Welfare Implications of Trade Liberalization and Fiscal Reform: A Quantitative Experiment

Transcription

1 Welfare Implications of Trade Liberalization and Fiscal Reform: A Quantitative Experiment Sunghyun H. Kim and M. Ayhan Kose October 23 Abstract This paper studies the welfare implications of revenue-neutral trade liberalization and fiscal reform programs for developing economies using a multi-sector dynamic general equilibrium model of a small open economy. We analyze how different combinations of tariffs on imported consumption goods, intermediate inputs, and capital goods and taxes on consumption, labor income and capital income affect the transitional and long-run welfare. We report three main findings. First, trade liberalization programs financed by consumption and labor income taxes tend to result in substantial welfare gains, but financing the lost tariff revenue through capital income taxes can have an adverse impact on welfare. Second, a significant fraction of welfare changes is due to transitional effects stemming from the allocation of resources in response to changes in tariffs and taxes. Third, trade liberalization and fiscal reform programs often translate into much larger welfare gains in countries that are more open to international financial markets. JEL classification: F3; F36; F43; H2; H2; H5. Key Words: tariff; tax; revenue neutral; financial liberalization; dynamic model. We thank Tim Kehoe, Gajendran Raveendranathan, Jack Rossbach, and two anonymous referees for their helpful comments that significantly improved the paper. We also thank the seminar participants at Boston University, Clark University, Suffolk University, the AEA Meetings, the Econometric Society Meetings, and Economics Alumni Conference at the University of Iowa for their suggestions. Ezgi Ozturk provided excellent research assistance. This paper was supported by Faculty Research Fund, Sungkyunkwan University, 22. The views expressed in this paper are those of the authors and do not necessarily represent those of the IMF or IMF policy. Corresponding Author. Department of Economics, Sungkyunkwan University, Seoul, Korea and Department of Economics, Suffolk University, Boston, MA. shenrykim@skku.edu. Research Department, International Monetary Fund; akose@imf.org.

2 Introduction Trade liberalization has been a central component of economic reform programs in a number of developing economies since the mid 98s. Trade liberalization is theoretically associated with better allocation of resources and improved growth prospects, but its implementation presents serious fiscal challenges to many of these countries where import tariffs often constitute a major source of government revenue. 2 Tariffs are relatively easy to collect whereas it is costly to expand fiscal revenues through taxation of domestic resources because developing countries often lack the necessary capacity to effectively monitor, administer, and collect taxes. Recognizing their severe budgetary consequences, trade liberalization programs have often been complemented with fiscal reform initiatives. 3 In light of these observations, we ask a fundamental question: What are the welfare implications of trade liberalization and fiscal reform programs in developing countries? In order to answer this question, we undertake a quantitative experiment and study the welfare effects of revenue-neutral changes in policies involving a rich menu of tariffs and taxes. Specifically, we employ a multisector dynamic general equilibrium model of a small open economy to evaluate the transitional and long-run changes in welfare in response to adjustments in tariffs (on imported consumption goods, intermediate inputs and capital goods), and taxes (on consumption, labor income and capital income). We report three main results. First, revenue-neutral trade liberalization and fiscal reform programs can lead to sizeable welfare gains depending on changes in taxes and tariffs. A fullfledged trade liberalization, i.e., elimination of all tariffs, is associated with welfare gains of up to 2.8 percent of lifetime consumption when the lost tariff revenue is financed by a consumption or labor income tax. In contrast, if taxes on capital income are increased to compensate the lost tariff revenue, this can translate into smaller welfare gains or outright welfare losses. These findings are intuitively appealing as they emphasize the magnitude of the dynamic efficiency gains stemming from capital accumulation over time. Second, the welfare implications of various types of trade liberalization and fiscal reform programs we analyze indicate that financing through capital income taxes is the least preferred fiscal tool to recover the lost tariff revenues. In contrast, financing through consumption taxes is the best fiscal policy tool. Irrespective of taxes used to finance lost tariff revenues, the elimination of tariffs on imported factors of production, i.e., capital goods and intermediate inputs, results in the largest welfare gains implying that it is the most preferred dimension of trade liberalization. The removal of tariffs on imported consumption goods generates the smallest welfare gains irrespective of the type of financing employed. These results are robust to a wide range of sensitivity experiments. Third, the welfare implications of liberalization and reform programs depend on a country s degree of access to international financial markets. We find that trade liberalization and fiscal reform The wave of unilateral trade liberalizations started with the Uruguay Round in 986 and then continued with the proliferation of preferential trade agreements (Kose and Prasad, 2). The fraction of countries with a liberalized trade regime increased from roughly 3 percent in 985 to about 7 percent in 28. The number of preferential trade agreements has skyrocketed over the same period going up to roughly 7 from just. 2 Tariff revenues account for more than 25 percent of total tax revenue in many low-income countries (Kubota, 999 and World Bank, 29). In contrast, only a minor fraction of tax revenues in the core OECD countries is due to tariffs (IMF, 29). 3 Caprio, et al. (998), Ebrill, et al. (999), ATPC (24), and Bilal, Dalleau and Lui (22) discuss experiences of countries that have difficulties in implementing the joint trade liberalization and fiscal reform programs. 2

3 programs result in larger welfare gains in economies that have a higher degree of access to international financial markets. These results collectively emphasize the importance of complementarities between trade and financial integration in the context of the liberalization and reform programs. Despite the rigorous policy debate on the welfare implications of trade liberalization and fiscal reform programs, the literature has yet to study these issues employing modern quantitative experiments in the context of rich dynamic general equilibrium models. The welfare implications of tax policies have been studied extensively, but there have been only a few papers analyzing the joint implications of tax and tariff policies. Moreover, a handful of previous studies have examined the macroeconomic effects of such liberalization and reform programs using mostly static models and simple empirical methods. 4 Our multi-sector dynamic general equilibrium model allows us to examine a number of critical factors the earlier studies have not been able to account for. First, the dynamic nature of our model allows us to analyze the intertemporal effects associated with the dynamics of physical and financial assets in response to trade liberalization and fiscal reform programs. Second, we are able to evaluate the welfare changes stemming from the transitional dynamics in addition to those associated with the pre- and post-reform equilibria. Models with static environments can analyze only the welfare changes between the pre- and post-reform equilibria. Third, our small open economy model has a rich production structure as it imports capital goods and intermediate inputs, and employs them to produce export goods. Given the rapid growth of global manufacturing chains involving the trade of intermediate inputs and capital goods across borders, this is another crucial factor necessary to accurately assess the implications of trade liberalization in developing countries. Moreover, we study the implications of the degree of financial openness for the welfare effects associated with the liberalization and reform programs. Although there is a large literature emphasizing the importance of complementarities between trade and financial integration, previous studies on the fiscal effects of trade liberalization fall short of analyzing this dimension. Given the significant role international financial markets play in financing the budget shortfalls of developing country governments, it is natural to evaluate the macroeconomic implications of policy changes in economies with different access to global capital markets. In section 2, we present the details of our model. Section 3 describes the model parameters, calibration and solution of the benchmark model for a specific small open developing economy. Section 4 examines the welfare implications of trade liberalization and fiscal reform. In Section 5, we examine how the welfare results change when we vary the degree of access to international financial markets. In Section 6, we study the robustness of our results. Section 7 concludes the paper. 2 Model We construct a dynamic general equilibrium model of a small open economy that captures the main structural characteristics of a typical developing country. The model provides a laboratory 4 For studies using theoretical models, see Osang and Pereira (996), Keen and Ligthart (22) and Konan and Maskus (2). Clarete and Whalley (987) use a simple static model to study commodity and trade taxes. Anderson (999) provides a static model to characterize welfare improving trade reform. Francois and Reinert (997) provide a comprehensive summary of several studies on the implications of trade liberalization. In a related paper, Choudhri, Faruqee, and Tokarick (2) consider the short-run and long-run effects of trade liberalization using a dynamic model, but their analysis abstracts from the impact of liberalization on fiscal balances. Using a simple endogenous growth model, Naito (26) discusses optimal import tariff and consumption tax combinations. 3

4 environment in which we are able to conduct computational experiments to evaluate the welfare implications of various combinations of tax and tariff policies. The model allows for interactions across different agents, including households, firms, and the government. Households consume three types of goods exportable (x), importable (m) and nontraded goods (n). Their labor income and capital income are subject to taxes, and they also pay taxes on their consumption. Firms produce two types of goods exportable and nontraded goods using labor and capital. We assume that the capital goods used for the production of exportable goods are imported whereas capital for the production of nontraded goods is domestically produced. This is a natural assumption given that many developing countries use imported capital goods, such as machinery, to produce and export manufacturing products while they often produce nontraded goods, such as services, using domestic capital. Imported consumption, intermediate input, and capital goods are subject to tariffs. The government must finance an exogenous stream of expenditures through revenues from domestic taxes and tariffs on imported goods. The benchmark model incorporates both current account and financial account transactions by allowing households to borrow and lend in international financial markets using one-period risk-free bonds. This property of incomplete access of households to international financial markets is a good characterization of financial markets in most developing economies. We experiment with different degrees of access to international financial markets, and also analyze the case of financial autarky where the current account is balanced every period. 2. Households We reduce the three-good optimization problem into a single good problem by defining the composite consumption good c t with price p t. A representative household solves X max β t U t, where U t = t= c θ t ( h xt h nt ) θ σ, () σ subject to budget constraint ( τ lt )[p nt w nt h nt + p xt w xt h xt ]+[( τ n kt )r nt + τ n kt δ n] p nt k nt +[( τ x kt )r xt + τ x kt δ x] k xt + p nt T t + B t = (+τ ct )p t c t + p nt i nt +(+t xt )i xt + R t B t+, (2) where w xt,r xt,i xt,h xt,k xt (w nt,r nt,i nt,h nt,k nt ) are wage rate, rental rate, investment, labor input and capital of exportable sector (nontraded sector). B t+ denotes the quantity of discount bonds purchased in period t maturing in t + with price R t. T t is the net transfers from government in a lump-sum fashion, τ denotes tax rates (τ l = tax on labor income, τ n k = tax on capital income in the nontraded sector, τ x k = tax on capital income in the exportable sector, and τ c = consumption tax), and t refers to tariff rates (t x = tariff on imported capital good, t v = tariff on imported intermediate good). 5 As usual, no Ponzi game condition applies (note that no Ponzi game condition never binds in equilibrium). Bonds are priced in terms of import good and the lump-sum transfers are priced in terms of nontraded good. The model also includes a depreciation tax allowance, τ kt δk t. All prices 5 Note that subscript t denotes the time as usual and the regular t corresponds to tariff rates. 4

5 are normalized in terms of the import good. Therefore, p x (p n ) denotes the price of the export (nontraded) good in terms of the import good and the composite good price, p t,canbeinterpreted as the CPI or real exchange rate. Note that i xt,k xt,andv t are all imported goods. Under the case of financial autarky, B t is set to zero in every period. Capital accumulation equations are µ int k n,t+ = ( δ n )k nt + k nt φ, (3) k nt µ ixt k x,t+ = ( δ x )k xt + k xt φ, (4) where δ is depreciation rate and φ is adjustment cost function (φ >, φ > ). The composite good c t consists of consumption of three types of goods, c m,c n, and c x ; h i c t = b m c γ mt + b n c γ nt + b x c γ γ xt,b m + b n + b x =. (5) Total expenditure on consumption can be expressed as the sum of expenditure on each good: k xt p t c t =(+t ct )c mt + p nt c nt + p xt c xt, (6) where t c is the tariff on imported consumption good. Maximizing (5) subject to (6) yields the equilibrium expressions of relative demand for each consumption good and the price of the composite consumption good: 2.2 Firms c mt c t = b c nt c t = b c xt c t = b p t = γ m γ n γ x µ +tct µ pnt p t µ pxt p t p t γ bm( + t ct ) γ γ γ, (7) γ, (8) γ, (9) γ + b γ γ n pnt γ + b γ γ x pxt γ γ. () The production functions for nontraded and exportable goods are y nt = F (k nt,h nt )=A nt knth α ( α) nt () h y xt = G(k xt,h xt,v t )=A xt h μ xt m (k xt) ψ +( m)(v t ) ψi μ ψ (2) where v t denotes the imported intermediate goods and t vt is the tariff rate on the imported intermediate goods. A nt and A xt refer to levels of sector-specific productivity. They are assumed to be constant at one since we consider a deterministic environment. The no-profit conditions imply that y nt = r nt k nt + w nt h nt (3) p xt y xt = r xt k xt + p xt w xt h xt +(+t vt )v t, (4) 5

6 2.3 Government Government revenues can be divided into tax income and tariff income. The government budget constraint is τ ct p t c t + τ lt (p nt w nt h nt + p xt w xt h xt ) +τ n kt (r nt δ n )p nt k nt + τ x kt (r xt δ x )k xt +t ct c mt + t xt i xt + t vt v t = G mt + p nt G nt + p nt T t, (5) where G m and G n denote exogenous government spending on the imported and nontraded goods, respectively. We can combine the government s budget constraint with the consumer s budget constraint and construct the two simplified aggregate budget constraints for the nontradable and tradable sectors as follows: 6 y nt = c nt + i nt + G nt, (6) p xt y xt + B t = c mt + p xt c xt + i xt + v t + G mt + R t B t+. (7) 3 Calibration and Solution Our calibration exercise has two steps. First, we calibrate the benchmark model to capture the main features of the Korean data. Korea is a natural choice as it is a small open economy with a large traded good sector. Moreover, it has reasonably good data series and has been quite active in liberalizing its trade regime for the past three decades. The second step of our calibration involves an extensive sensitivity analysis which we present later in the paper. This allows us to examine how our main conclusions vary depending on the changes in the parameters of the benchmark model. Table reports the parameter and steady state values of the benchmark model. 7 We fix the value of β at.96 to match the steady state level of the annual world real interest rate at 4 % which is the average rate calculated using the U.S. three-month T-Bill rate deflated with CPI inflation. 8 The share of consumption in Cobb-Douglas utility, θ, is set to match the ratio of working hours to total discretionary time from the Korean data. 9 The risk aversion parameter σ is set to 2.6 to match the intertemporal elasticity of substitution in the Korean data (see Kim and Chang, 28), which is close to the average value reported in the panel study by Ostry and Reinhart (992). The share parameters in the CES consumption function (bm, bn, bx) are set at (.3,.52,.8) to match the sectoral data for consumption in Korea, which are collected from the annual reports 6 We present the analytically tractable closed-form solutions for the first order conditions, steady state and the linearized first-order conditions in the working paper version (see Kim and Kose, 2). 7 The parameter values we employ here are similar to those used in earlier papers calibrated for the Korean data, see Park and Shin (2), Kim (27), Kim and Chang (28), and Kim et al. (2). 8 The average real interest rate in Korea is higher, around 6 percent (Park and Shin, 2). In the benchmark model, we assume that Korea has access to international financial markets and borrow at the world interest rate. In Section 6, we provide a sensitivity analysis assuming that Korea faces a higher interest rate due to its country risk premium. 9 Park and Shin (2) compute the average share of labor activity by dividing the average monthly working hours in manufacturing sector (28 hours) with total hours (72 hours) using the Korean data for the period 97 to

7 on the survey of service industries by the Bank of Korea. The value of γ issetat.782tomatch the elasticity of substitution in the aggregate consumption function at.3 which is very close to the value used by Mendoza (992) and Ostry and Reinhart (992). A lower γ implies that consumption responds more to the changes in relative prices. We set the annual capital depreciation rate at % according to Kim et al. (2), which is also a commonly used value in the literature. The elasticity of the marginal adjustment cost function η of the exportable and nontraded sectors is set to 3, to match the volatility of sectoral investment in Korean data. Labor share in the export good production μ is set at.37 following Kim and Ahn (25). Others have used numbers ranging from.2 to.45 (Kouparitsas, 997). The share of capital in the imported intermediate good m is set at.55 (Park and Shin, 2). The elasticity of substitution between capital and the imported intermediate good ψ issetat.35followingkose (22). The capital share in the non-traded good sector, α, is set at.34 to match the average labor income share in services sector in the Korean data during the 27 period. Table 2 reports the weighted mean tariff rates of Korea and average of developing countries in selected regions, collected from the World Bank Development Indicators. 2 Tariff rates in Korea decreased from % in 22 to 7% in 27. We observe a similar decline in all regions. In particular, the Asia-Pacific and Middle East-North Africa regions exhibit more than 5% decline in tariff rates over the same period. For the simulation of our model, we set the steady state tariff rate at % which is the weighted mean tariff rate of Korea in 22. We assume that the tariff rate is same across different types of imported goods (consumption, capital, and intermediate goods) to simplify the interpretation of our findings. We select the steady state tax rates that match the data on tax revenue shares of Korea. Since the tax base is determined by the deep parameters of the model, the effective tax rates that match the tax revenue share data can be easily determined within the model. We follow the method in Mendoza, et al. (994) to calculate the effective tax rates which reflect governments policies on tax credits, deductions, and exemptions as well as information on statutory tax rates. The effective tax rates are consistent with the concept of aggregate tax rates at the national level and with the assumption of representative households in our model. Specifically, we first compile the data on actual tax payments on consumption, labor income and capital income and then calculate tax revenue shares by dividing each of tax payments by total tax revenues. We use the data from the Bank of Korea and National Income Accounts and Revenue Statistics by the OECD. We calculate the consumption tax payment by adding general taxes on goods and services, excise taxes and import duties. One problem in measuring labor and capital income tax payments is that the Korean data does not provide a breakdown of income tax revenue according to its sources (whether it is from labor or capital income). Therefore, we first measure general income tax rate of households assuming that all sources of the household income are taxed at the same rate. We use the Korean Standard Industry Classification (KSIC) in which the industries are classified at the 2 digit level with 36 sectors. We define tradables as the manufacturing sectors where the share of total trade in domestic production is greater than 3 percent. Nontradables include the agricultural and service sectors and the manufacturing sectors where the share of total trade is below 3 percent. Importables refer to the manufacturing sectors where the share of imports on domestic consumption is greater than 5 percent. Exportables are defined as the manufacturing sectors where the share is below 5 percent. Consumption shares of import, nontraded and export goods are 2%, 68% and %, respectively. This number is close to the one used in Jonsson (25). The average labor share in services sector ranges from 4% in the 97s to 65% in the 2s in Korea (Kim and Chang, 28). 2 The weighted mean tariff rate is the average of effectively applied rates weighted by the product import shares of each partner country. 7

8 Households overall income is estimated by taking the sum of their wages and salaries, operating surplus and net property income. We calculate the effective income tax rate by dividing household income tax payment by the overall household income, which is around.7% in 25. Labor income tax payment is calculated by multiplying wages and salaries plus all social security contributions by income tax rate. Capital income tax payment is the sum of all corporate tax payments, including taxes on immovable properties, and financial and capital transactions. The calculated tax revenue shares in Korea are on average 37%, 35% and 28% for consumption, labor income and capital income tax, respectively. The effective tax rates that match these tax revenue shares are τ c =2%,τ l =.5%, τ k = 35%, for consumption, labor income and capital income tax, respectively. Table 2 shows the range of effective tax rates for the G7 and OECD countries reported in Mendoza, et al. (994) and Carey and Tchilinguirian (2). These tax rates are in the range of 5 to 33%, 9 to 48%, and 6 to 5% for consumption, labor and capital income taxes, respectively. Compared to these numbers, labor income tax rate in Korea appears to be quite low, while consumption and capital income tax rates are within the range. We set the government spending to 9% of GDP. This allows the government budget to be balanced under the steady state tax and tariff rates (and zero net lump-sum transfers). The initial asset holding position is set to zero and p x is set to one. 3 We analyze how changes in tax and tariff rates affect the steady state and transitional dynamics of the model economy (between the pre- and post-reform states). We assume that changes in tax and tariff rates are permanent implying that we only consider time-invariant one-time changes in these variables. In order to solve the model, we employ a linear approximation method based on a double shooting algorithm as in Mendoza and Tesar (998) and Gorodnichenko et al. (22). 4 The double shooting algorithm ensures that the value of the post-reform steady state asset holding position is consistent with the debt-accumulation dynamics of the pre-reform equilibrium. 5 4 Implications of Trade Liberalization and Fiscal Reform Trade liberalization is defined as the elimination of tariffs (from % to zero). Tariffs are levied on all three imported goods consumption, capital and intermediate goods. We conduct four experiments to analyze different aspects of trade liberalization: () removal of tariffs on the imported consumption goods; (2) removal of tariffs on the imported intermediate goods; (3) removal of tariffs on the imported capital goods; (4) removal of tariffs on all three imported goods. In order to simplify the exercise and provide as much intuition as possible, each of these liberalization options is accompanied by changes in one of the three tax rates: consumption tax, labor income tax, and 3 These two parameters are freely selected. In a small open economy model, the price of exportable good is exogenously determined. However, depending on the specific value of p x, the consumption share parameters should be adjusted to match the consumption share data. 4 It is possible to solve the model using any standard nonlinear equation solver. For example, Conesa, Kehoe and Ruhl (27) and Kehoe and Ruhl (29) employ nonlinear methods. We check the magnitude of the approximation errors due to our solution method. We feed the approximate solutions to the original nonlinear first order conditions to measure the approximation errors. We find that the mean approximation errors are rather small suggesting that our method produces reasonably accurate solutions. We also check the minimum truncation date which ensures convergence of the solution. Our simulations indicate that the minimum truncation date is between 85 4 periods. We discuss the details of our solution method in Kim and Kose (23). 5 This step is necessary because we use a linear approximation method. In a nonlinear solution, the steady state is unique and does not depend on the transitional path. See Kim and Kose (23) for the dynamic properties of asset holdings in open economy models. 8

9 capital income tax. As we already noted, we consider only one-time permanent changes in tax and tariff rates. Our benchmark experiment involves the case of a tariff cut financed by lump-sum taxes. This experiment enables us to isolate the effects of tariff liberalization from the effects of fiscal reform associated with changes in tax rates. We solve the benchmark model to derive the pre-reform steady state values of the model economy. Then, we use the shooting algorithm to compute the necessary changes in the post-reform tax rates that satisfy the intertemporal government budget constraint, following a tariff cut. 6 Table 3 reports the results of these experiments. When we eliminate all three tariffs, therateofconsump- tion tax should be increased from 2% to 6.7%, labor income tax from.5% to 4.8%, and capital income tax from 35% to 52.5% to satisfy the intertemporal government budget constraint. Capital income tax needs to be increased by a larger amount because the tax base for capital income tax is small relative to others. This is consistent with the observation that the capital income tax base is smaller than the labor income or consumption tax bases in most developing countries. When we remove the tariffs on only one type of imported goods, tax rates need to be increased by a smaller amount compared to the case of removal of tariffs on all three types of imported goods. Using the post-reform tax rates, we analyze the dynamic responses of the economy to the trade liberalization and fiscal reform programs. To compute the welfare effects, we measure the constant percentage change in permanent consumption that leaves the households indifferent between the lifetime utility in pre- and post-reform states, including the transitional period (Lucas, 987). We assume that labor input is fixed at the pre-reform steady state. We decompose the welfare changes into the long-run (steady state) effects and (short-run) transitional effects. The longrun welfare effects measure differences in utility between the pre- and post-reform steady states whereas the transitional effects measure changes in welfare during the transitional period from the pre-reform equilibrium to the post-reform steady state equilibrium. We now analyze each of the cases by considering the corresponding impulse responses and welfare effects. We provide a detailed explanation of the first case, but we keep the discussions of the other three cases relatively shorter for brevity. 4. Removal of Tariffs on Imported Consumption Goods We first analyze the case when the government removes tariffs on imported consumption goods, t c =. Figure shows the impulse responses of the benchmark model when a tariff cut is financed by four different types of taxes. We first discuss these impulse responses to analyze the macroeconomic implications of trade liberalization and fiscal reform programs. We then turn our attention to the welfare effects of these programs. 4.. Impulse Responses Financing through lump-sum taxes does not create an additional distortion to the economy (other than the usual income effects) since all three tax rates remain the same. The impulse responses have intuitively appealing explanations. The removal of tariffs on the imported consumption goods reduce the effective price of imported goods and increases their consumption by more than % 6 We keep the government spending constant at the pre-reform level and calculate the necessary changes in the tax rates that balance the intertemporal budget constraint assuming that the government has a life span of 5 years. Since the government focuses on balancing its intertemporal budget constraint with time-invariant tax rates, we assume that lump sum taxes/transfers are used to ensure the balance of fiscal account every period. 9

10 while leading to a decline in the consumption of the exportable and nontradable goods by about % in the long-run. This is due to the substitution effect among the three types of consumption goods. The decline in the consumption of the nontraded good reduces investment, labor input, and output in the nontraded sector by around %. In a small open economy model, the price of the export good, p x,isfixed, and hence, a rise in the relative price p x (over p n ) boosts all factor inputs thereby increasing output in the exportable sector (by around 6% in the long run). The balance of trade initially worsens as the imported consumption goods increase in the short run but it improves into surplus in the long run as output rises. The net asset position, therefore, shows the economy is borrowing throughout the simulation period: the increase in imported consumption goods is partially financed by borrowing from abroad. The real exchange rate (P ) decreases (domestic currency depreciates) as the nontraded good price, p n,andthetariff rate, t c, decline, as shown in equation (). When the removal of tariffs isfinanced by a consumption tax or labor income tax, the impulse responses of sectoral variables display qualitatively similar patterns to those under the case of lump-sum tax financing, but there are some quantitative differences. The impulse responses under the consumption and labor income tax financing are quite similar to each other since both of these taxes affect the Euler equation associated with consumption (and tilt the marginal conditions with respect to consumption vs. leisure choice) in a parallel fashion. Compared to the case of financing through lump-sum taxes, consumption of all three goods responds negatively to an increase in the consumption or labor income tax rate. Financing through labor income taxes leads to a slightly smaller impact on consumption compared to the case of financing through consumption taxes. The decline in consumption has, in turn, an adverse impact on output, investment and labor input in both sectors. There are substantial changes in the impulse responses (relative to previous cases) when the government finances the tariff cut with an increase in capital income tax. Unlike the cases of consumption and labor income tax financing, consumption of export and nontraded goods decreases by 5% and 3% in the long run, respectively, and consumption of imported goods increases by 7% (but this increase is still smaller than % under the case of consumption or labor income tax financing). The lower demand for nontradable goods reduces the nontradable sector output by more than 3%, which then leads to a decline in labor input and investment in that sector. The increase in capital income tax lowers investment in the export sector. Firms use less capital, but employ more labor and intermediate inputs. Both labor and intermediate inputs increase initially due to the substitution effect. However, in the long run, the decline in investment (capital) constrains the production capacity and output of the exportable sector declines. Unlike the case of consumption and labor income tax financing, financing through capital income taxes implies that households accumulate international bonds. This is driven by the fact that after-tax returns on domestic investment declines due to the increase in capital income tax rates, raising the relative return on international bonds (interest rates on these bonds are fixed, as usual in these types of models). The trade balance initially improves due to the substantial contraction in investment, but registers a deficit in the long-run with the decline in aggregate output Welfare Effects Table 4 summarizes the welfare implications of different combinations of tariff and tax policies. When the tariffs on imported consumption goods are removed (t c =), welfare gains are.64% of lifetime consumption under the lump-sum tax financing. With the consumption and labor income

11 tax financing, welfare gains are about.33% and.26%, respectively. These results suggest that the welfare gains from tariff cuts outweigh the losses associated with the increase in tax rates. However, when the removal of tariffs isfinanced by an increase in capital income tax, there are welfare losses (-.64%). This finding implies that adverse effects of the increase in capital income taxes dominates the beneficial effects of tariff cuts on consumption goods. At a mechanical level, this is not an unexpected result: changes in the tax on capital income influences the intertemporal marginal condition for production while changes in consumption and labor income taxes affect only the intratemporal conditions on consumption/leisure choice. We also decompose the total welfare effects into those associated with the transitional and long-run changes (see Table 4). In the case of consumption and labor income tax financing, welfare changes in the long run are negative (.4.%), but the transitional effects are positive (.37%). Households do not have to sacrifice consumption in order to increase investment as they utilize international financial markets. Investment can be financed by international borrowing, so called the intertemporal smoothing effect. Hence, the aggregate consumption and utility increase during the transitional period, which in turn result in transitional welfare gains. However, as households accumulate foreign debt in the long-run, the trade balance needs to register a surplus to satisfy interest payments on the accumulated debt (so called the wealth redistribution effect). 7 Both consumption and leisure decrease in the long-run (as shown by the impulse responses) and this translates into long-run welfare losses. Financing through capital income tax leads to different welfare outcomes. The decrease in investment lowers output and consumption over time, which then leads to both transitional and long-run welfare losses (.3% and.5%, respectively). In the long-run, the decline in investment reduces the production capacity of the economy, and as a result, consumption also decreases. These losses outweigh the welfare gains associated with the wealth redistribution effect stemming from the accumulation of international bonds. 4.2 Removal of Tariffs on Imported Intermediate Inputs Figure 2 presents the impulse responses (again with four tax financing schemes) when the government removes tariffs on imported intermediate inputs (t v =). First,wefocusonthecaseof financing through lump-sum taxes. The removal of tariffs on imported intermediate inputs lowers their effective cost, and leads to an increase in the use of intermediate inputs. Investment in the exportable sector also significantly rises due to the complementarity of the two factor inputs. While the labor input in the exportable sector (h x ) decreases initially due to the substitution effect, it increases over time (9% in the long-run) as more exportable goods are produced. As a result, output of the exportable good sector increases in the long-run. As resources shift from the nontradable sector to the exportable sector, investment, labor input and output in the nontradable sector decline in the long-run. Consumption of the export and import goods increases while consumption of nontraded goods slightly decreases over time. The price of nontraded goods, p n, increases as a result of the rise in the production of the exportable goodwhichinturnlowersitseffective relative price, (p x /p n ). The increase in imports (i x, c m and v) is financed by borrowing from abroad (as shown in the rise in international borrowing to 3% of GDP in the long-run). The increase in imports in the 7 The terms "international smoothing effect" and "wealth redistribution effect" are also used by Mendoza and Tesar (998).

12 initial period leads to a trade deficit (more than 2% of GDP), but the balance of trade improves into a surplus as output of the exportable good sector (y x ) increases over time. One notable difference from the case of a tariff cut on the consumption good (t c =)isthattherealexchangerate(p ) increases in this case. Domestic currency appreciates because the production of nontraded goods declines pushing up their relative price. When we change the source of financing to the consumption and labor income taxes, we observe that the impulse responses are qualitatively similar to those in the previous sub-section with the removal of tariffs on consumption goods. The increases in consumption and labor income taxes lead to smaller (but still positive) effects on consumption, investment, labor input and output, compared to the case of financing through lump-sum taxes. The case of capital income tax financing is also associated with an increase in investment in the exportable sector (but only % compared to 2% in the lump-sum tax financing case) while it lowers investment in the nontradable sector. The removal of tariffs on intermediate inputs (t v =)results in welfare gains of about.46% of lifetime consumption under the case of lump-sum tax financing. These gains are more than two times larger than those associated with the removal of tariffs on consumption goods (.64%). As one would expect, when a tariff is levied on a production factor, it generates larger distortions than when a tariff is imposed on a consumption good. Therefore,the removal of such distorting tariffs translates into much larger welfare gains. Financing through consumption and labor income taxes generates welfare gains of.22% and.7%, respectively. Even under financing through capital income taxes, welfare gains remain at.58%. The decomposition of welfare gains shows that the transitional welfare changes are all positive due to the intertemporal smoothing effect. The long-run welfare effects are mostly positive because consumers enjoy interest income from the accumulation of international bonds (and benefit from positive wealth redistribution effects). The utility gains from the increase in consumption outweighs the losses from the decline in leisure. However, with financing through capital income taxes, the increase in the long-run consumption is not large enough to translate into long-run welfare gains. 4.3 Removal of Tariffs on Imported Capital Goods Figure 3 shows the impulse responses when tariffs on imported capital goods are removed (t x =). Since both intermediate inputs and capital goods enter the production function of the exportable sector in a similar fashion, the impulse responses to the removal of tariffs on intermediate inputs and capital goods display similar patterns. Therefore, we focus only on differences between the two cases. First,the removal of tariffs on the imported capital goods increases investment in the exportable sector (by 23% in the initial period and 35% in the long-run). This increase is much larger than that in the case of the removal of tariffs on intermediate inputs (5% initially, and 2% in the long-run). The increase in intermediate inputs (2% in the long-run) is smaller than that in the case of the removal of tariffs on intermediate inputs (3% increase). Second, the amount of intermediate inputs used in production initially decreases, but then picks up over time. Unlike the case of the removal of tariffs on intermediate inputs in which the marginal product conditions change immediately (in the initial period), the removal of tariffs on capital goods affects the marginal product condition after a period later (when the change in investment in the initial period is used for production). Therefore, in the initial period, intermediate inputs decrease while resources shift from the exportable sector to the nontraded sector due to the substitution effect. 2

13 Third, the impulse responses in all four financing schemes are quite similar to each other since the removal of tariffs on capital goods affects the first order conditions for investment and capital in a similar fashion that an increase in capital income tax does. Financing through capital income taxes in this case simply replaces distortions associated with tariffs on capital goods and changes only the magnitude of responses (without generating additional price effects). The tariff on imported capital goods distorts the intertemporal efficiency in production, and therefore, the removal of such tariffs produces large welfare gains (see Table 4). Even under the case of financing through capital income taxes, the welfare gains are sizable at.3%. Most of these welfare gains stem from the positive transitional effects. In the long-run, the losses from the decrease in leisure outweighs the gains from the increase in consumption (despite the positive wealth redistribution effect), and translates into a long-run steady state welfare loss in all cases of financing. The welfare gains from the removal of tariffs on intermediate inputs is slightly lower than the gains under the case of elimination of tariffs on imported capital goods. 4.4 Removal of all three tariffs The removal of all three tariffs generates large welfare gains (3.5% with the lump-sum tax financing). Even with financing through increases in distortionary taxes, welfare gains are sizable: 2.76%, 2.6% and.6% with financing through consumption, labor income, and capital income taxes, respectively. Most welfare gains are from the positive transitional changes in welfare associated with the intertemporal smoothing effect. This implies that the removal of international borrowing and lending channel changes these results as we explain in the next section. We have three main results regarding the welfare effects of the revenue-neutral trade liberalization and fiscal reform programs. First, the revenue-neutral trade liberalization generates large welfare gains when the lost tariff revenue is compensated by an increase in consumption or labor income taxes. The removal of all three tariffs can generate welfare gains in the order of % of lifetime consumption. However, when the lost revenue is financed by an increase in capital income tax, welfare gains are relatively small (around %). When the tariffs on consumption goods are removed (and financed by capital income taxes), there are welfare losses (around.6% of permanent consumption). Second, when financed by specific taxes, financing through consumption taxes is the best fiscal option, followed by a labor income tax, in all cases of elimination of tariffs. However, the differences in welfare gains across these options are small. Financing through capital income taxes is the least preferred fiscal tool to recover the lost revenue. Lastly, in all cases of financing of trade liberalization programs, the removal of tariffs on imported capital goods is the most preferred policy option. The removal of tariffs on imported capital is welfare improving even if this type of liberalization program is financed through capital income taxation. The elimination of tariffs on consumption goods generates the smallest welfare gains in all cases. 5 Financial Autarky In previous sections, we assume that the model economy has incomplete access to international financial markets as households can buy and sell only one-period risk-free international bonds. In this environment, most of the welfare gains associated with a trade liberalization program result 3

14 from the transitional effects stemming from households ability to conduct international borrowing and lending. In this section, we relax the assumption of international borrowing and lending, and examine a model with no access to international financial markets. In particular, we assume that the economy is under a regime of financial autarky implying that households cannot conduct international borrowing and lending, and only barter trade is allowed. Impulse responses of tariff-tax experiments under financial autarky differ in some important dimensions from those under incomplete access to financial markets (bond economy). 8 For example, in all cases we examine, the removal of tariffs is associated with a smaller increase in investment in the exportable sector under autarky than that in the bond economy. Since households cannot borrow in international financial markets, an increase in investment requires a reduction in consumption, especially during the transition period. Hence, the elimination of asset trade has an adverse impact on both investment and consumption. Impulse responses also suggest that firms employ less capital but demand more labor under financial autarky than in the bond economy because of the substitution effects between capital and labor. This, in turn, reduces leisure during the transition period. The declines in consumption and leisure translate into lower welfare during the transition period. We report the welfare gains associated with the model under financial autarky in Table 4 (lower panel). The results reflect the changes observed in the impulse responses. Except in the case of financing through capital income taxes, the transitional changes in welfare are all negative whereas the steady state welfare changes are mostly positive. Under financial autarky, the economy has no adverse effects from wealth redistribution as observed in the bond economy (lower consumption due to the accumulation of debt), which makes a positive contribution to the long-run steady state welfare gains. In the case of financing through capital income taxes, the increase in capital income tax reduces investment in both sectors, which allows consumption to rise during the transitional period. Hence, the transitional welfare gains are larger compared to the other cases of financing. However, the decline in investment lowers the steady state output and consumption, which in turn dampens the long-run welfare gains. In sum, trade liberalization leads to smaller changes in welfare under the case of financial autarky than it does under the bond economy. 9 Forexample,withtheremovalofallthreetariffs, the welfare gains (with the lump-sum tax financing) under financial autarky is 2.92%, while the gains are 3.49% in the bond economy. The elimination of international borrowing and lending channel reduces the potential welfare gains from trade liberalization by around 2 3% in most cases. The welfare ranking of tax-tariff policies under financial autarky is the same as the one in the bond economy. We consider some additional experiments where the access to international financial markets is limited due to transaction costs. Specifically, we introduce bond holding costs into the model to study the welfare implications of limited borrowing in international financial markets. The bond ς holding costs we consider take the following form: 2 (B t+) 2. We simply add these costs to the budget constraint in equation (2). Changes in the value of ς (from zero to positive values) measure how the degree of access to international financial markets varies from bond economy to financial autarky. We report our results in Table 5. Irrespective of the mode of financing, the economies with more limited borrowing experience 8 These impulse responses are presented in the working paper version (see Kim and Kose, 23). 9 The only exception is the removal of tariffs on consumption goods financed by capital income taxes. In this case, the welfare losses in the bond economy are smaller than that under financial autarky. 4