Trade liberalization, Openness and Growth - HEC Lausanne · Using data on factor endowments from UNCTAD, we can determine the « revealed » factor intensity of each product by taking

1

Trade liberalization, Openness and Growth

Table of Contents

1. The gains from trade ........................................................................................................................ 3

1.1 Comparative advantage in Ricardo’s model .............................................................................. 3

1.2 Comparative advantage with two factors: Heckscher-Ohlin ..................................................... 6

1.2.1 Concepts .............................................................................................................................. 6

1.2.2 Measurement ....................................................................................................................... 8

1.2.3 Do countries specialize or diversify ? ............................................................................... 10

1.3. Monopolistic competition and product variety ....................................................................... 13

1.3.1 Homogenous firms ............................................................................................................ 13

1.3.2 Heterogeneous firms ......................................................................................................... 15

2. The openess-growth debate between 1995 and now...................................................................... 16

2.1 Trade liberalization in a Solow model ..................................................................................... 16

2.2 The « East Asian Miracle » ...................................................................................................... 17

2.3 The cross-sectional evidence ................................................................................................... 19

2.3.1 Trade openess index of Sachs-Warner .............................................................................. 19

2.3.2 The critique of Rodriguez et Rodrik ................................................................................. 23

2.4 The evidence in panel .............................................................................................................. 25

3. Structural adjustment and resource allocation ............................................................................... 34

References .......................................................................................................................................... 36

List of tables

Table 1: Productivity and endowment data in a Ricardian example ................................................... 3 Table 2: candidate equilibrium for the integrated world economy ...................................................... 5

Table 3: Estimated TFPG in the East Asian Miracle ......................................................................... 18 Table 4: Estimation results of Young (1993) ..................................................................................... 19 Table 5: Growth and openness: Cross-section regression results 1980-90 ........................................ 21 Table 6: TFP growth regressions with various openness indices ...................................................... 22

Table 7: Regression results of Rodrik Rodriguez with the decomposed SW index .......................... 23 Table 8: Trade liberalization dates in Wacziarg and Welsh .............................................................. 25 Table 9: Details of trade reforms by country ..................................................................................... 26 Table 10: SW results reproduced ....................................................................................................... 27 Table 11: SW results on a different time period ................................................................................ 28

2

Table 12: Growth and openness in panel ........................................................................................... 29

Table 13: Growth regressed in the liberalization indicator ................................................................ 32 Table 14: Growth regressed on tariff changes ................................................................................... 33 Table 15: Approach 1 with instrumental variable.............................................................................. 34

List of figures

Figure 1: Production Possibility Frontier and autarky equilibrium : Portugal ..................................... 4 Figure 2: Production Possibility Frontier and autarky equilibrium : Great Britain ............................. 4 Figure 3: The gains from trade............................................................................................................. 5

Figure 4: The gains from specialisation ............................................................................................... 6 Figure 5_ Factor Endowment and PPF: The Rybczynski Theorem .................................................... 7 Figure 6: Factor Endowment and Comparative Advantage: The Heckscher-Ohlin Theorem ............. 7

Figure 7: Factor endowments and revealed factor intensities .............................................................. 9 Figure 8: Comparative advantage and the survival of exports .......................................................... 10 Figure 9: Fuel exports and GDP volatility ......................................................................................... 10

Figure 10: Export concentration and the level of income .................................................................. 11 Figure 11: Concentration: Individual country trajectories ................................................................. 11

Figure 12: The export re-concentration at high levels of income ...................................................... 12 Figure 13: What are the closing export lines? ................................................................................... 12 Figure 14: Effect of an increase in the custom duty rate on capital equipment in the Solow model . 17

Figure 15: Average growthe for closed and open economies ............................................................ 20 Figure 16: convergence among closed economies............................................................................. 20

Figure 17: Convergence among open economies .............................................................................. 21 Figure 18: Growth and import tariffs ................................................................................................. 24

Figure 19: Growth and NTB coverage ratios ..................................................................................... 24 Figure 20: Time profile of growth around liberalization year ........................................................... 30

Figure 21: Time profile of investment around the liberalization year ............................................... 30 Figure 22: Decomposition of productivity growth : within-sector vs. Structural adjustment ........... 35 Figure 23: Correlation between productivity and variation in employment per sector ..................... 35

3

1. The gains from trade

Traditional trade theory suggests that:

o Trade among countries generates efficiency gains for all countries, irrespective of their level of

productivity; in other words, all countries, in general gain from trade. Absolute productivity

determines income levels, not trade patterns.

o Countries specialize according to their comparative advantage, generating efficiency.

1.1 Comparative advantage in Ricardo’s model

In the simplest version of the Ricardian model (not to be confused with the Ricardo-Viner model as

we will see later, and which has nothing to do!), the assumptions are

• Two countries (Portugal et GB)

• Two sectors (wine and drape)

• Only one factor of production (labor), perfectly mobile between the two sectors

• Constant returns to scale

• No transport costs

• No government intervention

• Perfect competition (price = cost)

• At the same price, consumers share their budget equally between wine and drape

Comparative advantage is determined by the relative productivity of labor (the only production

factor) in the two sectors, as illustrated in Table 1.

Table 1: Productivity and endowment data in a Ricardian example

Productivity Endowments

(labor)

Wine Drape

Portugal 8 4 5

UK 1 2 20

Note that the example’s assumptions imply that Portugal is everywhere more productive than the

UK; however, it is relatively more productive in wine. The maximum combination of goods that

Portugal can produce in autarky (i.e. without international trade) is shown by its production

possibility frontier, illustrated in Figure 1. The same thing for the UK is shown in Figure 2.

4

Figure 1: Production Possibility Frontier and autarky equilibrium: Portugal

Figure 2: Production Possibility Frontier and autarky equilibrium: Great Britain

In order to illustrate the gains from trade, we « guess » an integrated world equilibrium with free

trade and show that (i) it is a viable equilibrium in the sense that both countries have balanced trade

and supply equals demand in each industry at the global level, and (ii) every country is better off

than under autarky. The equilibrium is characterized in Table 2.

In that equilibrium, Portugal specializes completely in wine, the UK in drape, the relative price of

the two products is just one (meaning that one liter of wine trades for one meter of drape) and each

country splits its consumption half-half between wine and drape. Thus Portugal imports 20 units of

drape and export 20 units of wine, while the UK does the reverse.

Wine

Drape

40

20

PPF (Production Possibility Frontier)

Point of Consumption in Autarky

Indifference Curve

Wine

Drape

20

40

PPF

Point of Consumption in Autarky

Indifference Curve

5

Table 2: candidate equilibrium for the integrated world economy

The fact that Portugal can consume 20 units of drape plus some wine shows that it gains from trade,

since under autarky if it wanted to consume 20 units of drape it would have had to put all its

manpower in that sector, leaving no wine to be produced. Conversely for the UK. So everyone is

better off as a result of specialization.

More generally, gains from trade arise whenever international relative prices, shown by the

(absolute value of the) slope of the blue line in Figure 3, differ from domestic relative prices, shown

by the absolute value of the slope of the PPF.

Figure 3: The gains from trade

While Figure 3 shows the gains from pure trade, additional gains can be generated by specializing

the production structure in order to take advantage of comparative advantage, as shown in Figure 4.

The process of moving from the autarky equilibrium to the specialized equilibrium where the

utilization of resources generates the maximum efficiency gains at world price is called “structural

adjustment” (SA). It is the essence of the policies pushed by the IMF and the World Bank in Latin

American and African countries in the 1980s and 1990s. What the SA reforms entailed is the

closure of industries geared towards import substitution in order to free up resources to be invested

in export-oriented industries. While the principle was sound, as we will discuss later on in this

Wine Drape Wine Drape

Portugal 40 0 20 20

UK 0 40 20 20

Total 40 40 40 40

Production Consumption

World Price Line

Wine

PPF

Drape

Indifference Curves

export

import

6

chapter numerous difficulties stood in the way of demonstrating that this would generate a growth

dividend, and even today there is controversy about whether SA really pushed resources (in

particular employment) toward more “efficient” sectors.

Figure 4: The gains from specialization

1.2 Comparative advantage with two factors: Heckscher-Ohlin

While the Ricardian model is a classical model where labor is the only productive factor, the same

story can be told in a so-called “neoclassical” model where technology combines labor and capital

(or unskilled and skilled labor, or any combination of factors for that matter) to produce output.

1.2.1 Concepts

The first step of the analysis here consists of relating factor endowments with the PPF (the PPF

being drawn in product space, not in factor-endowment space); this is done by the Rybczynski

theorem, which says that if there is an inflow of a given factor of production, the PPF “inflates”

asymmetrically in the direction of the sector that uses that factor most intensively. Quite a

mouthful. What it means is that if there is an inflow of capital and steelmaking is capital-intensive,

the PPF will inflate more in the direction of steelmaking than in the direction of, say, textiles, which

are not capital intensive. This is shown in Figure 5.

Indifference Curves

Wine

PPF

Drape

Lines of World prices

Point of Production

7

Figure 5_ Factor Endowment and PPF: The Rybczynski Theorem

From then on the story is pretty much the same. If a country has lots of capital (because previous

generations saved a lot), it will have a PPF looking like the blue one in Figure 5 and will have a

comparative advantage in capital-intensive industries. Moving from autarky to free-trade as shown

by the flue arrow in Figure 6, it will specialize in those industries and, in so doing, reap efficiency

gains.

Figure 6: Factor Endowment and Comparative Advantage: The Heckscher-Ohlin Theorem

Note that this is frankly a bit of a fairy tale. We are assuming here that for every job destroyed in a

comparative-disadvantage sector by SA, a new job will appear all of a sudden by some miracle in a

comparative-advantage sector. This is because the HO model is, by construction, a full-employment

model. Outside of this fairy tale, in the real world, there are numerous barriers to the movement of

factors and things just don’t that way. People who lose their jobs are typically located in declining

PPF initial

Steel

The impact of foreign investments

The impact of immigration

Textile

“trade triangle”

Steel

Textile

Indifference curves

Relative Price on world market (textile less expensive)

Point of consumption after structural adjustment

Autarky relative price

Point of production after structural adjustment

PPF

8

regions where “sunrise” industries don’t locate. Many of them have firm-specific or no skills,

making them poor hires for the sunrise industries. And so on.

1.2.2 Measurement

When there are more goods than factors of production, in general the direction of trade (who

exports what) is not determined.1 However, overall, countries tend to export goods whose factor

intensity corresponds more or less to their factor endowments.

The traditional measure of comparative advantage is the Balassa index of revealed comparative

advantage, calculated as follows: let inx be country i’s exports of product n, ix the total exports of

country i, nx world exports of product n, and x world exports. The Balassa index is then:

/

/

in iin

n

x x

x x

The problem with this index is that it assumes that if country i exports product n, it has a

comparative advantage in this product; but the index does not use the factor endowment of country

i.

Here is another approach. Using data on factor endowments from UNCTAD, we can determine the

« revealed » factor intensity of each product by taking the average endowments of the countries that

are exporting it. If i is the endowment of capital of country i, the capital intensity of the good n is

approximated by a weighted average of the capital endowments of the countries exporting it:

n in ii

where in , the weights, are modified versions of the Balassa index made to add up to one, i.e.

1ini .2 The advantage of this normalization is that it allows us to represent national factor

endowments and the products’ revealed intensities in the same space, i.e. to superimpose them on

one picture. If countries export products whose intensities correspond more or less to their factor

endowment, the cloud of points representing their export portfolio should be relatively concentrated

around their endowments. That’s just what we observe in Figure 7, where the horizontal axis

measures capital intensity/endowments (in dollars of capital per worker), the vertical axis measures

human-capital intensity/endowments (in years of education per worker), and each circle

corresponds to a product, with circle size proportional to the product’s dollar export value for the

country in question. The black cross is the country’s endowment. In the case of Costa Rica in 1993,

the capital and HK’s intensity of its biggest export product (bananas) corresponded to a factor

endowment slightly lower in terms of capital and human capital than its own; in other words, it was

slightly under-selling itself. The reverse was clearly true for Pakistan in 2003-5.

1 Can be found in trade flows with a continuum of goods as in Dornbusch Fisher Samuelson model. 2 Caution: If some countries subsidize the exports of products that do not correspond to their comparative advantage,

the computation is distorted (e.g. agricultural products to Europe). We must therefore correct for this bias in the

computations.

9

Figure 7: Factor endowments and revealed factor intensities

Costa Rica 1993 Pakistan 2003-5

Figure 7 (ctd)

Tunisie 2003-5 Tunisie : new export products

Interestingly, exported products that are distant from a country’s factor endowment (hence do not

correspond to a clear comparative advantage) seem to survive less on export markets, as shown by

the slightly negative correlation visible in Figure 8.

02

46

810

12

Re

vea

led

Hu

man

Cap

ital I

nte

nsity

Inde

x

0 50000 100000 150000 200000Revealed Physical Capital Intensity Index

Endowment point

02

46

810

12

Re

vea

led H

um

an

Capita

l In

tensi

ty In

de

x

0 50000 100000 150000 200000Revealed Physical Capital Intensity Index

02

46

810

12

Rev

eale

d H

uman

Cap

ital I

nten

sity

Inde

x

0 50000 100000 150000 200000Revealed Physical Capital Intensity Index

02

46

810

12

Re

vea

led

Hu

man

Cap

ital I

nte

nsity

Inde

x

0 50000 100000 150000 200000Revealed Physical Capital Intensity Index

10

Figure 8: Comparative advantage and the survival of exports

1.2.3 Do countries specialize or diversify?

The Ricardian model suggests that countries should specialize in their comparative advantage rather

than diversify. But specialization in raw materials, for example, can be synonymous of "imported

volatility", because the prices of raw materials fluctuate widely. This is shown in Figure 9 where the

horizontal axis measures the share of hydrocarbons in country exports and the vertical axis

measures the coefficient of variation (standard error/mean) of GDP over a seven-year period. The

positive correlation between the two is pretty obvious.

Figure 9: Fuel exports and GDP volatility

23

45

67

2 2.2 2.4 2.6 2.8 3(mean) std_dist_1

(mean) length Fitted values

Length of trade relationship and distance to CA

GABGABGABGABGABGABGABGAB

0.2

.4.6

.81

Coef

fici

ent

of

var

iati

on o

f G

DP

, 2000-2

007

0 20 40 60 80 100Fuel share in exports

11

Recent studies also suggest that the decline in the volatility of GDP observed in recent decades in

the United States is largely linked to the diversification of the economy (in services).

Interestingly, generally the concentration of exports follows a non-monotone path as countries

develop: first diversification, then reconcentration. We measure the concentration of exports in

similar way as we measure the concentration in income, by three indices: (i) Gini, (ii) Herfindahl,

and (iii) Theil. Here we considered the index of Theil, whose formula is:

1

lni i

i

x xT

n x x

(1)

Figure 10: Export concentration and the level of income

And the reconcentration occurred in the individual trajectories of countries:

Figure 11: Concentration: Individual country trajectories

24

68

0 20000 40000 60000 80000GDP per cap, 2005 PPP dollars

Theil index Theil index, Uganda

Fitted values

More concentrated than

predicted

Less

concentrated

than predicted

Uganda

2000 Predicted

Uganda

2010

IRL

ESP

GRC

GBR23

45

15000 20000 25000 30000 35000 40000GDP per capita, PPP (constant 2005 international $)

12

Which countries are those that re-concentrate?

Figure 12: The export re-concentration at high levels of income

How can we explain the reconcentration? Essentially the inertia of trade flows, export lines that are

closed have factor intensities corresponding to weaker endowments than those of countries that

close. For example, the average of trade lines closed by the EU corresponds to the combined

endowment of human and physical capital of Indonesia. These lines should be long gone, but they

remain open by inertia.

Figure 13: What are the closing export lines?

In short, the theory seems to stick quite well with empirical observation, although the "content

factors" does appear to explain only a small part of international trade. We therefore need other

models to have a more complete view of its determinants.

34

56

7

The

il in

de

x

0

100

02

00

03

00

04

00

05

00

0

num

ber

of e

xport

ed p

rod

ucts

0 20000 40000 60000GDP per capita PPP (constant 2005 international $)

Active lines - quadratic Active lines - non parametric

Theil index - non parametric Theil index - quadratic

# active export lines

Theil index

24

68

10

12

0 50000 100000 150000 200000

Product intensities Country endowments

Capital

Human capital

13

On the other hand, so far everything discussed was essentially static: allocative efficiency

considerations tell us nothing about the growth. In the models of endogenous growth, growth is

mainly due to innovation; international trade plays only an indirect role (i.e. through innovation).

So we will discuss the relationship between trade and growth from an essentially empirical point of

view, except a small detour to the Solow model in Section 3

1.3. Monopolistic competition and product variety

1.3.1 Homogenous firms

The monopolistic competition model The Heckscher-Ohlin model explains trade by differences in factor endowments. It cannot explain

the trade between countries with similar endowments, and even less intra-industrial trade. We will

now focus on an alternative model proposed by Krugman (1980), called « monopolistic

competition».

The ingredients for a model of monopolistic competition are:

o Product differentiation that generates a finite elasticity for each firm

o Economies of scale

The gains of trade in the MC model come from competition, which compresses margins and prices.

Consider the following example from Krugman, Obstfeld and Mélitz (2012), pp 168-177

Let S be the volume of national trade, which we take as exogenously given (independent of the

prices of the firms active in this market) which is of course unrealistic but simplifies the analysis

greatly. Let n be the number of firms active in the market, b a parameter of demand (linear), Qi the

quantity sold per company i, pi its price and p the average price in the market.

Total cost is the sum of a fixed cost F and a marginal cost c:

i iC F cQ

which gives average cost of :

i

i

FAC c

Q (2)

Demand function facing firm i:

1

i iQ S b p pn

(3)

In a «symmetric equilibrium» where all firms set the same price ip p , it can be seen from (3)

that /iQ S n ; market shares are equal.

Optimal pricing by profit-maximizing firms equalizes marginal cost and marginal revenue. To

derive marginal revenue, invert (3) to get the demand price:

1 i

i

Qp p

bn bS (4)

14

Revenue is price multiplied by quantity

2

i ii i i

Q Qp Q pQ

bn bS (5)

and marginal revenue is the derivative of revenue w.r.t. quantity:

1 2

1

i

ii

i i

p

ii

QRM p

bn bS

Q Qp

bn bS bS

Qp

bS

(6)

Marginal cost is simply c. Optimal pricing is therefore

ii

Qp c

bS (7)

Or

"mark-up"

ii

Qp c

bS (8)

In the symmetric equilibrium where all firms adopt the same price, Q = S/n, optimal pricing

simplifies to

1

ip p c ibn

(9)

In this equilibrium, average cost is found by substituting Q = S/n in (2), which yields

Fn

AC cS

(10)

With free entry, profits must be zero, which means that price has to equal average cost:

1 Fn

c cbn S

(11)

or

2 Sn

bF (12)

which determines the number of firms compatible with zero profits in the market (no incentive for

additional entry). In this model, gains from trade arise because of

o Economic integration that creates a bigger market

o Increasing competition, reducing margins

15

This can be seen by « merging » two countries with equal size S as part of a big-bang trade-

liberalization experiment.

Effects of trade liberalization

We do the comparison that is commonly done in international trade between an equilibrium in

autarky and an equilibrium with free trade where all barriers are eliminated. The effect is illustrated

in a numerical example in the excel file ‘Exemple concurrence monopolistique.xlsx.’ Suppose that

the two countries are of equal size and that there is no transportation cost. Then their combined size

is ' 2S S , so the total number of firms is

' 2

' 2 1.414S S S

n nbF bF bF

and the equilibrium price and quantity are

1

''

p c pbn

' 2

' 2 2' 2

S S SQ Q Q

n nn

So :

o The total number of varieties available to any consumer in the two-country area increases (there

are fewer in each country but consumers have access to both)

o The equilibrium price is lower, and so are profit margins (not shown but easy to calculate)

o Output per firm increases.

Damn it, everything is fine in this world?

1.3.2 Heterogeneous firms

Note that the trade liberalization induces firm exit, since ' 2 2n n n . In a symmetric

equilibrium, which firms will exit is indeterminate. But suppose now that potential entrants differ in

their marginal cost ci, in accordance with new “heterogeneous-firm” models. Those with marginal

cost higher than the « choke price » don’t enter.

Intercept of demand facing each firm: Using (3), Qi = 0 implies

choke 1p p

bn

And the slope of the demand curve is:

1i

i

dp

dQ bS

16

Trade liberalization means that chokep goes down as n goes up, while the slope becomes “less

negative” as S goes up. Thus, under the effect of an increase in S and the induced increase of n, the

demand curve rotates anticlockwise.

o The demand increases for big firms with low marginal cost

o But it decreases for firms with higher marginal cost, which leads to the exit of some firms.

2. The openness-growth debate between 1995 and now

2.1 Trade liberalization in a Solow model

Everything that we saw at the beginning of this chapter was static. Is there any reason to think that

trade liberalization could accelerate the growth? Yes if trade liberalization affects the price of

capital goods, for example. To see this, we take the Solow model and assume that the domestic

price of capital goods (the capital) is:

* 1K K Kp p t

where *

kp is the world price of one unit of capital (one « machine ») and kt is the customs duty on

imported capital. Assume, to simplify, that * 1kp , that is if we measure the capital in dollars, then

one unit is worth a dollar. Rewrite the law of motion of capital as :

1K K

I IK K K

p t

Then we have :

1 ˆ

1 ˆ1

1.

1

K

K

K

IK

tdkk kL

dt L

I KkL

t L L

sk n kt

A high customs duty therefore lowers the curve in Figure 14; the steady state (the intersection of the

curves, that are respectively representing the first term on the right of the equation above, and the

second term) moves to the left (at a lower level of capital per worker) and the rate of growth during

the transition to steady state, slows.

17

Figure 14: Effect of an increase in the custom duty rate on capital equipment in the Solow model

In the particular case of capital goods, the link between trade liberalization is therefore direct (and

obvious). This explains the 'climbing' structures of tariffs, prevailing in most developing countries:

low or zero tariff on capital goods, moderate tariff rates on intermediate products used as inputs in

the industry, and the highest rates on consumer goods.

2.2 The « East Asian Miracle »

Empirically, what can we say about the relationship between trade and growth? The « East Asian

miracle » is the title of a World Bank report published in 1994 and dedicated to the spectacular

growth of the Asian tigers (compared to other continents, in particular Africa and Latin America).

This report had considerable visibility although it was highly controversial.

The approach was a “growth accounting” one based on a Cobb-Douglass production function:

ln

it it it it

it it

Y K L H

y Y

(0.13)

Log-linearizing gives an estimable growth equation

it it it it ity k h u (14)

18

where ui is the error term. Let ei be the residual of the estimation in (14), i.e.

ˆˆ ˆit it it it ite y k h (0.15)

We will give a name to this residual: TFP (Total Factor Productivity). Taking first differences (i.e.

growth rates, since everything is in logs) we define Total Factor Productivity Growth (TFPG) as

1it it it ite e e TFPG (0.16)

This gives us a decomposition of sources of growth in two components:

o Accumulation, i.e. what is predicted by (14),

o TFPG or improved Efficiency (the residual)

This decomposition is very important. If accumulation (especially capital) is the dominant

contribution to growth, the recipe for economic policy is the "mobilization of savings" for

investment. This can be done - and has been historically - abruptly by taxing agriculture to generate

the resources needed for investment. Extreme cases: the Soviet Union under Stalin. It is also what

inspired many economic policies in Africa.

On the other hand, if the TFPG is dominant, then is something else. The problem is that as the

TFPG is a residual, by definition it is not known what it is, and we could put what we want as

interpretation.

Table 3: Estimated TFPG in the East Asian Miracle

There is clearly a difference in the nature of growth between the SE Asia and the remainder (AL

and ASS). The TFPG is dominant in Asia, not elsewhere. Explanation: trade openness that forces

local businesses to restructure and improve the efficiency.

Unfortunately, the same year when the preliminary draft of the « East Asian Miracle » circulated,

Alwyn Young published a paper that showed the labor factor was improperly measured

(underestimated) for SEA (South East Asia) countries in the report of the Bank, the residual

measured correctly was a bit smaller for these countries. Even more of a miracle!

Average TFP 1970-90 (% per year)

Taiwan 3.76 Hong-Kong 3.64 Korea 3.10 Japan 3.48 Thailand 2.49 Singapore 1.19 Malaysia 1.07

Latin Am. 0.13 Afr. sub-sah. -0.99

19

Table 4: Estimation results of Young (1993)

Source : Young, 1993

2.3 The cross-sectional evidence

2.3.1 Trade openness index of Sachs-Warner

Idea: correlate the growth in the period 1980-90 with a measure of trade openness. Binary measure:

either open or closed country. « Closed » if one or more of the following criteria are satisfied:

1. Average tarif greater than 40%

2. Rate of coverage of non-tariff barriers (quotas etc.) greater than 40% of imports

3. Black market currency premium greater than 20% during the decade

4. Export State Monopoly

5. Socialist Economy

SW found a strong correlation between growth and their measurement of the opening. Already in

descriptive statistics, the difference is clear:

20

Figure 15: Average growthe for closed and open economies

Source: Sachs and Warner (1995)

In addition there is convergence among the open countries but not closed countries:

Figure 16: convergence among closed economies

Source: Sachs and Warner (1995)

21

Figure 17: Convergence among open economies

Source: Sachs and Warner (1995)

The cross-section regression results confirm the descriptive statistics (table 5).

Table 5: Growth and openness: Cross-section regression results 1980-90

Source: Sachs and Warner (1995)

22

Edwards (1998) attempted to show that the results of Sachs and Warner were robust and not the

effect of a particular approach. It includes all of the openness measures (Sachs and Warner and

other) and systematically explores the correlation between these measures and the TFPG.

OPEN Sachs-Warner

WDR Openness Index of the World Bank (composite)

LEAMER Residual of an equation of openness

BLACK Black market premium on currencies

TARIFF Average import tarif

QR Rate of coverage of quantitative trade barriers

HERITAGE Trade-distortion perception index

CTR Revenue on import taxes in proportion of the value of imports

WOLFF Another residual of a regression of openness

SW results are robust; several other similar exercises give the same results

Table 6: TFP growth regressions with various openness indices

Source: Edwards 1998.

23

Basically, the message is that regardless of the measure of openness that we take, the correlation

with the TFPG seems well-established. The message of the East Asian Miracle was fundamentally

correct even if the measures are different good this is.

2.3.2 The critique of Rodriguez and Rodrik

Rodriguez and Rodrick (2001) show the opposite. They do an exercise of brutal deconstruction of

all this econometrics, in particular of the econometrics of Sachs Warner.

1 si tariffs < 40% & NTB < 40% et pas SOC

0 sinonSQT

1 si BMP < 20% & pas de MON

0 sinonBM

Table 7: Regression results of Rodrik Rodriguez with the decomposed SW index

24

Figure 18: Growth and import tariffs

Source: Rodriguez and Rodrik (2001)

Figure 19: Growth and NTB coverage ratios

Source: Rodriguez and Rodrik (2001)

So what explains the differences of TFPG, is not so much trade policy stricto sensu, but rather

macroeconomic policy (the overvaluation of the exchange rate measured by the premium on

currencies) and export monopolies. But, what country had exchange rates overvalued in the 1980s?

Latin America. Which country had export monopolies? Africa.

25

2.4 The evidence in panel

All first generation studies were cross-sectional. Wacziarg and Welsh (2008) remake the estimates

in panel data by carefully identifying the date of trade liberalization (while SW did not date, since

they were using a cross-sectional over a decade). Employing a panel allows to use the fixed-effects

estimator (dummy variables that capture country-invariant characteristics over time). The effect is

much better identified, as it is "within-country" that is to say, it filters the heterogeneity between

countries due to unrelated trade openness factors.

Table 8: Trade liberalization dates in Wacziarg and Welsh

26

Table 9: Details of trade reforms by country

27

Results: Reproducing exactly the exercise of SW, they find the same findings:

Table 10: SW results reproduced

28

On the other hand, when running the same regressions on another time period (the 90s), nothing

stays significant:

Table 11: SW results on a different time period

What to make of it? The answer comes with panel regressions where the fundamental explanatory

variable is the date of trade liberalization; the date of the liberalization contains additional

information that is not distorted by other unexplained differences between countries (since we use

the difference in time for each country).

29

Table 12: Growth and openness in panel

Once doing this, the results become correct for all periods – far more convincing. Figure 20

displays the results in a more intuitive way. Time is normalized to be zero in the year of trade

liberalization for each country (so if Colombia liberalizes in 1995, 1994 = -1, 1995 = 0, 1996 =

1 for Columbia; if Chile liberalizes in 1970, 1969 = -1 etc.). Each point on the curve is the average

of the sample growth at t = -10, t = -9, etc. We observe an acceleration of growth of about 1.5

percentage points around the year zero.

30

Figure 20: Time profile of growth around liberalization year

Figure 21 shows the same finding for investment. We observe a spectacular rise in the rate of

investment after the liberalization.

Figure 21: Time profile of investment around the liberalization year

On the other hand, the identification problem remains still unsolved in Wacziarg and Welsh due to

the fact that the trade reforms were often implemented at the same time as reform packages that

affected several other sectors of the economy (macroeconomic stabilizations, privatizations,

31

governance reforms, etc.) and oftentimes also coincide with changes in government. So: is it really

the trade liberalization causing the effects or other simultaneous developments?

Estevadeordal and Taylor (2009) revisited the question in the different ways where the second one

is interesting in itself to understand for the used methodology.

Approach 1 (« simple differences »)

They regress the change in growth on the change in tariffs in a panel of countries—a standard

technique. With i representing a country, t the time, itg the growth of country i at time t,

,it it ith zx a country-specific vector (human capital and characteristics of governance), and it the

average of tariffs of country i in time t.

, 1it it i tg g g (17)

And the same for other variables put in differences. The equation becomes

0 1 , 1 2 3 ln 1it i t it it itg g u x α (18)

Approach 2 (« differences in differences »)

E&T use as natural experiment the liberalization implemented by a number of countries during

trade negotiations in Uruguay (the « Uruguay Round » that took place between 1986 and 1994).

Certain countries liberalized their tariffs; they form the « treatment group », other countries that

didn’t liberalize are put in the « control group ». Again, the sample structure is a panel, but now the

estimation technique is called « differences in differences ». This term expresses that we compare

the performance before and after a certain date where the treatment starts (the first difference), but

for two groups, the treatment and the control group (second difference). This estimation technique

is commonly used in medical sciences.

With iD being a dummy variable marking belonging to the treatment group and tT the treatment

period (after the Uruguay round); so

1 if 1994

0 if < 1994t

tT

t

The basic equation becomes

0 1 2 3 0 4

_

it i t i t i it

Treatment effect

g D T D T u x α (19)

And the coefficient 3 gives the treatment effect. We can also re-write (19) in a simpler way with

fixed effects for countries and years:

Treatment effect

it i t i t itg D T u (20)

32

Finally a third way of writing and estimating this equation consists of defining two long periods (

0 1975 1989t and 1 1990 2004t ), which gives us a two-period panel, and taking the change

between those two periods :

1 0, ,i i t i tg g g (21)

Which yields

00 1 , 2 3i i t i i ig g D u x α (22)

The basic results of the Diff-in-Diff approach (DD) are displayed in Table 13. The first column uses

the average of tariffs for all goods as regressor of interest (« liberalizer indicator »); the second

column uses the average of tariffs only on consumption goods, the third uses the average on tariffs

on equipment goods and the fourth uses the average tariff on intermediate goods. We find that the

coefficients are significant and estimated more precisely for the equipment goods than for

consumption goods. However, the effects are rather weak.

Table 13: Growth regressed in the liberalization indicator

33

The results of the first approach are very similar, but with the opposite sign since lower tariffs

accelerate growth:

Table 14: Growth regressed on tariff changes

Again, the coefficients are very small (signaling a very weak effect) and only significant at the 10%

level (signaling that the effect is not well measured). On the other hand, the coefficient on the tariffs

for equipment good is two times higher than the coefficient for consumption goods, which is in line

with the basic growth model of section 2.1.

Endogeneity

The two approaches face similar problems, firstly endogeneity and secondly selection. They only

handle the first problem, where the problem is that the variation in tariffs and the variation in

growth could be explained by the same omitted variable, for example a change in government.3

The instrumental variable is the interaction between two things:

1. The intensity of the Great Depression in the observed countries

2. The level of tariffs in the countries before the Uruguay round.

The idea of the first element is that countries that suffered more in the Depression have more than

others lost the faith in liberalism and adopted more protectionist policies afterwards, which could

3 In the second approach, we face a selection problem. The approach relies on the hypothesis that the decision to take

the treatment is uncorrelated with the potential effect of the treatment. In fact, if the countries that liberalized were

systematically more likely to benefit from the treatment, we cannot use the equation (19) to deduct that the same effect

would have worked in the countries that were not treated. Therefore we have to control for this selection effect that is

always present when the treatment is not given at random, but this control is not done here.

34

have survived until today and resulted in less willingness for a liberalization. The idea of the second

element is that for liberalizing, countries must have entered the Uruguay round with high tariffs

(otherwise no need for liberalization). So, low intensity of the Great Depression × high level of

tariffs predict a strong liberalization in the Uruguay round.

Table 15: Approach 1 with instrumental variable

We note that the effect is now stronger and more significant (it’s the « second stage » that we care

about; we have -0.05 now vs. -0.03 before, and the effect is significant at 5%).

3. Structural adjustment and resource allocation

Is this the end of the debate? Not yet. In a recent paper, McMillan and Rodrik (2010) have

decomposed the growth of productivity and shown a result opposing the message of the beginning

of the course:

o The productivity growth in a sector is comparable across countries ; particularly there is no

substantial difference between Africa and America as before

o In contrast, in favor of a structural adjustment: in these two regions resources have moved from

sectors with high productivity growth towards sectors with low productivity growth.

Supposing that the productivity of the manufacturing sector was a weighted average of the

productivity of several sectors.

t j jj

q q

with 1jj . We can express its variation, 1t t tq q q as

Croissance "within" Structural adjustment small--we ignore it!

j j j j j jj j jq q q q

35

Representing the first term in grey and the second in black in averages per region, McMillan and

Rodrik (2011) obtain in Figure 22:

Figure 22: Decomposition of productivity growth: within-sector vs. Structural adjustment

Source : McMillan and Rodrik (2011).

The grey component doesn’t really vary from one region to the other. However, the black part

really makes a difference. The structural adjustment has moved resources to the wrong place! The

case of Argentina is particularly interesting (Figure 23).

Figure 23: Correlation between productivity and variation in employment per sector

Source: McMillan and Rodrik (2011).

36

References

Edwards, Sebastian (1998); “Openness, Productivity and Growth: What Do We Really Know?”;

Economic Journal 108, 383-98.

Estevadeordal, Antoni, and Alan Taylor (2009), “Is the Washington Consensus Dead? Growth,

Openness, and the Great Liberalization, 1970s-2000s”; IDB working paper IDB-WP-I38;

Washington, DC: Inter-American Development Bank.

McMillan, Margaret S. and Dani Rodrik, “Globalization, Structural Change and Productivity

Growth,” Working Paper No. 17143, NBER (http://www.nber.org/papers/w17143), June 2011.

Rodrik, Dani, and F. Rodriguez (2001), “Trade Policy and Economic Growth: A Skeptic's Guide to

the Cross-National Evidence”; in Ben S. Bernanke and Kenneth Rogoff, editors, NBER

Macroeconomics Annual 2000, Volume 15, p. 261 – 338; Boston, MA: National Bureau of

Economic Research.

Sachs, Jeffrey, and Andrew Warner (1995), “Economic Reform and the Process of Global

Integration”; Brookings Papers on Economic Activity 26, 1-118.

Wacziarg, Romain, and K. Welch (2008), “Trade Liberalization and Growth: New Evidence”;

World Bank Economic Review 22, 187-231.

The World Bank (1993), The East Asian miracle : economic growth and public policy; Washington,

DC: The World Bank.