investment and entrepreneurship, and (b) too much

production diversification in the modern sector.

Empirical evidence

It is difficult to provide direct empirical evidence for the model we have developed here

because much of our story has to do with outcomes that are not observed: the failure to develop

non-traditional activities because of inadequate incentives to invest in learning what one is good

at producing. Our argument rests on the idea that the returns to entrepreneurship of this

particular kind are easily competed away by free entry. Looking for systematic evidence that

successful investments are rapidly copied is a self-defeating strategy because there shouldn’t be

much evidence of this sort to the extent that our model does capture an important part of reality.

Entrepreneurial initiatives of this kind should tend to remain episodic, almost random events—

not systematic ones. We shall discuss a number of such cases below. Similarly, if we were to

learn that many successful new firms from developing countries operate with technologies that

are hard to copy or have devised successful strategies of product differentiation (with protection

against imitative entry)

12, this apparently contradictory finding may in fact be quite consistent

with our model. After all, a direct implication of our argument is that only investments that

provide such protection will be undertaken in equilibrium.

So we are forced to take an indirect route to the empirical evidence. We proceed by

providing support separately for three propositions that we consider to be building blocks for our

argument. First, there is a large element of uncertainty as to what a country will be good at

producing, once we move beyond broad aggregates such as “labor-intensive manufactures” to

specific products. Second, there are significant difficulties entailed in importing technology offthe-

shelf, with successful local adaptation requiring considerable domestic tinkering. Third,

domestic imitation often proceeds quite rapidly when the first two difficulties are overcome,

See for example the discussion of Spain’s sparkling wine and publishing industries in Guillén (2001, chap. 4).

bidding away the rents of the early incumbents. None these propositions is individually

controversial. As we highlight below, it is easy to find broad support for all of them in the

literatures on international trade, technology transfer, and economic history.

Predicting comparative advantage

The factor-endowments model is reasonably good at predicting the broad structure of

comparative advantage for developing countries. Leamer’s (1984) classic work, for example,

shows that the pattern of global commodity trade is well explained by the distribution of resource

endowments across countries. In this work, Leamer aggregates traded products into 10

commodities (petroleum, raw materials, forest products, tropical agriculture, animal products,

cereals, labor intensive, capital intensive, machinery, and chemicals) and factors of production

into 11 resources (capital, three types of labor, four types of land, coal, minerals, and oil). Wood

and Mayer (1999) and Mayer and Wood (1999) show that the developing countries’ pattern of

specialization--labor-intensive manufactures versus natural resource-based products--depends

critically on their ratio of human capital endowment to land.

However, for entrepreneurs trying to decide what they should invest in, these kinds of

findings are hardly helpful. Consider the information technology sector in India, which

represents a shining example of technological success in a low-income country. The industry

has grown from a very modest base in the early 1980s to export more than $6 billion of software

by 2000. Yet India is a country that one would have hardly expected to have a comparative

advantage in a technology-intensive sector (Kapur 2002). India ranks low in terms of

conventional indicators of IT penetration, has failed to develop leadership in other high-tech

sectors, and is relatively well endowed in unskilled (rather than skilled) workers. Until recently,

government policies have not been particularly friendly to investors in that sector. Yet

Bangalore-based companies like Infosys and Wipro have managed to create successful business

models that have been widely emulated not only by other local entrepreneurs, but also by foreign

companies investing in India.

After the fact, it is not difficult to enumerate some of the features that account for this

success: the time-zone difference that allows the processing to be done in Bangalore before the

West Coast of the U.S. is back at work in the morning, the linkages with the Indian diaspora in

Silicon Valley, the facility with the English language, the establishment of the Indian Institutes

of Technology, and even the departure of IBM in the late 1970s which gave Indian engineers a

head-start with UNIX-based software development (Kapur 2002). Ex ante, however, few of

these advantages were as visible. As Narayana Murthy, who founded Infosys with six other

software professionals in 1981, once put it to an interviewer, “[our] objective was to conduct an

experiment

….” (Nasdaq International Magazine 1999, emphasis added).

Even if we consider IT in India as an anomaly, the predictions of the factor-endowments

are too coarse to have much operational value. Knowing that Bangladesh’s comparative

advantage lies in labor-intensive manufactures and not in high-tech machinery is useful for sure,

but that still leaves hundreds, if not thousands, of different types of activity up for grabs. The

six-digit Harmonized Schedule (HS), which most countries use to assess customs duties,

comprises around 5,000 different commodity groups. The United States reports its trade

statistics in an even more detailed form, using a 10-digit system resulting in about 8,000

commodity groups. These statistical conventions reflect the tremendous variety of products that

are “out there” and which newcomers could in principle invest in. Neither trade theorists nor

management consultants can be good guides as to whether Bangladesh should produce hats or it

should produce bed sheets instead. There is much randomness in the process of discovering

what one can be good at. Evenson and Westphal (1995) summarize the firm-level studies thus:

...the body of case study research and anecdotal evidence includes numerous cases of

failure to achieve the minimum mastery needed to attain the levels of productivity

expected when the physical investment was undertaken. It also includes numerous cases

of unforeseen success in achieving sufficient mastery to exceed the expected levels of

productivity. In the former cases there is no technological development to benefit from

subsequent investments in implementing the same or similar technology. In the latter

cases there is technological development so that subsequent investments are implemented

with increasing efficiency due to the spillovers from previous experience. (1995, 2262-

63)

As it turns out, Bangladesh is very good at producing hats—more specifically “hats and

other headgear, knitted or from textile material not in strips” (HS 650590)—which constitute

Bangladesh’s third most important export item to the U.S. after men’s cotton shirts and trousers.

And it is not very good at producing bedsheets—specifically “bedsheets, pillowcases and bed

linen (incl. sets) – woven, not printed – cotton” (HS 630231)—of which it exports only a

miniscule amount. Is this a predictable result of innate comparative advantage? Consider

Pakistan, which is not too dissimilar to Bangladesh in its economic circumstances. Pakistan

exports a large quantity of bedsheets, but few hats. Since these commodities are fairly

standardized and labor-intensive, it is difficult to believe that the resource endowments and cost

structures of the two countries predispose them in any predictable way to specialize in one but

not the other. More likely, existing patterns of specialization are the consequence of historical

accidents and serendipitous choices by entrepreneurs.

The case of hats and bedsheets constitutes not the exception but the rule. Table 1 lists the

top 25 export items for Bangladesh and Pakistan in the U.S. market, to make the point that the

export structures of these two similar countries are surprisingly different at the level of 6-digit

HS categories. In fact, only six products are in the top 25 for both countries (these are identified

by bold letters). Among Pakistan’s major exports, curtains (HS 630392) and soccer balls (HS

950662) are some others that Bangladesh exports either very little or none at all.

We repeat the exercise in Tables 2 and 3 for two other pairs of countries in different parts

of the world and at different levels of incomes: Honduras versus the Dominican Republic and

Taiwan versus South Korea. Once again, these are pairs of countries that would be expected to

have fairly similar patterns of specializations. At a sufficiently aggregated level, that is certainly

true. Honduras and the Dominican Republic concentrate on garment exports, while Taiwan and

Korea focus on computer products. But at the level of individual products, there are again

striking differences. In 2000, the Dominican Republic exported $119 million worth of footwear

uppers (HS 640610) to the U.S. while Honduras exported none. Honduras is a major exporter of

ignition wiring sets (HS 854430) whereas the Dominican Republic barely exports any. Taiwan

exported $279 million worth of bicycles to the U.S. against Korea’s $623

thousand. Korea

exports a lot of air conditioning machines; Taiwan very few. The overlap among the top 25

exports of these two pairs is somewhat larger than in the case of Bangladesh/Pakistan, but far

from overwhelming—9 out of 25 for Honduras/Dominican Republic and 10 out of 25 for

Taiwan/South Korea.

Suppose we take export performance in different product categories as a measure of a

country’s relative productivity across these categories. Then these statistics reveal that the range

of products countries end up becoming good at producing is quite narrow. Exports are typically

highly concentrated, even in rich and relatively diversified countries. This is shown in Figure 1,

which displays the cumulative share of the top export items for the six countries mentioned

Edward N. Wolf reports a related finding in chapter 11 of Gomory and Baumol (2000). Wolf shows that the

industrial structures of the

advanced industrial countries have not converged even though aggregate total factor

productivity and relative factor endowments both have. Patterns of specialization are divergent, and once set,

remain stable.

above. The figure also includes German exports for comparison purposes. The top 4 export

commodities account for around 40 percent of South Korea’s and Honduras’ total exports to the

U.S. market, and 30 percent of the other countries’. The cumulative share of the top 25 export

items ranges from a high of 83 percent for Honduras to a low of 43 percent for Germany, with

the figure for all others lying above 60 percent (save for Taiwan).

The message we take from these numbers is twofold. First, for all economies except

possibly the most sophisticated, industrial success entails concentration in a relatively narrow

range of high-productivity activities. Second, the specific product lines that eventually prove to

be hits are typically highly uncertain and unpredictable.

Absorbing foreign technology

Neoclassical models of trade and economic growth presume not only that production

functions are known, but that they are identical across countries. Absent government-imposed

barriers, once an innovator figures out how to produce something, the same technology quickly

becomes available for adoption by followers in other countries.

14 In reality, the problem of

mastering modern production techniques is greatly complicated by the

tacit elements in

technology. This aspect of technological development is well recognized in discussions of

technology transfer by development economists and economic historians. In a survey for the

Handbook of Development Economics, for example, Evenson and Westphal express the point

clearly:

There have been some interesting departures from these assumptions in the recent literature. For example Basu

and Weil (1998) and Caselli and Coleman (2000) both focus on differences that arise in per-capita GDP due to

differences in “appropriate technologies.” However, neither paper considers the role played by uncertainty in

technology adoption.

… much of the knowledge about how to perform elementary processes and about how to

combine them in efficient systems is tacit, not feasibly embodied and neither codifiable

nor readily transferable. Thus, though two producers in the same circumstances may use

identical material inputs in conjunction with equal information, they may nonetheless

employ what are really two distinct techniques owing to differences in understanding of

the tacit elements. … Even supposing that [currently existing techniques] represent

optimal solutions for the circumstances in which they are respectively used, it does not

follow that they must necessarily be optimal with respect to different circumstances

where they have not been previously tested…. Once technology is understood in these

more complex terms, it is quite obvious that investments in technology are made

whenever it is newly applied, regardless of the novelty of the application. Learning about

technology and problem solving using the knowledge acquired in mastering technology

are not costless, even if the choices made in realizing the technique to be used are

identical in generic terms to choices previously made elsewhere…. A stream of

investments over time is typically required to overcome tacitness and thus achieve

mastery. Not only is much technology tacit, so too is much knowledge about the

specifics of local circumstances and about the ways that differences in circumstances

affect the productivity of particular techniques. Tacit knowledge can only be acquired

through investments in learning.… (Evenson and Westphal 1995, 2212-2214)

This lengthy quote expresses well many of the points we have tried to capture in our

stylized model above. Entrepreneurs in developing countries who start to produce standardized

goods that have long been on the scene in more advanced countries are forced to make

technological investments even though the goods in question may not be new in any sense.

These investments require discovering the tacit elements of technology and adapting them to the

local environment. That in turn entails a process of experimentation and learning. As Lall puts

it, “enterprises may not be able to predict if, when, how, and at what cost they would learn

enough to become fully competitive, even when the technology is well known and mature

elsewhere” (2000, 17). There is really no such thing as off-the-shelf technology.

One of the best-studied cases of technology transfer in economic history concerns the

cotton textile industry. Cotton mills in Britain’s Lancashire were the world leader in this

industry until well into the 20

th century. The diffusion of technology from Lancashire to other

regions—India, Japan, the United States—makes for a fascinating story about the role of local

capabilities in recipient countries. Lancashire’s technological mastery resided in mule spinning.

It gave rise not only to British dominance in world markets for cotton textiles, but also to the

establishment of a capital-goods export industry which sold textile machinery and expertise to

other countries. Lancashire firms such as Platt would supply foreign buyers a full technological

package, complete with equipment, know-how, training, and even managers and skilled workers.

Thanks to this, countries around the world had in principle access to the “same” technology that

was available to Lancashire textile producers. In the case of India, for example, an important

textile producer in its own right, Clark and Woolcott (2001, 10-11) note that “up to at least the

1940s there is no sign of any Indian lag in the types of machinery employed compared to the

advanced economies.” Yet productivity and profitability in Indian mills remained a fraction of

the levels in Lancashire, a failure that Clark and Woolcott attribute to India’s “inability to

employ the technology effectively” (2000, 10-11). Saxonhouse and Wright have studied the

history of the cotton textile industry in detail by analyzing the records of the textile machinery

exporters from Lancashire. They too note the heterogeneity in performance across countries

despite the access to identical technologies, and they observe: “the diversity of experience among

the newly emerging industries make it clear that progress was a two-sided affair, a mutual

adaptation between machines and local conditions” (2000, 10).

Japan, which started out way behind India, was able to overtake India in the early part of

the 20

th century and become a strong competitor of Britain. Japanese progress was based not on

the British mule-spinning technology, but on American-style ring spinning. The story of Japan’s

cotton textile industry is an intricate one involving both entrepreneurial ingenuity and state

initiative. The beginning of the industry can be dated to 1872 when the Meiji government took

over a small private mill with the intent of making into a model factory that would inspire private

investment in similar enterprises. While not a huge commercial success, the factory proved to be

a major popular attraction: “Such large crowds gathered outside to gawk at the mill that it

decided to charge a fee of ten sen for a tour” (Fletcher 1996, 53). The government started more

model factories, purchased spindles from England, and sold them to local entrepreneurs on very

easy terms (providing 10-year loans at zero interest) (Fletcher 1996, 54). The state-initiated

projects were unable to operate the imported technology efficiently, and after the 1880s the

government ceased direct assistance to cotton spinning. It was private firms with independent

operations and using variants of ring-spinning technology that came to dominate the Japanese

industry. Saxonhouse and Wright attribute Japan’s success to the Japanese producers’ skill in

adapting the American techniques to “an entirely different economic and cultural environment”

(2000, 10).

Japan’s experience with steel had many of the same elements. Early, “off-the-shelf”

imports of foreign technology during the Meiji period did not produce results until Japanese

engineers were able to adapt production processes to local conditions. Odagiri and Goto (1993,

90-91) describe the Meiji government’s efforts. In 1874, the Meiji government overrode a local

engineer’s more modest plans, and built a large steel furnace in Kamaishi using British

equipment and engineers. The plant had to be closed down after 196 days of operation because

of problems with local supplies of fuel: charcoal availability was inadequate, and when the fuel

source was switched to coke, the coke that could be obtained proved to be of poor quality. The

plant was restarted several years later after a smaller furnace based on local technology was

introduced. When in 1896 the government decided to build an integrated steel mill in Yawata,

“the lesson of Kamaishi appears to have been forgotten” (Odagiri and Goto 1993, 91). A

German company was hired to make the plans, equipment was brought in from Germany, and

German engineers and managers were put in charge. But the quality of coke in Japan differed

from that in Germany, making steel production very difficult. The mill was forced to shut down

within a year. Once again, the plant had to be reconfigured by domestic engineers to make the

furnaces more appropriate to Japanese coke.

The more recent history of East Asian and Latin American countries in the postwar

period is also replete with experiences that highlight the importance of local adaptation of

technology. Amsden (1989, 278) relates the story of Korean shipbuilder Hyundai, which

established its first shipyard in the early 1970s using imported designs from a Scottish firm

Scotlithgow. Because Scotlithgow had small capacity, the Scottish firm’s technique relied on

building the two halves of the ship separately, and then putting them together. When Hyundai

followed the same approach, trying to replicate all specifications in detail, it turned out that the

Korean workers could never make the two halves fit just right. There was more to the

technology than following the blueprint; the tacit knowledge required in building the ships could

not be transferred. Hyundai eventually became the world’s largest shipbuilder, but only after it

made a large investment in in-house design and research capabilities.

The Hyundai story is broadly indicative of Korea’s experience with regard to technology

imports: “It is a striking fact that formal purchase of technology in complete packages through

such means as turnkey plant contracts and licensing, plus their functional equivalent—direct

foreign investment—accounts for only a modest share of the technology that has been mastered

in Korea…” (Evenson and Westphal 1995, 2263-64). Similarly, in Taiwan, “activities such as

imitating, copying, or limited improvement of the existing foreign product (i.e., various reverse

engineering tactics), were the major sources of acquiring foreign technologies” (Hou and Gee in

Nelson, 390).

Even in technologically lagging countries, significant amount tinkering normally takes

place in order to adapt imported techniques. In Argentina, “many of the newly erected firms

found themselves needing to gradually develop ‘in house’ technological capabilities in areas

such as product design, production engineering, industrial planning, and organization where

‘from the shelf’ technology was not particularly well suited, given the highly idiosyncratic nature

of the local production structure” (Katz and Bercovich in Nelson 455; see also Katz and

Kosacoff 2000). These instances show how local firms have to engage in a series of

improvements to enhance productivity even when the technology is standardized and

implemented in the form of turnkey plants (see also Kim in Nelson 365). According to Evenson

and Westphal (1995, 2241): “Plants established under turnkey projects … often continue years

later to produce well under their design capacity owing to insufficient local effort to develop the

requisite production capabilities.”

Domestic spillovers and imitation

There is a large body of anecdotal evidence on the process of technology diffusion from

successful first-comers to later imitators. In their review of literature on technology transfer,

Evenson and Westphal (1995) cite several cases. They discuss the case of a successful foreignowned

farm machinery producer in Brazil, which progressively lost market share to indigenous

producers as the latter first imitated and then adapted the multinational’s models, making them

better suited to local conditions. Within twenty years, local firms came to dominate the market

for all but the most complex models (Evenson and Westphal 1995, 2257). Another case involves

rice-threshing technology:

… the key activity enabling Philippine rice producers to benefit from rice threshing

technology developed in Japan was the adaptive invention of a prototype thresher at

IRRI. Using this prototype, local inventors made the specific adaptations required to

enable the economic use of threshers in the many different circumstances in which they

are now used in the Philippines. (ibid, 2261)

Note that the International Rice Research Institute (IRRI) is a non-profit, public entity. Had a

private producer instead played IRRI’s role, it would have been unable to appropriate much of

the social returns due to the rapid entry of imitators.

The diffusion occurs typically, but not exclusively, through the turnover of skilled

workers and managers who have acquired the requisite expertise on the job. Indeed, the

poaching of such employees by later entrants is one of the most important competitive threats

that pioneering firms face. The early Japanese experience with cotton textiles again provides an

interesting illustration. A private enterprise, Osaka Spinning Co., Ltd., became so profitable in

the 1890s that its success “produced fifteen firms more or less modeled upon that firm”

(Yonekawa 1982, 29). The high priority that these early investors attached to retaining their

skilled workers is evident from the statutes of the Japan Spinners Association—the industry

association that was established in 1882 by the director of one of the early government-owned

mills: “The most specific provisions [in the Association’s regulations] … aimed at preventing

firms from raiding skilled workers from each other and from hiring troublemakers” (Yonekawa

1982, 56).

The practice of raiding pioneering firms’ workers has not lost its attractions over the

years. In Korea, for example, it is the main channel through which technology has diffused

within the domestic economy. According to Dahlman et al., when they establish new lines of

production Korean firms, “simply offer higher salaries to engineers and other technicians from

other local firms, technicians already skilled in the installation and operation of the new process

being adopted” (1985, 45).

15 Most late entrants poach personnel from incumbents—initially

state enterprises and later on other private enterprises—to set up their own facilities (Kim 1993,

365). In Taiwan, labor mobility appears to have played a similar role, particularly in diffusing

technology from the local subsidiaries of foreign firms to other domestic firms (Hou and Gee

1993, 389). In the Indian IT industry, poaching of workers by late entrants and employees

leaving to start their own companies are among the main headaches faced by established firms

(Lateef 1997, Forbes Global 1998).

Consider the garment industry in Bangladesh, which is a well-studied case of industrial

take-off (Rhee 1990). At the origin of the Bangladeshi garment “miracle” lies a largely

serendipitous investment made by a local entrepreneur in a joint venture with Daewoo of Korea.

As Rhee makes clear in his account, few local investors had shown interest in garments, and

Daewoo did not hold very high hopes for the project either. (Daewoo’s interest in investing in

Bangladesh derived from other reasons). But once the venture, called Desh, proved successful,

imitation was very rapid: “…115 of the initial corps of 130 workers trained by Daewoo in Korea

left Desh at different times following the end of the Desh-Daewoo agreement to set up their own,

often competing, garment exporting firms” (Rhee 1990, 341). [See the story of the Mohammadi

company in the paper.] As a result, the industry grew from a handful of factories in 1979 to

more than 700 exporters by 1985 (ibid). The spread of know-how from the workers initially

trained by Desh was critical—much more so than the availability of capital and other resources.

Once it became common knowledge that Bangladesh had a profitable market niche in garments,

the industry took off like a mushroom. But note as well that, as our previous review of export

statistics for Bangladesh, has highlighted, the consequence of all this was not a generalized

As Dahlman et al. note, “such footloose personnel can speed diffusion considerably, but they also discourage

investments by firms in their human capital” (1985, 45).

enhancement of industrial capacity in manufactures, or even in all types of labor-intensive

manufactures. The spillovers to other types of manufacturing seem to have been very limited,

and Bangladeshi exports remain extremely undiversified.

Or consider the Colombian cut flower industry, the story of which is told in Rhee and

Belot (1990). How did Colombia get from nowhere to being the largest supplier of cut flowers

to the U.S.? Once again, it was a single entrepreneur that was responsible for it all. Thomas

Kehler, an American businessman in Colombia, happened to be looking for a business

opportunity. He and some partners did a feasibility study, jointly put up an initial investment of

$100,000, and set up Floramerica in 1969. By 1986, annual sales had risen to $50 million, and

this produced a tremendous demonstration effect:

Other companies have copied Floramerica’s production and marketing methods; this

dynamic process has been facilitated by the movement of key staff who embodied the

know-how accumulated at Floramerica… David Cheever, one of the founders of

Floramerica, left the company after two years to become a consultant to the growing

Colombian flower industry. By assisting in the start-up of many new flower companies,

he has been a driving force in diffusing know-how accumulated in Floramerica

throughout the country. Second, two salesmen at Floramerica’s Miami sales office left

the company ….” (Rhee and Belot 1990, 31).

By 1990, there were about 250 flower export firms in Colombia, the vast majority of which were

fully Colombian-owned (ibid).

Policy issues

We have focused here on an often-neglected aspect of economic development. Learning

what one is good at producing is an important determinant of structural change, but it is also one

that is unlikely to be adequately provided under laissez-faire. The social returns to such learning

For more examples of how competitive industries are often started as a result of efforts by individual

entrepreneurs who combine vision and expertise, see Rhee and Belot (1990). Rhee and Belot call these “catalysts.”

are likely to be much larger than the private returns, as successful “discoveries” of what can be

produced at low cost can be easily imitated in general.

Our model is highly stylized in order to make these points clearly. However, it is less

interesting to use it as a tool for specific policy design. It makes very many assumptions that

simplify the issues but that do not capture important aspects of the world. For example, it does

not include non-traded goods or intermediate inputs; it takes prices to be given independently of

output; it assumes no need for capital and has no financial imperfections. Optimal policies will

be highly sensitive to the assumptions adopted in these areas. Instead, we will use our model to

reinterpret some of the standard instruments that governments have used in order to promote

self-discovery.

The key policy recommendation that comes out of our framework is that laissez-faire

leads to underprovision of innovation and governments need to play a dual role in fostering

industrial growth and transformation. They need to encourage entrepreneurship and investment

in new activities ex ante, but push out unproductive firms and sectors ex post. This is of course

easier said than done. The specifics of how this can be managed is likely to differ considerably

from country to country, depending on administrative capability, the prevailing incentive regime,

the flexibility of the fiscal system, the degree of sophistication of the financial sector, and the

underlying political economy. Governments without adequate capacity to exercise leadership

over their private sectors are likely to mess things up rather than do better. But there are

examples to suggest that the job can be done.

In attempting to promote innovation, governments have used a variety of instruments

such as trade protection, public sector credit, tax holidays, and investment and export subsidies.

Clearly, all appropriate policy interventions need to increase the expected pay-off to innovation.

However, interventions typically create other distortions. For example, they may lower

θ ~ by

making less productive activities privately profitable, inefficiently increasing the heterogeneity

of the modern sector. Moreover, if the instrument does not adequately discriminate between

innovators and copycats, it will promote early entry, thus limiting the benefits to innovators

while increasing the social cost of the intervention since copycats will get part of the resources

transferred.

Interventions can be further classified in two groups, depending on whether they

compensate innovators in case they fail (i.e. in case they draw a low

θ ), or increase the payoff in

case they are successful. The first type of interventions is likely to create moral hazard, but the

second type will not help those who lack the resources to finance activities in period 1.

Thus, for example, temporary trade protection is far from an ideal instrument. It may

increase expected profits of innovators, but it does so only for firms selling in the local market.

Moreover, since it does not discriminate between innovators and copycats, it promotes early

entry, thus lowering the expected payoff to innovation while inefficiently channeling resources

to copycats. Moreover, as protection gets extended to intermediate goods, it will make

downstream activities less competitive. As a consequence, innovation will tend to focus on

domestic markets, instead of new export activities. Since domestic markets are small relative to

world markets, the social returns to innovation will be likewise diminished. Hence, trade

protection is not an efficient way of promoting self-discovery

17.

Export subsidies avoid the anti-export bias of trade protection, but do not discriminate

between innovators and copycats. They also lower

θ ~ , promoting excessive diversification. One

More convoluted arguments in favor of trade protection would need to be based on the positive effects that having

a captive domestic markets may have on innovation for exports. It is the local market that generated the monopoly

profits that sustain innovation that can eventually become export competitive. However, there are bound to be many

firms that are able to survive behind high protective barriers without ever having a chance to become exporters.

advantage of exports subsidies is that they can be relatively good at discriminating between

successful and unsuccessful performers ex post. Low-cost producers are more likely to incur the

sunk costs of exporting. Therefore providing subsidies contingent on exporting can allow policy

makers to sort out firms and sectors that have high productivity from those that don’t. Moreover,

the fact that they appear explicitly in the budget makes it more likely that they will be granted

only on a temporary basis. This strategy was actively used by the South Korean and Taiwanese

governments during their industrial drives in the 1960s and 1970s. New international

agreements in the context of the World Trade Organization have made such subsidies illegal.

By contrast, public sector credit or guarantees operate by transferring part of the risk of

failure to the government: if the project is successful, the loan gets repaid; if it is not, then the

firm will default. Credit has the advantage that it can be made discretionary and thus, it can be

targeted on innovators and not on copycats. In fact, in Latin America during the heyday of

industrial policy, it was common for development banks to require that no other domestic

supplier exist before granting financial support. When market reforms were introduced in the

early 1990s, this policy was seen as particularly inadequate: it limited entry and competition.

However, in the context of our model, this is precisely what is needed: public resources should

be concentrated on the first entrants. Moreover, judgment can be used in choosing ideas that

promise the greater social benefits. However, with discretion come problems of political

influence, corruption or at least moral hazard.

Policy instruments can also differ with respect to elements that are required to impose

discipline. For example, tariff protection and export subsidies need to be taken away at some

point in time. The latter will be under fiscal and international pressure to disappear. The former

may require more political effort. Firms used to government loans need to be let go: at some

point in time loans cannot be rolled over forever and the firm needs to be required to return the

loan, which implies that it is able to convince private financial markets of its viability or declare

default. A brief summary of the costs and benefits of these different types of policies is

presented in the table below.

Trade

protection

Export

subsidies

Government

loans and

guarantees

Increases payoff

innovation

by…

…increasing

the returns to

success and

lowering

θ ~

…increasing

the returns to

success and

lowering

θ ~

…lowering the

losses in case of

failure and

lowering

θ ~

Can

discriminate

between

innovators and

copycats

No No, but better

at rewarding

high

productivity

activities

Yes

Other

distortions to

innovation

Biased against

export activities

Distorts risk

assessments

(moral hazard)

Action required

to impose

discipline

Lowering

tariffs

Lowering

subsidies

Cut funding

What stands out in many discussions of East Asia is how governments in the region were

unusually good at supplying the requisite discipline:

Where Korea differs from other developing countries in promoting big business was the

discipline the state exercised over these

chaebols by penalizing poor performance and

rewarding only good ones… The government as the controller of commercial banks was

in a powerful position to punish poorly managed firms by freezing bank credit. As a

result only three of the largest 10

chaebols in 1965—Samsung, Lucky-Goldstar, and

Ssangyong—remained on the same list 10 years later. Similarly, seven of the largest 10

in 1975 remained on the same list in 1985.” (Kim 1993, 363).

The Korean government was quick to shelve its plans for supporting particular firms or

industries when new information suggested that productivity would lag (Westphal 1981, 34).

The development of textile industry in Taiwan in the 1950s provides a particularly clear example

of the strategy suggested by our framework. The Taiwanese government subsidized entry into

the industry by supplying inputs and spinning mills, providing working capital, imposing import

restrictions, and buying up the resulting production. Local production grew spectacularly as a

result. But the government also subsequently restricted entry and tried to prune the nonproductive

firms (see Evans 1995, 57, and Wade 1990, 79). Japan used a similar combination of

state promotion/protection followed by rationalization in the computer industry (Evans 1995,

101).

Consider on the other hand Latin America during its import-substituting industrialization

(ISI) period. Latin American ISI produced many successful firms, but also an industrial structure

that was too diversified—too many low productivity firms alongside the high performers (see for

example Dahlman and Frischtak 1993, 424-5). Discipline was to come to Latin America in the

1990s in the form of trade openness, and many of the low-productivity firms were eventually

driven out. Countries such as Argentina, Brazil, and Chile deepened their specialization in

capital-intensive, natural resource based industries, while others like Mexico and the smaller

Central American countries increased their focus on assembly industries servicing the U.S.

market (Katz 2001). But, as our framework suggests, openness and institutional reform were not

enough to spark a significant new wave of entrepreneurship and investment in non-traditional

activities.

A crude, but useful characterization of the policy environments in East Asia and Latin

America, as viewed from the perspective of our framework, would be as follows. East Asian

governments provided their firms during the 1960s and 1970s with both promotion (the carrot)

and discipline (the stick). Against this benchmark, Latin American industrial performance has

fallen short because of varying shortcomings. Under ISI, Latin America was marked by plenty

of promotion, but too little discipline. In the 1990s, Latin America has considerable discipline

(provided through competitive markets and open trade), but too little promotion. Interestingly,

growth was faster in the 1960’s and 1970’s than in the 1990’s.

REFERENCES

Acemoglu, Daron, Philippe Aghion, and Fabrizio Zilibotti, “Distance to Frontier, Selection, and

Economic Growth,” NBER Working Paper No. 9066, July 2002.

Amsden, Alice H., Asia’s Next Giant: South Korea and Late Industrialization, Oxford University

Press, New York and Oxford, 1989.

Basu, Susanto, and David N. Weil, “Appropriate Technology and Growth,” Quarterly Journal of

Economics, November 1998.

Caselli, Francesco, and Wilbur John Coleman II, “The World Technology Frontier,” NBER

Working Paper 7904, September 2000.

Clark, Gregory, and Susan Wolcott, “One Polity, Many Countries: Economic Growth in India

1873-2000” in Dani Rodrik, ed., Analytical Case Studies on Growth, 2001, forthcoming.

Dahlman, Carl, Bruce Ross-Larsen, and Larry Westphal, “Managing Technological

Development: Lessons from the Newly Industrializing Countries,” World Bank Staff Working

Paper No. 717, 1985.

Easterly, W., N. Loayza and P. Montiel, “Has Latin America’s Post-Reform Growth Been

Disappointing?”

Journal of International Economics, Supplement 43(3-4), 1997, 287-312.

Evenson, Robert, and Larry E. Westphal, “Technological Change and Technology Strategy,” ch.

37 in Jere Behrman and T.N. Srinivasan, eds., Handbook of Development Economics, vol. 3A,

Amsterdam, North-Hollan, 1995, pp. 2209-2229.

Fernández-Arias, E., and P. Montiel, “Reform and Growth in Latin America: All Pain, No

Gain?” Research Department Working Paper 351. Washington, DC: Inter-American

Development Bank, Research Department, 1997.

Fletcher, W. Miles III, “The Japan Spinners Association: Creating Industrial Policy in Meiji

Japan,” Journal of Japanese Studies, 22:1, Winter 1996, 49-75.

Forbes Global, “Big Bang in Bangalore,” 11/2/1998.

Gomory, Ralph E., and William J. Baumol, Global Trade and Conflicting National Interests, The

MIT Press, Cambridge, MA, 2000.

Guillén, Mauro F., The Limits of Convergence: Globalization and Organizational Change in

Argentina, South Korea, and Spain, Princeton University Press, Princeton, N.J., 2001.

Hoff, Karla, “Bayesian Learning in an Infant Industry Model,” Journal of International

Economics, 43, 1997, 409-436.

Hou, Chi-Ming, and San Gee, “National Systems Supporting Technical Advance in Industry:

The Case of Taiwan,” in Richard R. Nelson, ed., National Innovation Systems: A Comparative

Analysis, Oxford University Press, New York and Oxford, 1993, 384-413.

Kapur, Devesh, “The Causes and Consequences of India’s IT Boom,” India Review, vol. 1, no.

2, April 2002, 91-110.

Katz, Jorge, and Bernardo Kosacoff, “Technological Learning, Institution Building and the

Macroeconomics of Import Substitution,” chap. 2 in Enrique Cardenas, Jose Antonio Ocampo,

and Rosemary Thorp, eds., An Economic History of Twentieth-Century Latin America, vol 3:

Industrialization and the State in Latin America, The Postwar Years, Palgrave, Hampshire, 2000.

Katz, Jorge M., Structural Reforms, Productivity and Technological Change in Latin America,

Economic Commission for Latin America and the Caribbean, Santiago, Chile, 2001.

Lall, Sanjaya, “Technological Change and Industrialization in the Asian Newly Industrializing

Economies: Achievements and Challenges,” in L. Kim and R.R. Nelson, eds., Technology,

Learning and Innovation, Cambridge University Press, New York, 2000, 13-68.

Lateef, Asma, “Linking up with the global economy: A case study of the Bangalore software

industry,” New Industrial Organization Programme, DP/96/1997, International Institute for

Labour Studies, International Labor Organization, 1997.

Leamer, Edward E., Sources of Comparative Advantage: Theory and Evidence, MIT Press,

Cambridge, MA, 1984.

Lora, Eduardo, “Structural Reforms in Latin America:What Has Been Reformed

and How to Measure it”, Research Department Working Paper 350. Washington, DC: Inter-

American Development Bank, Research Department, 1997.

Lora, Eduardo, “Structural Reforms in Latin America: What Has Been Measured and How to

Measure It,” Research Department Working Paper 466, Washington, DC: Inter-American

Development Bank, December 2001.

Lora, E., and F. Barrera, “A Decade of Structural Reforms in Latin America: Growth,

Productivity and Investment are Not What They Used to Be.” Research Department Working

Paper 350. Washington, DC: Inter-American Development Bank, Research Department, 1997.

Lora, Eduardo and Ugo Panizza, “Structural Reforms in Latin America under Scrutiny,” mimeo,

Inter-American Development Bank, 2002.

Mayer, Jörg, and Adrian Wood, "South Asia's Export Structure in a Comparative Perspective,"

IDS, Sussex, 1999, 1-45.

http://www.ids.ac.uk/ids/publicat/wp/wp91.pdf

Morley, Samuel A., Roberto Machado, and Stefano Pettinato, “Indexes of Structural Reform in

Latin America,” ECLAC, Santiago, Chile (LC/L.1166), January 1999.

Mukand, Sharun, and Dani Rodrik, “In Search of the Holy Grail: Policy Convergence,

Experimentation, and Economic Performance,” January 2002.

Nasdaq International Magazine, #24, September 1999.

Nelson, Richard R., “On Technological Capabilities and their Acquisition,” in R. Evenson and

G. Ranis, eds., Science and Technology: Lessons for Development Policy, Westview Press,

Boulder, CO, 2000, 71-80.

Odagiri, Hiroyuki, and Akira Goto, “The Japanese System of Innovation: Past, Present, and

Future,” in Richard R. Nelson, ed., National Innovation Systems: A Comparative Analysis,

Oxford University Press, New York and Oxford, 1993, 76-114.

Rauch, James E., and Joel Watson, “Entrepreneurship in International Trade,” University of

California at San Diego Economics Discussion Paper 2001-20, December 2001.

Roll, Richard, and John Tallbott, “Why Many Developing Countries Just Aren’t,” unpublished

paper, November 13, 2001.

Saxonhouse, Gary, and Gavin Wright, “Technological Evolution in Cotton Spinning, 1878-

1933,” unpublished paper, Stanford University, January 2000.

Stiglitz, Joseph E., “Learning to Learn, Localized Learning and Technological Progress,” in P.

Dasgupta and P. Stoneman, eds., Economic Policy and Technological Performance, Cambridge

University Press, Cambridge, 1987, 125-153.

Westphal, Larry E., “Empirical Justification for Infant Industry Promotion,” World Bank Staff

Working Paper No. 445, March 1981.

Wood, Adrian, and Jörg Mayer, "Africa's Export Structure in a Comparative Perspective," IDS,

Sussex, September 1999, 1-38.

http://www.ids.ac.uk/ids/global/pdfs/wood2a.pdf

Yonekawa, Shin-ichi, “The Growth of Cotton Spinning Firms: A Comparative Study,” in A.

Okochji and S. Yonekawa, eds., The Textile Industry and its Business Climate, the International

Conference on Business History 8, University of Tokyo Press, Tokyo, 1982, 1-44.

Zeira, Joseph, “Investment as a Process of Search,” Journal of Political Economy, vol. 95, no. 1,

1987, 204-210.

Table 1: Comparison of top 25 export items to the U.S in 2000: Bangladesh and Pakistan (US$ million)

A. Bangladesh

HS Code and Product Bangladesh exports Pakistan exports

620520 - MENS/BOYS SHIRTS - WOVEN - COTTON 236.2 34.5

620342 - MENS/BOYS TROUSERS, OVERALLS AND SHORTS - WOVEN - COTTON 183.9 70.9

3 650590 - HATS AND OTHER HEADGEAR - KNITTED OR FROM TEXTILE MATERIAL NOT IN STRIPS 175.1 0.7

4 620462 - WOMENS/GIRLS TROUSERS, OVERALLS AND SHORTS - WOVEN - COTTON 148.5 17.2

5 030613 - SHRIMPS AND PRAWNS - FROZEN 145.2 7.5

6 620630 - WOMENS/GIRLS BLOUSES, SHIRTS AND SHIRT-BLOUSES - WOVEN - COTTON 141.4 11.4

7 611030 - SWEATERS, SWEATSHIRTS AND WAIST-COATS - KNITTED - MAN-MADE FIBRES 123.3 5.6

611020 - SWEATERS, SWEATSHIRTS AND WAIST-COATS - KNITTED - COTTON 99.4 284.1

9 620193 - MENS/BOYS ANORAKS, SKI AND WIND JACKETS AND SIMILAR ARTICLES - WOVEN - MAN-MADE FIBRES 82.0 2.6

620343 - MENS/BOYS TROUSERS, OVERALLS AND SHORTS - WOVEN - SYNTHETIC FIBRES 61.4 20.0

11 610821 - WOMENS/GIRLS BRIEFS AND PANTIES - KNITTED - COTTON 51.5 0.7

12 620293 - WOMENS/GIRLS ANORAKS, SKI AND WIND JACKETS AND SIMILAR ARTICLES - WOVEN - MAN-MADE FIBRES 48.7 2.4

13 620920 - BABIES GARMENTS (INCL COATS AND SNOWSUITS) AND CLOTHING ACCESSORIES - WOVEN - COTTON 36.9 5.5

14 621111 - MENS/BOYS SWIMWEAR - WOVEN 35.7 0.7

621142 - WOMENS/GIRLS COVERALLS, SMOCKS AND GARMENTS NES - WOVEN - COTTON 35.7 24.9

16 620640 - WOMENS/GIRLS BLOUSES, SHIRTS AND SHIRT-BLOUSES - WOVEN - MAN-MADE FIBRES 34.1 0.7

17 620821 - WOMENS/GIRLS PYJAMAS AND NIGHTDRESSES - WOVEN - COTTON 31.6 3.6

18 620530 - MENS/BOYS SHIRTS - WOVEN - MAN-MADE FIBRES 31.3 2.4

19 620463 - WOMENS/GIRLS TROUSERS, OVERALLS AND SHORTS - WOVEN - SYNTHETIC FIBRES 31.3 6.8

20 630622 - TENTS - SYNTHETIC FIBRES 31.2 0.6

21 620452 - WOMENS/GIRLS SKIRTS AND DIVIDED SKIRTS - WOVEN - COTTON 25.5 4.5

22 611120 - BABIES GARMENTS (INCL HOSIERY) AND CLOTHING ACCESSORIES - KNITTED - COTTON 23.7 11.7

23 620469 - WOMENS/GIRLS TROUSERS, OVERALLS AND SHORTS - WOVEN - TEXTILE NES 23.2 0.9

24 621040 - MENS/BOYS SNOWSUITS AND SIMILAR GARMENTS NES - WOVEN - COATED FABRICS 20.6 0.9

610510 - MEN'S/BOYS SHIRTS - KNITTED - COTTON 20.6 172.4

SUB-TOTAL 1878.1 693.4

OTHERS 540.3 1474.0

TOTAL (ALL PRODUCTS) 2418.4 2167.4

B. Pakistan

HS Code and Product Pakistan exports Bangladesh exports

611020 - SWEATERS, SWEATSHIRTS AND WAIST-COATS - KNITTED - COTTON 284.1 99.4

610510 - MEN'S/BOYS SHIRTS - KNITTED - COTTON 172.4 20.6

3 630260 - COTTON TERRY TOWELS AND HOUSEHOLD LINEN OF COTTON TERRY FABRICS 138.0 19.6

4 570110 - CARPETS - WOOL/ FINE HAIR - KNOTTED 100.8 0.0

5 630710 - INDUSTRIAL SHOP TOWELS (FLOOR/DISH CLOTHS, DUSTERS AND CLEANING CLOTHS) 85.9 11.1

620342 - MENS/BOYS TROUSERS, OVERALLS AND SHORTS - WOVEN - COTTON 70.9 183.9

7 420310 - ARTICLES OF APPAREL OR CLOTHING (EXCEPT GLOVES) - LEATHER 70.3 0.0

8 630210 - BEDSHEETS, PILLOWCASES AND BED LINEN (INCL SETS) - KNITTED OR CROCHETED 50.9 0.0

9 520819 - COTTON (>85%) FABRICS - WOVEN NES - UNBLEACHED - <200 G/M2 39.1 0.0

10 610910 - T-SHIRTS, SINGLETS AND OTHER VESTS - KNITTED - COTTON 34.7 14.9

620520 - MENS/BOYS SHIRTS - WOVEN - COTTON 34.5 236.2

12 520812 - COTTON (>85%) FABRICS - PLAIN WEAVE - UNBLEACHED - 100-200 G/M2 32.6 2.7

13 630221 - BEDSHEETS, PILLOWCASES AND BED LINEN (INCL SETS) - WOVEN, PRINTED - COTTON 31.3 4.2

14 630222 - BEDSHEETS, PILLOWCASES AND BED LINEN (INCL SETS) - WOVEN, PRINTED - MAN-MADE FIBRES 30.5 0.0

15 711319 - ARTICLES OF JEWELLERY - PRECIOUS METALS (OTHER THAN SILVER) 25.7 0.0

621142 - WOMENS/GIRLS COVERALLS, SMOCKS AND GARMENTS NES - WOVEN - COTTON 24.9 35.7

17 520512 - COTTON (>85%) YARN - SINGLE, UNCOMBED 714.29-232.56 DECITEX - NON-RETAIL 24.2 0.0

18 901890 - INSTRUMENTS AND APPLIANCES USED IN MEDICAL, SURGICAL OR VETERINARY SCIENCES NES (INCL PARTS) 24.1 0.0

19 520839 - COTTON (>85%) FABRICS - WOVEN NES - DYED 1 COLOR - <200 G/M2 23.4 0.2

20 950662 - OTHER INFLATABLE BALLS FOR SPORTS 23.2 0.0

21 630392 - CURTAINS (INCL DRAPES), INTERIOR BLINDS AND BED VALANCES - WOVEN - SYNTHETIC FIBRES 22.6 0.1

22 630231 - BEDSHEETS, PILLOWCASES AND BED LINEN (INCL SETS) - WOVEN, NOT PRINTED - COTTON 21.8 0.0

23 521021 - COTTON (<85%) /MAN-MADE FABRICS - PLAIN WEAVE - BLEACHED - <200 G/M2 21.4 0.0

24 630790 - MADE UP ARTICLES OF TEXTILE MATERIALS NES (INCLUDING DRESS PATTERNS AND SHOE LACES) 20.3 0.7

620343 - MENS/BOYS TROUSERS, OVERALLS AND SHORTS - WOVEN - SYNTHETIC FIBRES 20.0 61.4

SUB-TOTAL 1427.7 690.7

OTHERS 739.8 1727.7

TOTAL (ALL PRODUCTS) 2167.4 2418.4

Source: Trade Data Online from Industry Canada (http://strategis.ic.gc.ca/sc_mrkti/tdst/engdoc/tr_homep.html)

Table 2: Comparison of top 25 export items to the U.S in 2000: Honduras and Dominican Republic (US$ million)

A. Honduras

HS Code and Product Honduras exports

Dominican Republic

exports

610910 - T-SHIRTS, SINGLETS AND OTHER VESTS - KNITTED - COTTON 499.1 173.7

611020 - SWEATERS, SWEATSHIRTS AND WAIST-COATS - KNITTED - COTTON 458.7 118.1

620342 - MENS/BOYS TROUSERS, OVERALLS AND SHORTS - WOVEN - COTTON 142.6 493.0

4 610711 - MEN'S/BOYS UNDERWEAR - KNITTED - COTTON 133.1 123.7

621210 - BRASSIERES 128.0 144.6

6 610990 - T-SHIRTS, SINGLETS AND OTHER VESTS - KNITTED - TEXTILE NES 122.6 20.8

7 610510 - MEN'S/BOYS SHIRTS - KNITTED - COTTON 121.0 14.0

8 620520 - MENS/BOYS SHIRTS - WOVEN - COTTON 112.1 21.6

9 090111 - COFFEE - NOT ROASTED, NOT DECAFFEINATED 97.2 3.2

10 030613 - SHRIMPS AND PRAWNS - FROZEN 83.6 0.4

11 080300 - BANANAS, INCLUDING PLANTAINS - FRESH OR DRIED 76.1 2.8

620462 - WOMENS/GIRLS TROUSERS, OVERALLS AND SHORTS - WOVEN - COTTON 74.7 247.8

13 854430 - IGNITION WIRING SETS AND OTHER WIRING SETS USED FOR VEHICLES, AIRCRAFT OR SHIPS 57.9 0.0

14 621010 - RAINCOATS AND GARMENTS NES MADE UP OF TEXTILE FELTS AND NONWOVEN TEXTILE FABRICS 54.8 7.6

240210 - CIGARS, CHEROOTS AND CIGARILLOS (CONTAINING TOBACCO) 49.8 191.5

610822 - WOMENS/GIRLS BRIEFS AND PANTIES - KNITTED - MAN-MADE FIBRES 49.6 37.5

620343 - MENS/BOYS TROUSERS, OVERALLS AND SHORTS - WOVEN - SYNTHETIC FIBRES 47.9 194.5

18 611030 - SWEATERS, SWEATSHIRTS AND WAIST-COATS - KNITTED - MAN-MADE FIBRES 45.8 23.8

610821 - WOMENS/GIRLS BRIEFS AND PANTIES - KNITTED - COTTON 45.4 36.3

20 620530 - MENS/BOYS SHIRTS - WOVEN - MAN-MADE FIBRES 39.0 13.6

21 610462 - WOMENS/GIRLS TROUSERS, OVERALLS AND SHORTS - KNITTED - COTTON 34.7 19.6

22 030611 - ROCK LOBSTER AND OTHER SEA CRAWFISH - FROZEN 26.4 0.0

23 080719 - MELONS OTHER THAN WATERMELONS - FRESH 22.7 5.9

24 610342 - MEN'S/BOYS TROUSERS, OVERALLS AND SHORTS - KNITTED - COTTON 20.7 18.2

25 620711 - MENS/BOYS UNDERWEAR - WOVEN - COTTON 20.1 2.9

SUB-TOTAL 2563.7 1915.2

OTHERS 526.3 2468.8

TOTAL (ALL PRODUCTS) 3090.0 4384.0

B. Dominican Republic

HS Code and Product

Dominican Republic

exports Honduras exports

620342 - MENS/BOYS TROUSERS, OVERALLS AND SHORTS - WOVEN - COTTON 493.0 142.6

2 901890 - INSTRUMENTS AND APPLIANCES USED IN MEDICAL, SURGICAL OR VETERINARY SCIENCES NES (INCL PARTS) 339.9 0.0

620462 - WOMENS/GIRLS TROUSERS, OVERALLS AND SHORTS - WOVEN - COTTON 247.8 74.7

620343 - MENS/BOYS TROUSERS, OVERALLS AND SHORTS - WOVEN - SYNTHETIC FIBRES 194.5 47.9

240210 - CIGARS, CHEROOTS AND CIGARILLOS (CONTAINING TOBACCO) 191.5 49.8

610910 - T-SHIRTS, SINGLETS AND OTHER VESTS - KNITTED - COTTON 173.7 499.1

621210 - BRASSIERES 144.6 128.0

8 711319 - ARTICLES OF JEWELLERY - PRECIOUS METALS (OTHER THAN SILVER) 142.3 0.0

9 610711 - MEN'S/BOYS UNDERWEAR - KNITTED - COTTON 123.7 133.1

10 640610 - PARTS OF FOOTWEAR - UPPERS (EXCLUDING STIFFENERS) 119.2 0.0

611020 - SWEATERS, SWEATSHIRTS AND WAIST-COATS - KNITTED - COTTON 118.1 458.7

12 853620 - AUTOMATIC CIRCUIT BREAKERS - VOLTAGE NOT EXCEEDING 1,000 VOLTS 82.4 0.0

13 170111 - CANE SUGAR - RAW 78.0 4.4

14 720260 - FERRO-NICKEL 70.5 0.0

15 650590 - HATS AND OTHER HEADGEAR - KNITTED OR FROM TEXTILE MATERIAL NOT IN STRIPS 57.0 1.1

16 620333 - MENS/BOYS JACKETS AND BLAZERS - WOVEN - SYNTHETIC FIBRES 51.5 2.8

17 853690 - OTHER APPARATUS - FOR SWITCHING, PROTECTING OR CONNECTING ELECTRIC CIRCUITS - NOT EXCEEDING 1,000 V 51.1 0.0

18 850440 - ELECTRIC STATIC CONVERTERS (INCL POWER SUPPLIES, RECTIFIERS AND INVERTERS) 48.3 0.0

19 611592 - SOCKS AND STOCKINGS - KNITTED - COTTON 46.4 11.2

20 620463 - WOMENS/GIRLS TROUSERS, OVERALLS AND SHORTS - WOVEN - SYNTHETIC FIBRES 42.3 13.2

21 611120 - BABIES GARMENTS (INCL HOSIERY) AND CLOTHING ACCESSORIES - KNITTED - COTTON 38.8 18.0

610822 - WOMENS/GIRLS BRIEFS AND PANTIES - KNITTED - MAN-MADE FIBRES 37.5 49.6

610821 - WOMENS/GIRLS BRIEFS AND PANTIES - KNITTED - COTTON 36.3 45.4

24 611241 - WOMEN/GIRL SWIMWEAR - KNITTED - SYNTHETIC FIBRES 34.6 4.2

25 610832 - WOMENS/GIRLS PYJAMAS AND NIGHTDRESSES - KNITTED - MAN-MADE FIBRES 33.5 3.2

SUB-TOTAL 2996.4 1687.1

OTHERS 1387.6 1402.9

TOTAL (ALL PRODUCTS) 4384.0 3090.0

Source: Trade Data Online from Industry Canada (http://strategis.ic.gc.ca/sc_mrkti/tdst/engdoc/tr_homep.html)

Table 3: Comparison of top 25 export items to the U.S in 2000: Taiwan and South Korea (US$ million)

A. Taiwan

HS Code and Product Taiwan exports South Korea exports

847130 - PORTABLE COMPUTERS (WEIGHT 10-KG OR LESS) WITH AT LEAST A CPU, A KEYBOARD AND A DISPLAY 3796.1 320.7

847330 - PARTS AND ACCESSORIES OF AUTOMATIC DATA PROCESSING MACHINES (INCL COMPUTERS) AND UNITS THEREOF 3769.2 2824.3

854213 - MONOLITHIC DIGITAL INTEGRATED CIRCUITS - METAL OXIDE SEMICONDUCTORS (MOS TECHNOLOGY) 3343.0 6525.7

847160 - INPUT OR OUTPUT UNITS FOR COMPUTERS AND OTHER DATA PROCESSING MACHINES 1220.7 2236.1

854230 - MONOLITHIC INTEGRATED CIRCUITS NES (OTHER THAN DIGITAL) 1151.4 586.9

6 852510 - TRANSMISSION APPARATUS - FOR FAX, TELEVISION AND RADIO TRANSMITTERS 908.7 37.4

853400 - PRINTED CIRCUITS 895.9 226.9

8 847180 - OTHER UNITS (EXCL STORAGE AND INPUT/OUTPUT UNITS) OF AUTOMATIC DATA PROCESSING MACHINES 892.7 112.3

9 851750 - MODEMS AND OTHER APPARATUS FOR CARRIER-CURRENT LINE SYSTEMS OR FOR DIGITAL LINE SYSTEMS NES 614.4 43.0

10 850440 - ELECTRIC STATIC CONVERTERS (INCL POWER SUPPLIES, RECTIFIERS AND INVERTERS) 486.8 62.4

11 852390 - PREPARED UNRECORDED MEDIA FOR SOUND RECORDING OR OTHER PHENOMENA (EXCL CINEMATOGRAPHY) NES 447.7 6.7

611030 - SWEATERS, SWEATSHIRTS AND WAIST-COATS - KNITTED - MAN-MADE FIBRES 385.7 366.5

13 731815 - OTHER BOLTS OR SCREWS NES (WITH OR WITHOUT THEIR NUTS OR WASHERS) - IRON OR STEEL 318.4 20.6

14 846591 - SAWING MACHINES FOR WORKING WOOD, CORK, BONE. HARD RUBBER, HARD PLASTICS OR SIMILAR HARD MATERIALS 286.6 0.0

15 871200 - BICYCLES AND OTHER CYCLES 278.8 0.6

16 848180 - TAPS, COCKS, VALVES AND OTHER SIMILAR APPLIANCES, NES 277.7 74.4

17 950691 - ARTICLES AND EQUIPMENT FOR EXERCISE, GYMNASTICS AND ATHLETICS 271.0 4.1

18 731814 - THREADED SELF-TAPPING SCREWS - IRON OR STEEL 260.9 3.9

847170 - STORAGE UNITS FOR COMPUTERS AND OTHER DATA PROCESSING MACHINES 229.9 826.8

20 853669 - ELECTRICAL PLUGS AND SOCKETS - FOR VOLTAGE NOT EXCEEDING 1, 000 VOLTS 215.4 18.0

21 940320 - METAL FURNITURE NOT FOR OFFICE USE 206.2 14.5

22 731816 - NUTS - IRON OR STEEL 203.0 13.7

847150 - DIGITAL PROCESSING UNITS WITH OR WITHOUT REST OF SYSTEM 202.2 1245.3

24 940360 - WOODEN FURNITURE FOR OTHER USE 193.1 8.5

852540 - STILL IMAGE VIDEO CAMERAS AND OTHER VIDEO CAMERA RECORDERS 189.4 271.2

SUB-TOTAL 21045.0 15850.5

OTHERS 19469.2 24449.8

TOTAL (ALL PRODUCTS) 40514.2 40300.3

B. South Korea

HS Code and Product South Korea exports Taiwan exports

854213 - MONOLITHIC DIGITAL INTEGRATED CIRCUITS - METAL OXIDE SEMICONDUCTORS (MOS TECHNOLOGY) 6525.7 3343.0

2 870323 - MOTOR VEHICLES - SPARK IGNITION - CYLINDER CAPACITY 1500-3000 CC 4281.3 0.4

3 852520 - TRANSMISSION/RECEPTION APPARATUS - FOR CB/AMATEUR RADIOS, FAX, CELLULAR PHONES AND THE LIKE 2916.3 58.2

847330 - PARTS AND ACCESSORIES OF AUTOMATIC DATA PROCESSING MACHINES (INCL COMPUTERS) AND UNITS THEREOF 2824.3 3769.2

847160 - INPUT OR OUTPUT UNITS FOR COMPUTERS AND OTHER DATA PROCESSING MACHINES 2236.1 1220.7

847150 - DIGITAL PROCESSING UNITS WITH OR WITHOUT REST OF SYSTEM 1245.3 202.2

847170 - STORAGE UNITS FOR COMPUTERS AND OTHER DATA PROCESSING MACHINES 826.8 229.9

8 271000 - PREPARATIONS OF/NON-CRUDE PETROLEUM OILS AND OILS OBTAINED FROM BITUMINOUS MINERALS 599.9 1.1

854230 - MONOLITHIC INTEGRATED CIRCUITS NES (OTHER THAN DIGITAL) 586.9 1151.4

10 870322 - MOTOR VEHICLES - SPARK IGNITION - CYLINDER CAPACITY 1000-1500 CC 542.8 0.0

11 851650 - MICROWAVE OVENS - DOMESTIC 486.8 0.6

611030 - SWEATERS, SWEATSHIRTS AND WAIST-COATS - KNITTED - MAN-MADE FIBRES 366.5 385.7

847130 - PORTABLE COMPUTERS (WEIGHT 10-KG OR LESS) WITH AT LEAST A CPU, A KEYBOARD AND A DISPLAY 320.7 3796.1

852540 - STILL IMAGE VIDEO CAMERAS AND OTHER VIDEO CAMERA RECORDERS 271.2 189.4

15 841510 - AIR CONDITIONING MACHINES (AIR CONDITIONERS) - WINDOW OR WALL TYPES, SELF-CONTAINED 262.6 1.2

16 620530 - MENS/BOYS SHIRTS - WOVEN - MAN-MADE FIBRES 257.9 18.0

17 852190 - VIDEO RECORDING OR REPRODUCING APPARATUS - EXCLUDING MAGNETIC TAPE-TYPE 235.9 16.5

853400 - PRINTED CIRCUITS 226.9 895.9

19 852990 - PARTS (EXCLUDING AERIALS) FOR RADIO, TELEVISION, RADAR AND OTHER SIMILAR APPARATUS 187.4 123.0

20 401110 - NEW PNEUMATIC TIRES OF RUBBER - FOR USE ON MOTOR CARS (INCL STATION WAGONS AND RACING CARS) 181.1 43.6

21 852110 - VIDEO RECORDING OR REPRODUCING APPARATUS - MAGNETIC TAPE TYPE 172.5 6.3

22 870899 - OTHER MOTOR VEHICLE PARTS NES 148.2 150.7

23 481011 - PAPER (WOODFREE) - COATED - FOR WRITING/PRINTING - <150 G/M2 143.8 2.8

24 841430 - COMPRESSORS OF A KIND USED IN REFRIGERATING EQUIPMENT 139.0 17.2

25 847149 - OTHER COMPUTERS OR DIGITAL AUTOMATIC DATA PROCESSING UNITS PRESENTED IN THE FORM OF SYSTEMS 131.1 87.5

SUB-TOTAL 26117.3 15710.6

OTHERS 14183.1 24803.6

TOTAL (ALL PRODUCTS) 40300.3 40514.2

Source: Trade Data Online from Industry Canada (http://strategis.ic.gc.ca/sc_mrkti/tdst/engdoc/tr_homep.html)

Figure 1: Export concentration in selected countries

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

no. of 6-digit HS products

cumulative shares of exports to the U.S. in 2000

Honduras

Dominican Rep.

Bangladesh

Pakistan

Taiwan

S. Korea

Germany

KAKASHEV

15 Mart 2012 Perşembe

ECONOMIC DEVELOPMENT AS SELF-DISCOVERY

Hiç yorum yok:

Yorum Gönder

Popüler Yayınlar

Sayfalar