On the distributional effects of trade policy: Dynamics of household saving and asset prices

The Quarterly Review of Economics and Finance 49 (2009) 944–970

Contents lists available at ScienceDirect

The Quarterly Review of Economics and Finance journal homepage: www.elsevier.com/locate/qref

On the distributional effects of trade policy: Dynamics of household saving and asset prices Luis San Vicente Portes ∗ Department of International Business, School of Business, Montclair State University, 1 Normal Ave., Upper Montclair, NJ 07043, United States

a r t i c l e

i n f o

Article history: Received 22 August 2006 Received in revised form 30 January 2008 Accepted 4 October 2008 Available online 17 October 2008 JEL classification: E60 F13 F40 Keywords: Trade policy Trade liberalization Inequality Agriculture Heterogeneous agents

a b s t r a c t This paper studies the effect of trade liberalization on inequality. We develop a theoretical framework that generates economy-wide distributions of wealth and income for different levels of trade protection. The model unambiguously determines the effect of liberalization on inequality; and rationalizes why larger inequality can be the outcome of a welfare enhancing policy, as households reduce their buffer savings when liberalization lowers the price of food. The framework reconciles the increase in inequality, the fall in the value of land, and farmers’ opposition to freer trade, that have featured in different liberalization episodes. We also present empirical support for the model’s predictions. © 2008 The Board of Trustees of the University of Illinois. Published by Elsevier B.V. All rights reserved.

1. Introduction Who gains and who loses from trade liberalization? Larger inequality is often cited as its possible cost. Though for the claim to be true, first, causality needs to be established, and second, the increase in inequality ought to reflect a welfare reduction. Whether liberalization leads to larger inequality, and whether changes in inequality are negatively related to changes in well-being, are open questions. Empirical studies suggest that trade liberalization is linked to increases in per capita income, implying that the country as whole gains. But are these gains evenly distributed or does anybody lose? From ∗ Corresponding author. Tel.: +1 973 655 2126; fax: +1 973 655 4456. E-mail address: [email protected]. 1062-9769/$ – see front matter © 2008 The Board of Trustees of the University of Illinois. Published by Elsevier B.V. All rights reserved.

doi:10.1016/j.qref.2008.10.001

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

945

an empirical standpoint the distributional effects from liberalization are uncertain. In cross-country regressions openness to trade promotes economic growth, and growth itself reduces absolute poverty; although no systematic effect on inequality is found.1 Our working notion of inequality concerns the size distribution of income, which looks at individuals disaggregated by income level. This definition is preferred when assessing the welfare effects of alternative policies since it accounts for agent specific characteristics, and allows for more than one channel to affect an individual at a point in time.2 This approach goes beyond trade models that often focus on the effect of trade liberalization on the functional distribution of income (the share of national income corresponding to the factors of production). The effects of trade liberalization occur at both the microeconomic and macroeconomic level. Often, individual-specific responses reshape the aggregate features of the economy over time. The consensus in the literature is that a shift towards free trade affects the relative price of goods and factors of production; reduces government revenue; changes incentives to investment and innovation; and modifies an economy’s exposure to domestic and foreign shocks.3 While all these channels can in principle alter the distribution of wealth and of income, the focus of this paper is on the dynamic effects of liberalization from changing the relative price of goods. In particular, we concentrate on the impact of lower tariffs on households’ consumption-saving decision, which in turn affects the aggregate distribution of wealth and income. By centering our attention in the households’ inter-temporal decisions, our insights mostly concern the liberalization effects on capital income by means of precautionary savings and changes in asset prices. Accounting for the effect of trade policy on labor income is beyond the scope of this paper.4 The contribution of this paper is to develop a theoretical framework that (1) generates economywide distributions of wealth and income for different levels of trade protection, and (2) assesses the associated changes in welfare and inequality. The proposed model provides new insights into the trade–inequality relation since it also allows to determine a household’s stance on free trade depending on its specific characteristics, such as relative wealth (poor or rich), source of income (labor, capital or both), and portfolio composition (degree of diversification). An important outcome from the framework is that it shows the conditions under which trade liberalization leads to larger inequality, to a decline in the value of land, and sets off the opposition of farmers to free trade. Moreover, the model unambiguously determines the effect of liberalization on inequality, and can rationalize why larger inequality could be the outcome of a welfare enhancing policy. One end of the model is to explore whether higher protection on agricultural goods relative to non-agricultural goods observed in international trade, has long-term effects on capital accumulation and inequality.5 “Artificially” high food prices can potentially affect households’ saving behavior. Then, if in principle food is necessary for survival, higher agricultural protection can raise the relative price of food, distorting consumption-saving decisions. That is, higher food prices can lead to larger precautionary savings particularly at low levels of income, thus reducing wealth and income inequality.6 The proposed model is a two-sector small-open economy. On the household side we use a HuggettAiyagari style heterogeneous-agent, incomplete-markets environment, where households’ savings generate endogenous distributions of income and wealth. Household heterogeneity arises due to

1

See Berg and Krueger (2003); Bhagwati and Srinivasan (2002); Dollar and Kraay (2001), and Edwards (1997). See Adelman and Robinson (1989). 3 Goldberg and Pavcnik (2004) do an extensive survey of the literature on trade, inequality and poverty. Bannister and Thugge (2001), and Winters, McCulloch, and McKay (2002) describe in detail the channels through which trade policy may affect poverty. 4 An important part of the literature attempts to measure and model the effect of liberalization on labor income, mainly in terms of the skill premium (see Feliciano, 2001; Panday, 2003). In an influential paper Krusell, Ohanian, Ríos-Rull, and Violante (2000) attribute most of the observed variation in skill premium to capital–skill complementarities, rather than increased trade. 5 The ratio of the tariff on agriculture to non-agriculture goods is greater than 1 in approximately 70% of the observations in a sample of countries over the 1988–1999 period. For a description of the data see Appendix C. 6 As an illustration, in the year 2000 in Mexico, the tortilla share of expenditure in the lowest income quartile was three times that of the highest quartile (Mckenzie, 2002). In the same year, the most-favored nation tariff on corn (basis of tortillas) was 198%, and corn accounted for more than half of Mexico’s imports of grain (Jank, Fuchsloch, & Kutas, 2001). 2

946

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

different labor income histories as households are hit by an idiosyncratic shock to earnings every period. To incorporate the notion of a subsistence floor, a key element of the analysis is that households need to satisfy a minimum consumption level of agricultural goods. Within the model, subsistence implies that the share of food in household spending is larger in low income households than in wealthier ones.7 The production side of the model consists of an agricultural sector and a non-agricultural sector; and both goods can be traded in international markets. In the model, land is a specific factor in the production of the agricultural good. Agricultural production requires land, labor and capital; while nonagricultural production only utilizes labor and capital. Labor and capital are mobile across sectors; and capital is internationally mobile as well.8 In the model the government levies taxes on consumption, labor income and capital income; imposes ad valorem tariffs on imports; and can borrow domestically to finance its expenditures. Our findings can be arranged in several dimensions. First, model simulations suggest that upon liberalization inequality can increase permanently, since the bottom 30% of the population reduce their asset holdings by more than 1% as the price of food falls. Second, the welfare change from liberalization ranges from a 1.33% permanent gain in non-agriculture consumption to a 12% permanent loss depending on the household’s wealth, productivity and portfolio of assets. Third, we find that if the government has to compensate the loss in tariff revenue from eliminating tariffs, raising consumption taxes is the better instrument, relative to labor income or capital income taxes. And fourth, we present empirical support for the model’s predictions. We find that relative tariffs are positively and significantly associated with higher savings and income at the lower deciles of the distribution of income, particularly so in low income countries that are net importers of food. One of the main insights of the paper is that even though trade liberalization might lead to greater inequality, there are welfare gains for the majority of the population. This result can be explained as follows. Inequality is a relative (cross-sectional) measure of household heterogeneity; while the welfare measure represents (absolute) dynamic improvements in household well-being, which accounts for mobility within the distribution of wealth over time.9 In terms of welfare, liberalization has a positive and a negative side in the model. On one hand, free trade lowers the relative price of food, but on the other, a lower price of food leads to a lower price of land as the agricultural sector contracts; and thus to a potential capital loss. The interaction of these factors determines the overall effect of liberalization on a household’s well-being and saving behavior. Poor households experience large welfare gains, while wealthier households with portfolios biased toward land are worse-off since the lower relative price of the agriculture good does not compensate for the capital loss. For instance, a household with ‘zero’ wealth and low productivity experiences a welfare gain equivalent to a permanent 1.33% increase in non-agriculture consumption. In contrast, a household in the top wealth decile and low productivity could experience a welfare change ranging from a −11.75 to a 0.82% equivalent change of permanent non-agriculture consumption depending on its portfolio. The model suggests that the average gain from liberalization is equivalent to a permanent 1% increase in non-agricultural consumption, provided that the government does not make up for the lost revenue by raising other taxes. The intuition behind permanently larger inequality is that as domestic prices converge to international prices, the lower relative price of food triggers households’ buffer savings to decline as they are less likely to hit the subsistence bound; less resources are needed to buy a unit of food. Inequality increases because the bulk of the effect occurs in low wealth households, who use precautionary 7 Obiols-Homs and Urrutia (2005); Alvarez-Peláez and Díaz (2005), and Chatterjee and Ravikumar (1999) study the importance of a minimum consumption requirement for growth and inequality as countries develop. These studies show that in the presence of a subsistence floor, the households’ intertemporal elasticity of substitution is increasing in wealth. 8 Models with traded and non-traded goods usually define the traded good as a composite of imports and exports. Hence, commercial policy is defined as a common tax on all international trade. In such framework neither differential protection nor the distinction between tariffs and export taxes can be studied separately. 9 In a similar vein, Flinn (2002) finds that even though the distribution of wages in the U.S. is more disperse than in Italy; the long term distribution of welfare in the U.S. has a higher mean and smaller dispersion than that of Italy. This being primarily explained by the higher mobility in the U.S. labor market.

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

947

savings to smooth consumption in times when the subsistence floor is likely to bind. From an economywide perspective this leads to a 0.6% decline in aggregate asset holdings. This effect is relatively small because most of the change in the pattern of saving occurs at the bottom of the distribution. In the long-run, savings decline by 1.5% in the bottom decile, while in the top decile savings decline by 0.2%. In the absence of the subsistence floor, we find that trade liberalization leads to a 0.80% average welfare gain despite no change in long-term inequality. In this case the households’ intertemporal elasticity of substitution is constant, and the lower relative price of the agriculture good does not change the consumption-saving decision. Therefore, in the long-run the economy returns to the pre-reform distribution of income and wealth.10 To close the study, we test the model’s predictions using a panel of countries that cover the 1988–1999 period. In line with the theoretical results, we find a positive and significant relation between relative tariffs and the income and saving of the poor; particularly so in low income countries or in countries that are net importers of food. The paper is organized as follows. Section 2 describes the model and calibration exercise, Section 3 presents the results, Section 4 provides an empirical test of the model’s predictions, and Section 5 concludes. 2. The model The model environment consists of households, a representative firm, the government and the external sector. The economy is characterized as a small-open economy that takes good prices and the interest rate as given from international markets. In the model full tariff pass-through implies that domestic prices may be different from international prices so long the country is a net importer of either or both goods. Hence trade policy might change the relative price of the goods and distort the allocation of resources within the economy.11 Throughout the model, the non-agriculture good is taken as the numeraire. For this purpose let ∗ ≡ p∗A /p∗O be the international relative price of the agricultural good in terms of the non-agriculture good. If the country were a net importer of both goods, tariff pass-through implies that the internal relative price would be given by  = (1 +  A )/(1 +  O )∗ ; where  A is the tariff rate on the agriculture good and  O is the tariff rate on the non-agriculture good.12 In the model households can accumulate wealth in the form of any of the following assets: • Physical capital (K): can be rented out for the production of either good at rate r and depreciates at rate ı. • Land (L): can be bought at price pL and may be rented out for the production of the agriculture good at rate rL . • Government bonds (B): by no arbitrage yield a rate of return equal to the international interest rate r∗. • Foreign bonds (B∗ ): yield the international interest rate r ∗ .

10 Krebs, Krishna, and Maloney (2005) study the effect of trade reform on income risk based on survey data for Mexico for the period between 1987 and 1998. Using an endowment economy with incomplete markets and idiosyncratic risk they estimate the cost of trade reform to be in the order of 0.98% of lifetime consumption; particularly in sectors with high import penetration. While Krebs et al. do not address the distributional effects of trade reform, their work and this paper provide complementary views of the welfare implications of liberalization as one focuses on labor market risk and the other on price-driven consumptionsaving decisions. Specifically, throughout the analysis we hold constant workers’ productivity and the transition probabilities between states in the labor market. 11 The model assumes long-run pass-through of tariffs into internal prices. Feenstra (1989) reports some evidence on full pass through in Japanese imports of motorcycles into the U.S. market. Nicita (2004) reports full pass through in imported food products and other manufacturing products in Mexico. 12 To facilitate interpretation, note that by using the internal (after-tariff) price of a unit of non-agriculture good as the numeraire, implies that if the household works one unit of time, the wage rate tells how many units of the non-agriculture good can be bought.

948

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

In equilibrium all assets yield the same risk free return so the households’ portfolio does not need to be specified. However, for simulation purposes we have to establish the share of land in the household’s portfolio as trade policy can generate unexpected capital gains/loses (see Section 3.3.2). At a given point in time household wealth is represented by at ≡ (Kt + pL,t Lt + Bt + Bt∗ ). 2.1. Households There is a continuum of infinitely-lived households, taken to be of measure one, who are ex ante identical but ex post heterogeneous due to different histories of labor income. Their preferences are defined over agriculture goods (cA ) and non-agriculture goods (cO ), and need to satisfy a subsistence level in the consumption of the agriculture good (sA ). Households supply labor inelastically and are assumed to be potentially credit constrained and cannot borrow. Subject to their initial wealth, their objective is to maximize the expected discounted utility from consumption.13 To introduce agent heterogeneity, every period t each household is assumed to face an uninsurable productivity shock εt ∈ E, which evolves according to a p-state first order Markov process with a p × p transition matrix . Let E ≡ [ε1 ε2 . . . εp ] be a 1 × p vector that represents the set E; then each row of  represents the probability distribution over E such that for any state j, ij = P(εi |εj ) ≥ 0 for i = 1, . . . , p;

p

and  = 1. Normalizing working hours to one, at time t a household’s labor income is given by i=1 ij wεt , where w is the wage rate. This implies that from an economy-wide perspective the aggregate labor supply is given by N = ∞ E  , where ∞ is the invariant distribution implied by . Let A ⊆ + be the set of possible values for household wealth; then at time t a household’s state is given by (εt , at ), where εt ∈ E and at ∈ A. Taking as given taxes on labor income ( n ), taxes on capital income ( k ), taxes on consumption ( c ), tariffs on agricultural goods ( A ) and non-agriculture goods ( O ); prices (r, r ∗ , rL , pL , w, ∗ ), their initial wealth (a0 ) and productivity (ε0 ), the households’ problem is max{cA,t ,cO,t } E0

∞ 

ˇt u(cA,t , cO,t ; sA )

t=0

subject to (1 +  c )(cA,t + cO,t ) + at+1 ≤ [1 + (1 −  k )r ∗ ]at + (1 −  n )wεt , ∀t, where ˇ ∈ (0, 1) is the discount factor and at+1 ≥ 0. 2.2. Firms The production side of the model consists of one competitive firm that takes as given domestic prices and can produce any combination of agriculture and non-agriculture goods by renting capital (K) and hiring labor (N). Capital and labor are mobile across sectors and are paid a rental rate r and a wage rate w. In addition to capital and labor, agricultural production also requires land. The rental rate of land (L) is rL . Let YA,t ≡ fA (KA,t , NA,t , Lt ) and YO,t ≡ fO (KO,t , NO,t ) represent the production functions of the agricultural good and of the non-agricultural good, respectively. Each sector’s production function is assumed to exhibit constant returns to scale, to satisfy the Inada conditions and that each input’s marginal product is increasing in the other arguments; that is, the crossed-derivatives are positive.14 The firm’s problem at time t is max{KA,t ;KO,t ;NA,t ;NO,t ;Lt } fA (KA,t , NA,t , Lt ) + fO (KO,t , NO,t ) − w(NA,t + NO,t ) − r(KA,t + KO,t ) − rL Lt subject to KA , KO , NA , NO , L ≥ 0 for every period t.

13

See Ríos-Rull (1995) for a review of the literature on heterogeneous agents models. If there were more than one firm, because of constant returns to scale the firms would only be scaled versions of each other. Therefore, assuming only one firm is without loss of generality. 14

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

949

One feature of the economy is that since in equilibrium the rental rate of capital only depends on the international interest rate and the depreciation rate (r = r ∗ + ı), the capital–labor ratio in the non-agriculture sector, and thus the wage rate, are pinned down by the world interest rate (r ∗ ). This means that any change in trade policy (affecting ) does not affect the wage rate nor the return on capital; although it would reallocate resources between sectors, and change the price and the rental rate of land to equalize returns across all assets (see Appendix B). 2.3. Aggregation From an economy-wide perspective, aggregate labor is given by N = ∞ E  ; aggregate capital (K) is determined by the firm’s demand of capital; and the supply of land (L¯ ) is fixed. However, on a given date, the state of the economy is characterized by how agents are positioned across levels of asset holdings and individual shocks. For this purpose, let A denote the Borel sets that are subsets of A (asset space) and let E be the set of all subsets of E (productivity space). Then, letting (X, ) = (A × E, A × E) be the product space, we can define a probability measure on (X, ) such that  :  → [0, 1], represents the distribution of households in the state space. In the model,  is the basis for computing economy-wide variables such as aggregate asset holdings At and total consumption of each good (CA,t , CO,t ) at any point in time. 2.4. Government The government is assumed to consume a constant amount of the non-agricultural good (G) every period and makes no transfers. To finance its consumption the government levies taxes and can borrow domestically. Indirect taxes from international trade are collected as long as there are imports of either commodity. If the country is a net importer of the agricultural good (i.e. CA,t > YA,t ), let MA,t ≡ (CA,t − YA,t ) denote the agricultural trade deficit; similarly let MO,t ≡ [CO,t + Kt+1 − (1 − ı)Kt + G − YO,t ] denote non-agriculture imports. Based on the above, government debt evolves according to the following equation: Bt+1 +  k r ∗ At +  n wN +  c (CA,t + CO,t ) + IMA ,t

 A p∗A pO

MA,t + IMO ,t

 O p∗O pO

MO,t = (1 + r ∗ )Bt + G,

where the initial amount of debt, B0 , is given; and IMA ,t and IMO ,t are indicator functions that respectively take a value of one if there are agricultural or non-agricultural imports, and zero otherwise.15 2.5. Equilibrium A stationary equilibrium for this economy is a set of taxes ( k ,  n ,  c ,  A ,  O ), a set of decision rules {cA (ε, a), cO (ε, a), a (ε, a)}, a set of prices (w, r ∗ , r, rL , pL , ∗ ), aggregate level of asset holdings (A), government debt (B), net foreign assets (B∗ ), and effective labor (N), and a probability measure  such that 1. The decision rules solve the households’ problem. 2. The firm’s problem is solved. 3. The market for savings clears:

 K + B + B∗ + pL L =

a(x) d = A. X

15

By taking a value of zero when the economy is a net exporter of either good, IMA ,t and IMO ,t prevent the possibility of  A

and  O being interpreted as export subsidies.

950

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

4. The government budget constraint is satisfied and debt remains bounded:  k r ∗ A +  n wN +  c (CA + CO ) + IMA

 A p∗A pO

MA + IMO

 O p∗O pO

MO = r ∗ B + G

. 5. The probability measure , is a stationary distribution consistent with the transition probability matrix  and the savings decision rules:

 () =

P(x, ) d,

for all  ∈ ,

X

where P(x, ) is the probability that a household with state x, will have a state next period lying in the set .16 6. The goods market clears:



 cO (x) d + ıK + G = YA + YO + r ∗ B∗ ,

cA (x) d +

 X

X

where the left-hand side represents aggregate spending and the right-hand side aggregate income in the economy. Outside of the steady state, equilibrium is also defined by conditions 1–6, except that B and  need not be constant over time as implied by conditions 4 and 5. Government debt should evolve according to the process described in Section 2.4 , and  must be consistent with  and with the households’ decision rules. Note that because of free capital mobility, condition 6 holds outside of the steady state since for any K0 = / K, the level of capital in the economy automatically jumps to its steady state value K; which is pinned down by r − ı = r ∗ . This implies that investment is equal to ıK for every t. 2.6. Welfare To conduct the welfare analysis we compute the constant percentage increment in non-agricultural consumption under no reform (NR) that in expectation renders a given household (with state x0 = {ε0 , a0 }) indifferent between the status quo and trade liberalization. Assuming that the economy is in the stationary equilibrium, and that there is an unannounced permanent change in trade policy towards liberalization at time 0, the welfare change is the x0 that solves: E0

∞  t=0

NR NR

ˇt u( c , c ; sA , x0 ) = E0 A,t O,t

∞ 

ˇt u(cA,t , cO,t ; sA ).

t=0

To quantify the welfare effect of liberalization reported in Section 3.3 , we simulated forward the ‘life’ of many households with the same initial level of wealth and productivity. Then, as an approximation to the expectation, averaged their discounted utilities and found the x0 that equalized the average discounted utility under both regimes. To exploit the heterogeneity implied by the model, we evaluate households with ‘zero’ wealth, bottom decile wealth, between 1st and 2nd decile wealth, median wealth, between 8th and 9th decile wealth, and those in the top wealth decile for the four levels of productivity. In short, the bottom, the middle and the top of the distribution.

16

Hopenhayn and Prescott (1992) and Huggett (1993) show that there exists a solution to this equilibrium condition.

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

951

From a population-wide perspective, when all households are equally weighted, the average welfare change from liberalization is the that satisfies the following equation:

 E0 X

∞  t=0

t

NR NR



ˇ u( c , c ; sA , ) d = A,t O,t

E0 X

∞ 

ˇt u(cA,t , cO,t ; sA ) d.

t=0

In this case is also interpreted as the constant percentage increase in non-agricultural consumption that gives the same expected utility with and without reform. If the government were to change the tax mix to compensate for the lost tariff revenue by raising other taxes, a similar welfare analysis between the status quo and the new policy regime can be performed. Furthermore, the average welfare change provides a selection criteria to choose among alternative policies. A “utilitarian” social planner would choose the post-liberalization regime that maximizes the average welfare change. 2.7. Parametrization and solution method In order to solve the model and find the economy’s stationary equilibrium with and without trade liberalization, we use the standard techniques for solving models with incomplete markets and heterogeneous agents. Appendix A, outlines the steps of the solution algorithm.17 To solve the model numerically we assume the following functional forms in the household’s and firm’s problem. • Households’ utility function: u(cA,t , cO,t ; sA ) = ([(cA,t − sA ) c 1− ] O,t

1−

)/(1 − ) with > 0.

• Agriculture production function: YA ≡ fA (KA,t , NA,t , Lt ) = L NA KA1−− . • Non-agriculture production function: YO ≡ fO (KO,t , NO,t ) = K ˛ N 1−˛ . O

O

To parameterize the model, we had the following criteria in mind: (1) to choose a country that by way of example would emphasize the mechanisms the model intended to shed light on, and (2) to choose a country for which enough data was available to calibrate all the parameters to provide a realistic representation of the economy. The first restriction would ideally provide a low income country for which the role of subsistence could be better understood; although often, obtaining specific parameters (or data) for such countries is hard. The candidate that better satisfied these conditions was Mexico. Hence, the model is calibrated to resemble the Mexican economy. Mexico applies relatively high tariffs on agriculture and non-agriculture imports, exhibits differential protection towards agricultural goods, and over the last decade has been a net importer of both types of goods (INEGI, 2004a).18 The model is calibrated such that every model period represents 1 year. The calibration exercise focuses on matching as close as possible the comparable features of the model economy to the Mexican economy; using parameter values estimated for Mexico when available. Table 1 contains the benchmark calibration.

17

See Ríos-Rull (1995) for a summary description of these methods. To compare inequality across steady states, and to evaluate the role of differential protection and the effect of liberalization, the model economy needs a way to finance its imports either by (1) borrowing from abroad, (2) being a net importer of one good and a net exporter of the other good, or (3) holding a positive net foreign asset position. However, option 1 is not feasible in the steady state unless the country runs a Ponzi scheme; option 2 does not characterize Mexico’s pattern of trade; and 3 does not characterize Mexico foreign asset position (since it is not likely to be in the steady state). In the model, imports of both goods are financed through a positive net foreign asset position (option 3). This assumption is without loss of generality since our concern is on the effect of relative price changes on capital accumulation, rather than on the financing of the trade deficit. If we assumed that agricultural imports were financed via non-agricultural exports, the effect of agricultural liberalization on savings and inequality would not change, though it would leave out of the analysis the effect of the tariff on the non-agriculture good. Furthermore, a positive foreign asset position that finances the imports of both goods allows for tariff pass-through on both agriculture and non-agriculture goods, representing the more general case. 18

952

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

Table 1 Model parametrization.

Notes. This table reports the parameter values (with sources or targets) used for solving the model.

We use standard values from the literature for the coefficient of relative risk aversion ( ), the depreciation rate (ı) and the discount factor (ˇ), see Cooley (1995). The international interest rate (r ∗ ) is chosen to match the current account relative to GDP. On the fiscal side, for the year 2001, we estimate effective (average) tax rates for consumption, labor income and capital income in Mexico based on the methodology proposed by Mendoza, Razin, and Tesar (1994).19 The average ad valorem tariff rates on agricultural and non-agricultural products are taken from the World Trade Organization (2004) for the year 2001. The debt to GDP ratio for Mexico is taken from the World Bank’s “World Development Indicators” database for the same year.

19 The methodology requires the breakdown of collected taxes on income, profits and capital gains by individuals and by corporations; as well as employers’ social security contributions. Such data is not available for Mexico, so these proportions were substituted by OECD averages from 2001.

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

953

Technology in the non-agriculture sector is calibrated based on Gollin (2002), whose estimates for the capital share of income in developing countries is in the order of 30%. The studies on agricultural technology by Manuelli and Seshadri (2004), and Soloaga (2000), suggest that the land share of income in agriculture is around 20%, and the labor share of income is between 25 and 40%. The transition between productivity states is parametrized based on Budar-Mejía and García-Verdú (2003). In their study they estimate transition probabilities between formal employment, informal employment, unemployment and out-of-the-labor force in Mexico for the period between 1994 and 2001. In the model, we interpret idiosyncratic shocks as different productivity states which can be matched to formal employment, informal employment, unemployment and out-of-the-labor force. The values of the elements of the productivity vector E are such that average productivity is equal to 1 and yield a Gini coefficient of income similar to that observed in Mexico.20 In regard to preferences, the weight on agricultural consumption () is such that the composition of the trade balance in terms of agriculture and non-agriculture goods in the model matches that of the Mexican economy in the year 2002. The subsistence level (sA ) was calculated based on WorldBank’s (2004) estimates of the food based poverty line for Mexico. For the calculation, we used the value of the subsistence basket for the year 2002 and multiplied it by Mexico’s population, and then expressed it as a proportion of GDP in that year. This is an approximation to Mexico’s share of output that was used to satisfy ‘subsistence’ consumption. For the year 2002, the estimated share of Mexico’s GDP used to satisfy subsistence was in the order of 6.7%. The equivalent figure in terms of the agriculture good after integrating across the model population is the value of sA reported in Table 1. To estimate the welfare effect of free trade we need to assume a portfolio that accounts for the exposure to changes in land prices due to liberalization. To determine the share of land in the households’ portfolio by decile, we use INEGI (2000, 2002, 2004b) household surveys of income and expenditure. Although neither asset holdings nor financial income are included in the surveys, we use the reported share of agriculture in entrepreneurial rent to calibrate the benchmark portfolio.21 3. Results In this section we report the effect of trade liberalization in the model economy on prices, production, aggregate variables (composition of output, inequality, asset holdings), and household-specific welfare. Under both the subsistence and the no subsistence scenarios, the exercise is conducted in three steps. First, without any change in fiscal or trade policy we compute the model’s steady state. We call this the no reform economy. Second, we eliminate tariffs on both goods and name the new steady state the reform economy with liberalized trade. In this step it is assumed that neither debt nor taxes change to make up for the lost tariff revenue, and thus government spending falls. In the final step, the government compensates lost revenue by increasing either capital income, labor income or consumption taxes so that the path of government expenditures (G) remains constant at the pre-reform level. Unless otherwise noted, all reported results compare the no reform versus the reform economy (i.e. the one in which the government neither increases debt nor raises other taxes to compensate for the lost tariff revenue).22 Throughout, effective labor (N) and the supply of land (L¯ ) are normalized to 1. The results on prices and production presented in the next subsection are common to both subsistence scenarios.

20 In addition to these two conditions, two additional restrictions are necessary to uniquely identify the productivity vector E. The first is to assume that productivity during unemployment and out-of-the-labor force is alike, although each state has its own transition probabilities. The second restriction is given by the ratio of earnings in the formal sector to the informal sector. Based on survey data from Mexico for the period 1994–1997, Cruz (2000) reports this ratio at approximately 3.5. 21 Like in the model, this implicitly assumes that the return on land is the same as that of other assets; and as explained before, the portfolio composition in terms of the other assets (K, B, B∗ ) need not be established as they all yield the same return. 22 In Section 3.3.2 we discuss alternative tax policies to compensate for the loss of tariff revenue.

954

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

Table 2 Relative prices.

Agriculture inputs  = PA /PO World int. rate r ∗ Wage rate w Rental rate capital r Rental rate land rL Price land pL Purchasing power w/PA r ∗ PA rL /PA

Before liberalization

After liberalization

0.619 0.035 0.986 0.135 0.026 0.753

0.583 0.035 0.986 0.135 0.019 0.555

1.591 0.057 0.043

1.692 0.060 0.033

Notes. This table reports selected relative prices before and after liberalization. Table 3 Production.

Labor NA NO Capital KA KO Land L

Before liberalization

After liberalization

0.05 0.95

0.04 0.96

0.39 2.96

0.29 3.01

1.00

1.00

Notes. This table reports the allocation of resources by sector before and after liberalization.

3.1. Prices and production Due to differential protection towards agricultural goods, the immediate effect of trade liberalization is a decline in the relative price of the agricultural good as it converges to the international relative price (recall  = (1 +  A )/(1 +  O )∗ > ∗ ). Table 2, shows that after liberalization the rental rate of capital and the wage rate do not change as they are pinned down by the international interest rate. However, due to the decline in the price of the agriculture good, capital and labor are reallocated to the non-agriculture sector (Table 3), leading to a fall in the price and in the rental rate of land.23 The intuition for the decline in the price and rental rate of land, is that since land is specific to the production of agricultural goods and less labor and capital are allocated into agriculture, the rental rate of land – determined by its marginal productivity – falls. A lower ‘dividend’ on land requires a lower price of land in order to align its return to that of the other assets. This suggests one channel through which liberalization affects the distribution of wealth and income as land owners incur a capital loss and a permanent fall in land income. This finding is summarized in the following proposition.24 Proposition 1. Let pL,t ( A ,  O ) be the price of land in period t. Suppose fA (KA,t , NA,t , Lt ) describes the production technology in the agriculture sector and it is such that the crossed-derivative of its arguments is positive. Also suppose that there is full pass-through of tariffs from international prices to domestic prices and that the economy is a net importer of agricultural goods. Then, if  A >  O ≥ 0, for any ˆ A ≥ 0 and ˆ O ≥ 0 such that  A −  O > ˆ A − ˆ O ≥ 0 then pL,t ( A ,  O ) > pL,t (ˆ A , ˆ O ). That is, the price of land decreases whenever policy shifts towards less differential protection (including liberalization).

23 This prediction is consistent with evidence from Latin America, where land prices have fallen in periods of trade liberalization (see World Bank, 2003). 24 Note that the proposition is not constrained to the stationary equilibrium.

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

955

Table 4 Selected variables (steady state). Data (2002)

GDP Agr. sector Non-agr. sector Consumption Investment Gov. spending Net exports Agr. net. exp. Non-agr. net. exp. Gov. debt Tax revenue Tariff revenue Gini coefficient

– YA /GDP YO /GDP C/GDP I/GDP G/GDP NX/GDP NXA /GDP NXO /GDP B/GDP R/GDP TR/ R Income

– .090 .910 .690 .207 .121 −.019 −.002 −.017 .213 .158 .037 .5100

No subsistence

Subsistence

Before lib.

After lib.

Before lib.

After lib.

1.00 .090 .910 .639 .229 .151 −.019 −.002 −.017 .213 .161 .018 .5083

.990 .067 .933 .649 .227 .150 −.025 −.026 .001 .215 .160 .000 .5083

1.00 .090 .910 .639 .229 .151 −.019 −.002 −.017 .213 .161 .018 .5009

.990 .067 .933 .647 .227 .150 −.024 −.022 −.002 .215 .160 .000 .5019

Notes. This table reports the long-run (steady state) values for a series of macroeconomic variables before and after liberalization under both subsistence scenarios. The table also reports the variables’ values for Mexico in 2002 that served as the basis of the calibration exercise.

Proof.

See Appendix B. 

Proposition 1 is reminiscing of the Stolper-Samuelson theorem, which positively relates the change in the relative price of a good to the relative price of the input intensively used in its production. Moreover, consistent with the distributional implications of the specific factors model (where land is specific to agriculture), the supply side of the model suggests that after liberalization land-owners are worse-off while workers and capital owners are better-off; this is so since w/pA and r ∗ /pA increase, and w/pO and r ∗ /pO do not change; while rL /pA and rL /pO decrease, in addition to the decrease in the price of land (see Table 2).25 3.2. Macroeconomic overview As a result of the calibration exercise, under both subsistence scenarios the model closely matches some features of the Mexican economy (Table 4). Also, regardless of the assumption about subsistence both economies exhibit some anticipated effects of liberalization. Namely, the consumption share of GDP increases, the government share of GDP falls (when the lost tariff revenue is not compensated for), the trade deficit increases, tax revenue falls, and the composition of the trade deficit shifts towards more imports of the agricultural good. Furthermore, in the economy with no subsistence non-agricultural net exports become positive. Also, relative to the pre-reform level, the model predicts a 1% decline in GDP (measured in units of the non-agriculture good). This occurs because after liberalization the expansion of the non-agriculture sector is not enough to compensate for the decline in the value of agricultural output, which results from less production and lower (relative) price.26 However, the main difference between the two economies is that in the long-term, without subsistence trade liberalization has no effect on wealth, income or consumption inequality—measured either by the Gini coefficient or the coefficient of variation. In contrast, when households need to satisfy a minimum consumption requirement of food, all measures of inequality increase (Table 5). The reason for such difference is that when subject to satisfying the subsistence level, the households’ spending share in agricultural goods is decreasing in wealth (Fig. 1).27 That is, poor households

25 In the Heckscher-Ohlin model, such redistribution of income would arise when the international price of the good is lower than the domestic (autarky) price. In the model, differential protection drives the wedge between these prices. 26 Even though the level of outstanding government debt is the held constant before and after liberalization, its share of GDP increases upon liberalization. The reason for this is that GDP (measured in units of the non-agriculture good) declines as agricultural production falls, and becomes less valuable as its relative price decreases. 27 In the no subsistence economy preferences are homothetic, while in the subsistence economy they are not.

956

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

Table 5 Change in long-term inequality. No subsistence

Gini coefficient Wealth Income Consumption Coefficient of variation Wealth Income Consumption

Subsistence

Before lib.

After lib.

Before lib.

After lib.

.5692 .5083 .2269

.5692 .5083 .2269

.5625 .5009 .2085

.5628 .5019 .2151

.7484 .6648 .2395

.7484 .6648 .2395

.7358 .6522 .2267

.7378 .6535 .2268

Notes. This table reports the long-run Gini coefficient and the coefficient of variation as measures of wealth, income and consumption inequality before and after liberalization under both subsistence scenarios.

Fig. 1. Consumption profiles: share of agriculture in consumption. Notes. The left (right) panel shows the ratio of agriculture consumption to total consumption for different levels of asset holdings under no subsistence (subsistence).

spend a larger fraction of their income on the agriculture good than wealthier ones.28 Hence, the higher relative price of food prior to liberalization, pushes low income households closer to subsistence levels to which they respond by self-insuring through precautionary savings. Once the price distortion is eliminated, and domestic prices converge to international prices, poor households reduce their buffer savings – as they are less likely to hit the subsistence bound – leading to a decline in aggregate asset holdings (Table 7) and to an increase in wealth and income inequality (Table 5).29 Table 6 reports the change in asset holdings by decile in the long-run upon liberalization. In the presence of subsistence, the negative relation between the share of food in spending and wealth, leads to a larger decline in asset holdings at the lower deciles as the price of food falls. While households in the bottom decile would reduce their steady state asset holdings – and therefore capital income – by 1.47%, agents in the top decile would lower their savings by 0.19%. Since the greater part of this effect concentrates on low wealth agents, Table 7 shows that at the aggregate level asset holdings only decline by 0.6%. 28 This behavior is consistent with the empirical findings by Mckenzie (2003), where the food share of expenditure of Mexican households with lower level of education – presumably poorer – is bigger than that of more educated households; and by Nicita (2004), who finds that 50% of the consumption of poor households in Mexico is on food. 29 Nicita (2004) reports that in Mexico in the 1990s, the relative price of non-animal agricultural products declined after the integration with the U.S. and Canada through the NAFTA.

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

957

Table 6 Saving: change in asset holdings (% of no-reform). Decile range

No subsistence Subsistence

0–1

1–2

2–3

3–4

4–5

5–6

6–7

7–8

8–9

9–10

0.00 −1.47

0.00 −1.22

0.00 −0.99

0.00 −0.84

0.00 −0.73

0.00 −0.65

0.00 −0.57

0.00 −0.49

0.00 −0.38

0.00 −0.19

Notes. This table reports the long-term change in asset holdings by decile, relative to pre-liberalization levels for both subsistence scenarios.

In contrast, in the model with no subsistence, the fraction of consumption that corresponds to the agricultural good is constant for every level of wealth (Fig. 1). This arises because under no subsistence, households’ preferences are homothetic and agents always wish to consume a constant fraction of their income on each good regardless of the relative price. Hence, the long-run distribution of wealth is preserved since no technological factor is affected after liberalization. Furthermore, in terms of the economy’s asset structure, in the long-run households make up for the decline in the capital stock and for the fall in the value of land by holding more foreign bonds to restore asset holdings to pre-reform level (Table 7). 3.3. Welfare From the household perspective, upon liberalization there are positive and negative news. The model predicts that free trade would lead to a fall in the relative price of food, but also to a reduction in the value of land. Hence the welfare change associated to freer trade is given by the interaction of both effects, which is in turn determined by the household’s characteristics. Households that do not hold land are expected to achieve the larger gains since they are safe from the capital loss and face a lower price of the agricultural good. On the other hand, a rich household whose wealth is only held as land would experience large capital loses. Based on the above, the welfare analysis draws on the following dimensions: household heterogeneity (productivity and wealth), share of land in portfolio, and whether households are subject to subsistence in the consumption of the agricultural good. The subsection also explores the welfare implications of raising either the capital income, the labor income, or the consumption tax rates to compensate for the loss of tariff revenue after liberalization. 3.3.1. Household heterogeneity A valuable insight from the model is that it allows to quantify the welfare implications from liberalization on households with different characteristics. The starting point for such differentiation is based on the households’ state: productivity and wealth. By simulating the utility paths of households who differ in productivity and in wealth, one can determine who wins and who loses from trade liberalization. The selection of households consists of households with ‘zero’ wealth, bottom decile wealth, between 1st and 2nd decile wealth, median wealth, between 8th and 9th decile wealth, and top wealth decile for each of the four levels of productivity. Table 7 Asset holdings. No subsistence

K L (value) B B∗ Total assets

Subsistence

Before lib.

After lib.

Before lib.

After lib.

3.35 0.75 0.31 0.79 5.21

3.29 0.55 0.31 1.05 5.21

3.35 0.75 0.31 0.78 5.20

3.29 0.55 0.31 1.01 5.17

Notes. This table reports the long-run composition of asset holdings before and after liberalization under both subsistence scenarios.

958

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

Table 8 Welfare and income changes from liberalization (% of no-reform co |% no-reform income). Wealth

No subsistence Zero 1st decile Btw. 1st and 2nd decile Median Btw. 8th and 9th decile 9th decile Subsistence Zero 1st decile Btw. 1st and 2nd decile Median Btw. 8th and 9th 9th decile

Productivity ε1

ε2

ε3

ε4

1.03|– 0.69|– 0.59|– 0.67|– 0.75|– 0.57|–

1.03|– 0.57|– 0.48|– 0.62|– 0.73|– 0.54|–

1.03|– 0.55|– 0.47|– 0.62|– 0.73|– 0.54|–

1.03|– 0.53|– 0.45|– 0.61|– 0.73|– 0.54|–

1.22|– 0.77| −0.47 0.65|−1.10 0.66|−1.85 0.62|−0.95 0.34|−0.51

1.40|– 0.75| −1.66 0.64|−3.35 0.68|−3.48 0.64|−1.26 0.35|−0.64

1.42|– 0.74| −1.86 0.63|−3.67 0.68|−3.61 0.64|−1.28 0.35|−0.64

1.45|– 0.74| −1.86 0.63|−3.67 0.69|−3.61 0.64|−1.28 0.35|−0.64

Notes. This table reports the equivalent permanent change in non-agriculture consumption and the long-term change in household income from liberalization for different wealth–productivity combinations under both subsistence scenarios. For every wealth–productivity pair we simulated the ‘life’ of 5000 agents over a 500 period horizon, and then computed the change in non-agricultural consumption under no-reform that would render the households indifferent between liberalization and the status quo. The change in household income represents the percentage difference between the post- and pre-reform capital and labor income implied by the corresponding stationary distributions. In the model without subsistence there is no change in income since the long-run distribution of income and wealth do not change.

For this exercise we simulated the ‘life’ of 5000 agents over a 500 period horizon for every chosen wealth–productivity combination (x0 ). To carry out the analysis we assumed the survey-based portfolio described in Section 2.7 for each of the selected households.30 For the welfare evaluation we conduct a form of an equivalent variation that tells us by how much should non-agriculture consumption change in the no reform economy to render the household indifferent between liberalization and the status quo. 

NR − s ) [(1 + )c NR ] For this purpose, the no reform utility function becomes u(·) = {(cA,t x0 O,t A

1− 1−

}

/(1 −

). 31 Measured as the lifetime percentage change in non-agricultural consumption, Table 8 shows the welfare effect that the selected household would experience after liberalization ( A =  O = 0). The model predicts that the largest gains would accrue on households with ‘zero’ wealth; since by definition do not experience any capital loss. Another insight is that the welfare gain is decreasing in wealth. This suggests that as wealth increases, the benefit from the fall in the relative price of food decreases as the capital loss associated with the decline in the price of land increases. For example, in the model with no subsistence the gain for a household with high productivity (ε1 ) and ‘zero’ wealth is in the order of 1.03%; the estimated welfare gain for a household with the same productivity but in the top decile is 0.57%. In the economy with subsistence, the corresponding changes are 1.22 and 0.34%. The analysis also shows that in the model without subsistence the welfare change is weakly increasing in productivity. Since productivity is positively related to labor income, the latter acts as a buffer on the capital loss yielding a larger gain to those with higher productivity, who are in a higher income path. For instance, the welfare gain for a household with median wealth and high productivity (ε1 ) is approximately 0.67%; while the gain for a household with the same wealth but low productivity (ε4 ) is 0.61%.

30 Section 3.3.3 presents the results of a separate exercise in which the selected households only hold land or no land at all, rather than the diversified portfolio. 31 We base the welfare evaluation in terms of the non-agriculture good for the following reasons. First, in the model the nonagriculture good serves as the numeraire; second, it is the good in which households spend most of their income (see Fig. 1); and third, it allows to compare the subsistence and no subsistence scenarios without any interaction with the subsistence level (sA ).

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

959

In the subsistence scenario the relation between the welfare change and productivity is nonmonotonic. For ‘zero’ wealth households, which do not experience any capital loss, the welfare gain is decreasing in productivity. Those with low productivity gain the most from a lower price of food as they are closer to subsistence, and spend a larger share on food. Next, we find that since households in the bottom of the distribution exhibit the highest proportion of land in their portfolios, they experience the largest capital loss relative to their wealth. Hence, those with the better prospects of rebuilding their wealth (i.e. those with high productivity) gain the most. Those with lower earnings prospects, gain less as the capital loss pushes them closer to subsistence. Finally, for those in the upper half of the distribution, who are portfolio-wise less exposed to the capital loss, the benefits from the lower price of food are decreasing in productivity. That is, liberalization benefits more those with low productivity as they spend a larger share of their income on the agricultural good. One difference between the subsistence scenarios is that for households with ‘zero’ wealth, without subsistence, the gain from the lower relative price of food is independent of productivity; while with subsistence the gain is decreasing in productivity. The reason for this is that, as established before, in the model with subsistence the effect liberalization is larger on those who spend proportionately more on food, while with no subsistence is proportional for all levels of productivity, as food’s share of spending is independent of income or wealth. Intuitively, if there is no capital loss, under no subsistence households go through the same income path under the pre- and post-reform economies (conditional on their productivity) and the welfare measure only represents the change from one relative price of food to another. In this case, movements in the relative price of food only change the households’ static decision without affecting the consumption-saving decision. With subsistence, however, larger earnings imply: (1) that the household is further from subsistence consumption and therefore experiences a smaller gain from the fall in the price of the agricultural good; and (2) larger intertemporal elasticity of substitution thus affecting the consumption-saving decision.32 For example, when subject to subsistence the approximate welfare gain for a household with ‘zero’ wealth and low productivity (ε4 ) is 1.45%; while the welfare gain for a household with ‘zero’ wealth and high productivity (ε1 ), thus further from the subsistence bound, is 1.22%. With no subsistence, ‘zero’ wealth households would approximately experience a 1.03% gain regardless of their productivity. To capture the effect of liberalization on both welfare and inequality, Table 8 also reports the long-term change in household income associated to lower precautionary savings in the model with subsistence. In the model without subsistence there is no change in income since the long-run distribution of income and wealth do not change; as lower food prices only change the households’ intra-temporal consumption mix. With subsistence, lower food prices also lead to a decrease precautionary savings, which underlie the increase in inequality. The table reports the long-run change in income (sum of capital and labor income) from liberalization for the selected wealth–productivity combinations. Consistent with the decline in precautionary savings (see Table 6), the drop in income is larger at the bottom of the distribution. Across the selected wealth deciles the drop in income is decreasing in productivity (recall, ε1 > ε4 ) since for a given wealth decile the fall in capital income is constant for all productivity levels, while labor income is increasing in productivity. Hence, low productivity workers experience a larger drop in total income as capital income represents relatively larger share of their income. 3.3.2. Revenue-neutral liberalization The analysis thus far suggests that the welfare effects from liberalization vary according to the households’ characteristics. Nonetheless, by assigning equal weight on every household, we can estimate the average welfare change from an economy-wide perspective.33 Table 9a shows that the average

32 Under no subsistence, preferences are homothetic and the intertemporal elasticity of substitution is constant. In the presence of subsistence, preferences are no longer homothetic and the intertemporal elasticity of substitution is increasing in wealth (see Alvarez-Peláez and Díaz, 2005). 33 To compute the average welfare change from liberalization, the ‘life’ of 50,000 households was simulated for 500 periods. The households were distributed over productivity–wealth space based on the pre-reform stationary distribution, and the share of land in their portfolio (by decile) followed the benchmark calibration.

960

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

Table 9a Average welfare change from alternative policies (% of no-reform co ).

No subsistence Subsistence

Liberalization

Revenue compensating tax rate

( A ,  O ) = (0, 0)

 k = 12.5% (13%)

 n = 14.7%

 c = 8.9%

0.80 1.00

0.09 −5.17

0.20 −5.07

0.27 −4.98

Notes. This table reports the average equivalent permanent change in non-agriculture consumption for different fiscal regimes. The first column reports the welfare change from liberalization if the government does not compensate the loss in tariff revenue. The second, third and fourth columns report the welfare change from liberalization and the required tax rate on capital income, labor income and consumption when the government raises either of these taxes to compensate the revenue loss from liberalization. With subsistence, the revenue compensating tax rate on capital is 13%.

welfare gain from liberalization is equivalent to a permanent 0.8% increase in non-agricultural consumption in the model without subsistence, and to a 1% increase in the model with subsistence. This difference is mainly explained by the larger gains from poor households in the subsistence scenario relative to comparable households in the no subsistence economy, and by both the distribution of wealth and income being positively skewed. If the government has to maintain a constant path of expenditures (G), and the fall in tariff revenue must be compensated by increasing either consumption taxes ( c ), labor income taxes ( n ), or capital income taxes ( k ), model simulations suggest that for either of the subsistence scenarios consumption taxes appear as the preferred instrument.34 However, while in the no subsistence economy replacing tariff revenue with higher consumption taxes leads to an average welfare gain of 0.27%, it would imply a 4.98% average welfare loss in the model with subsistence. The opposite welfare effect from revenue-neutral liberalization between subsistence scenarios is due to the interaction of the income and substitution effects from lower food prices, and the ‘wealth’ effect associated to higher taxes (of either form) that resource-wise pushes households closer to the subsistence bound. In the no subsistence economy the income and substitution effects dominate; while in the subsistence economy the wealth effect prevails. The negative welfare effect of liberalization under subsistence is consistent with previous studies that show that liberalization combined with a revenue-neutral rise in consumption taxes could be welfare reducing. In particular, revenue-neutral higher consumption taxes can reduce welfare when one allows for non-traded goods (Anderson, 1996; Keen & Ligthart, 2002), intermediate goods and imperfect competition (Keen and Ligthart, 2002), or a large informal sector (Emran and Stiglitz, 2005). Together with the latter reference, the distinct effect of liberalization on welfare under each subsistence scenario strengthens the case for careful revenue substitution prescriptions for developing countries as they likely to be better represented by the subsistence economy. Even though consumption taxes are on average the preferred instrument to maintain the prereform stream of government spending, intuitively one can expect different households to favor one instrument or another based on their wealth–productivity state. For instance poor households would rather have the burden on the rich, and thus favor capital income taxes; while wealthy households might favor labor income taxes. Based on the set of previously selected households, Table 9b reveals that higher consumption taxes are not always preferred for compensating the lost tariff revenue. For instance, ‘zero’ wealth households and those on the first wealth decile – except the ones with high productivity who have a positive savings outlook – would favor raising capital income taxes; median wealth households favor consumption taxes; and wealthier households would prefer higher labor income taxes. Because of the different distributional burden, households’ characteristics determine their preferred instrument when faced with alternative tax regimes. As one moves along the distribution of wealth, the preferred instrument shifts from taxes on capital income, to consumption taxes, to taxes on labor income in both subsistence and no subsistence scenarios. However, as noted for the average

34 Previous studies with heterogeneous households have found consumption taxes as the preferred instrument when evaluating tax reform (e.g. Domeij and Heathcote, 2004).

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

961

Table 9b Welfare change for revenue-neutral liberalization (% of no-reform co ). Wealth

Productivity ε1 Zero 1st decile Btw. 1st and 2nd decile Median Btw. 8th and 9th decile 9th decile Productivity ε2 Zero 1st decile Btw. 1st and 2nd decile Median Btw. 8th and 9th decile 9th decile Productivity ε3 Zero 1st decile Btw 1st and 2nd decile Median Btw. 8th and 9th decile 9th decile Productivity ε4 Zero 1st decile Btw. 1st and 2nd decile Median Btw. 8th and 9th decile 9th decile

No subsistence

Subsistence

 k = 12.5%

 n = 14.7%

0.42 0.03 −0.17 −0.35 −0.83 −1.51

0.37 0.05 −0.03 0.11 0.29 0.17

0.58 0.07 −0.10 −0.22 −0.68 −1.40

 c = 8.9%

 k = 13%

 n = 14.7%

 c = 8.9%

0.50 0.17 0.06 0.14 0.22 0.04

−4.85 −5.35 −5.56 −5.81 −6.39 −7.13

−4.93 −5.33 −5.41 −5.33 −5.28 −5.47

−4.77 −5.20 −5.30 −5.28 −5.31 −5.57

0.37 −0.06 −0.12 0.08 0.28 0.16

0.50 0.04 −0.04 0.10 0.20 0.01

−4.52 −5.20 −5.39 −5.62 −6.21 −7.00

−4.77 −5.35 −5.42 −5.31 −5.25 −5.45

−4.61 −5.22 −5.32 −5.27 −5.30 −5.57

0.58 0.05 −0.10 −0.21 −0.67 −1.39

0.37 −0.08 −0.13 0.08 0.28 0.16

0.50 0.02 −0.06 0.09 0.20 0.01

−4.51 −5.20 −5.39 −5.61 −6.21 −7.00

−4.76 −5.35 −5.42 −5.31 −5.25 −5.45

−4.60 −5.23 −5.32 −5.27 −5.30 −5.57

0.60 0.05 −0.10 −0.19 −0.65 −1.37

0.37 −0.10 -0.15 0.07 0.28 0.15

0.50 0.00 −0.08 0.08 0.20 0.01

−4.46 −5.18 −5.37 −5.58 −6.18 −6.98

−4.73 −5.36 −5.42 −5.30 −5.24 −5.45

−4.57 −5.23 −5.33 −5.27 −5.30 −5.57

Notes. This table reports the welfare change from liberalization and the required tax rate on capital income, labor income and consumption when the government raises either of these taxes to compensate the forgone tariff revenue. For every wealth–productivity pair we simulated the ‘life’ of 5000 agents over a 500 period horizon, and then computed the change in non-agricultural consumption under no-reform that would render the households indifferent between revenue-neutral liberalization and the status quo.

welfare effects, under subsistence any revenue-neutral choice leads to a negative welfare change for all households. 3.3.3. Welfare bounds As previously argued, the extent to which a household benefits form liberalization depends on its exposure to the fall in the price of land. Households that do not hold land will be better-off from the lower price of food and safe from the capital loss. In contrast, a household whose portfolio is only made up by land would experience a 26% capital loss (Table 2). Table 10, presents the upper and lower bounds that a particular household would experience from liberalization depending on the share of land in their portfolio. For households that do not hold land, the welfare gain represents the effect of the lower relative price of the agriculture good over their lifetime and marks the upper bound. The model suggests that such gain is independent of human and financial wealth in the model without subsistence, and is approximately equivalent to a 1.03% permanent increase in the consumption of the non-agriculture good. With subsistence, the gain is decreasing in wealth and ranges from 0.79 to 1.33%.35 When a household’s financial wealth is only held as land, liberalization can lead to significant welfare loses. For these households the positive effect of a lower price of the agricultural good 35 Since under subsistence the share of the agriculture good in spending is decreasing in wealth, poor households benefit the most from the lower price of food (see Section 3.3.1).

962

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

Table 10 Welfare bounds: share of land in portfolio (% of no-reform co ). Wealth

Productivity ε1 1st decile Btw. 1st and 2nd decile Median Btw. 8th and 9th decile 9th decile Productivity ε2 1st decile Btw. 1st and 2nd decile Median Btw. 8th and 9th decile 9th decile Productivity ε3 1st decile Btw 1st and 2nd decile Median Btw. 8th and 9th decile 9th decile Productivity ε4 1st decile Btw. 1st and 2nd decile Median Btw. 8th and 9th decile 9th decile

No subsistence

Subsistence

Only land

No land

Only land

No land

0.06 −1.11 −3.77 −8.03 −11.16

1.03 1.03 1.03 1.03 1.03

0.01 −1.19 −3.86 −7.90 −10.94

1.18 1.13 1.03 0.89 0.79

−0.30 −1.64 −4.43 −8.72 −11.84

1.03 1.03 1.03 1.03 1.03

-0.31 −1.68 −4.49 −8.57 −11.60

1.31 1.23 1.10 0.93 0.82

−0.37 −1.71 −4.49 −8.78 −11.89

1.03 1.03 1.03 1.03 1.03

−0.34 −1.71 −4.52 −8.59 −11.63

1.31 1.24 1.10 0.93 0.82

−0.44 −1.81 −4.61 −8.90 −12.01

1.03 1.03 1.03 1.03 1.03

−0.39 −1.79 −4.62 −8.70 −11.73

1.33 1.26 1.12 0.94 0.82

Notes. This table reports the bounds on the equivalent permanent change in non-agriculture consumption from liberalization for different wealth–productivity combinations under both subsistence scenarios. For every wealth–productivity pair we simulated the ‘life’ of 5000 agents over a 500 period horizon and assumed that either all their wealth was held as land (Only land) or that they held no land at all (No land). Then, we computed the change in non-agricultural consumption under no-reform that would render the households indifferent between liberalization and the status quo.

is outweighed by the capital loss. Under both subsistence scenarios, the welfare loss is decreasing in productivity since more productive households are on higher income paths. Nonetheless, a high productivity household in the top wealth decile would experience a welfare loss of approximately 11.16% of non-agricultural consumption under no subsistence, and 10.94% when subject to subsistence. Depending on household productivity, wealth, and share of land in the household’s portfolio, the welfare change from liberalization in the model with no subsistence lies between −12.01 and 1.03%; and in the model with subsistence the welfare change lies between −11.73 and 1.33%. 3.4. Dynamics of inequality So far it has been argued that liberalization implies: (1) a reduction in the price of food, which is particularly beneficial in the model with subsistence, and (2) a capital loss associated with land tenure. The greater of these effects determines in turn whether a household would support or not opening to trade. This subsection analyzes the impact of the capital loss on inequality at the time of reform and its posterior evolution; that is, the transition between steady states. The model suggests that in the long-run under no subsistence inequality is unchanged, while under subsistence it increases. Under the benchmark calibration, on impact, trade liberalization is associated with a 1.2% increase in income inequality. This is so because the incidence of the capital loss associated with the fall in the price of land is larger on low wealth households, who hold a larger share of land in their portfolios. Since wealthier households hold proportionately less land are relatively unaffected, hence overall inequality increases. Given that wages and asset returns are unchanged after liberalization, as they

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

963

Fig. 2. Dynamics of income inequality: change in coefficient of variation after liberalization. Notes. The left (right) panel shows the percentage change in the coefficient of variation of income in the short-term, medium-term and long-term after liberalization under no subsistence (subsistence). The ‘benchmark portfolio’ is survey-based, while the ‘common portfolio’ assumes that all households hold the same proportion of land in their portfolios.

are pinned down by the international interest rate, larger wealth dispersion implies larger income dispersion.36 This way, we know the starting and finishing point of income inequality once reform takes place. What happens in between is plotted in Fig. 2, which shows the change in the coefficient of variation during the transition between steady states. Starting from the no reform stationary equilibrium, suppose that in both scenarios the government announces an unexpected permanent change in trade policy that liberalizes the imports of both types of goods at t = 0. At that moment income inequality increases as the value of land falls, but as households adjust their behavior to the new policy regime and start to build back their savings inequality decreases. Under no subsistence, the coefficient of variation eventually returns to the pre-reform level. However, under subsistence, the coefficient of variation is larger than the pre-reform level as households reduce their steady state asset holdings, as they are less likely to hit the subsistence floor once the internal price of food falls. Lower buffer savings of the poor could lead to larger inequality in the long-run. In the short-term, as argued, the effect of liberalization on inequality is determined by the households’ exposure to changes in the price of land. If alternatively, all households held a common share of land in their portfolio, a proportional capital loss associated with the fall in the price of land would imply that wealthier households experience (absolute) larger reductions in asset holdings than the poor; thus shifting the distribution of wealth leftward and reducing inequality. The transition path under such assumption is represented by the dashed line in Fig. 2, which converges to the long-term distribution of income from below. 4. Empirical validation In this section we verify the extent to which the model’s predictions are supported by evidence from a panel of countries, while controlling for the model’s environment. In particular, we focus on the effect of the relative tariff on agricultural goods on the income and savings of agents at different points in the distribution of income. The model predicts a positive relation

36 The increase in income inequality predicted by the model is qualitatively consistent with estimates for Mexico from Dollar and Kraay (2001) and the World Bank (2004). The World Bank reports a 1.8% increase in income inequality between 1992 and 2000.

964

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

between relative tariffs and (capital) income, as particularly low income households increase their buffer savings to insure against subsistence consumption. Hence, we would expect a positive effect of relative tariffs on income and savings at the bottom of the distribution, particularly so in low income countries that are net importers of agricultural goods. To map the model into the data, we control for income per capita, for net imports of agricultural goods, for country characteristics, and for relative tariffs. The dataset draws on the UN-WIDER (2007) World Income Inequality Database; tariffs are compiled from the World Bank’s Trade and Production Database and from Feenstra, Lipsey, and Bowen (1997) World Trade Flows 1970–1992; the composition of trade flows are taken from the World Bank’s World Development Indicators; and real GDP per capita and the consumption share of real GDP per capita at purchasing power parity are taken from Heston, Summers, and Aten (2002). The dataset is described in detail in Appendix C. To compute the mean income and mean consumption by decile within each country, we follow Dollar and Kraay (2002), so that inter-decile mean income (consumption) is given by the decile’s share of income (consumption) multiplied by mean income (consumption), divided by 0.1. Then, our proxy for inter-decile savings is given by the difference between log income and log consumption.37 Based on the above, the econometric specification for quantifying the effect of trade protection on inter-decile log income (yd) and on inter-decile log savings (sd) is given by p

nif

mi = ˇ0 + ˇ1 yi + ˇ2 i + ˇ3 i i + ˇ4 i i + Xi + i ,

(1)

where mi = [yd1,i yd1−2,i . . . yd9,i sd1,i sd1−2,i . . . sd9,i ]. In the specification above, subscript i identifies the country; m represents the dependent variable being tested; y is the log of real per capita GDP;  is the (gross) relative tariff (i.e. 1 +  A /1 +  O ); p is an interaction dummy variable for low-income countries; nif is an interaction dummy variable for net importers of food; X is a vector of country-specific characteristics based on the World Bank’s country grouping classification, which captures whether the observation belongs to a low income, lower-middle income or upper-middle income country, with high income countries serving as the control group; and is a zero mean disturbance.38 To quantify the effect of trade liberalization on inequality, we estimate the parameters in Eq. (1). Once we control for income level, pattern of trade and country-specific characteristics, the estimates of ˇ2 , ˇ3 and ˇ4 measure the change on inter-decile income, inter-decile savings and inequality from increasing the (gross) relative tariff of the agricultural good. Table 11 presents the OLS estimates of the coefficients of Eq. (1), with mean inter-decile income used as the dependent variable. The estimates suggest that larger (lower) relative tariffs are significantly associated with higher (lower) income in low income countries and/or countries that are net importers of food.39 Like in the theoretical model, these effects decrease as we move up on the distribution. The low income country effect is only significant in the first decile of the distribution, and the net importer of food effect becomes insignificant from the seventh decile on. All else equal, higher (lower) relative tariffs increase (decrease) the income of the poor, thus reducing (increasing) inequality. By way of example, one could forecast the effect of liberalization on mean inter-decile income in Mexico using the regression model as follows. In a country that is a net importer of food, a 6% reduction in the relative tariff from the pre-reform level to free trade would be associated with a 1.93, 1.32, and

37 We use the suggested proxy for savings because the standard definition of saving (i.e. current disposable income minus spending) does not allow for the log of dissaving. In the sample there are several observations for which mean inter-decile consumption is larger than mean income. Dissaving can be explained by households running down their assets or by borrowing. While the former is accounted for in the model, the latter is not. 38 We use the gross relative tariff rather than the ratio of the tariffs for two reasons. First, the ratio  A / O is undetermined for a country that imposes no tariffs on its imports. And second, for consistency between predictions derived from the theoretical and empirical models. Recall that the gross relative tariff serves as a gauge of the distortion of trade policy on international prices in the theoretical model. 39 The unitary income elasticity of the income at the bottom of the distribution is consistent with Dollar and Kraay (2002) estimates.

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

965

Table 11 Distributional effects of trade policy: income (Robust standard errors in parenthesis). Log income

Intercept

ln(per cap GDP)

Rel Tariff (RT)

RT ∗ low-income dummy

RT ∗ net-imp food dummy

1st Decile Btw. 1st and 2nd Deciles Btw. 2nd and 3rd Deciles Btw. 3rd and 4th Deciles Btw. 4th and 5th Deciles Btw. 5th and 6th Deciles Btw. 6th and 7th Deciles

−3.454∗ (2.001) −2.056 (1.190)∗ −1.061 (0.847) −0.595 (0.645) −0.358 (0.511) −0.304 (0.367) −0.093 (0.258)

1.122∗∗∗ (0.181) 1.066∗∗∗ (0.113) 1.003∗∗∗ (0.081) 0.984∗∗∗ (0.062) 0.988∗∗∗ (0.050) 1.014∗∗∗ (0.035) 1.014∗∗∗ (0.026)

0.621 (0.560) 0.445 (0.300) 0.344 (0.235) 0.270 (0.184) 0.164 (0.144) 0.034 (0.103) −0.010 (0.077)

4.429∗∗ (2.088) 1.988 (1.416) 0.988 (1.110) 0.376 (0.873) 0.166 (0.700) 0.139 (0.538) −0.179 (0.406)

0.321∗∗∗ (0.082) 0.220∗∗∗ (0.054) 0.168∗∗∗ (0.043) 0.133∗∗∗ (0.034) 0.101∗∗∗ (0.028) 0.078∗∗∗ (0.021) 0.040∗∗∗ (0.015)

Notes: This table reports the estimated coefficients and robust standard errors (in parenthesis) of equation 1, for the dependent variable reported on the first column.∗∗∗ (∗∗ ) (∗ ) represent 1% (5%) (10%) significance. Observations: 209.

Fig. 3. Inter-decile saving and tariffs. Notes. This figure plots our measure of savings controlled for real per capita GDP and country-specific characteristics against the (gross) relative tariff for the first decile, and third, fifth and seventh inter-deciles. Observations corresponding to the bottom decile are represented by the thickest markers; higher inter-deciles are represented by progressively thinner markers.

1% reduction in the income of the first, second and third inter-deciles.40 In the theoretical model, the change in capital income directly corresponds to the change in savings or asset holdings; therefore, the predicted change in capital income from lower buffer savings for the first three inter-deciles from the theoretical model (Table 6) are: 1.47, 1.22, and 0.99%, respectively. These values fall within the 95% confidence interval implied by the regression estimates.41 Another way of verifying the theoretical model’s predictions is by directly determining the effect of the relative tariff on savings. Using the same dataset we approximate mean savings at the interdecile level as the difference between log mean income and log mean consumption. When conducting this test, the sample contracts to only 34 contemporaneous observations on income and consumption

40 From the the theoretical model calibration, trade liberalization would imply moving from a (1 +  A )/(1 +  O ) = 1.245/1.171 = 1.06 (gross) relative tariff to a regime with  A =  O = 0, where the (gross) relative tariff is equal to (1 +  A )/(1 +  O ) = 1. This represents a 6% reduction in the (gross) relative tariff upon liberalization. 41 In the dataset used to estimate the regression model there is no distinction between capital and labor income, therefore changes in capital income are not directly controlled for. Hence, part of the difference between the theoretical model and the regression predictions can be due to changes in labor income.

966

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

shares by decile, reducing the statistical reliability of the test. Nevertheless, Fig. 3 plots our measure of savings controlled for real per capita GDP (y) and country-specific characteristics (X), against the (gross) relative tariff () for the first decile, and third, fifth and seveth inter-deciles. Observations corresponding to the bottom decile are represented by the thickest markers; higher inter-deciles are represented by progressively thinner markers. In line with the theoretical model, the positive relation between savings and relative tariffs weakens in the upper deciles of the distribution.42 These tests suggest that low income agents in low income countries, or in countries that are net importers of food, exhibit behavior that resembles that implied by the theoretical model, in which households self-insure against subsistence consumption in the presence of higher relative tariffs. In the upper part of the distribution as the precautionary motive fades, higher relative agricultural tariffs are neutral to income and saving. 5. Conclusion Among the different channels through which trade liberalization can affect the distribution of wealth and income, we explore the effect of freer trade on the relative price of goods. Specifically, we show that while lower relative agriculture protection reduces the price of food, it is also associated with a fall in the value of land. A lower price of food can modify saving behavior in a way that leads to larger inequality in the long-term, as households reduce their buffer savings to insure against hitting a subsistence floor. However, in the short-term the impact of liberalization on inequality is primarily determined by the capital loss associated to the fall in the value of land. The higher the share of land in a household’s portfolio the larger the capital loss due to freer trade. When land tenure is more concentrated among the poor (rich) inequality increases (decreases), as a larger share of the incidence of a lower price of land falls on them. We also find that the larger welfare gains are realized at the bottom of the distribution. Based on the benchmark calibration, in spite of the fall in the price of land, there are welfare gains for all the population. In the extreme case of a household’s portfolio comprising only land, the welfare loss can be equivalent to a permanent 11.73% reduction in non-agriculture consumption. On the other hand, the largest welfare gain for a household that does not own land could be equivalent to a permanent 1.33% increase. In the absence of subsistence, trade liberalization does not have long-term distributional effects, although there significant welfare changes. Under this scenario, the welfare effect from liberalization ranges from −12.01 to 1.03% permanent change in non-agriculture consumption depending on the households’ wealth, productivity and land tenure. From an utilitarian planner’s perspective, model simulations suggest that the average gain from liberalization is 1% with subsistence and 0.8% with no subsistence; and that raising consumption taxes are the preferred instrument for compensating the loss of tariff revenue. The proposed framework reconciles the increase in inequality, the fall in the price of land, and the opposition to freer trade by small farmers, that have featured in different liberalization episodes. Furthermore, in terms of the motivation put forward in the introduction, we can answer that under the light of the model, on average, everybody gains from liberalization; that the poor are bigger winners among the winners; and that in the short-term land owners could lose. When we test these theoretical findings we find support of the model’s predictions, and hope this sheds some light into conflicting empirical evidence on the issue. Above all, one of the main findings is that increasing inequality may be the outcome of the agents’ response to a welfare enhancing policy. In future, it would be important to account for skill differences in the labor force, to explicitly model the effects of trade policy in the non-traded sector of the economy, and to validate further the model’s predictions within richer empirical frameworks.

42 Not reported OLS estimates from the small sample suggest that at the bottom of the distribution higher relative tariffs imply higher savings, particularly so in low income countries and/or in countries that are net importers of food. We find that the coefficient estimates for these variables are statistically significant at the 5% level in the first decile; and for net importers of food the effect remains significant at the 10% level within the second decile.

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

967

Acknowledgements I would like to specially thank Jonathan Heathcote, Mark Huggett, and Anna Maria Mayda for their encouragement, insights and long discussions throughout the completion of the paper. I also wish to thank the editor and two anonymous referees for very useful comments; as well as seminar and conference participants at different academic venues. Appendix A. Solution algorithm The algorithm for computing the equilibrium of the calibrated economy for a given set of tariff rates consists of three steps. 1. Solving the firm’s problem Given the tax rates, the tariff rates, the international interest rate, the international relative price of the agricultural good, and the exogenous supply of labor and land, the firm’s production plan consists of the labor and capital allocated for the production of each good. Such allocation in turn determines the wage rate, the rental rate of land and the price of land. 2. Solving the households’ problem We exploit the recursive form of the households’ problem to numerically solve it by iterating on the Euler equation. The method requires a grid on the state space (productivity and wealth combinations), and an initial guess of the derivative of the value function with respect to wealth at every grid point. Then one finds a decision rule that satisfies the Euler equation for next period wealth given current wealth and productivity, and the initial guess is updated based on the problem’s envelope condition. This step is repeated until the derivative of the value function approximately converges. When the decision rule is not an element of the grid we approximate it by linear interpolation. 3. Computing the stationary distribution43 For any level of trade protection the state of the economy is characterized by the distribution of households over the state space. As discussed in Section 2.5 , the solution to the households’ problem and the properties of the productivity process guarantee a unique stationary distribution. Furthermore, any initial distribution converges to the stationary distribution. To compute the stationary distribution we use the invariant densities over shocks and we approximate distribution functions over assets by a piecewise linear function over the wealth space. The grid for the distribution functions should be finer than that used to solve the household problem. The algorithm to compute the stationary distribution consists of two steps: (1) initializing the piecewise distribution functions (one for each productivity state), and (2) iterating the distribution functions until they approximately converge. The distribution functions are updated by identifying the source of the current mass on a given grid point based on the transition probabilities and the decision rules; that is, by determining the previous period set of states consistent with the grid point. Appendix B. Proof of proposition 1 To show that the price of land (pL ) is decreasing in differential protection the proof is structured in three steps. Step 1 shows that since liberalization leads to a decline in the relative price of the agricultural good (), labor (NA,t ) and capital (KA,t ) are reallocated from the agricultural sector to the non-agricultural sector. In step 2 it is shown that such shift translates into a decline in the marginal productivity of land, which along the lower price of food, leads to a fall in its rental rate (rL ). Step 3 shows that since in equilibrium all assets yield the same return (r ∗ ) the price of land falls (pL ) to restore the equilibrium return across assets. 1. The agriculture sector contracts

43

This description closely follows that in Ríos-Rull (1997).

968

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

Consider the following transformation to the firm’s problem. max{KA,t ;KO,t ;NA,t ;NO,t ;Lt } FA (KA,t , NA,t , Lt ) + NO,t fO (kO,t ) − w(NA,t + NO,t ) − r(KA,t + KO,t ) − rL Lt subject to KA,t , KO,t , NA,t , NO,t , Lt ≥ 0 where kO,t ≡ KO,t /NO,t . With competitive factor markets the first order conditions at an interior solution are: 

∂FA (·) = w, ∂NA,t

(B.1)



∂FA (·) = r, ∂KA,t

(B.2)



∂FA (·) = rL , ∂Lt

(B.3)

fO (kO,t ) = r,

(B.4)

fO (kO,t ) − kO,t fO (kO,t )

= w.

(B.5)

By no arbitrage the rental rate of capital should equal the international interest rate plus depreciation (r = r ∗ + ı). Hence Eq. (B.4) implies that the international interest rate (r ∗ ) uniquely pins down the capital–labor ratio in the non-agriculture sector (kO,t ). Since labor is mobile across sectors, given kO,t Eq. (B.5) determines the economy-wide wage rate (w). Since the wage rate (w) and the rental rate of capital (r) are determined by the international interest rate (r ∗ ), Eqs. (B.1) and (B.2) imply that a fall in the relative price of the agriculture good () must be offset by an increase in the marginal product of labor (∂FA (·)/∂NA,t ) and capital (∂FA (·)/∂KA,t ). Diminishing marginal productivity imply that capital (KA,t ) and labor (NA,t ) in the agriculture sector must fall, and the agriculture sector contracts. 2. The rental rate of land decreases as the internal relative price of agriculture falls Under the assumption that the crossed-derivatives of the production function in the agriculture sector are positive (i.e. ∂2 FA (·)/∂L∂KA , ∂2 FA (·)/∂L∂NA > 0), Eq. (B.3) implies that as the relative price of the agriculture good () falls, along with the use of capital (KA,t ) and labor (NA,t ) in the sector, so does the marginal productivity of land and hence its rental rate (rL ). 3. The price of land falls as the rental rate of land falls Since in equilibrium the market for land has to clear and all assets yield the same return (r ∗ ), as the rental rate of land falls (rL ) falls, the price of land (pL ) declines in order to restore the equilibrium return.



[1 + (1 −  k )r ∗ ] = 1 + (1 −  k )



rL . pL

Thus as the rental rate of land falls, so does the price of land. Appendix C. Dataset empirical validation To verify the model’s predictions, in Section 4 we report estimates of the effect of the relative tariff on income and savings by decile. The data used in the analysis combines previously used datasets compiled by UN-WIDER (2007); Feenstra et al. (1997), the World Bank, the United Nations, and Heston et al. (2002). The data on inequality is drawn from UN-WIDER (2007), World Income Inequality Database (V 2.0b) which comprises the inequality data from Dollar and Kraay (2002). The dataset includes inequality indicators such as income Gini coefficients, and income and consumption shares by decile. We compiled tariff data from the World Bank’s Trade and Production Database and from Feenstra et al. (1997). The Trade and Production Database reports manufacturing tariff rates at the 3-digit ISIC level compiled by the World Trade Organization and the United Nations; while Feenstra et al. (1997) includes 4-digit Harmonized System from the TRAINS dataset. The observations from these datasets were combined and reconciled by converting the 4-digit Harmonized System data into 3-digit ISIC

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

969

revision 2. To conduct the equivalence we took the simple average over the 4-digit Harmonized System tariff lines that comprised the 3-digit ISIC revision 2 tariff rate. Then, the relative agriculture tariff was defined as the ratio of the tariff on food (ISIC rev. 2 code 311) divided by the average of the tariffs on all other manufactured goods.44 Additional variables such as net imports of agriculture and non-agriculture goods, and Purchasing Power Parity real per capita income and consumption were taken from the World Bank’s World Development Indicators and Heston et al. (2002), respectively. The resulting dataset is an unbalanced panel consisting of 204 countries. The number of tariff rate observations is 264; however, contemporaneous tariffs and decile income shares reduce the sample to 96 observations. To increase the sample size for conducting the econometric analysis we matched tariff rates from time t − 4 to time t − 1 to time t inequality observations.45 The countries used in the analysis are: Argentina, Australia, Austria, Bangladesh, Bulgaria, Bolivia, Brazil, Canada, Switzerland, Chile, Cameroon, Colombia, Germany, Denmark, Ecuador, Ethiopia, Finland, France, Great Britain, Greece, Guatemala, Hungary, Indonesia, Ireland, Italy, South Korea, Sri Lanka, Latvia, Moldova, Mexico, Malaysia, Netherlands, Norway, Nepal, New Zealand, Peru, Philippines, Poland, Portugal, Romania, Spain, Sweden, Thailand, Trinidad and Tobago, Uruguay, United States, and Venezuela. References Adelman, I., & Robinson, S. (1989). Income distribution and development. Handbook of development economics (Vol. 2). Elsevier Science Publishers. Anderson, J. (1996). Trade reform with a government budget constraint. NBER working paper 5827. National Bureau of Economic Research. Alvarez-Peláez, M., & Díaz, A. (2005). Minimum consumption and transitional dynamics in wealth distribution. Journal of Monetary Economics, 52, 633–667. Bannister, G., & Thugge, G. (2001). International trade and poverty alleviation. IMF working paper 01/54. International Monetary Fund. Berg, A., & Krueger, A. (2003). Trade, growth, and poverty: A selective survey. IMF working paper 03/30. International Monetary Fund. Bhagwati, J., & Srinivasan, T. (2002). Trade and poverty in the poor countries. American Economic Review, 92, 180–183. Budar-Mejía, O., & García-Verdú, R. (2003). A dynamic model of formal and informal aggregate labor force participation. Mexico: Bank of Mexico. Chatterjee, S., & Ravikumar, B. (1999). Minimum consumption requirements: Theoretical and quantitative implications for growth and distribution. Macroeconomic Dynamics, 3, 482–505. Cooley, T. (1995). Frontiers of business cycle research. Princeton University Press. Cruz, Y. (2000). Transición Entre Estados Ocupacionales (1994–1997): La Importancia del Trabajo por Cuenta Propia en los Hogares Mexicanos. Gaceta de Economía, 10. Dollar, D., & Kraay, A. (2001). Trade, growth, and poverty. Development Research Group. The World Bank. Dollar, D., & Kraay, A. (2002). Growth is good for the poor. Journal of Economic Growth, 7, 195–225. Domeij, D., & Heathcote, J. (2004). On the distributional effects of reducing capital taxes. International Economic Review, 45, 523–554. Edwards, S. (1997). Trade policy, growth and income distribution. American Economic Review, 87, 205–210. Emran, M., & Stiglitz, J. (2005). On selective indirect tax reform in developing countries. Journal of Public Economics, 89, 599–623. Feenstra, R. (1989). Symmetric pass-through of tariffs and exchange rates under imperfect competition: An empirical test. Journal of International Economics, 27, 25–45. Feenstra, R., Lipsey, R., & Bowen, H. (1997). World trade flows 1970–1992 with production and tariff data. NBer working paper 5910. National Bureau of Economic Research. Feliciano, Z. (2001). Workers and trade liberalization: The impact of trade reforms in Mexico on wages and employment. Industrial and Labor Relations Review, 55, 95–115. Flinn, C. (2002). Labour market structure and inequality: A comparison of Italy and the U.S. Review of Economic Studies, 69, 611–645. Goldberg, P., & Pavcnik, N. (2004). Trade, inequality, and poverty: What do we know? Evidence from recent trade liberalization episodes in developing countries. Brookings Trade Forum, 223–269. Gollin, D. (2002). Getting income shares right. Journal of Political Economy, 110, 458–474.

44 The larger dataset is the World Bank’s Trade and Production Database, which only reports tariffs from the manufacturing sector, which includes processed food among all other manufactured goods. The 3-digit ISIC codes for manufacturing industries range from ISIC 311 (processed food) to ISIC 390. 45 When more than one past tariffs observations were available (for example, t − 1 and t − 2) the closest to time t was selected. Also, for years in which more than one tariff rate was reported in the World Bank’s Trade and Production Data Base, the UN tariff was selected over the WTO, since the most observations (209) in the dataset were from the UN relative to the WTO (57).

970

L. San Vicente Portes / The Quarterly Review of Economics and Finance 49 (2009) 944–970

Heston, A., Summers, R., & Aten, B. (2002). Penn world table version 6.1. Center for International Comparisons at the University of Pennsylvania (CICUP). Hopenhayn, H., & Prescott, E. (1992). Stochastic monotonicity and stationary distributions for dynamic economies. Econometrica, 60, 1387–1406. Huggett, M. (1993). The risk free rate in heterogeneous-agent incomplete-insurance economies. Journal of Economic Dynamics and Control, 17, 953–969. INEGI. (2000). Encuesta Nacional de Ingresos y Gasto de los Hogares. Mexico: Instituto Nacional de Estadística Geografía e Inform ática. INEGI. (2002). Encuesta Nacional de Ingresos y Gasto de los Hogares. Mexico: Instituto Nacional de Estadística Geografía e Inform ática. INEGI. (2004a). Sistema de Cuentas Nacionales de México: Cuenta de Bienes y Servicios (Vol. I). Mexico: Instituto Nacional de Estadística Geografía e Informática. INEGI. (2004b). Encuesta Nacional de Ingresos y Gasto de los Hogares. Mexico: Instituto Nacional de Estadística Geografía e Informática. Jank, M., Fuchsloch, I., & Kutas, G. (2001). Agricultural liberalization in multilateral and regional trade negotiations. Inter-American Development Bank. Keen, M., & Ligthart, J. (2002). Coordinating tariff reduction and domestic tax reform. Journal of International Economics, 56, 489–507. Krebs, T., Krishna, P., & Maloney, W. (2005). Trade policy, income risk, and welfare. NBER working paper 11255. National Bureau of Economic Research. Krusell, P., Ohanian, L., Ríos-Rull, J. V., & Violante, G. (2000). Capital–skill complementarity and inequality: A macroeconomic analysis. Econometrica, 68, 1029–1053. Manuelli, R., & Seshadri, A. (2004). Frictionless technology diffusion: The case of tractors. Madison: Mimeo, University of Wisconsin. Mckenzie, D. (2002). Are tortillas a Giffen good in mexico? Economics Bulletin, 15, 1–7. Mckenzie, D. (2003). The consumer response to the Mexican Peso Crisis. Mimeo: Stanford University. Mendoza, E., Razin, A., & Tesar, L. (1994). Effective tax rates in macroeconomics: Cross-country estimates of tax rates on factor income and consumption. Journal of Monetary Economics, 34, 297–323. Nicita, A. (2004). Who benefited from trade liberalization in Mexico? Measuring the effects on household welfare. World Bank policy research working paper 3265. The World Bank. Obiols-Homs, F., & Urrutia, C. (2005). Transitional dynamics and the distribution of assets. Economic Theory, 25, 381–400. Panday, P. (2003). An empirical analysis of the determinants of earning and employment. Journal of Economic Development, 28, 35–47. Ríos-Rull, J. (1995). Models with heterogeneous agents. In Thomas Cooley, frontier of business cycle research. Princeton University Press. Ríos-Rull, J. (1997). Computation of equilibria in heterogeneous agent models. Lecture Notes. Univerisity of Pennsylvania. Soloaga, I. (2000). The treatment of non-essential inputs in a Cobb-Douglas technology: An application to Mexican rural household level data. Policy research working paper 2499. The World Bank. UN-WIDER. (2007). World income inequality database (V 2.0b). World Institute for Development Economics Research. Winters, A., McCulloch, N., & McKay, A. (2002). Trade liberalization and poverty: The empirical evidence. Discussion paper 88. University of Sussex. World Bank. (2003). Land policies for growth and poverty reduction. The World Bank/Oxford University Press. World Bank. (2004). Poverty in Mexico: An assessment of conditions, trends and government strategy. The World Bank. World Trade Organization. (2004). Country profiles. Mexico: Online Database, World Trade Organization.