Trade liberalization, product variety and growth in a small open economy: a quantitative assessment

Journal of International Economics 56 (2002) 247–272www.elsevier.com/ locate /econbase

Trade liberalization, product variety and growth in asmall open economy: a quantitative assessment

a , b*Thomas F. Rutherford , David G. TarraDepartment of Economics, University of Colorado, Boulder, CO 80309-0256, USA

bMailstop No. MC3-303, The World Bank, 1818 H St., N.W., Washington, DC 20433, USA

Received 31 August 1999; received in revised form 1 February 2001; accepted 9 April 2001

Abstract

We develop a numerical growth model that quantifies the welfare effects of tradeliberalization. Additional intermediate input varieties provide the engine of growth anddramatically magnify the welfare gains from trade liberalization. In our central model, a10% tariff cut leads to a 10.6% estimated gain in Hicksian EV. Systematic sensitivityanalysis shows that there is virtually no chance of a welfare increase less than 3%, but a6.6% chance of a welfare gain greater than 18%. We show that complementary reforms arecrucial to fully realize the potential gains from the trade reform. 2002 Elsevier ScienceB.V. All rights reserved.

Keywords: Trade liberalization; Product variety; Growth; Computable general equilibrium

JEL classification: F12; F17; F13

1. Introduction

International trade economists have typically argued that an open trade regime isvery important for economic development. This view has been based partly onneoclassical trade theory, which generally finds that a country improves its welfarefrom trade liberalization, partly on casual empirical observation that countrieswhich remain highly protected for long periods of time appear to suffer

*Corresponding author. Tel.: 11-202-473-7677; fax: 11-202-522-1159.E-mail address: [email protected] (D.G. Tarr).

0022-1996/02/$ – see front matter 2002 Elsevier Science B.V. All rights reserved.PI I : S0022-1996( 01 )00121-0

248 T.F. Rutherford, D.G. Tarr / Journal of International Economics 56 (2002) 247 –272

significantly and partly on several empirical papers (e.g., Dollar, 1992; Edwards,1993; Sachs and Warner, 1995) that find that trade or trade liberalization is

1beneficial to growth. A troubling problem is that the numerical modelingestimates of the impact of trade liberalization have generally found that tradeliberalization increases the welfare of a country by only about one-half to one

2percent of GDP, gains which are very small in relation to the paradigm. Theconsistently small estimated welfare gains in constant returns to scale (CRTS)models of trade liberalization came to be known as the ‘Harberger constant.’Although increasing returns to scale (IRTS) models (such as Harris, 1984) haveproduced gains up to 10% of GDP, the estimates remain less than convincing for a

3strong version of the paridigm. For many years authors have claimed that thewelfare gains from trade liberalization would be much larger if the dynamicimpact of trade liberalization were taken into account, but we are only beginning

4to develop such models.There are endogenous growth models, such as Young (1991), that show that if

trade leads to specialization in products without productivity enhancing charac-teristics and there are no spillovers from trade, openness can be immizerizing. AndGrossman and Helpman (1990) have shown that trade can induce shifts between

1Of course, all aspects of the paradigm that trade or trade liberalization leads to faster growth havebeen subject to criticism. See, for example, criticisms by Rodrik (1992) and Harrison and Hanson(1999). Importantly, causality has been questioned in ordinary least squares results, see Rodriguez andRodrik (2000). However, after developing an instrumental variable for trade, Frankel and Romer(1999) find that ordinary least squares does not overestimate the positive impact of trade on growth.

2Examples of constant returns to scale models with estimates of welfare gains from tradeliberalization of less than one percent of GDP include: de Melo and Tarr (1990, 1992, 1993); Harrisonet al. (1993a,b, 1997a,b); Morkre and Tarr (1980, 1995); and Tarr and Morkre (1984).

3Estimates from some of the IRTS based models have been controversial, since in the tradeliberalization counterfactual other behavorial assumptions were simultaneously modified (see Harrisonet al., 1993a, 1997a).

4Keuschnigg and Kohler (1996) and Rutherford and Tarr (1997) have developed CRTS Ramsey typemodels. Their results show that a comparative static model may be a close approximation to the annualwelfare gains from trade liberalization in a dynamic model, if the dynamic model does not have anincreasing returns to scale sector. The only numerical application of an endogeous growth model weknow is Connolly (1999). Using a quality ladder model, she estimates that the Southern country willgain from 20 to 29% of GDP from an exogenous shift from autarchy to 13% imports of intermediates.Some numerical general equilibrium modelers have produced comparative ‘steady state’ estimates ofthe welfare gains which are two to four times the comparative static estimates of their models (e.g.,Harrison et al., 1996, 1997a; Francois et al, 1996; and Baldwin et al., 1997). These are multi-sectorquantifications of the Baldwin (1989) ‘medium term growth bonus,’ which hold the rental rate oncapital constant and allow the capital stock to vary. Harrison et al. (1996, 1997a) and Rodrik (1997)have explained, however, that these estimates overestimate the gains from trade liberalization in a fullydynamic representation of their models because they fail to adjust for the foregone consumption cost ofachieving the higher capital stock. Nonetheless, the estimates for Hicksian equivalent variation remainless than 5% of GDP, except for the Baldwin, Francois and Portes paper; and Rodrik (1997) hasestimated that after adjusting for the foregone consumption cost of investment, the estimated equivalentvariation in the Baldwin, Francois and Porter paper would also be less than 5%.

T.F. Rutherford, D.G. Tarr / Journal of International Economics 56 (2002) 247 –272 249

the manufacturing and research and development sectors that can either speed upor slow down worldwide growth. On the other hand, there are many endogenousgrowth models (for example, Romer, 1990; Romer and Rivera-Batiz, 1991;Grossman and Helpman, 1991; Segerstrom et al., 1990) that show that tradeliberalization unambiguously increases economic growth. Even these latter models,however, have not shown that trade liberalization has a large positive impact onwelfare. That is, due to the complexity of the models the theoretical literature hasnecessarily focused on a comparison of the steady-state growth paths. Thetheoretical literature does not derive the dynamic adjustment path, the time it takesto converge to the new steady-state path, and does not evaluate the adjustmentcosts or foregone consumption necessary to achieve the new higher steady stategrowth path. That is, even if the theoretical endogenous growth literature hasshown that trade liberalization results in a new higher steady-state growth path,there could be very small welfare gains if it takes a long time to reach the newgrowth path and there are large adjustment costs in the transition. For example,Connolly (1999) has developed a quality ladder numerical model of North–Southtrade in which the South learns how to imitate and innovate by trade. She showsthat after the South opens to trade with the North, the North loses due to transitioncosts despite having a larger long run growth rate. Although our paper incorporatessome additional dimensionality relative to the theoretical literature the principalcontribution of our paper is not theoretical. Rather we derive the dynamicadjustment path which allows us to numerically evaluate the welfare effect andadjustment costs during the transition.

International trade can have an impact on the total factor productivity of acountry through a number of mechanisms (see Grossman and Helpman, 1991 for ageneral theoretical discussion). Our approach follows from Romer (1994) whoemphasized that proper modeling of the impact of trade protection on the reductionin the number of available varieties is crucial to properly quantifying the welfareimpact of trade liberalization. The crucial idea of our model is that imports of alarger variety of intermediate inputs allow producers to increase productivitythrough selection of intermediate inputs that more closely match their productionrequirements. Support comes from several sources: Caballero and Lyons (1992)who show that productivity increases in industries when output of its inputsupplying industries increases. Coe et al. (1997) find that foreign research anddevelopment increases domestic total factor productivity and that the positivespillover effect of foreign research and development is stronger the more open the

5economy is to intermediate inputs. Finally, Feenstra et al. (1999) show thatincreased variety of exports in a sector increase total factor productivity in mostmanufacturing sectors in Taiwan (China) and Korea, and they have some evidencethat increased input variety also increases total factor productivity.

5But see Keller (1998) for an opposing view.


We develop a dynamic small open economy model defined over a 54 yearhorizon, from 1997 to 2050 with terminal constraints which approximate aninfinite horizon. There are two sectors X and Y. The Y sector produces goods fordomestic and export markets under constant returns to scale (CRTS). Inputs into Yare labor and a pure intermediate good X. The good X is produced by both foreignand domestic firms under the large group monopolistic competition assumption

6and IRTS. We employ the by now standard assumption that inputs of X affect theproduction of Y according to a Dixit–Stiglitz function. This means that additionalvarieties of X reduce the cost of producing Y.

The only tax distortion in the economy in the benchmark data set is a twentypercent tariff on imports. We first construct a steady state growth path with whichwe can compare results of counterfactual experiments. We then reduce the tariff toten percent and compare all variables to their values in the benchmark steady-state.We construct a series of counterfactual scenarios to determine the sensitivity of theresults to tax, macro and financial policies, as well as to different tariff cuts andparameter specification. We evaluate the welfare consequences of a change inpolicies, i.e., we report the Hicksian equivalent variation for the infinitely-livedrepresentative agent.

Some of our most important results are as follows: with lump sum revenuereplacement, reducing a tariff from 20 to 10% produces a welfare increase (interms of Hicksian equivalent variation over the infinite horizon) of 10.6% of thepresent value of consumption in our central model. Systematic sensitivity analysiswith over 34,000 simulations in this model showed that there is virtually nochance of a welfare gain of less than 3% but a 6.6% chance of a welfare gainlarger than 18%. We investigated several modeling variants and performed asensitivity analysis on all the key parameters and found that the welfare estimatesfor the same tariff cuts ranged up to 37.4% with capital flows, and down to 4.7%with our most inefficient replacement tax – a tax of capital. The latter resultshows, importantly, that even the very inefficient tax on capital is superior to thetariff. Doubling the size of the tariff cuts, to better reflect the substantialliberalizations of many developing countries in the past 30 years, resulted inroughly doubling the estimated welfare gains.

Large welfare gains in the model arise because the economy benefits fromincreased varieties of foreign X which dominate the decrease in varieties ofdomestic X. In order to assess the importance of variety gains, we perform thetariff reform in the same fully dynamic model, except that we assume that the X

6Our analysis can be viewed as an extension of Ethier (1982) and Markusen (1989, 1991). Markuseninvestigated the implications of the substantial trade in imported intermediate inputs using static andtwo period models. In an earlier model of ours, Rutherford and Tarr (1999), we choose a functionalform such that product varieties did not lead to a productivity boost in final demand. As a result thewelfare gains were considerably smaller. This paper relaxes that assumption and adapts the model sothat it may still produce a steady state growth path.


sector is subject to constant returns to scale and perfect competition; thenadditional varieties do not increase total factor productivity. In this model the‘Harberger constant’ reemerges, as welfare gains are about 0.5% of the presentvalue of consumption.

Our results illustrate the crucial importance of complementary reforms to fullyrealize the potential gains from the trade reform. Notably, with the ability to accessinternational capital markets, the gains are roughly tripled. Moreover, use ofinefficient replacement taxes will significantly reduce the gains. These combinedresults show that complementary macroeconomic, regulatory, and financial marketreforms to allow capital flows and efficient alternate tax collection are crucial torealize the potentially large gains from trade liberalization.

The remainder of this paper is organized as follows. Section 2 outlines essentialfeatures of our model. Section 3 presents results and sensitivity analysis withrespect to model structure and parameter specification. Section 4 concludes bybriefly examining the results in the light of econometric evidence on the impact oftrade liberalization on growth and discusses the role for optimal subsidies in thismodel.

2. Model formulation

We consider a two sector economy. The X sector, which is composed of bothdomestic and foreign firms, produces intermediate goods under increasing returnsto scale (IRTS) and imperfect competition. The Y sector produces exports and finalgoods for the domestic market under perfect competition and constant returns toscale (CRTS) in its two inputs – labor and X, where X is a Dixit–Stiglitzaggregate of the available varieties. Markups on goods in the IRTS sector arebased on the Chamberlinian large group assumption – that is, the elasticity ofdemand facing the representative firm is equal to the compensated elasticity ofsubstitution between varieties. Final demand arises from an infinitely-livedrepresentative agent who is at the margin indifferent between an additional unit ofconsumption and an additional unit of investment. In this section we outline thekey features of the model in terms of the objectives and constraints facing variousagents.

2.1. Consumer behavior

The intertemporal utility function of the infinitely lived representative consumeris the discounted sum of the utility of consumption over the horizon:

1 /rt rU 5 O D C (1)S tDt

In this equation the parameter r controls the intertemporal elasticity of


7substitution and D is the single period discount factor. Aggregate consumption ina given period (C ) is a Cobb–Douglas aggregate of consumption of domestic andt

imported final goods:

a 12aD DC 5 CD CM (2)t t t

We assume that imported final goods cannot be produced in the home market, dueto technical limitations of the domestic final goods sector. The agent’s inter-temporal and within-period consumption decisions are weakly separable. Thus, thetypical static first order condition applies on consumption decisions within a timeperiod, given a decision on how much to spend on consumption in any period. Inthe standard manner, the intertemporal decision is based on the maximization ofthe utility function subject to the constraint that the present value of consumptionequals the present value of income:

1 /rt a r (12a ) rD Dmax U 5 O D CD CMS t t Dt

s.t. (3)D M ¯(1 1 t ) O p CD 1O p CM 5O w L 1 PC S t t t tD t 0

t t t

D MIn this expression, p is the price of domestic inputs, p is the cost of importedt t

final goods, gross of tariff, w is the wage rate, and all prices are defined in presentt

value terms, discounted to period 0 (51997). The right side of the constraint,8which is the present value of income, includes the present value of wage income

together with profits from existing capital stocks and patents. t is the consump-C

tion tax, discussed below. In a steady-state equilibrium, there are no pure profits,but along an adjustment path moving to a new steady-state there may be returnsassociated with existing capital and markups over marginal cost. In other words,pure profits and losses are only associated with current (extant) firms. All firmsformed during the model horizon earn zero economic profit.

2.2. Government revenue and expenditure

The government purchases domestic final output (GD ) and imported finalt

output (GM ) to assure that the benchmark steady state level of public provision ist

maintained, i.e.]a 12aD DGD GM 5G (4)t t t

7The intertemporal elasticity of substitution s 5 1/(1 2 r). See Table 1 for the assumed values ofT

elasticities in different sectors in our central elasticity scenarios.8Note that population is fixed over the time horizon. Economic growth results solely from

productivity improvements due to the accumulation of varieties, and the real wage increases over timerelative to the prices of domestic output and imports.


The public sector budget constraint (which determines the replacement tax rate) isthen written in present value terms as:

D M M M C K YO p GD 1 p GM 5O T (t ) 1 T (t ) 1 T (t ) 1 T (t ) (5)s d s dt t t t t t t C t K t Yt t

In this equation the sum on the left represents the cost of public expenditures, andthe sum on the right represents the imposition of an equal-yield constraintasserting that any change in tariff rates must be compensated by a permanentchange in one of three alternative domestic tax instruments: consumption taxes,

Mcapital income taxes and final output taxes, respectively. The tariff rates (t ) aret

exogenously specified policy variables, whereas the tax rates on consumption,capital income and output are determined endogenously to assure that the

9government budget constraint is satisfied.

2.3. Sales and production of the final good

Good Y is produced as differentiated products for sale in the domestic andinternational markets. A constant elasticity of transformation (CET) functionshows the transformation possibilities in a given period between domestic (D ) andt

export (E ) sales for a given composite output level (Y ). The shares of sales att t

home and abroad are determined by relative prices given that firms produce thefinal good to maximize profit subject to the CET constraint:

11h(11h) /h h / (11h)D E ]t th] ]Y 5 h 1 (1 2h ) (6)F S D S D Gt D D¯ ¯D E

¯ ¯In this equation parameters D and E are the base year (1997) levels of output tothe domestic and export markets, and h is the baseline value share of domesticD

sales in total sales (the base year production level is scaled to unity) and h is theelasticity of transformation.

Production of the Y composite in any period (we drop the time subscript for theremainder of this subsection) is associated with a nested production function basedon inputs of labor (L) and differentiated intermediate inputs (x ). Given prices ofi

intermediate goods and labor, the aggregate production sector operates so tominimize the costs of producing a given output subject to the constraint:

n n a /rN a /r D F] ]12a r 12a r rY 5L Ox 5L Ox 1Ox (7)F GF Gi iD iF

i51 i51 i51

In this function, the intermediate inputs and labor enter in a Cobb–Douglasaggregate with value shares determined by base year demands. It is evident fromthe production function that we have firm level rather than national product

9In any equilibrium only one of the replacement tax instruments is non-zero, depending on thescenario-specific replacement tax option.


differentiation. Since costs of varieties can differ by foreign or domestic origin, atthe second level, within the intermediate input (X) nest, we account for substitu-tion between domestic and foreign varieties according to a constant-elasticity-of-substitution aggregation. The inputs of intermediates from domestic and foreignfirms represent the effective supply of X from these firm types. The effectivesupply of all type j firms is described by:

n 1 /rj

r r (12r ) /r1 /r ˜X ; Ox 5 n x 5 n X j [ hD,Fj (8)S D s dj ij j j j ji51

˜in which x 5x (by symmetry) is output of a representative type j firm, X 5 n xij j j j ij

is the total output from type j firms. Holding total output from type j firms(12r ) /r 1 / (s 21)constant, effective supply from type j firms increases with n 5 n ,j j

which is the ‘variety effect multiplier.’ The multiplier increases with n andj

increases as s decreases toward 1 (s . 1). (The second equation in this expressionreflects our assumption of symmetric firm structure.) We then may express theaggregate production function as:

r r] a /r12a 12r 12r˜ ˜Y 5L n X 1 n X (9)f gD D F F

Following Romer and others, we assume that the value share of X in aggregateproduction is related to the elasticity of substitution between varieties so thata 5 r, which implies that the elasticity itself is defined by the value share ass 5 1/(1 2 a). Making this substitution, we have the following expression foraggregate output:

a a] ]12a 12a˜ ˜Y 5 n L X 1 n L X (10)s d s dD D F F

2.4. Market clearance conditions

Output of the good Y supplied to the domestic market can be consumed orinvested. Investment in X and X sectors involve foregone consumption ofD F

domestic output. The market clearance for macro output sold in the domesticmarket is given by:

D DD 5 CD 1 GD 1 O b B 1 I 1 n a 1 a x (11)f s dgt t t j jt jt jt j jt jtj[hD,F j

This equation states that domestic output is purchased by households (CD ),t

government (GD ) and firms. In turn, the demand for domestic output from firmstDderives from four sources: (i) inputs to blue-print design (b B ), (ii) inputs toj jt

physical capital formation (I ), (iii) recurring fixed costs (n a ) and (iv) variablejt jt jD Dcosts of production (n a x ). The term a is a unit demand function forjt jt jt jt

domestic inputs to production, the value of which depends on relative prices inperiod t.


Both firm types (domestic and foreign) are treated symmetrically, although weadopt parameters reflecting a relatively larger share of domestic inputs for thedomestic firm. Domestic firms are assumed to make investments in plant andequipment, whereas foreign firms who generally import all the key componentsinvest solely in distribution facilities such as warehouses and transportationequipment.

The corresponding supply–demand balance for imported goods is as follows:

M MM 5 CM 1 GM 1 O b B 1 n a x (12)s dt t t j jt jt jt jtj[hD,F j

Thus, imports enter into final demand by consumers and government andintermediate demand by firms. Imported inputs may be required to establish a new

M Mfirm (b B ), and they may also enter into the cost of production (n a x ). Thej jt jt jt jtMterm a is a unit demand function for imported inputs to production, the value ofjt

which depends on relative prices in period t.

2.5. Capital stock evolution

In our model, capital is firm-specific following installation, and investment ratesmay fall to zero as a consequence of unanticipated changes in policy parameters.Following a standard Solow growth model, investment in period t produces a unitof additional capital in the following year which may be used for production in thefuture. Physical capital stock depreciates at a constant geometric rate:

K 5 l K 1 I j [ hD,Fj (13)jt11 jt jt

whereas the number of blueprints (5number of firms) of each type are permanent-ly in the market after they are produced:

n 5 n 1 B j [ hD,Fj (14)jt11 jt jt

The demand for capital equals the demand per firm times the number of firms:KK 5 a n j [ hD,Fj (15)jt jt jt

KIn this equation the term a is a unit demand function for capital inputs tojt

production, the value of which depends on relative prices in period t.

2.6. Firms and production varieties

Our model is one of a small open economy. In particular, we assume that thesmall open economy has only a negligible impact on the number of varietiesavailable on world markets, and the cost of blueprints for foreign firms. In general,we observe that there are many more varieties of products available on worldmarkets than are available in the small open economies. Accordingly, we assume


that the decision facing foreign firms is how many of the products for whichblueprints already exist can be profitably introduced in the local economy. Thus,the fixed costs of initial product development are a much smaller component of thecost of production for foreign firms than for domestic firms – and a larger fractionof total fixed costs are associated with recurring fixed costs of selling in thedomestic market.

An equivalent and perhaps easier way to think of this model is to assume thereare two types of domestic firms selling intermediate products: producers anddistributers. The distributers import the product from abroad under license at worldmarket prices. The distributers incur once and for all costs to adapt the product tothe domestic market and ongoing distribution costs, where the latter have bothfixed cost and variable cost components.

In sector X there is a one to one correspondence between firms and productvarieties. The production of good X involves both fixed and variable costs.Variablecosts include inputs of domestic and imported goods and capital. These inputs areaggregated in a Cobb–Douglas production function. For a single firm, we have:

M D KM D Ku u uj j jx 5 f A A A j [ hD,Fj (16)s d s d s djt f jt jt jt

k kin which A 5 x a k [ hM,D,Kj, and the budget shares for imported, domesticjt jt jtM D Kand capital inputs sum to unity (u 1u 1u 5 1) – marginal costs of productionj j j

exhibit constant returns to scale. Most of the costs for foreign firms selling in theDdomestic market are associated with capital services and imported goods (i.e., u F

is small). In this setting the foreign firm’s (distributer’s) production costs may beinterpreted as the cost of maintaining imported inputs and the maintenance of adistribution system within the country.

D D M M Kmin p a 1 p a 1 r at jt t jt jt jt(17)M D KM D Ku u uj j js.t. f a a a 5 1s d s d s dj jt jt jt

Firm owners choose inputs which minimize costs of production, taking prices ofdomestic, imported and capital inputs as given. Producer inputs are chosen to

D Mminimize the cost of production: in which p is the price of domestic inputs, p ist t

the cost of imported inputs, gross of tariff, and r is the rental price of capitaljt

services for type j firms in period t.Fixed costs in the production of X have two components: (i) the present value of

‘overhead,’ i.e., the present value of the recurring fixed costs involving onlydomestic inputs which is incurred in every period that the firm operates (a ), andj

(ii) ‘setup cost,’ a one time research and development (or blueprint) cost that mustbe incurred in order to design and market a new product. We assume that foreignfirms (or distributers) sell products which have been designed abroad, so theirsetup costs represent only the cost of adapting an existing design to the domesticmarket. Consequently, we assume that blueprint costs are only 10% of the total


Dfixed costs for foreign firms, but 90% for domestic firms. The parameters b andjM

b (defined above) are calibrated to be consistent with these proportions, but arej10varied in the sensitivity analysis. We assume in the present model that there is no

international trade in blueprints. Hence domestic firms may not license designs but11must purchase resources to develop new products from scratch.

The model is deterministic and firms have perfect (point) expectations of futureprices. Hence, a new firm will enter at time t if and only if the present value ofmarkup revenue over marginal costs into the future is equal to or greater than thefixed costs of initial product development plus the present value of the fixed costsof operation (for foreign and domestic firms). It is possible to interpret thisdecision using Tobin’s q theory (see Baldwin and Forslid, 1999). The rate ofinvestment in blueprints occurs to the point that the stock market value of the netincome (i.e., the present value of net surplus) equals the replacement costs, namelythe marginal cost of a blueprint, since R&D is perfectly competitive. We introducea state variable for each firm type which tracks the present value of future markupearnings; this effectively treats the human capital embodied in blueprint designs inthe same analytic framework as is conventionally applied to physical capitalformation. The free-entry assumption assures zero profit over the infinite horizon,and the time path of future prices affect not only investment activity but thedecisions by firms to enter markets and undertake product development. Optimi-zation over the infinite horizon applies not only to consumers and competitive

12firms, but also to the managers of monopolistically competitive firms.The production function for the final good Y listed above is perfectly symmetric

with respect to its use of intermediate inputs from domestic and foreign firms, i.e.,we have firm level product differentiation, with no brand or national preferences.Varieties of different vintages are equally preferred but differentiated. In thisframework, all domestic firms that operate sell the same quantity of output and

10The key value shares in our model are as follows: intermediate use in final output 0.67; imports inintermediate (final) demand 0.5 (0); share of imports in the marginal costs of foreign (domestic) firms0.75 (0.0); blueprint cost share of total fixed costs for foreign (domestic) firms 0.1 (0.9).

11We believe that this assumption allows for a greater range of model variants and impacts. Thereason is that, as discussed above, we can think of our model equivalently as one where foreign firmsare replaced with domestic firms who only distribute foreign products. Domestic distributers are thenlicensed distributers of the foreign product, i.e., domestic licensees are permitted. The key difference inour model between domestic firms and either foreign firms or domestic distributers, is that domesticfirms produce the product and blueprint domestically and thus have a different cost structure. If weallowed domestic firms to purchase blueprint designs, we would substitute a payment to foreigners forthe use of domestic resources to generate a blueprint. As a consequence, comparative advantage wouldplay less of a role since domestic firms would be less dependent on domestic factor costs. Moreover,we would not be able to model the effect of spillovers of foreign varieties on domestic blueprint costsas we do below.

12Although the underlying logic is unchanged, this achieves a considerable simplification over ourprevious model (Rutherford and Tarr, 1999) in which we tracked the level of investment for allvintages through the model horizon.


Table 1Equilibrium variables

U Utility level of the representative agent

C Aggregate final consumption in period tt

CD , CM Final consumption of domestic and imported goodst tD Mp , p , w Present value price of domestic and imported goods and wagest t t,

r present value of the rental rate on capital for firm type jjt

GD , GM Government purchases of domestic and imported final goodst tM MT (t ) Tariff revenue in period t, a function of the exogenous tarifft

rateC C K K Y YT (t ), T (t ), T (t ) Tax revenue associated with replacement taxes ont t t

consumption, capital income and output, respectivelyY , D , E Period t aggregate output; domestic supply; and exportst t t

D Fx , x Output per firm by domestic and foreign firms in period tt t

n Number of type j firms producing in period tjt

I , K , B Investment, capital stock and number of new blueprints,jt jt jt

respectively, associated with type j firms in period t.D K Ma , a , a Demand per unit of output of type j firms in period t forjt jt jt

domestic inputs, capital and imported inputs, respectively (forvariable costs)

Note: The subscript t denotes the period t in all cases; j refers to either foreign or domestic firms.

their varieties sell for the same markup-inclusive price. Likewise all foreign firmswhich operate sell the same quantity at the same price. Domestic and foreign firmsenter symmetrically in the final goods production function so the derived demandfor domestic and foreign intermediates is symmetric; but their products remaindifferentiated so their prices may therefore differ. Since foreign and domestic firmsare treated differently regarding their cost structures, their prices usually differ.

The model is formulated using the GAMS-MPSGE software developed byRutherford (1995). Variable definitions are summarised in Table 1.

3. Model results

We consider a 54-year model horizon, defined over the years 1997–2050.Initially there is only one distortion in the economy: a 20% tariff on imports ofboth goods X and Y. In order to establish a point of reference we calibrate a modelto a ‘benchmark’ steady-state growth path with assumed baseline data shown inTable 3. As is standard in these models, we have assumed that the number of firmsis a continuous variable and we have taken the number of firms of each type to bean index equal to unity in the initial year of the benchmark steady state growthpath. For details on the translation of input data to model parameters, seeAppendix B of Rutherford and Tarr (1998).

In our central counterfactual scenarios, we reduce the tariff from 20% to 10% onan ad valorem (net) basis and compare the results in all scenarios to the benchmark


steady-state equilibrium with the initial tariff in place. Unless otherwise indicated,all key variables are reported as a percentage of their values in the benchmarksteady-state equilibrium.

In our central model we assume: (1) there are no spillovers from the entry offoreign firms on the costs of domestic intermediate producers; (2) that entry of anadditional foreign or domestic firm conveys a positive externality (or spillover) byreducing the cost of a unit of aggregate intermediate products for the final goodsindustry (as explained in Section 2.3); (3) the lost tariff revenue is replaced by atax on consumption. (Given the absence of a labor–leisure choice, this isequivalent to a Lump Sum tax.); and (4) the country has difficulty accessinginternational capital markets so it faces a balance of trade constraint in eachperiod, i.e., the value of its exports must equal the value of its imports (both atworld prices) in each period. Assumption (4) can arise because the country hasimposed restrictions on financial flows, or because macroeconomic conditions inthe country are such that it cannot attract international investors. Assumptions(1)–(4) are relaxed in Sections 3.2–3.5, respectively.

In all scenarios we present the Hicksian equivalent variation (EV). The EV isbased on the intertemporal utility function optimized over the 54-year modelhorizon, with an approximation for the infinite horizon. (See Rutherford and Tarr(1998), Appendix A) for details.) We present EV in percentage terms, where thedenominator is the present value of benchmark consumption over the infinitehorizon. In the figures, we present the 54-year time path for the key variables(relative to the steady-state). The variables we report are as follows: Fig. 1: Finalconsumption, composite of domestic and imported final goods; Figs. 2 and 3:Number of domestic and foreign firms, respectively. (See Rutherford and Tarr(1998) for figures for several other variables.)

3.1. Tariff reduction with central assumptions

We first consider the scenario in which we cut all tariffs from 20 to 10%. In thisscenario, Hicksian equivalent variation (EV) increases by 10.6% of the presentvalue of consumption over the infinite horizon. Results for all model variants arereported in Table 2.

What is driving these results is the following. The removal of the tariff onimported intermediates results in an increase in the tariff ridden demand curve forimports and an increase in the price foreign firms receive for their products. Thisincreases the present value of quasi-rents for foreign firms. Entry by foreign firmsoccurs in any period until the present value of the quasi-rents are driven down tothe one time start up costs of establishing a domestic presence for the foreign firmplus the present value of the fixed costs of operating the domestic subsidiary. Afterabout 10 years, the number of foreign varieties stabilizes for the duration of themodel horizon at about 30% more than in the steady state (see Fig. 3). Theincrease in imports, however, results in a substitution effect that reduces the


Fig. 1. Consumption path following tariff reform.

Fig. 2. Domestic firms following tariff reform.


Fig. 3. Foreign firms following tariff reform.

demand for and price of domestic varieties; this shuts down investment and firmcreation for the domestic variety for a period of about 6 years.

Although the increase in foreign varieties has the impact of decreasing thedemand for domestic varieties of the intermediate good X, the domestic industryeventually stabilizes after 8 years at about 90% of its steady state value, rather than

Table 2aEstimated welfare and growth effects of tariff reduction

Model EV` G2010 G2050 Gterm

1. Central assumptions 10.6 2.6 2.2 2.12. Spillovers for domestic 10.9 2.6 2.2 2.1

Intermediate producers3. Constant returns to scale 0.5 2.0 2.0 2.04. Output tax replacement 5.2 2.4 2.1 2.05. Capital income tax replacement 4.6 2.4 2.1 2.06. Capital flows 37.4 4.2 2.7 2.2

EV`, Hicksian equivalent variation over the infinite horizon as a percent of the present value ofbenchmark steady state; G2010, Average consumption growth 1997–2010; G2050, Average consump-tion growth 1997–2050; Gterm, Terminal growth rate (from 2049 to 2050).

a Parameter choices for all models are shown in Table 3 and footnote 10. Unless otherwise indicated,all models include: lump sum replacement taxes, no spillover effect of new foreign varieties on thedomestic cost of new blueprints, and period by period balance of trade constraint.


progressively going into demise. The principal reason for this is that the marginalproductivity of domestic X in Y production increases as use of domestic X in Ydeclines. The entire labor force is employed in the production of Y, and it is notpossible to reduce labor usage applied to domestic X without also reducing laborusage in imported X. Thus, as domestic X declines due to substitution towardcheaper imported X, its marginal product increases to eventually arrest the furtherdecline of the domestic X industry.

The transitional dynamics of the model in the early years are dominated by theincrease in the number of foreign firms. The increase in the number of firms has animmediate impact on the labor productivity in the final goods sector inducingoutput of the final goods sector to increase in the first year. This immediateincrease in productivity and output of the final goods sector allows the economy tosatisfy two constraints painlessly: (1) the economy is able to invest more inintermediate goods (it takes capital to produce the foreign variety) withoutreducing consumption in the short run relative to the steady state. Although theeconomy faces the problem of determining the optimal tradeoff between consump-tion and investment, the tradeoff is within the framework of an expanded choiceset relative to the steady state. In fact, the economy consumes about 7% more inthe initial years compared with the benchmark steady state (see Fig. 1); and (2)despite the period by period balance of trade constraint, the economy is able toimport more foreign intermediate varieties, without reducing its imports of finalgoods. The economy meets its balance of trade constraint by exporting more of thefinal good. Thus, reducing the tariff does not result in any adjustment costs in thismodel except for the losses that accrue to the specific capital owners in the

13domestic intermediate goods sector. One possible way to dampen the productivi-ty surge would be to assume that capital is acquired at increasing marginal costs,see Rutherford and Tarr (1999).

Due to an increase in the rental rate on capital during the transition, during the54-year model horizon the economy grows at a rate (2.1%), which is above itssteady-state growth rate (2%). In the long run, however, the domestic interest rateand the international interest rate converge asymptotically, so that the growth rate

14of the economy converges back to the original 2% steady-state growth rate. Thus,

13In her quality ladder model, Connolly (1999) has also found that the productivity boost fromopening trade allows the small developing economy to increase consumption in the early years despiteincreasing investment expenditures. Thus, our result of the expansion of consumption and investmentcan also occur in a quality ladder model, but in a quality ladder model it would presumably bedependent on the positive spillovers from trade that Connolly assumes.

14Although they converge asymptotically, the domestic and international rental rates can differ by asubstantial amount during the transition due to the differentiation of domestic and imported goods.Savings and investment are characterized by an infinitely lived representative agent with a constantdiscount rate. The international interest rate and the discount factor together determine the long-rungrowth rate in this model.


the gains we estimate in this model are a combination of level effects andtransitional dynamics; they are not the result of a higher steady-state growth path.

Since our model employs the Chamberlinian large group assumption, themarkup over fixed costs remains unchanged, so there are no rationalization gains.Thus, these calculations show that the Ethier–Dixit–Stiglitz characterization ofproduction, where additional varieties lowers costs, is sufficient to generate thelarge welfare gains and increase in per-capita income.

3.2. Tariff reduction with spillovers to the costs of domestic competitors

There are two kinds of spillovers in this model. Most importantly, an additionalintermediate variety creates an externality for the final good sector by reducing thecost to the final good sector of a composite unit of intermediate inputs. In thisscenario only, we allow for a second type of spillover: an increase in the numberof foreign firms decreases the blueprint costs for domestic firms, with a spilloverelasticity, g, is equal to 4%. (The cost of domestic blueprints is proportional to

2ghte where h equals the proportional change in the number of imported varieties.)t

In the results reported in Table 2, we take g 50.04 so that a 100% increase in thenumber of foreign varieties relative to the steady state, reduces the blueprint costsof domestic firms by roughly 4% relative to the steady state blueprint costs. Whenwe cut the tariff on both final and intermediate imports from 20% to 10%, startingin the initial period, the equivalent variation of this scenario is 10.9% of thepresent value of consumption. In the sensitivity analysis reported in Table 3 wesimulate several values of g up to 0.3. EV increases to 12.4 with g 50.3. Thegreater the spillover elasticity on the blueprint costs of domestic firms, the fasterthe domestic industry recovers. Relative to the benchmark, the number of domesticfirms in the year 2050 as a function of g is: 24% (g 50); 11% (g 50.04); 15%(g 50.2); and 19% (g 50.3).

Thus, spillovers on the blueprint costs of domestic firms does have a positiveimpact on the Hicksian equivalent variation, but the impact is smaller than theimpact of several other elasticity and share parameters reported in Table 3. Thereason for the somewhat muted impact of this parameter is that what is mostimportant to the welfare increase is the externality (or other spillover) from thenumber of varieties rather than the geographic source of them. With blueprintspillovers, once the profitability of investment is restored for the domestic industryproducing good X, the growth of new foreign varieties falls relative to nospillovers on domestic blueprint costs – so there are more domestic firms andvarieties, but fewer foreign varieties. The loss of productivity due to the loss offoreign varieties mutes the gain in welfare due to the increased number ofdomestic varieties and the lower costs of creating them. As explained in theintroduction, in quality ladder models it is possible to exclude all spilloverexternalities, so that in quality ladder models a greater range of results is derivedfrom modifying spillover assumptions.


Table 3Piecemeal sensitivity analysis

a bParameter Parameter value Hicksian equivalent variationwith corresponding parameter

Upper Intermediate LowerUpper Intermediate Lower

g 0.3 0.1 0.04 12.4 11.8 10.6h 2.0 1.5 0.5 17.1 14.1 7.0DX

1/(1 2 r) 0.7 0.6 0.3 11.6 11.1 9.7a 0.4 0.6 0.8 11.0 10.8 12.5G 0.03 0.025 0.01 14.6 12.4 7.9R 0.06 0.055 0.04 9.3 9.9 13.0u 0.6 0.4 0.25 13.6 7.9 4.6i

Mt 60% 40% 5% 59.5 32.6 24.0u 0.7 0.4 0.1 4.3 5.9 10.6F

Key: Parameter definitions and their values in the central simulation: g, 0.04, elasticity of blueprintcosts with respect to the number of foreign firms. h , 1.0, elasticity of transformation in aggregateDX

production (domestic versus exports). 1 /(1 2 r), 0.5, intertemporal elasticity of substitution. a, 0.66,intermediate share of aggregate cost. G, 0.02, baseline growth rate. R, 0.05, baseline interest rate. u ,i

M0.5, import share of intermediate inputs. t , 20%, tariff rate on imports (in the benchmark steady state)which is exogenously changed to 10%. u , 0.1, the share of the fixed costs of foreign firms representedF

by blueprint costs.a The piecemeal sensitivity analysis employs central values for all parameters (see below) other than

the tested parameter, lump sum tax replacement and period by a period balance of trade constraint.b Hicksian equivalent variation over the infinite horizon as a percent of the present value of

consumption in the benchmark steady state.

3.3. Constant returns to scale (no variety multiplier)

In this model, we replace increasing returns to scale and imperfect competitionin the intermediate sector with constant returns to scale and perfect competition ina homogenous intermediate good. Although decisions by consumers and investorsoptimize the consumption-capital stock choice, only total output of the inter-mediate is important, so there is no productivity boost from additional varieties.

The Harberger constant reemerges, as equivalent variation for this scenariodrops to 0.5% of the present value of consumption over the infinite horizon.Transitional dynamics are then more painful, since in order to finance theadditional investment in the earlier period, consumption falls in the early yearsrelative to the steady state.

3.4. Impact of alternate replacement taxes

In the scenarios in this subsection, we use two alternate taxes as the replacementtax for the lost tariff revenue: a tax on output; and a tax on capital. Theconsumption tax is a Lump Sum distortionless tax in our model, but a tax ondomestic output discriminates against domestic output in favor of imports, and


against final output in favor of intermediate production. As a result of the relativeinefficiency of the domestic output tax, the equivalent variation is reduced to again of 5.2% of the present value of consumption.

Finally, the tax on capital produces a consumption path that is inferior to the onewith an output tax, but preferable to the benchmark path with the tariff in place,i.e., the tax on capital is the most inefficient of our three replacement taxes, but isbetter than a tariff in our model. In our model, the intermediate good X usescapital intensively since only intermediate production uses capital as a primary

15input and only final goods use labor as a primary input. A tax on capital thendiscourages the introduction of new varieties since it discourages the investmentrequired for the introduction of new varieties of products and discourages theproduction of intermediates relative to the production of final goods. The economyloses the productivity boost from the varieties and the gain in equivalent variationis 4.6% of the value of consumption. These results illustrate the importance ofefficient tax replacement. With inappropriate replacement tax mechanisms, thegains from trade liberalization can be drastically cut.

3.5. Capital flows

In this scenario, the tariff is reduced from 20% to 10%, but the country isassumed to be able to borrow on international capital markets provided theborrowing is repaid within the model horizon. Thus, the period by period balanceof trade constraint is replaced by the constraint that the present value of its exportsless imports is zero over the model horizon, but all other assumptions in the centralmodel remained unchanged. In this scenario Hicksian equivalent variation is37.4% of the present value of benchmark consumption over the infinite horizon.There is an initial jump in consumption of about 23% relative to the steady state.

Why is the increase in EV with capital flows more than three times the EV valuewithout capital flows? The ability of the country to borrow and run trade deficitsallows the country to pay for additional imports which are are used to finance thestart up capital of new foreign firms and to pay for the additional imports offoreign varieties. Foreign firms increase by almost 100% of their steady state valuewith capital flows by the year 2003, as opposed to an increase of about 35% oftheir steady state value without capital flows. The larger increase in foreign firms

16leads to a considerably larger increase in labor productivity, consumption andthe wage rate in the early years.

What is interesting is that there is a much larger increase in the number of

15Each good is used as an intermediate in production of the other good, and thus both goods use bothfactors of production indirectly.

16The latter occurs because consumers optimize consumption over the model horizon subject to theirlifetime income constraint or permanent income. With the ability to borrow on international capitalmarkets, consumers can smooth consumption more easily.


domestic firms with capital flows – in fact, the number of domestic firms increasesdramatically relative to the steady state value by the end of the model horizon.This is explained by a real exchange rate effect. The capital inflows in the earlyyears result in less real exchange rate depreciation in the early years of the modelcompared with no capital flows (in fact, they result in real exchange rateappreciation in the first 2 years); but the capital outflows in the later years of themodel result in a strong real exchange rate depreciation in those years (seeRutherford and Tarr, 1998, Fig. 2). The steeper real exchange rate depreciationwith capital flows in the years following 2020 (compared with central assump-tions) raises the costs of importing foreign varieties in those years, and shiftsdemand toward domestic varieties. Domestic agents, who fully anticipate futurereal exchange rate movements, recognize profit opportunities and begin to investby the year 2003. Among the models we consider, it is perhaps ironic that themodel with capital flows, where we see the largest initial increase of foreign firms,ultimately leads to by far the strongest resurgence of domestic firms, evencompared to the model with spillovers on the costs of domestic competitors.

3.6. Sensitivity analysis

In order to assess the robustness of the estimates, we executed 34,385simulations of the central model. In each simulation, values for the eight keyparameters of the model were drawn randomly from uniform probability

17distributions. In this model, we assume the tariff rate is cut from 20% to 10%with lump sum tax replacement, there are no spillovers of foreign firm entry on thecosts of domestic producers, and there is a period by period balance of tradeconstraint.

In Fig. 4, we display the sample distribution for Hicksian EV. The mean of theresulting sample distribution is an increase in EV equal to 10.5% of the presentvalue of consumption, which is approximately the point estimate in the centralcase. The sample variance is 23.8. Based on the sample distribution, there is a 95%probability that the welfare gain exceeds 4.5%. We observe that the distribution isskewed to the right. Notably, there is virtually no chance of a gain in Hicksian EVof less than 3%, but there is a 6.6% chance of a gain in Hicksian EV greater than18%.

In order to better understand the impact of the parameters in our model we havealso performed ‘piecemeal’ sensitivity analysis. Central parameter values areemployed in all scenarios, except for the parameter subjected to piecemealsensitivity analysis. In Table 3 we present these results. We also show the impactof different tariff changes. Except for the row where the benchmark tariff is

17The lower and upper values of the uniform probability distributions for the parameters are:0.02 , g , 0.08; 0.5 ,h , 1.5; 0.3 , 1/(1 2 r) , 0.7; 0.6 , a , 0.72; 0.01 , g , 0.03; 0.04 , r ,DX

0.06; 0.3 ,u , 0.7; 0.85 ,u , 0.95. See Table 3 for parameter definitions.i F


Fig. 4. Sample frequency distribution of welfare gains (sample size 34,385).

changed, in all scenarios the tariff rate is reduced from 20 to 10%. Lump sum taxreplacement and period by period balance of trade constraint is also assumed. Wereport results for a high, low and some intermediate value of the parameter thattypically is different from the parameter value employed in our central simulation.We present the Hicksian equivalent variation (with the approximation for theinfinite horizon) as a function of the parameter value.

The result from row 1 reports the impact of spillovers of domestic blueprintcosts discussed in Section 3.2. The results in row 2 and 3 indicate that, as is typicalin comparative static models, the more elastic are the substitution possibilities, thegreater the gains. In particular, with a larger elasticity of transformation, theeconomy is able to export more in response to a real exchange rate depreciationfollowing tariff reduction, which allows it to pay for more imports; and theadditional intermediate imports provide a productivity boost through the additional

18varieties. The impact of this parameter is quite strong.For the share of intermediates in final production (a), there are offsetting

effects: a higher share of intermediates in final production should increase thenumber of varieties because again the proportional change induced by a 10% tariffcut should induce a larger absolute increase in the number of varieties; but the

18Barro and Sala-i-Matin (1995, p.79) indicate that values of the intertemporal elasticity ofsubstitution significantly different from our central value are inconsistent with cross-country data in aRamsey model.


larger is a, the smaller is the variety multiplier for any number of varieties. Thisexplains why EV as a function of a is not monotonic in the range of our centralelasticity value.

In Rutherford and Tarr (1998, Appendix A), we show that as the rental rate oncapital and the benchmark growth rate approach each other, a given permanentincrease in consumption over the infinite horizon yields a larger EV. In rows 5 and6, we see that EV increases as (R–G) decreases.

The larger the share of imports in intermediate use (u ), the larger the EV gains.i

The reason is that for the given 10% cut in the tariff, the same proportional effecton the share of imports, generates more imported varieties, with the consequentproductivity boost, when this share is high. Although this share has a strongimpact on welfare, information on this parameter is relatively good.

We have simulated the impact of changing the tariff to 10% from an initial tariffrate of 60, 40 and 5%. The results are roughly proportional to the size of the tariffchange, e.g., the cut from 40 to 10% is about three times the EV as the cut from 20to 10%. Comparative static models produce the result that the welfare gainsincrease more than proportionately with the size of the tariff cut primarily becausethe area of the Harberger triangles increases more than proportionately with thetariff cut; but the results in this paper are based primarily on the externality fromthe number of varieties which in these simulations do not increase more thanproportionally with the tariff cut.

Finally, we simulate a change in the share of fixed costs that foreign firms mustpay for blueprint costs (as opposed to the present value of recurring fixed costs).EV declines significantly as the share of blueprint costs increases. Since blueprintcosts must be paid in the initial period that the firm is formed, they entail initialcapital costs to a much greater extent than recurring fixed costs. This drives up therental rate on capital in the early years and slows the formation of firms.

4. Extensions and conclusions

4.1. Are the welfare estimates too big?

Although estimates of equivalent variation have been widely seen as too small,some may question whether our estimates are too large. To put these numbers inperspective, in Rutherford and Tarr (1998, Appendix A), we have analyticallyderived the relationship between a permanent increase in the steady state growthrate and equivalent variation. A welfare gain of between 10 and 35% ofconsumption corresponds to a permanent increase in the growth rate of between0.4 and 1%. A policy induced change in the growth rate of this magnitude is quiteplausible in the context of the actual long term per capita growth rates over the25–30 year period beginning in 1962. For example, Sachs and Warner (1995)estimate that open economies have grown about 2.45% faster than closed


economies, with even greater differences for open versus closed economies amongdeveloping countries. They note that trade liberalization is often accompanied bymacro stabilization and other market reforms, and their open economy variable canbe picking up these other effects as well. But they argue that trade liberalization isthe sine qua non of the overall reform process, because other interventions such asstate subsidies often are unsustainable in an open economy. Moreover, Frankel andRomer (1999) have shown that adjusting for the simultaneity bias in cross countryregression studies such as Sachs and Warner, does not reduce the estimated impactof openness on growth.

These econometric estimates suggest that our central estimate of equivalentvariation, which corresponds in our central model to a growth rate change of 0.4%,may still be too small. But larger tariff changes than our 10% cut produce largerwelfare gains.

4.2. Rationalization gains and their impact

Suppose the markup over fixed costs for intermediate firms is endogenous.Then, typically, additional competition from new firms results in a reduction in themarkup of price relative to marginal costs. A reduced markup means that inequilibrium there are rationalization gains as firms operate at a lower point on theirdownsloping average cost curve. Since the first order effect of the tariff change isto increase the quasi-rents for foreign firms, and the rationalization is a secondorder effect of entry on the markup, there should still be a net increase in thenumber of firms even with the small group assumption; hence welfare gains fromadditional varieties should persist with endogenous markups. But, firms operatingat a larger scale means there will be fewer firms operating in equilibrium. Sincethe externality for the final goods sector is dependent on the number of firms, thereduced number of firms would reduce the welfare gains relative to our model.

On the other hand, the rationalization gains would be an additional welfarebenefit relative to what we have estimated in our model. Thus, there are competingwelfare effects of the small group assumption relative to our model.

4.3. Optimal subsidies and negative tariffs

Since additional varieties provide an externality, the optimal tariff in this modelis negative. Since it is additional varieties, not their geographic source that isimportant, a superior intervention instrument to a negative tariff would be asubsidy to the creation of a new firm, either domestic or foreign. We do notadvocate such a policy, however, since opportunistic behavior by firms or politicaleconomy considerations will make successful application of subsidies unlikely.For example, lobbyists will claim their industries quality as intermediate whenthey are final.


4.4. Conclusions

These results of this paper show that when we introduce an Eithier–Dixit–Stiglitz production structure into a numerical general equilibrium model, we mayvastly exceed the Harberger constant for the welfare impact. The results clearlysupport the paradigm that trade liberalization can lead to significant incomeincreases, and the results are consistent with estimated gains of trade liberalizationfrom cross-country growth regressions. But the results also illustrate the crucialimportance of complementary reforms to fully realize the potential gains from thetrade reform. Notably, with the ability to access international capital markets, thegains are more than tripled and inefficient replacement taxes will significantlyreduce the gains. The tariff in our model, however, is more inefficient that even themost inefficient alternate replacement tax.

Acknowledgements

We would like to thank Richard Baldwin, Glenn Harrison, Elias Dinopoulos,two anonymous referees of this journal and seminar participants at the conferencein Milan, Italy on Technology Diffusion and Developing Countries for helpfulcomments. Research support was provided by the World Bank under RPO No.68140, ‘The Dynamic Impact of Trade Liberalization in Developing Countries.’The views expressed are those of the authors and are not necessarily those of theWorld Bank.

References

Baldwin, R., 1989. The Growth Effects of 1992. Economic Policy 0 (9), 247–281.Baldwin, R., Francois, J., Portes, R., 1997. The costs and benefits of eastern enlargement: the impact on

the EU and central Europe. Economic Policy, 127–170.Baldwin, R., Forslid, R., 1999. Putting growth effects in CGE models. In: Baldwin, R., Francois, J.

(Eds.), Dynamic Issues in Applied Commercial Analysis. Cambridge University Press, Cambridge.Barro, R., Sala-i-Matin, X., 1995. Economic Growth. McGraw-Hill, New York.Caballero, R., Lyons, R., 1992. The case for external economies. In: Cukierman, A., Hercowitz, Z.,

Leiderman, L. (Eds.), Political Economy, Growth and Business Cycles. MIT Press, Cambridge MA.Coe, D., Helpman, E., Hoffmaister, E., 1997. North–South R&D spillovers. The Economic Journal 107,

134–149.Connolly, M., 1999. North–South technological diffusion: a new case for dynamic gains from trade.

Duke University Economics Department Working Paper No. 99-08, available at http: / /www.econ-.duke.edu.

Dollar, D., 1992. Outward-oriented developing economies really do grow more rapidly: Evidence from95 LDCs, 1976–1985. Economic Development and Cultural Change 40 (3), 523–544.

Edwards, S., 1993. Openness, trade liberalization, and growth in developing countries. Journal ofEconomic Literature 31 (3), 1358–1393.


Ethier, W.J., 1982. National and international returns to scale in the modern theory of internationaltrade. American Economic Review 72, 389–405.

Feenstra, R., Madani, D., Yang, T.H., Liang, C.Y., 1999. Testing endogenous growth in South Koreaand Taiwan. Journal of Development Economics 60, 317–341.

Francois, J., McDonald, B., Nordstrom, H., 1996. The Uruguay Round: A numerically based qualitativeassessment. In: Martin, W., Winters, L. (Eds.), The Uruguay Round and the developing countries.Cambridge University Press, Cambridge, UK.

Frankel, J., Romer, D., 1999. Does trade cause growth. American Economic Review 89 (3), 379–399.Grossman, G., Helpman, E., 1991. Innovation and Growth in the Global Economy. MIT Press,

Cambridge, MA.Grossman, G., Helpman, E., 1990. Comparative advantage and long run growth. American Economic

Review 80, 796–815.Harris, R., 1984. Applied general equilibrium analysis of small open economies with scale economies

and imperfect competition. American Economic Review 74 (5), 1016–1032.Harrison, A., Hanson, G., 1999. Who gains from trade reform: some remaining puzzles. Journal of

Development Economics 59 (1), 125–154.Harrison, G.W., Jones, R., Kimbell, L., Wigle, R., 1993a. How robust is applied general equilibrium

analysis? Journal of Policy Modeling 15 (1), 99–115.Harrison, G.W., Rutherford, T.F., Tarr, D.G., 1993b. Trade reform in the partially liberalized economy

of turkey. The World Bank Economic Review 7 (2), 191–217.Harrison, G.W., Rutherford, T.F., Tarr, D.G., 1996. Increased competition and completion of the market

in the European Union. Journal of Economic Integration 11 (3), 332–365.Harrison, G.W., Rutherford, T.F., Tarr, D.G., 1997a. Quantifying the Uruguay round. Economic Journal

107, 1405–1430.Harrison, G.W., Rutherford, T.F., Tarr, D.G., 1997b. Economic implications for Turkey of a customs

union with the European Union. European Economic Review 41, 861–870.Keller, W., 1998. Are international R&D spillovers trade related: analyzing spillovers among randomly

matched trade partners. European Economic Review 42, 1469–1481.Keuschnigg, C., Kohler, W., 1996. Austria in the European Union: dynamic gains from integration and

distributional implications. Economic-Policy 0 (22), 155–190.Markusen, J.R., 1989. Trade in producer services and other specialized inputs. American Economic

Review 79, 85–95.Markusen, J.R., 1991. First mover advantages, blocked entry and the economics of uneven develop-

ment. In: Helpman, E., Razin, A. (Eds.), International Trade and Trade Policy. MIT Press,Cambridge MA.

de Melo, J., Tarr, D., 1990. Welfare costs of US quotas in textiles, steel and autos. Review ofEconomics and Statistics 72 (3), 489–497.

de Melo, J., Tarr, D., 1992. A General Equilibrium Analysis of US Foreign Trade Policy. MIT Press,Cambridge MA.

de Melo, J., Tarr, D., 1993. Industrial policy in the presence of wage distortions: the case of the USauto and steel industries. International Economic Review 34 (4), 833–851.

Morkre, M., Tarr, D.G., 1980. Effects of Restrictions on United States Imports: Five Case Studies andTheory. Federal Trade Commission, Washington, D.C.

Morkre, M., Tarr, D.G., 1995. Reforming Hungarian agricultural trade policy: a quantitative evaluation.Weltwirtschaftliches Archiv 131 (1), 106–131.

Rodriguez, F., Rodrik, D., 2000. Trade Policy and Economic Growth: A Skeptics Guide to theCross-National Evidence, mimeo, available at: http: / /ksghome.harvard.edu/ |.drodrik.academic.ksg.

Rodrik, D., 1992. The limits of trade policy reform in developing countries. Journal of EconomicPerspectives 6 (1), 87–105.

Rodrik, D., 1997. Discussion of Baldwin, Francois and Portes, Economic Policy, 170–173.Romer, P.M., 1990. Endogenous technological change. Journal of Political Economy 98 (5), 71–102.


Romer, P.M., 1994. New goods, old theory and the welfare costs of trade restrictions. Journal ofDevelopment Economics 43 (1), 5–38.

Romer, P., Rivera-Batiz, L., 1991. Economic integration and endogenous growth. Quarterly Journal ofEconomics 106 (2), 531–555.

Rutherford, T., 1995. Extensions of GAMS for complementarity problems arising in appliedeconomics. Journal of Economic Dynamics and Control, 1299–1324.

Rutherford, T., Tarr, D.G., 1997. Regional Trading arrangements for Chile: do the results differ with adynamic model?, Paper presented at the Allied Social Science Meetings, New Orleans, LA., mimeo,The World Bank, Washington, D.C.

Rutherford, T., Tarr, D.G., 1998. Trade liberalization and endogenous growth in a small open economy,Policy and Research Working Paper Number 1970, The World Bank, Washington D.C.

Rutherford, T., Tarr, D.G., 1999. Blueprints, spillovers and the dynamic gains from trade liberalizationin a small open economy. In: Baldwin, R., Francois, J. (Eds.), Dynamic Issues in AppliedCommercial Analysis. Cambridge University Press, Cambridge, MA.

Sachs, J.D., Warner, A., 1995. Economic reform and the process of global integration, in: Brainard,W.C., Perry, G.L. (Eds.), Brookings Papers on Economic Activity, Washington D.C., Vol. 1, pp.1–117.

Segerstrom, P.S., Anant, T.C.A., Dinopoulos, E., 1990. A Schumpeterian model of the product lifecycle. American Economic Review 80, 1077–1092.

Tarr, D.G., Morkre, M., 1984. Aggregate Costs To the United States of Tariffs and Quotas On Imports.Federal Trade Commission, Washington, D.C.

Young, A., 1991. Learning by doing and the dynamic effects of international trade, Quarterly Journal ofEconomics 106 (2), 365–405.

Documents

Trade liberalization, product variety and growth in a small open economy: a quantitative assessment