Import Substitution and Economic Growth

Import Substitution and Economic Growth Mauro Rodrigues University de Sao Paulo, Brazil Abstract Despite Latin America s dismal performance between the 1950s and 1980s, the region experienced strong capital deepening. We suggest that these facts can be explained as a consequence of the restrictive trade regime adopted at that time. Our framework is based on a dynamic Heckscher-Ohlin model, with scale economies in the capital-intensive sector. Initially, the economy is open and produces only the labor-intensive good. The trade regime is modeled as a move to a closed economy. The model produces results consistent with the Latin American experience. Speci cally, a su ciently small country experiences no long-run income growth, but an increase in capital. JEL Classi cation: O41, F12, F43, E65 Keywords: Trade Policy, Growth, Latin America Department of Economics, University of Sao Paulo. Av. Prof. Luciano Gualberto 908, Sao Paulo, SP, 05508-900, Brazil. Phone: +55 11 3091-6069. Fax: +55 11 3091-6013. Email: mrodrigues@usp.br. I am specially indebted to Hal Cole, Matthias Doepke and Lee Ohanian for their help and encouragement. I also thank Andy Atkeson, Roger Farmer, Carlos Eduardo Goncalves, Bernardo Guimaraes, Miguel Robles and Jean-Laurent Rosenthal for valuable suggestions and comments. Financial support from Capes (Brazilian Ministry of Education) and Fundacao Instituto de Pesquisas Economicas (Fipe) is gratefully acknowledged. All errors are mine. 1

1 Introduction Between the 1950s and 1980s, Latin American countries pursued import substitution (IS) as a development strategy. A set of policies was implemented with the objective of developing an internal manufacturing sector. A crucial element of the IS strategy was to grant high levels of protection to domestic producers, largely closing these economies to international trade. Despite these e orts, most countries in the region were unable to gain any signi cant ground relative to the industrial leaders. In 1950, Latin America s GDP per capita was 27 percent of that of the U.S.; in 1980, this number was 29 percent. Figure 1 displays this pattern: relative income remained roughly stable throughout the 1950s, 1960s and 1970s, before collapsing in the 1980s. While other countries or regions also failed to develop during this period, a key distinctive feature of the Latin American case is that low growth went along with strong capital deepening. Table 1, based on Hopenhayn and Neumeyer (2004), illustrates this point. Although Latin America s growth rate in output per capita was similar to that of the U.S. between 1960 and 1985, capital-output ratio increased at a rate comparable to the fast-growing countries of East Asia. Furthermore, productivity measured as total factor productivity (TFP) grew at a slow rate especially relative to the U.S. In other words, Latin America fell behind the world technological frontier represented here by the U.S. during this period. 1 This paper suggests that a single explanation, namely the adoption of import substitution policies, can account for the unique combination of low productivity with strong capital deepening which characterized Latin America in that period. We argue that the sectors targeted by these policies were subject to economies of scale. Given the small size of the average Latin American country, these industries were unable to develop once restrictions to 1 This pattern was also identi ed by Syrquin (1986). Moreover, Cole et al. (2005) show that Latin America s capital-output ratio has approached that of the U.S. since the 1950s, even though there has been no catch up in terms of income per capita. 2

international trade were imposed. As a result, the region stagnated. The analytical framework presented here is based on a dynamic Heckscher-Ohlin model with two factors, capital and labor, and two sectors, a labor-intensive sector and a capitalintensive sector. Production of the capital-intensive good is subject to economies of scale. A country is assumed to be initially at an open-economy steady state in which it is fully specialized in the production of the labor-intensive good. IS is modeled as an unanticipated move to a situation where there is no trade in goods. The model is qualitatively consistent with important features of the Latin American experience. Speci cally, for a su ciently small country, IS will lead to no long-run growth in output per capita. Capital per capita will however increase and, as a result, measured TFP will fall. This result follows because Factor Price Equalization holds in the model. That is, the return on capital depends on the world capital stock, but not on the distribution of capital across countries. This implies that an open economy may display a relatively low capital stock in steady state. Closing the economy raises temporarily the return on capital, which leads to capital deepening in the long run. Nevertheless, if the economy is su ciently small, output per capita will not take o due to the lack of scale. Quantitatively, the model is able to approximate some key facts of the Latin American development experience. Speci cally, it implies that the average Latin American country exhibits high rates of capital deepening, but growth rates in output per capita that are similar to those of the U.S. Consequently, the model can produce low TFP growth rates as in the data. An important implication of this theory, which emphasizes scale e ects, is that larger countries should do relatively better than smaller countries after adopting IS policies. Such prediction is in line with the experience of Brazil and Mexico, the largest economies in the region. Between 1950 and 1980, Brazil increased from 17 to 28 percent of the U.S. GDP per capita, while Mexico increased from 25 to 34 percent. Figure 2 shows that these countries 3

signi cantly outperformed the average Latin American country during the IS years. This implication is also compatible with empirical evidence from Ades and Glaeser (1999) and Alesina, Spolaore and Wacziarg (2000), which have shown that, among countries that are relatively closed to trade, larger economies tend to grow faster and have higher per capita incomes. But when more open economies are considered, this size premium tends to disappear. Several scholars have considered Latin America s trade regime as a possible source for its lack of development (see for instance Balassa, 1989, Lin, 1989, Edwards, 1995, Taylor, 1998, Hopenhayn and Neumeyer, 2004, Manuelli, 2005, Cole et al., 2005). Speci cally, they argue that these economies have been largely closed to international trade, especially when compared with countries that successfully reduced their gap relative to the industrial leaders in a similar period. The present paper provides a dynamic general equilibrium model to explore this relationship. In addition, the study analyzes the implications of Latin America s trade regime for capital accumulation and aggregate productivity. The emphasis on scale is motivated by contributions of economists during the IS years, which identi ed problems such as small-scale and ine cient production plants, especially in consumer durables and capital goods sectors. For instance, in the 1960s Scitovsky (1969) pointed out that the automobile industry required plants with an annual output of 250,000 units in order to minimize costs. Nonetheless, there were eight Latin American countries with established automobile industries, whose combined production of around 600,000 cars and trucks was generated by 90 rms. 2 This study is related to the classical work of Johnson (1967), which was also motivated by the unsatisfactory results of IS policies. In the context of the static Heckscher-Ohlin model, the author has shown that, if the capital-intensive sector is protected by a tari, an exogenous increase in the capital stock may not lead to growth in national income. This 2 Other references include Balassa (1971), Baer (1972) and Carnoy (1972). 4

follows because capital accumulation shifts resources towards the protected sector, amplifying the ine ciency imposed by the tari. 3 The present paper generates similar results, but considers an alternative mechanism in which scale plays a crucial role. In addition, here capital accumulation occurs endogenously. The modeling strategy is closely related to dynamic Heckscher-Ohlin models, as in Baxter (1992), Ventura (1997), Atkeson and Kehoe (2000), Cunat and Ma ezzoli (2004), Bajona and Kehoe (2006a, 2006b) and Ferreira and Trejos (2006). These papers typically study economies composed of two standard competitive sectors (with di erent factor intensities) and two factors, capital and labor, where capital is reproducible. We add to them by considering a capital-intensive sector characterized by economies of scale and imperfect competition. Speci cally, the production structure is similar to Ethier (1982) and Helpman and Krugman (1985). While the labor-intensive good is produced by competitive rms using a standard constant returns to scale technology, the capital-intensive sector is composed of di erentiated variety producers, which behave as monopolistic competitors and operate a production function subject to a xed cost. Economies of scale arise from the expansion of the set of varieties available. Such structure is particularly useful, since it allows for Factor Price Equalization even in the absence of perfect competition. Furthermore, as in Grossman and Helpman (1989, 1990), Rivera-Batiz and Romer (1991) and Young (1991), growth takes place here through the introduction of new goods. The paper is organized as follows. Section 2 describes the model and analyzes the steady state of a closed economy. Section 3 introduces international trade and describes the steady state of an open economy. Section 4 assesses qualitatively the long-run e ects of the policy change on capital, income and consumption per capita. Section 5 extends the model to allow for long-run growth. Section 6 presents a quantitative application of the model. Section 7 concludes. 3 This argument was later formalized by Bertrand and Flatters (1971) and Martin (1977). 5

2 The Model Time is discrete and indexed by the subscript t, t = 0; 1; 2; ::: There is no uncertainty. There are two basic factors: capital and labor; and two sectors: a labor-intensive sector (sector 1) and a capital-intensive sector (sector 2). The labor-intensive sector is competitive and subject to constant returns to scale. The capital-intensive sector generates output by combining a set di erentiated varieties in a Dixit-Stiglitz fashion. Each variety is produced using capital and labor, but there is a xed cost. Variety producers behave as monopolistic competitors. This structure gives rise to scale economies in sector 2. Goods 1 and 2 (produced respectively by sectors 1 and 2) are combined to generate the nal good, which can be either consumed or invested. There is a measure N of identical in nitely-lived households, which is interpreted as the scale of the economy. Factors are fully mobile across sectors, but immobile across countries. There is no international borrowing or lending. The remainder of this section describes the basic features of the model, along with main analytical results. Details on rst-order and equilibrium conditions can be found in Appendix 1. 4 2.1 Production The nal good Y is produced by a competitive rm which combines the labor-intensive good (Y 1 ) and the capital-intensive good (Y 2 ) through a Cobb-Douglas production function Y t = Y $ 1t Y2t 1 $. The price of the nal output is normalized to 1; p 1 and p 2 denote respectively the prices of the labor- and the capital-intensive good. The production structure of the labor-intensive sector is also standard. Competitive rms operate a Cobb-Douglas technology, with capital (K 1 ) and labor (L 1 ) as inputs, that is, Y 1t = K 1 1t L 1 1 1t. 4 See Supplemental Materials for the appendices of this paper. 6

The capital-intensive sector is nonetheless subject to economies of scale. Production is modeled as a two-stage process. In the rst stage, a continuum of measure n monopolistic competitors use capital and labor to produce di erentiated varieties. In the second stage, a competitive rm combines those varieties to assemble good 2. For this last stage, we assume a CES production function with elasticity of substitution 1=(1 ), in which each variety enters symmetrically: 0 Zn t 1 Y 2t = @ x it di A 0 1= ; 0 < < 1 where x i is the quantity of variety i used in the production of Y 2. The second-stage rm chooses the quantity of each available variety, taking as given the prices of these varieties and the price of good 2. The solution of this problem yields the demand for a given variety i: x it = pit p 2t 1 1 Y2t (1) where p i is the price of variety i. In the rst stage, each variety i is produced using capital (K i ) and labor (L i ), which are combined through a Cobb-Douglas production function. Additionally, each producer has to pay a xed cost f in units of its own output. In other words, x it = K 2 it L1 2 it 1. f, where 2 > 1, i.e., sector 2 is more capital intensive than sector Variety producers behave as monopolistic competitors, that is, variety i s producer chooses p i and x i to maximize pro ts, subject to its own demand function (1). The solution of this problem yields a constant mark-up over the marginal cost ( ): p it = p t = 1 t = 1 r 2 t w 1 2 t 2 2 (1 2 ) 1 2 (2) 7

where r and w are respectively the rental rate of capital and the wage rate. The measure of available varieties n is determined by a free entry condition, which implies that pro ts are always equal to zero in the rst stage, that is, p t x it = t (x it + f). This condition, along with equation (2), imply that production x i is constant across time and varieties: x it = x = 1 f This means that expansion and contraction of sector 2 take place through changes in the number of varieties, but not through changes in output per variety. By symmetry, each variety uses the same quantity of inputs, i.e., K i = K 2 and L i = L 2 for all i. 2.2 Households There is a continuum of measure N in nitely-lived households. At any point in time t, each household is endowed with k t units of capital and one unit of time, which is supplied inelastically. Income can be used either for consumption or investment. For a given initial level of capital k 0, the representative household chooses sequences of consumption fc t g 1 t=0 and capital stock fk t+1 g 1 t=0 to maximize a time-separable lifetime utility, with discount factor and instantaneous utility function u(c), which is twice-di erentiable, strictly increasing and strictly concave. Capital depreciates at the constant rate. First-order conditions then yield the usual Euler equation u 0 (c t ) = u 0 (c t+1 )(r t+1 + 1 ). 2.3 Market clearing The following equations represent respectively the equilibrium conditions for the nal good, capital and labor markets: 8

Y t = N[c t + k t+1 (1 )k t ] Nk t = K 1t + n t K 2t N = L 1t + n t L 2t Focus is initially given to the closed-economy equilibrium. Therefore, domestic markets for goods 1 and 2 also clear. 2.4 Closed-economy steady state In what follows, variables with no time subscript denote their respective steady-state values. The choice of functional forms allows us to reach closed-form solutions for the main variables in steady state. Appendix 2 describes how to obtain expressions for number of varieties (n), capital per capita (k), output per capita (y) and consumption per capita (c). In particular, the impact of country size on n, k, y and c depends on parameter values. Assumption 1 guarantees that those variables are all increasing in N. Assumption 1 > (1 $) 2 1 $ 1 Consider an increase in N, which leads to a higher demand for each existing variety and, everything else constant, to positive pro ts. Assumption 1 ensures that entry (in the form of higher n) will restore zero pro ts in this situation. Speci cally, an increase in n has two opposite e ects on pro ts per variety. On one hand, it leads to higher prices and revenues per variety, given that the marginal product of each existing variety is an increasing function of n. 5 On the other hand, it also raises costs per variety: the expansion of the capital-intensive 5 From the production function of good 2, the marginal product of variety i is is increasing in n t. 9 nt R x it 0 1= 1 x 1, which it

sector (corresponding to the increase in n) is associated with an increase in k and, as a result, with higher wages. Assumption 1 guarantees that the demand for each variety is su ciently elastic, such that costs increase faster than revenues as entry takes place. Consequently, an increase in the number of varieties will eventually drive pro ts back to zero. 6 This result is summarized by the following proposition. Appendix 3 contains proofs for all analytical results of the paper. Proposition 1 Given Assumption 1, steady-state per capita output, capital and consumption are increasing functions of N in a closed economy. In other words, the presence of scale economies enables larger internal markets to support a higher number of varieties in steady state. As a result, they will display higher per capita income, capital and consumption when closed. All remaining variables can also be expressed in terms of N. For the remainder of the paper, z(n) will denote the steady-state value of a variable z in a closed economy. 3 Open-economy Equilibrium Trade is introduced in the model through a simple extension of the framework in section 2. In particular, there are now two countries, home and foreign, with sizes N and N respectively, where N + N = 1. These economies share all remaining parameters. Variables with an asterisk refer to the foreign country. Economies are allowed to trade good 1 and varieties costlessly. Good 2 and the nal good are then assembled internally by each country. The world market clearing condition for good 1 is therefore Y 1t + Y 1t = K 1 1t L 1 1 1t + K 1 1t L 1 1 1t, where Y 1 and Y 1 now stand for 6 Section 5 shows that Assumption 1 is necessary to guarantee the existence of a balanced growth path when country size grows at a constant rate. 10

the amounts of good 1 used by the home and the foreign country respectively, while their total production is denoted by their respective production functions. A measure of n (n ) varieties is produced by the home (foreign) country, such that there are now n + n varieties available to generate good 2 in each economy. Quantities of good 2 are then Y 2t = (n t + n t ) 1= x d t and Y2t = (n t + n t ) 1= x d t. 7 The home (foreign) country utilizes x d (x d ) units of each of the n + n varieties. Market clearing implies that x = x d t + x d t. The description of the open economy equilibrium is completed by the optimality conditions for each of the agents, along with market clearing conditions for factors and goods that are not traded internationally. Such equations are analogous to those of the closed-economy case, except that there are now two of each (one for each country). Nevertheless, optimality conditions for sector 1 rms and variety producers may not hold as equalities. The presence of international trade allows for the possibility of complete specialization of these economies either in the production of good 1 or in the production of varieties. Since sector 1 is competitive, it follows that: p 1t r 1 t w 1 1 t ; with equality if K 1 1t ; L 1t > 0 (3) 1 (1 1 ) 1 1 where the right-hand side is the marginal cost of a rm in sector 1. Similarly for a variety producer: p t 1 r 2 t w 1 2 t ; with equality if n 2 t > 0 (4) 2 (1 2 ) 1 2 Analogous conditions hold for the foreign country. For countries with su ciently low capital per capita, condition (4) does not bind and the economy fully specializes in the production of good 1. Likewise, condition (3) does not bind when capital per capita is su ciently high, so that the country only produces varieties. For intermediate values of 7 These equations were derived using the fact that production x i is constant across time and varieties. 11

capital per capita, the economy will produce some of good 1 and some varieties. This set of intermediate values is often referred as the cone of diversi cation by the international trade literature. If both home and foreign country capital stocks lie within the cone of diversi cation, conditions (3) and (4) bind for the two economies. Consequently, they will face the same factor prices r and w. 8 This result is commonly known as the Factor Price Equalization theorem. 3.1 Open-economy steady state In steady state, the rental rate of capital is determined uniquely by the discount factor and the depreciation rate. Since the countries share these parameters, they will display the same rental rate r = r = 1= (1 ). This implies that sectoral capital-labor ratios, k 1 and k 2, and therefore the wage rate are also equalized across countries. In other words, Factor Price Equalization holds in steady state. Proposition 2 establishes this result. Proposition 2 In an open-economy steady state, r, w, p 2, k 1, k 2, K 2 and L 2 are equalized across countries. Proposition 2 also implies that conditions (3) and (4) hold as equalities in steady state for both countries. In other words, their steady state capital stocks lie within the cone of diversi cation. Proposition 3 states that open-economy prices and factor intensities are the same as in a closed economy of size N + N = 1. Proposition 3 Open-economy steady state values for r, w, p, p 1, p 2, k 1, k 2, K 2 and L 2 coincide with their respective steady state values in a closed economy of size 1. 8 If equations (3) and (4) bind for both countries, then p 1t = r 1 rt 1 1 1 wt =[ 1 1 (1 1) 1 1 ] and p t = (1=)rt 2 wt 1 2 These two expressions imply that r t = rt and w t = wt. =[ 2 2 (1 2) 1 2 ] = (1=)r 2 t wt 1 1 1 2 t wt =[ 1 1 (1 1) 1 1 ] = =[ 2 2 (1 2) 1 2 ]. 12

The last proposition also establishes that the open-economy steady state satis es the equations that characterize the closed-economy steady state for N = 1. This result allows us to pin down world quantities in steady state. For instance, the world capital stock is equivalent to the capital stock of a closed economy of size 1. For this reason, z(1) denotes the world value of a given variable z in steady state. Nonetheless, country-level quantities cannot be determined by such a strategy, since it does not guarantee that domestic markets clear. In particular, although the world capital stock is uniquely determined, there are several ways to allocate this capital across countries. To see this, let k and k be respectively the home and foreign capital per capita, such that Nk + N k = k(1). Moreover, assume that these capital stocks lie within the cone of diversi cation, i.e. k; k 2 [k 1 (1); k 2 (1)]. 9 Then market clearing in home capital and labor markets implies that the number of varieties produced by the home country is: 10 n = N k k 1 (1) L 2 (1) k 2 (1) k 1 (1) The equation above shows how the patterns of production and trade are a ected by changes in the distribution of capital across countries. If k = k 1 (1) the lower limit of the cone of diversi cation then n = 0, so that the home country specializes in the production of good 1 and all varieties are generated by the foreign country. Since varieties are subject to a capital-intensive production process, higher values of k are then associated with a larger share of varieties being produced in the home country. When k reaches the upper limit of the cone of diversi cation k 2 (1), it follows that nl 2 (1) = N, i.e. all factors are allocated to the production of varieties. 9 In the lower limit of the cone of diversi cation, all factors are allocated to the production of good 1, i.e., K 1 = Nk and L 1 = N, so that k = K 1 =L 1 = k 1 (1). In the upper limit of the cone of diversi cation, all factors are allocated to the production of varieties, i.e., nk 2 = Nk and nl 2 = N, so that k = K 2 =L 2 = k 2 (1). Values in between these limits correspond to production structures that mix good 1 and varieties. 10 This equation can be obtained from the market clearing conditions in home capital and labor markets, that is, k 1 (1)L 1 + nk 2 (1) = Nk and L 1 + nl 2 (1) = N. 13

Furthermore, the distribution of capital across countries also determines the distribution of income and consumption in steady state. Countries with higher capital per capita will display higher consumption and income per capita. This follows from the fact that per capita income and consumption are linear functions of k in steady state, i.e. y = w(1) + rk and c = w(1) + (r )k. 4 Import Substitution This section uses the framework developed previously to analyze the impact of trade restrictions. Analytical results emphasize steady-state comparisons between the closed- and the open-economy equilibrium, especially regarding capital, income and consumption per capita. More precisely, we assume that the home country is initially open and fully specialized in the production of the labor-intensive good. In other words, initial capital per capita is given by k open = k 1 (1) the lower limit of the cone of diversi cation and initial income and consumption per capita are respectively y open = w(1)+rk 1 (1) and c open = w(1)+(r )k 1 (1). Import substitution is modeled as an unanticipated move to a closed-economy situation. The economy remains closed for a su ciently long period of time, so that it can reach its closed-economy steady state. In the new steady state, capital, income and consumption per capita are given respectively by k closed = k(n), y closed = y(n) and c closed = c(n). It is important to note that, in an open-economy situation, per capita income, consumption and capital depend on the size of the world economy (N + N = 1), but not on the size of the home market N. This stands in contrast with the closed-economy steady state, in which those objects are increasing functions of N. Proposition 4 analyzes the relationship between country size and the impact of import substitution on capital, income and consumption per capita. 14

Proposition 4 1. There exists a unique N k 2 (0; 1) such that k closed k open if N N k and k closed < k open if N < N k 2. There exists a unique N y 2 (0; 1) such that y closed y open if N N y and y closed < y open if N < N y 3. There exists a unique N c 2 (0; 1) such that c closed c open if N N c and c closed < c open if N < N c 4. N k < N y < N c There are two e ects of IS on capital per capita. On one hand, since the country now needs to produce varieties (which are capital intensive), capital has a tendency to increase in a closed economy. On the other hand, the smaller scale (relative to the open economy) forces capital to shrink. If N is su ciently low, the latter e ect dominates so that k falls. Similar e ects apply to income per capita. However, the reduction on scale implied by the trade restriction also produces a fall in the wage rate. 11 This brings an extra e ect into play, which contributes to decrease y. For this reason, when N is such that k open = k closed, closing the economy leads to a fall in long-run income per capita (part 4 of Proposition 4). Put di erently, if N = N y, the economy experiences capital deepening, but output per capita does not show any long-run growth. These results show how the model can qualitatively account for key features of the Latin American experience during the IS period. In particular, su ciently small countries (with size around N y ) display low growth in output per capita, but high growth in capital per 11 Appendix 2 demonstrates that the steady-state wage is an increasing function of N. In an open economy, the wage rate is that of a closed economy of size 1. In a closed economy, the relevant scale is the size of the home market. Therefore, the imposition of trade restrictions produces a fall in the wage. 15

capita. As a result, measured TFP falls. 12 Furthermore, larger countries can be more successful in reducing their gap relative to industrial leaders, which is consistent with the cases of Brazil and Mexico. Proposition 4 also allows us to analyze the e ects of IS on consumption per capita and therefore the long-run welfare implications of the model. Speci cally, N c > N y means that the size requirement to generate long-run growth in consumption is even more strict. This follows because at N = N y the economy reaches the same per capita income as initially, but needs a higher level of investment in steady state. In other words, growth in income per capita is not necessarily associated with long-run welfare gains. Corollary 1 analyzes the impact of the policy change on capital-output ratios: Corollary 1 For any N 2 (0; 1), k closed =y closed = k closed =y closed and k open=y open < k closed =y closed. According to Proposition 1, two closed economies will not display convergence in capital and output per capita, unless they are of equal size. Nonetheless, Corollary 1 establishes that convergence will happen in terms of capital-output ratios. More importantly, the result also implies that the home country will experience an increase in its capital-output ratio independently of its size. This means that the policy change will induce capital deepening, whether or not there are gains in terms of output per capita. 4.1 Implications for TFP Assume that TFP (denoted by A) is measured using a Cobb-Douglas production function for the aggregate economy, with the nal output Y as a function of aggregate stocks of capital and labor, K and N. More precisely, Y t = A t K t N 1 t, with 2 ( 1 ; 2 ). Measured TFP 12 Although Table 1 reports some TFP growth, the fall in TFP produced by the model is consistent with the data. The reason is that the model has no intrinsic source of long-run growth. In a growing economy, this corresponds to an increase in TFP at a rate lower than trend. Section 5 shows that the model can be easily extended to allow for long-run growth. 16

growth is then A b = by b k, where by = (1=T ) ln(y closed =y open ) and b k = (1=T ) ln(k closed =k open ) are respectively the annual growth rates of capital and output per capita over a time interval T. Clearly, TFP falls for countries with size around N y in Proposition 4, since they display increase in k, but relatively no change in y. One can show that the negative e ect on measured productivity occurs even in relatively large economies. Furthermore, the TFP loss is inversely related to country size (see Appendix 2 for derivation). The decrease in TFP is associated with the reduction of the set of varieties available to the country. Speci cally, the number of varieties is an increasing function of market size. Closing the economy diminishes the size of the market from 1 (size of the world economy) to N (size of the internal market). Since this e ect is stronger for smaller economies, they tend to experience sharper drops in TFP. This implication is related to Proposition 4, especially regarding the e ect on output per capita. More precisely, the trade restriction a ects output per capita in two opposite ways: it leads to an increase in the capital-output ratio (which contributes to raise y), but to a decrease in TFP (which contributes to lower y). The impact on y will therefore depend on country size. In particular, smaller countries are more likely to experience reductions in long-run output per capita, given that the TFP e ect tends to more than compensate the capital deepening e ect for these economies. 5 Balanced Growth Path This section extends the model to allow for long-run growth. Speci cally, country sizes now increase at the constant rate g(n) which is common to both economies, i.e., N t = N 0 [1+g(N)] t and N t = N 0 [1+g(N)] t. Initial sizes are normalized as N 0 +N 0 = 1. Assuming CRRA preferences, Proposition 5 shows that per capita output, capital and consumption 17

exhibit the same growth rate on a balanced growth path. Proposition 5 On a balanced growth path, per capita output, capital and consumption grow at the same rate, which is given by: g(y) = g(k) = g(c) = (1 )(1 $) (1 $ 1 ) (1 $) 2 g(n) (5) Notice that the number of varieties can continuously expand as a result of size growth, allowing per capita quantities to display a long-run trend. In the aggregate, this trend appears as TFP growth: ba = g(y) g(k) = (1 )(1 )(1 $) (1 $ 1 ) (1 $) 2 g(n) (6) Proposition 5 also shows that Assumption 1 is necessary for the existence of a balanced growth path. This assumption constrains the parameters of the model so that k, y and c display positive long-run growth rates. The balanced growth path can be solved as usual. First, detrend all variables by their respective long-run growth rates. Then recast the model in terms of detrended variables and solve for the steady state. This procedure yields a system of equations analog to that of previous case, in which there was no long-run growth. Consequently, Propositions 1 through 4 hold for the balanced growth path as well. However, these results now refer to the relationship between detrended variables and initial country sizes N 0 and N 0. 6 Quantitative Application So far we have analyzed mainly the model s long run implications, emphasizing comparisons of steady states and balanced growth paths. This section addresses the dynamic impact 18

of IS policies, from a quantitative standpoint. In particular, transition paths are computed, following the introduction of trade restrictions in the model. Additionally, the model s quantitative implications are evaluated, focusing mainly on the e ects on income per capita, capital-output ratio and TFP. Outcomes are compared with actual growth rates for Latin America during the period 1960-1985, as displayed in Table 1. The remainder of this section describes the choice of parameter values, then presents transition paths, and nally discusses numerical results from the quantitative exercise. 6.1 Parameter Values Capital shares ( 1 and 2 ) and the share of the labor-intensive good ($) are taken from Cunat and Ma ezolli s (2004) work on the dynamic Heckscher-Ohlin model, i.e., 1 = :17, 2 = :49 and $ = :61. Following Gollin (2002), = :30. The size of a country is measured by its stock of e ciency units of labor N = EL, where L is population and E is an index of e ciency. 13 Let g(e) be the growth rate of e ciency units and assume no population growth. 14 The aggregate production function can be then written as Y t = A t K t (E t L t ) 1. Although A and E play similar roles in this equation, they are fundamentally distinct in the model: while E follows an exogenous process, A is an outcome of the model. This distinction is explored here to calibrate the parameters g(e) and. More precisely, equations (5) and (6) imply that the long-run growth rates of output per capita (Y=L) and A are given respectively by: 13 Tre er (1993) has shown that, when the labor input is measured in this fashion, Factor Price Equalization is consistent with observed cross-country wage di erences. 14 Increasing the population growth rate from zero has no signi cant impact on our main quantitative results. 19

g(y =L) = g(a) = 1 + (1 )(1 $) (1 $ 1 ) (1 $) 2 (1 )(1 )(1 $) g(e) (1 $ 1 ) (1 $) 2 g(e) (7) Parameters g(e) and are calibrated such that their values replicate growth rates of Y=L and A for the U.S. economy between 1960 and 1985. In particular, the growth rate of productivity net of physical and human capital is used to measure g(a). The values g(e) = 1:27% and = :96 yield g(y=l) = 1:30% and g(a) = 0:02%, which are consistent with actual growth rates reported by Klenow and Rodriguez-Clare (1997). 15 Values for the discount factor and the depreciation rate are set respectively at = :96 and = :1. Moreover, u(c) = ln c. 6.2 The impact of IS policies We now consider the impact of IS policies on an average Latin American country. As in previous sections, the initial distribution of capital is set such that the home country is fully specialized in the production of the labor-intensive good. At time zero, barriers to international trade are erected. In 1948, the U.S. was responsible for 38.2 percent of Latin America s imports and 52 percent of the region s exports. 16 For this reason, the U.S. is chosen to represent the rest of the world. The size of an average Latin American country is estimated as follows. The e ciency gap E=E is calibrated such that the initial steady state matches Latin America s GDP per capita 15 These values are consistent with those shown in Table 1. Klenow and Rodriguez-Clare s (1997) paper is also the data source used by Hopenhayn and Neumeyer (2004), work in which Table 1 is based. 16 Bulmer-Thomas (2003). 20

relative to the U.S. in 1960 (26%). 17 Moreover, from 1960 population data, L=L = 6%. This implies that the size of a typical Latin American country, N=(N + N ), is approximately 2 percent. Including Western Europe and Japan in the rest of the world drives this number down to 1 percent. Focus is therefore given to these two values when assessing the e ects of IS on Latin America. In previous sections, IS is modeled as a move to fully closed economy. This assumption is useful to establish key analytical results, but it is too extreme compared with the actual Latin American experience. Although trade barriers were considerably high during the IS years, they were not fully prohibitive. Trade shares did not converge to zero. For this reason, the following quantitative exercises explore cases where the home country partially closes its economy to international trade. Speci cally, the home country is still assumed to be initially specialized in the production of the labor-intensive good. However, at time zero, the size of the rest of the world is reduced to a smaller yet still positive number, instead of zero as in previous sections. More precisely, let k and k be the initial capital per capita stocks of the home and the foreign country. Then the size of the rest of the world is reduced from N to N 0, resulting in the initial aggregate capital per capita stock (Nk + N 0 k )=(N + N 0 ). Given these initial conditions, the open economy of size N +N 0 is simulated to produce transition paths, holding all remaining parameters xed. 18 The new size of the rest of the world, N 0, is selected to reproduce certain values of the home country s trade share, in the new steady state. Details on the computational strategy are described in Appendix 4. 17 Data from Maddison (2003). The model predicts that initial relative income per e cient units of labor, y=y, is around 84%. We therefore set E=E = :31, so that relative income per capita is 26%. 18 N 0 is solely a function of N and the new trade share. More restrictive trade policies (with smaller new trade shares) require larger reductions in N 0. Moreover, larger economies will need a higher N 0, to generate the same fall in the trade share. 21

6.3 Transition Paths This subsection presents numerical examples of transitions paths produced by the imposition of IS policies in the model. The policy is implemented at t = 0. Before this instant, the economy is on its open-economy balanced growth path, in which it is fully specialized in the production of the labor intensive good. Given parameter values, the initial trade share is 78 percent. 19 To analyze the e ect of trade restrictions, N is kept constant at 0:02, while the size of the rest of the world is reduced to reproduce three scenarios: a fall in trade share to 70%, a fall in the trade share to 50%, and the case in which the home country becomes fully closed to trade (new trade share equal to zero), as in the previous sections of this paper. Other parameters are held xed. 20 Figure 3 displays transition paths produced by this exercise, with time series for capital per capita, income per capita and the number of varieties generated at the home country. To focus on the transition, all variables are detrended according to their long-run growth rates. Variables behave similarly for the three values of the trade share. At t = 0, the policy reduces the number of varieties available, forcing the home country to produce some of them internally. As a result, n jumps from zero to a positive number. Nonetheless, given that capital is initially low, shifting resources towards the capital-intensive sector the activity in which the economy does not have comparative advantage generates a sharp fall in output. As time passes and capital accumulates, the number of varieties produced domestically increases further. Output then follows an increasing path after the initial drop. The three cases in Figure 3 provide further intuition on the role of trade restrictions. Speci cally, the larger the fall in the trade share, the higher the need to produce varieties 19 It is easy to show that, in the initial steady-state, the trade share is 2(1 $). This will be the initial trade share in all the experiments below, since it does not depend on N. 20 The only di erence between these three cases is the size of the rest of the world, N 0, which is adjusted to match the di erent values of the new trade share. 22

at home. More closed economies will therefore display higher growth in n. This in turn corresponds to a stronger incentive to accumulate capital, resulting in higher long-run levels of k and y. Nonetheless, more closed economies will initially su er larger reductions in their set of available varieties, thus experiencing sharper drops in output at time zero. Figure 4 examines the impact of scale on the transition, considering our preferred values for country size N = 0:01 and N = 0:02 along with a larger economy, N = 0:05. In all three cases, the trade share is lowered from 78 to 50%. All remaining parameters are kept constant. 21 As the transition paths show, larger countries are able to produce more varieties internally, both at t = 0 and as time passes. As a result, output displays not only a smaller initial drop, but stronger subsequent growth. In the cases considered here, lack of scale prevents the smaller country (N = 0:01) from recovering its initial level of income per capita, whereas the larger economy (N = 0:05) is able to experience some long-run growth in y (the economy of intermediate size, i.e. N = 0:02, reaches about the same level of y as initially). Table 2 analyzes the welfare impact of trade restrictions, for combinations of trade shares and country sizes used above. Numbers represent how much one would need to decrease the level of steady state consumption (relative to the initial situation) to replicate the fall in lifetime utility induced by the policy change. The e ects are quite signi cant. For instance, when the trade share decreases from 78 to 70%, a country of size N = :02 experiences a fall in welfare equivalent to a permanent reduction of 6.7 percent in consumption. In addition, the table shows how welfare is a ect by the extent of the trade restriction. For a given country size, smaller trade shares tend to entail a higher level of distortion in the economy and therefore are associated with larger welfare costs. This happens even though more restrictive trade policies generate higher long-run output (as in Figure 3). The results 21 In this case, N 0 is adjusted according to N, since larger economies need a new rest of the world of larger size to generate a given reduction in the trade share. 23

also illustrate the role of scale: for a given trade share, larger economies su er lower drops in welfare. 6.4 Growth rates We now assess the quantitative impact of IS policies, especially regarding average growth rates of output per capita, capital-output ratio and measured productivity. The model s response is compared with data from the period 1960-1985, as shown in Table 1. 22 The new size of the rest of the world is selected to reproduce the average trade share of Latin America between 1960 and 1985. The Penn World Tables provide two measures for the average trade share: 46.9% (in current prices) and 53.7% (in constant prices). Table 3 reports results for these two values, as well as their average, 50.3%. Growth rates displayed in Table 3 were calculated as follows. Initially, an open economy is on a balanced growth path, with the home country fully specialized in the production of the labor intensive good. At t = 0, trade restrictions are imposed. The stationary version of model is then simulated, producing a path for detrended output per capita fey t g. Within 25 years of the policy change (corresponding to the period 1960-1985), the average annual growth rate of output per capita is therefore d Y=L = (ey 24 =ey 0 ) 1=25 [1 + g(y=l)] 1, where (ey 24 =ey 0 ) 1=25 1 represents the average transitional growth rate and g(y=l) represents the trend growth rate (as in equation (7)). A similar procedure yields the capital-output ratio growth rate d K=Y. 23 TFP is then measured using a Cobb-Douglas production function, as in Subsection 4.1. Given growth rates of output per capita and capital-output ratio, the TFP growth rate is d T F P = (1 ) d Y=L d K=Y. 24 22 In this paper, the trade restriction is implemented at once. Nonetheless, there is evidence suggesting that the policy change was gradual (see for instance Taylor, 1998). For this reason, we use average growth rates over a relatively long period of time (25 years) to compare the model s outcome with the data. 23 In this case, the trend growth is zero, since both capital and output grow at the same rate in the long run. 24 More precisely, T F P = Y K L = (Y=L)1 1 (K=Y ). By log-di erentiating this equation, one can obtain d T F P = 24

The model yields growth rates that approximate well those experienced by the average Latin American country. Speci cally, the model can roughly reproduce some quantitative features of the Latin American economy, i.e., a high rate of capital deepening, along with growth rates in output per capita that are similar to those of the U.S. For our preferred values for country size (0.01 and 0.02) and trade share, the model produces growth rates in output per capita between 1.25 and 1.36%, growth rates in the capital-output ratio between 1.34 and 1.50% and growth rates in TFP between 0.46 and 0.52%. In the data, these numbers are 1.33, 1.39 and 0.51%, respectively. In particular, the parametrization with N = 0:02 and a trade share of 50.3% produces rates that are quite close to the data. Table 3 also analyzes the situation where the country becomes totally closed (trade share equal to zero). In this polar case, the model tends to exaggerate the impact of IS policies: growth rates in output per capita and capital-output ratio are too high compared with the data. This follows because the trade policy forces the home country to produce all varieties that are consumed internally. Given that varieties are subject to a capital-intensive production process, there will be a strong incentive to accumulate capital. Moreover, driving the trade share to zero entails a higher level of distortion in the economy, which is re ected in lower TFP growth rates. To illustrate the impact of country size, the table reports results for N = 0:05 and N = 0:09. For a given trade share, an increase in size leads to higher growth in output per capita, roughly the same growth rate in the capital-output ratio and, as a result, higher TFP growth. This is consistent with the analytical results established in previous sections. (1 ) Y=L d K=Y d. Here, TFP measures the variation in output not accounted for by physical capital and raw labor. Since we assumed a production function of the form Y = AK (EL) 1, TFP embodies both terms A and E. More precisely, T F P t = A t E 1 t. 25

7 Concluding Remarks This paper was motivated by a set of peculiar facts that characterized the Latin American economy during the IS years. Speci cally, the region experienced relatively fast growth in its capital-output ratio, despite of having no signi cant gains in terms of income per capita. Furthermore, TFP growth was low, especially in comparison with the U.S. We developed a simple model that can qualitatively account for these facts. The analytical framework was based on a dynamic Heckscher-Ohlin model, with economies of scale in the production of the capital-intensive good. A country was assumed to be initially open to international trade and fully specialized in the production of the labor-intensive good. IS was modeled as a move to a closed-economy situation. Proposition 4 established the main analytical result of the paper. For a su ciently small country, closing the economy leads to no long-run growth in per capita income, but capital per capita increases. As a result, measured TFP falls. Intuitively, capital increases since the economy now needs to produce the capital-intensive good, but income may not grow due to the lack of scale. The model is also able to quantitatively account for some of these key facts. In particular, it predicts that the average Latin American country will exhibit a fast rate of capital deepening, but will show no signi cant growth in income per capita (relative to the U.S.). In addition, the model produces low productivity growth rates, consistent with those experienced by Latin America. Scale economies were introduced here as in the intraindustry trade literature, i.e. scale arises as access to a higher number of di erentiated varieties of a given good. This simple structure was quite useful to account for the puzzling weak relationship between capital accumulation and output growth observed in Latin America during the IS period. Nonetheless, such framework leaves an important aspect of the Latin American experience unexplained. 26

As highlighted in the Introduction, the scale problem in Latin America was particularly evident from the presence of small production units, especially in industries with large minimum e cient plant sizes. In the model, however, the production of each rm does not depend on the size of the market. Considering the e ects of IS policies on rm size can be an interesting extension of the paper. As pointed out by Holmes and Stevens (2005), this would require us to step out of the traditional intraindustry trade framework in favor of a richer production structure. But we leave this issue for future research. References Ades, A. F., Gleaser, E. L., 1999. Evidence on growth, increasing returns, and the extent of the market. Quarterly Journal of Economics 114, 1025-1045. Alesina, A., Spolaore, E., Wacziarg, R., 2000. Economic integration and political disintegration. American Economic Review 90, 1276-1296. Atkeson, A., Kehoe, P. J., 2000. Paths of development for early- and late-bloomers in a dynamic Heckscher-Ohlin model. Federal Reserve Bank of Minneapolis, Research Department Sta Report 256. Baer, W., 1972. Import substitution and industrialization in Latin America: experiences and interpretations. Latin American Research Review 7, 95-122. Bajona, C., Kehoe, T. J., 2006a. Demographics in dynamic Heckscher-Ohlin models: overlapping generations versus in nitely lived consumers. Federal Reserve Bank of Minneapolis, Research Department Sta Report 377. Bajona, C., Kehoe, T. J., 2006b. Trade, growth, and convergence in a dynamic Heckscher- Ohlin model. Federal Reserve Bank of Minneapolis, Research Department Sta Report 378. 27

Balassa, B., 1971. Patterns of economic structure, growth, and trade. In: Balassa, B. (Ed.) The Structure of Protection in Developing Countries. The Johns Hopkins Press, Baltimore. Balassa, B., 1989. Outward orientation. In: Chenery, H., Srinivasan, T. N. (Eds.), Handbook of Development Economics. North Holland, Amsterdam, vol. 2, pp. 1675-1689. Baxter, M., 1992. Fiscal policy, specialization, and trade in the two-sector model: the return of Ricardo? Journal of Political Economy 100, 713-744. Bertrand, T. J., Flatters, F., 1971. Tari s, capital accumulation and immiserizing growth. Journal of International Economics 1, 453-460. Bulmer-Thomas, V., 2003. The Economic History of Latin America since Independence. Cambridge University Press, Cambridge. Carnoy, M., 1972. Industrialization in a Latin American common market. The Brookings Institution, Washington, D.C. Cole, H. L., Ohanian, L. E., Riascos, A., Schimitz, Jr., J. A., 2005. Latin America in the rearview mirror. Journal of Monetary Economics 52, 69-107. Cunat, A., Ma ezzoli, M., 2004. Heckscher-Ohlin business cycles. Review of Economic Dynamics 7, 555-585. Edwards, S., 1995. Crisis and reform in Latin America: from despair to hope. Oxford University Press, New York. Ethier, W., 1982. National and international returns to scale in the modern theory of international trade. American Economic Review 72, 389-405. Ferreira, P. C., Trejos, A., 2006. On the output e ects of barriers to trade. International Economic Review 47, 1319-1340. 28

Gollin, D., 2002. Getting income shares right. Journal of Political Economy 110, 458-474. Grossman, G. M., Helpman, E., 1989. Product development and international trade. Journal of Political Economy 97, 1261-1283. Grossman, G. M., Helpman, E., 1990. Comparative advantage and long-run growth. American Economic Review 80, 796-815. Helpman, E., Krugman, P. R., 1985. Market structure and foreign trade: increasing returns, imperfect competition, and the international economy. The MIT Press, Cambridge. Holmes, T. J., Stevens, J. J., 2005. Does home market size matter for the pattern of trade? Journal of International Economics 65, 489-505. Hopenhayn, H. A., Neumeyer, P. A., 2004. Latin America in the XXth century: stagnation, then collapse. Working Paper. Johnson, H. G., 1967. The possibility of income losses from increased e ciency or factor accumulation in the presence of tari s. Economic Journal 77, 151-154. Klenow, P. J., Rodriguez-Clare, A. 1997. The neoclassical revival in growth economics: has it gone too Far? In: Bernanke, B., Rotemberg, J. NBER Macroeconomics Annual 1997. Cambridge: The MIT Press. Lin, C., 1989. Latin America vs. East Asia: a comparative development perspective. M. E. Sharpe, Armonk, N.Y. Maddison, A., 2003. The world economy: a millenial perspective. OECD, Paris. Manuelli, R. E., 2005. Growth in Latin America: models and ndings. Mimeo. Martin, R., 1977. Immiserizing growth for a tari -distorted, small economy. Journal of International Economics 7, 323-328. 29

Rivera-Batiz, L. A., Romer, P. M., 1991. Economic integration and endogenous growth. Quarterly Journal of Economics 106, 531-555. Scitovsky, T., 1969. Prospects for Latin American industrialization within the framework of economic integration: bases for analysis. In: IDB, The Process of Industrialization in Latin America. Inter-American Development Bank. Syrquin, M., 1986. Growth and structural change in Latin America since 1960: a comparative analysis. Economic Development and Cultural Change 34, 433-454. Taylor, A. M., 1998. On the costs of inward-looking development: price distortions, growth, and divergence in Latin America. Journal of Economic History 58, 1-28. Tre er, D., 1993. International factor price di erences: Leontief was right! Journal of Political Economy 101, 961-987. Ventura, J., 1997. Growth and interdependence. Quarterly Journal of Economics 112, 57-84. Young, A., 1991. Learning by doing and the dynamic e ects of international trade. Quarterly Journal of Economics 106, 369-405. 30

Table 1: Average Annual Growth Rates (1960-85) Y=L K=Y A Latin America 1.33% 1.39% 0.51% East Asia 4.74% 1.63% 2.83% Developed 2.40% 0.61% 1.50% World 2.24% 1.08% 1.24% U.S. 1.30% 0.56% 0.74% Legend for Table 1: Based on Hopenhayn and Neumeyer (2004). Y =L = output per capita; K=Y = physical capital-output ratio; T F P = total factor productivity. TFP is calculated using a production function of the type Y = AK L 1, with = :3. 31

Table 2: Welfare reduction induced by the trade restriction (%) N = 0:01 N = 0:02 N = 0:05 New trade share = 70% 8.0713 6.7075 4.8942 New trade share = 50% 9.5069 8.1633 6.3626 New trade share = 0 10.5813 9.2570 7.4749 Legend for Table 2: Numbers represent the percentage reduction in the level of consumption per capita (relative to the initial steady state) necessary to induce the same variation in welfare as the policy change. The nine cases above are combinations of three values of country size (.01,.02 and.05) and three values of the new trade share (70%, 50% and zero). In all cases, the initial trade share is 78%. 2

Trade Share = 53.7% Table 3 Growth Rates Implied by Model (%) Trade Share = 50.3% Trade Share = 46.9% Trade Share = 0.0% Y/L K/Y TFP Y/L K/Y TFP Y/L K/Y TFP Y/L K/Y TFP N=.01 1.24 1.33 0.47 1.26 1.41 0.46 1.28 1.48 0.45 1.51 2.20 0.40 N=.02 1.31 1.33 0.52 1.33 1.41 0.51 1.35 1.48 0.50 1.58 2.20 0.44 N=.05 1.39 1.32 0.58 1.41 1.40 0.57 1.43 1.48 0.56 1.66 2.20 0.50 N=.09 1.44 1.31 0.62 1.46 1.39 0.61 1.48 1.47 0.60 1.72 2.20 0.54 Data 1.33 1.39 0.51

Legend for Table 3: Average annual growth rates produced by the model over a period of 25 years, as a result of the imposition of trade restrictions. Y/L = income per capita; K/L = capital per capita; TFP = total factor productivity Numbers are reported for combinations of four different country sizes (.01,.02,.05 and.09) and four different trade shares (53.7%, 50.3%, 46.9% and zero). The data come from Hopenhayn and Neumeyer (2004) specifically from their Table 5. Their numbers represent average growth rates across Argentina, Bolivia, Brazil, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Guyana, Haiti, Honduras, Mexico, Nicaragua, Panama, Peru, Uruguay and Venezuela, over the period 1960-1985. TFP growth is calculated using a Cobb-Douglas production function, with capital and labor as inputs. The capital share is set at 0.30.

1994 1996 1998 2000 1992 1990 1988 1986 1984 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 Figure 1 GDP per capita relative to U.S. 1968 1970 1972 1974 1976 1978 1980 1982 Year 1966 1964 1962 1960 1958 1956 1954 1952 1950 Fraction of U.S. GDP per capita

Legend for Figure 1: Data from Maddison (2003). GDP per capita is PPP adjusted. Latin American GDP per capita is a population-weighted average of data from Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Haiti, Honduras, Mexico, Nicaragua, Panama, Paraguay, Peru, Uruguay and Venezuela.

2000 1998 1996 1994 1992 1990 0.35 0.30 0.25 0.20 0.15 Figure 2 GDP per capita relative to U.S. 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 Year Latin America Brazil Mexico 1968 1966 1964 1962 1960 1958 1956 1954 1952 1950 Fraction of U.S. GDP per capita

Legend for Figure 2: Data from Maddison (2003). GDP per capita is PPP adjusted. See Figure 1 legend for details on the Latin American average.

Figure 3: Transition Paths (different trade shares) 0.04 number of varieties 0.035 0.03 0.025 0.02 0.015 0.01 70% 0.005 50% 0 0 0 5 10 15 20 25 30 35 40 1.2 capital per capita 1.1 1 0.9 0.8 0.7 70% 0.6 50% 0 0.5 0 5 10 15 20 25 30 35 40 0.54 output per capita 0.52 0.5 0.48 0.46 70% 0.44 50% 0 0.42 0 5 10 15 20 25 30 35 40

Legend for Figure 3: Transition paths for number of varieties, capital per capita, and output per capita. An economy of size =.02 is considered under three scenarios: (i) a reduction of the trade share to 70%; (ii) a reduction of the trade share to 50%; and (iii) a reduction of the trade share to zero. In all cases, the initial trade share is 78%.

Figure 4: Transition Paths (different country sizes) 0.05 number of varieties 0.04 0.03 0.02 0.01 N=.01 N=.02 N=.05 0 0 5 10 15 20 25 30 35 40 0.9 capital per capita 0.8 0.7 0.6 N=.01 N=.02 N=.05 0.5 0 5 10 15 20 25 30 35 40 0.5 output per capita 0.49 0.48 0.47 0.46 0.45 N=.01 0.44 N=.02 N=.05 0.43 0 5 10 15 20 25 30 35 40

Legend for Figure 4: Transition paths for number of varieties, capital per capita, and output per capita. A reduction of the trade share from 78 to 50% is analyzed for three different country sizes: =.01, =.02 and =.05.