PRELIMINARY DRAFT: Please do not cite.

Revealed Comparative Advantage and Extensive Margins of

Heterogeneous Firms

Steven Husted٭ Isao Kamata

† Shuichiro Nishioka

‡

University of Pittsburgh University of Wisconsin West Virginia University

This version: May 4, 2013

Abstract

Doing business in foreign countries requires the significant amount of fixed and variable costs.

Productive firms that can cover these costs have opportunities in accessing foreign markets.

The number of exporters depends not only on the distribution of firms in that industry but also on

the range of productivity for successful exporters. This paper examines the productivity range of

successful exporters across countries and industries, and finds that the ranges differ

systematically according to revealed comparative advantage. Firms in the comparative advantage

sectors are likely to be successful in foreign markets. This tendency is stronger for a country

pairs involving one developed country and one developing country.

Keywords: Revealed comparative advantages; Product-level gravity model; Extensive margins of

heterogeneous firms

JEL Classification: F14

٭ Department of Economics, 4508 WW Posvar Hall, University of Pittsburgh, Pittsburgh, PA 15216

† Robert M. La Follette School of Public Affairs, 1225 Observatory Drive, University of Wisconsin–Madison,

Madison, WI 53706 ‡ Department of Economics, 1601 University Avenue, West Virginia University, Morgantown, WV 26506

1 Introduction

The idea that only the productive subset of firms are able to export becomes a standard approach

of International Trade (Melitz, 2003). Doing business in foreign countries requires the significant

amount of fixed and variable costs, which distinguish productive exporters from unproductive do-

mestic firms. Thus, international trade plays a crucial role in explaining the average productivity of

countries and industries since trade liberalization and growth of exporters move domestic resources

to productive exporters.

While the literature initiated by Bernard and Jensen (1999) found evidence that exporters

are superior in many productive characteristics than domestic firms in a single nation, there is

no systematic empirical evidence that examines how the ranges of exporters’productivities differ

across countries and industries. The productivity range in this paper is the log difference between

the productivity level of the leading firm and that of the least productive exporter. Theoretically,

Bernard, Redding, and Schott (2006) examined how the ranges of exporters’productivities differ

across a comparative advantage sector and a comparative disadvantage sector. They showed that

firms in the comparative advantage sector have better chances to be exporters. In other words, even

relatively unproductive firms can have better chances to be exporters in the comparative advantage

sector. This paper empirically examines the global variation of productivity ranges across countries

and industries for each product group. We will examine if we can find any systematic associations

between the productivity ranges and the comparative advantages. In particular, we first estimate

the range of productivity for each industry using the binary gravity equation from Helpman, Melitz,

and Rubinstein (2008, hereafter HMR). Then, we obtain the productivity range for each of bilateral

pairs of countries for each product from the estimation results. We are particularly interested in

investigating how the productivity ranges differ across comparative advantage and comparative

disadvantage sectors. For this purpose, we derive the measure of relative revealed comparative

advantage from Balassa (1965).

We find that the measures of relative revealed comparative advantage associate positively with

the corresponding productivity ranges across industries for each pair of an exporter country and

an importer country. The correlation is stronger for the pairs that involve one developed and one

developing countries. The results in this paper indicate a significant role played by comparative

advantage for foreign market entry. In addition, the new concept of productivity range that derived

1

based on the firm-heterogeneity model (Melitz, 2003) depends crucially on the traditional sense of

revealed comparative advantage (Balassa, 1965).

2 Productivity Range in Firm-Heterogeneity Model

2.1 Firm-Level Decision to Enter Global Markets

The empirical estimation we will examine in this paper is the product-level estimation of Helpman,

Melitz, and Rubinstein (2008). The gravity equation of HMR is useful for our purpose since it can

study firm-level decision on market entry without using firm-level data.

The representative consumer in country l determines the demand for each variety ω in prod-

uct i subject to the optimal expenditure (Y lit). The product-level utility function is a standard

CES (Constant Elasticity of Substitution) form: ulit =[∫ω∈Bli

[qli(ω)

]αi dω]1/αi where qlit(ω) is

the consumption of variety ω in product i chosen by the consumer, Bli is the set of varieties in

product i available in country l, and the product-specific parameter αi determines the elasticity

of substitution across varieties so that εi = 1/(1 − αi) > 1. Using the standard utility maxi-

mization problem, we can find the demand function for each variety: qli(ω) =[pli(ω)]

−εiY li

(P li )1−εi where

P li =[∫ω∈Bli

(pli(ω)

)1−εi dω]1/(1−εi).A firm in country k produces one unit of output with a cost minimizing combination of inputs

that costs cki , which is country- and industry-specific cost for unit production. Thus, the unit

cost of production is identical across firms in producer country k’s industry i. 1/aki is firm-specific

productivity measure (i.e., a firm with a lower value of aki is more productive and that with a higher

value of aki is less productive) whose product-specific cumulative distribution function Gi(aki ) has

a country-specific support [aki ,+∞].

Each variety ω is produced by a firm with productivity aki . In addition, if a firm seeks to

sell its variety in foreign country l, it has to pay two types of trade costs: one is a fixed cost of

serving country l (fkli >0) and the other is a variable transport cost (τkli > 1). Since the market is

characterized by monopolistic competition, a firm in country k with a productivity measure of aki

maximizes profits by charging the productivity-adjusted mark-up price: pki = cki aki /αi. If the firm

exports it to country l, the delivery price is pkli = τkli cki aki /αi.

As a result, the associated operating profit from the sales to country l is

2

πkli

(aki

)= (1− αi)Y l

i

(τkli c

ki /αiP

li

)1−εi(aki )

1−εi − fkli (1)

where the expected profit is a monotonically increasing function with respect to 1/aki for any pair

of an exporter country k and an importer country l.

Since the profits are positive in the domestic market for surviving firms (Melitz, 2003), all firms

are profitable in home country k. However, sales to an export market such as country l are positive

only when a firm is productive enough to cover both the fixed and variable costs of exporting.

Moreover, the positive observation of country-level exports of product i depends solely on the

most productive firm since the expected profit from equation (1) varies only with the firm-specific

productivity (1/aki ) in each industry. Now, let us pick the most productive firm in country k in

industry i whose productivity level is aki and define the following latent variable Zkli (aki ) as

Zkli (aki ) =(1− αi)Y l

i

(τkli c

ki /αiP

li

)1−εi (aki )1−εi

fklit. (2)

Equation (2) is the ratio of export profits for the most productive firm (see: equation (1)) to the

fixed cost of exporting good i to market l. Positive exports are observed if and only if the expected

profits of the most productive firms in industries are positive: Zkli (aki ) > 1.

Equation (2) provides the foundation for our empirical work. To estimate this equation, we

define fkli = exp(λiϕkli −ekli ) where ϕkli is an observed measure of any country-pair specific fixed trade

costs, and ekli is an error term. Using this specification together with the empirical specification

of variable trade costs: (1 − εi) ln(τkli ) = −γidkl + ukli where dkl is the log of distance between

countries k and l and ukli is an error term, the log of the latent variable zkli (aki ) = ln(Zkli (aki )) can

be expressed as

zkli (aki ) = βi + βki + βli − γidkl − λiϕkli + ηkli (3)

where βki is an exporter fixed effect that captures (1−εi) ln(cki ) and (1−εi) ln(aki ); βli is an importer

fixed effect that captures (εi−1) ln(P li ) and ln(Y li ); ηkli = ukli +ekli is random error; and the remaining

variables are captured in a constant term (βi).

We now define the indicator variable T kli to be 1 when country k exports product i to country

l and to be 0 when it does not. Let ρkli be the probability that country k exports product i to

3

country l conditional on the observed variables. Then, we can specify the following logit equation:

ρkli = Pr(T kli = 1|βi, βki , βli, dkl, ϕkli

)(4)

= Λ(βi + βki + βli − γdkl − λitϕkli + ηkli

)

where Λ is the logistic distribution function with time-invariant standard error σηi .1

2.2 Estimation Results

To estimate equation (4) for each product for year 2005, we employ data on bilateral trade across

127 countries2 (16,002 country pairs) for 144 3-digit differentiated products (SITC Rev.3). We

have 2,304,288 potential observations, of which 680,003 observations (29.5%) are non-zero values.

Total value of imports from these non-zero observations is 6,394 millions US dollars. We prepare

the following bilateral indexes for the estimation of equation (4): distance between two countries

(distancekl), dummy variables for common border (boerderkl), common language (languagekl), the

sizes of trade costs (trade_costkl), business start-up costs (startupkl), and legal origin (legalkl) are

developed from Head et al. (2010) and the World Development Indicators. We use the log of the

sum of each index from the two (exporter and importer) countries to create the business-related

cost variables (i.e., trade_cost, startup, and legal).

We report the estimation results of equation (4) for each of the 144 differentiated products in

Table 1. For each product, we have at most 16,002 observations. The median value of observations is

15,624, of which 30.3% are non-zero observations on average. Although we have 16,002 observations

for each product, we have to drop the observations if a country exports that product to all 1261See Baldwin and Harrigan (2011) for further discussions on the estimation method and strategy.2We mark * for 38 developed countries. We divide developed and developing countries according to the nominal

GDP per capita above and below $20,000. The 127 countries are Albania, Algeria, Argentina, Australia*, Austria*,Bahamas*, Bahrain*, Bangladesh, Belarus, Belgium*, Belize, Bhutan, Bolivia, Botswana, Brazil, Bulgaria, BurkinaFaso, Cambodia, Cameroon, Canada*, Central African Rep., Chile, China, Hong Kong*, Colombia, Comoros, CostaRica, Côte d’Ivoire, Croatia, Cyprus*, Czech Republic*, Denmark*, Dominica, Dominican Republic, Ecuador, Egypt,El Salvador, Estonia, Ethiopia, Fiji, Finland*, France*, Gabon, Gambia, Germany*, Ghana, Greece*, Guatemala,Guinea, Guyana, Honduras, Hungary, Iceland*, India, Indonesia, Ireland*, Israel*, Italy*, Jamaica, Japan*, Jordan,Kazakhstan, Kenya, Kuwait*, Latvia, Lebanon, Lithuania, Luxembourg*, Madagascar, Malawi, Malaysia, Maldives,Mali, Mauritania, Mauritius, Mexico, Mongolia, Morocco, Mozambique, Namibia, Netherlands*, New Zealand*,Nicaragua, Niger, Nigeria, Norway*, Oman, Pakistan, Panama, Paraguay, Peru, the Philippines, Poland, Portu-gal*, Qatar*, South Korea*, Romania, Russian Federation, Rwanda, Saudi Arabia*, Senegal, Singapore*, Slovakia,Slovenia*, South Africa, Spain*, Sri Lanka, Sudan, Suriname, Sweden*, Switzerland*, Syria, Thailand, Trinidad andTobago*, Tunisia, Turkey, Ukraine, United Arab Emirates*, United Kingdom*, Tanzania, Uruguay, USA*, Venezuela,Viet Nam, Yemen, Zambia, and Zimbabwe.

4

trading partners, imports that product from all 126 countries, or does not export or import the

product at all. For example, Japan exported “passenger vehicles”(SITC 781) to all 126 countries.

In this case, we cannot estimate the probability of exports for Japanese auto industry since the

observed probability is 100 percent. Given the large number of estimates we have for each product,

we do not report all the results. Instead, in the table we provide summary statistics for the estimated

coeffi cients, the proportion of coeffi cients that have the expected sign, and the proportion of those

that are statistically significant at the 5% level.

The probability of successful exports from country k to l (ρkli ) is negatively related to the log

of distance between two countries. 100 percent of product-level estimates carry negative signs and

are statistically significant for all years. Many of the coeffi cients on the business- and trade-related

variables are statistically significant with the predicted signs, supporting the important roles played

by trade costs in influencing the market entry.

The success for a firm from a country’s certain industry in exporting to a destination market

depends only on its firm-specific productivity. Using this property, we estimate the ranges of

productivities that enable firms to export from country k to l for each product i. We define the

lowest productivity level or the cut-off productivity level (1/akli ) as that point where a firm from

country k has zero profit in a foreign market l: πkli (akli ) = 0. Note that a firm with a higher value

of 1/aki is more productive. Thus, we should have 1/akli < 1/aki if firms in country k’s industry i

succeed in exporting to country l.

Now, by using equation (2), we can show that zkli (aki ) is a function of the log of the relative

productivity: zkli (aki ) = (εi − 1) ln(akli /a

ki

). Remember that ρkli is the predicted probability of

market access, which is estimated from equation (4). Let zkli (aki )/σηit = Λ−1(ρkli ) be the predicted

value of the log of the latent variable. Then, we can show the relationship between our estimates of

the log of the latent variable and the log difference between the highest and the cut-offproductivities:

zkli (aki )/σηi ≈ (εi − 1)[ln(

1/aki

)− ln

(1/akli

)]. (5)

According to the Melitz model, more firms will choose to enter an export market over time if

they are increasingly able to achieve positive profits in foreign markets. As we discussed above,

this could be due to any of a number of factors including rising standards of living throughout

the world, technological advances in transportation technology, or country-specific advances in

5

production technology at the industry level. While it is intuitive to understand that trade costs are

fundamental determinants of exporters’productivity range, it is not well-known what determines

the productivity range within a country across 144 products. To get some insights about the

cross-product characteristics, we examine the case of Thailand. Figure 1-1 shows the scatter plot

across the number of each industry’s trading partners against the average value of zkli (aki )/σηi for

k=Thailand across 126 trading partners. Since we estimate equation (4) from the zero-one data of

trade partners and zkli (aki )/σηi is its fitted value, it is not surprising to find the tight correlation

between these two numbers. Thus, whether or not an industry from an exporter country has larger

values of productivity range depends critically on the success of market access. We also present

the case for the United States in Figure 1-2 as an example of large developed countries. The figure

suggests that as countries develop, the number of trading-partner countries increase across all

products. Although the United States does not have a comparative advantage in apparel product

(i.e., "cotton fabrics, women" SITC 652), the United States exports this product to 112 markets.

This tendency that developed countries have more markets even for comparative disadvantage

sectors may involve the issue of product-differentiation in quality (Schott, 2004).

3 Relative Revealed Comparative Advantages

The traditional idea of the revealed comparative advantage by Balassa (1965) is a useful measure

to understand the structure of the commodity trade. Since the data on commodity trade is by far

the best available source of global data, this measure has been used to measure the competitiveness

of industries worldwide by development-related agencies and policy makers.

In particular, we prepare the well-known index for revealed comparative advantage proposed

by Balassa (1965):

RCAki =

∑lm

kli /∑

i

∑lm

kli∑

k

∑lm

kli /∑

k

∑i

∑lm

kli

(6)

where mkli is the value of country l’s imports of product i from country k.

If RCAki is greater than unity, the commodity i in country k is revealed to have a comparative

advantage in producing product i. Notice that the RCA index represents the structure of commodity

trade relative to the world average. For example, the RCA index for k=U.S. and i=Aircraft (SITC

6

792) is around 4. This simply means that the share of Aircraft in U.S. total exports is four times

greater than that of world trade. This simply means that the share of Aircraft in U.S. total exports

is four times greater than that of world trade.3

Table 2 reports the highest RCA index value for each of selected products across 127 exporter

countries. For example, Ivory Coast had the highest RCA value of “Chocolate and Cocoa” and

the Switzerland had the highest RCA value of “Watches & Clocks”. Table 3 reports the list

of the products and the number of their trading partners for the top and bottom ten products

for some selected developed and developing countries. According to Table 3, regardless of the

comparative advantage or disadvantage products, the United States export products to almost all

trading partners. This tendency holds not only for productive small-size country such as Austria

but also large-size newly developed country such as China. For the least developing countries such

as Bangladesh, they do not have any export markets for their comparative disadvantage products.

4 Productivity Range and Revealed Comparative Advantages

To examine the revealed comparative advantages bilaterally, we use equation (6) from exporter

country k and importer country l and develop the following relative revealed comparative advan-

tages:

ln(RRCAkli ) = ln(RCAki )− ln(RCAli) (7)

Figure 2 plots the log of productivity range to that of the relative revealed comparative advan-

tage for Thailand as an exporter and USA an importer. Here, we have 137 of 144 commodities. In

this figure, the positive number in the log of the relative revealed comparative advantage implies

that industry i in Thailand has relatively larger share in exports relative to that in the United

States. Interestingly, there is a clear positive and statistically significant relationship between these

two variables. It is worth emphasizing that while we estimate productivity ranges from product-

level data we develop the RRCA measures from export shares in each country relative to world

exports. This tendency is not limited to South countries’access to North markets. For example,

3The index has criticized because there is no theoretical foundation. For this aspect, Constinot et al. (2011)develops the RCA index from the Eaton and Kortum (2002) model by using the non-zero observations in trade. We,on the other hand, study the role of comparative advantage for the range of productivities for exporters.

7

even the access of Japan to Brazilian market, we find the similar positive relationship between these

two variables.

Finally, we estimate the following equation to investigate the association between the produc-

tivity ranges and the revealed comparative advantage for each exporter-importer pairs.

ln(RRCAkli ) = akl + bklzkli (8)

The results of estimating equation (8) are summarized in Table 4. In the table, we report

the average values of bkl for each subset of the trading partners according to the development

level. For example, we report the country-pairs with an exporter and an importer both from

developing countries in the intersection of the first column and the first raw. The average value of

the coeffi cients (bkl) estimated from the equation above is 0.117. The highest average belongs to

the subsets with an importer from developed countries and an exporter from developing countries.

The results indicate that the productivity ranges are strongly correlated with relative revealed

comparative advantages for this subset. Finally, the second panel in Table 4 reports the portion of

the statistically significant coeffi cients. Regardless of the development stages of the countries, most

of the estimated values are positive and statistically significant at the 5% confidence level.

5 Conclusions

This paper examined the productivity range of successful exporters across countries and industries

from data consist of 127 countries and 144 products. We found that the productivity ranges differ

systematically according to revealed comparative advantage. In particular, firms in the comparative

advantage sectors are likely to be successful in foreign markets. This tendency is stronger for a

country pairs involving one developed and one developing countries.

References

[1] Baldwin, Richard, and James Harrigan, 2011, "Zeros, Quality, and Space: Trade Theory and

Trade Evidence," American Economic Journal: Microeconomics, pp. 60-88.

8

[2] Costinot, Arnaud, Dave Donaldson, and Ivana Komunjer, 2011, "What Goods Do Countries

Trade? A Quantitative Exploration of Ricardo’s Ideas," Mimeo.

[3] Balassa, Bela, 1965, "Trade Liberalization and Revealed Comparative Advantage,"Manchester

School of Economic and Social Studies 33, pp. 99-123.

[4] Bernard, Andrew B., and J. Bradford Jensen, 1999, "Exceptional Exporter Performance:

Cause, Effect, or Both?" Journal of International Economics 47, pp. 1-25

[5] Bernard, Andrew B., Stephen Redding, and Peter K. Schott, 2007, "Comparative Advantage

and Heterogeneous Firms," Review of Economic Studies 74(1), pp. 31-66.

[6] Eaton, Jonathan, and Samuel Kortum, 2002, "Technology, Geography, and Trade," Econo-

metrica, pp. 1741-1779.

[7] Head, Keith, Thierry Mayer, and John Ries, 2010, "The Erosion of Colonial Trade Linkages

after Independence," Journal of International Economics, pp. 1-14.

[8] Helpman, Elhanan, Marc J. Melitz, and Yona Rubinstein, 2008, "Estimating Trade Flows:

Trading Partners and Trading Volumes," Quarterly Journal of Economics, pp. 441-487.

[9] Melitz, Marc J., 2003, "The Impact of Trade on Intra-industry Reallocations and Aggregate

Industry Productivity," Econometrica, pp. 1695-1725.

[10] Schott, Peter K., 2004, "Across-Product Versus Within-Product Specialization in International

Trade," Quarterly Journal of Economics, pp. 647-678.

9

10

Figures and Tables

Table 1. Logit Estimates for 144 Differentiate Products

Expected Sign match Sign match

signs (%) & 5% level

ln (distance ) - 100.0 100.0 -1.689 -0.950 -2.284 0.252

Border + 99.3 94.4 0.917 1.965 -0.074 0.330

Language + 100.0 100.0 0.879 1.765 0.390 0.176

Trade cost - 95.1 69.4 -1.404 1.155 -3.108 0.776

Startup - 95.1 67.4 -1.453 0.564 -3.269 0.809

Legal + 66.7 20.1 0.067 0.380 -0.387 0.138

Observations 15,624 16,002 8,274 1,194

% of non-zero observations 0.303 0.498 0.124 0.086

st errors of regression 0.597 0.644 0.496 0.026

McFadden r-squared 0.274 0.305 0.221 0.018

Median Min St.DevMax

Table 2. RCA Index for Selected Products (SITC. Rev 3. 3-digit)

2005

RCA Country

FISH,DRIED,SALTED,SMOKED 741.4 Maldives

CHOCOLATE,OTH.COCOA PREP 72.1 Côte d'Ivoire

MEDICAMENTS 9.3 Ireland

EXPLOSIVES & PYROTECHNICS 41.5 Zambia

LEATHER 217.0 Nigeria

PEARLS & PRECIOUS STONES 72.6 Botswana

SILVER & PLATINUM 48.0 South Africa

TELEVISION RECEIVERS 15.2 Slovakia

SOUND RECORDER 5.1 Indonesia

TELECOMM EQUIPMENT 4.9 Finland

TRANSISTORS 11.0 Philippines

PASS.MOTOR VEHCLS. 2.7 Slovakia

TRUNK & SUIT-CASES 5.0 Viet Nam

CLOTHNG (MENS) 35.8 Bangladesh

FOOTWEAR 20.1 Viet Nam

OPTICAL INSTRUMENTS 11.6 Rep. of Korea

WATCHES & CLOCKS 27.1 Switzerland

BABY CARRIAGE,TOYS, & GAMES 4.2 China

11

Table 3. Top 10 and Bottom 10 Products in RCA (Selected Countries, 2005)

3.1. Developed countries

Exporter: USA Exporter: Austria

Top 10 RCA Markets Products RCA Markets Products

1 4.23 111 ENGINES,MOTORS NON-ELECT 13.23 98 NON-ALCOHOL.BEVERAGE,NES

2 3.81 122 AIRCRAFT,ASSOCTD.EQUIPNT 11.27 53 RAILWAY TRACK IRON,STEEL

3 2.54 121 PREPRD ADDITIVES,LIQUIDS 6.74 100 GLASSWARE

4 2.35 124 MEASURE,CONTROL INSTRMNT 6.36 60 RAILWAY VEHICLES.EQUIPNT

5 2.34 124 MEDICAL INSTRUMENTS NES 3.80 75 WOOD, SIMPLY WORKED

6 2.25 124 ELECTRO-MEDCL,XRAY EQUIP 3.35 70 PAPER,PULP MILL MACHINES

7 2.17 92 EXPLOSIVES,PYROTECHNICS 3.23 85 WOOD MANUFACTURES, NES

8 2.14 106 WORKS OF ART,ANTIQUE ETC 3.20 74 MONOFILAMENT OF PLASTICS

9 2.01 125 MEDICINES,ETC.EXC.GRP542 3.11 80 METALWORKING MACHNRY NES

10 1.89 98 TOBACCO, MANUFACTURED 3.09 100 MUSICAL INSTRUMENTS,ETC.

Exporter: USA Exporter: Austria

Bottom 10 RCA Markets Products RCA Markets Products

1 0.07 122 FOOTWEAR 0.01 14 FISH,DRIED,SALTED,SMOKED

2 0.08 119 MENS,BOYS CLOTHNG,X-KNIT 0.04 51 PEARLS,PRECIOUS STONES

3 0.08 113 POTTERY 0.05 14 CINE.FILM EXPOSD.DEVELPD

4 0.09 117 WOMEN,GIRL CLOTHNG,XKNIT 0.06 38 SHIP,BOAT,FLOAT.STRUCTRS

5 0.12 119 TRUNK,SUIT-CASES,BAG,ETC 0.06 59 RADIO-BROADCAST RECEIVER

6 0.12 116 WATCHES AND CLOCKS 0.10 20 FURSKINS,TANNED,DRESSED

7 0.13 123 OTHR.TEXTILE APPAREL,NES 0.12 87 TRUNK,SUIT-CASES,BAG,ETC

8 0.13 110 WOMEN,GIRLS CLOTHNG.KNIT 0.12 39 SILVER,PLATINUM,ETC.

9 0.14 122 SOUND RECORDER,PHONOGRPH 0.13 64 TELEVISION RECEIVERS ETC

10 0.15 122 CLOTHNG,NONTXTL;HEADGEAR 0.13 95 PARTS,FOR OFFICE MACHINS

3.2. Developing countries

Exporter: Bangladesh Exporter: Mexico

Top 10 RCA Markets Products RCA Markets Products

1 31.48 87 MENS,BOYS CLOTHNG,X-KNIT 8.18 61 TELEVISION RECEIVERS ETC

2 25.50 93 OTHR.TEXTILE APPAREL,NES 5.46 67 METERS,COUNTERS,NES

3 23.66 77 MENS,BOYS CLOTHING,KNIT 4.92 95 ELECTR DISTRIBT.EQPT NES

4 14.67 81 WOMEN,GIRL CLOTHNG,XKNIT 3.54 58 RADIO-BROADCAST RECEIVER

5 11.44 72 WOMEN,GIRLS CLOTHNG.KNIT 3.53 49 GOODS,SPCL TRANSPORT VEH

6 8.99 44 LEATHER 2.95 90 MEDICAL INSTRUMENTS NES

7 8.36 88 TEXTILE ARTICLES NES 2.17 95 ALCOHOLIC BEVERAGES

8 4.87 79 CLOTHNG,NONTXTL;HEADGEAR 2.01 92 FURNITURE,CUSHIONS,ETC.

9 3.34 53 POTTERY 2.00 87 ROTATING ELECTRIC PLANT

10 2.98 66 TEXTILE YARN 2.00 72 LIGHTNG FIXTURES ETC.NES

Exporter: Bangladesh Exporter: Mexico

Bottom 10 RCA Markets Products RCA Markets Products

1 0.00 1 SILVER,PLATINUM,ETC. 0.01 43 METAL REMOVAL WORK TOOLS

2 0.00 1 EXPLOSIVES,PYROTECHNICS 0.01 4 FURSKINS,TANNED,DRESSED

3 0.00 9 PIGMENTS, PAINTS, ETC. 0.02 57 PRINTNG,BOOKBINDNG MACHS

4 0.00 4 ELECTRO-MEDCL,XRAY EQUIP 0.02 37 PAPER,PULP MILL MACHINES

5 0.00 3 METERS,COUNTERS,NES 0.02 43 AIRCRAFT,ASSOCTD.EQUIPNT

6 0.00 9 SOUND RECORDER,PHONOGRPH 0.02 36 TOBACCO, MANUFACTURED

7 0.00 4 OTH.POWR.GENRTNG.MACHNRY 0.03 11 RAILWAY TRACK IRON,STEEL

8 0.00 3 RADIO-BROADCAST RECEIVER 0.03 32 MACH-TOOLS,METAL-WORKING

9 0.00 12 DOM.ELEC,NON-ELEC.EQUIPT 0.03 27 MONOFILAMENT OF PLASTICS

10 0.00 21 PARTS,TRACTORS,MOTOR VEH 0.04 25 PEARLS,PRECIOUS STONES

12

Table 4. RRCA versus Productivity Range

Average of coefficients and standard deviations

South as an importer North as an exporter

coef (st. dev) coef (st. dev)

South as an exporter 0.117 (0.122) 0.203 (0.164)

North as an exporter 0.081 (0.069) 0.083 (0.076)

% of significance at the 5% level

South as an importer North as an exporter

% share obs % share obs

South as an exporter 53.8 2924 73.0 2113

North as an exporter 48.9 2577 47.1 1280

13