全球智库信息集成服务系统

Paper prepared for the conference "China's Integration into the World Economy" January 17, 1998, Harvard University. The authors would like to thank seminar participants at the IIE for their comments on an earlier version of this paper. The views expressed herein are purely the authors' and should not be attributed to any of their respective institutions.

Abstract

In this paper we examine three issues. The first is the path of China's nominal and real exchange rates since 1990. As it turns out, this is more complicated than is commonly assumed, with basic results exhibiting sensitivity to the exchange rate measure used. We conclude that while China did experience a large nominal depreciation, its much higher relative inflation eroded this devaluation, and in real terms the reminbi has actually appreciated during the 1990s. The Chinese devaluation was at best a contributing factor to the Asian financial crises, not their primary cause.

The second issue we examine is the impact of exchange rate changes to date on world trade. Using a computable general equilibrium (CGE) model we simulate the series of devaluations which have occurred throughout Asia and generate estimates of the impact of these exchange rate changes on the volume and partner country and sectoral composition of world trade. In the case of the US, for example, we find that the exchange rate changes to date could add more than $50 billion to the US trade deficit. The impact would be felt throughout the traded goods sector, with the light manufacturing and machinery sectors hit particularly hard.

With respect to China, we find that a modest 5-10 percent devaluation of the reminbi would be sufficient to re-establish the status quo ante in light of the recent devaluations elsewhere in the region. At the same time, given the relatively minor impact on China of the events observed thus far, and the likelihood that a devaluation in China could spark another round of financial turbulence in the region, this result surely argues for restraint on the part of the Chinese.

Finally, we apply the results of these simulations to econometric models of trade conflict. We find that the projected trade flow changes will increase the likelihood of bilateral friction between the US and its Asian trade partners. The impact may be felt acutely in Korea, where we estimate that the results in train significantly increase the likelihood of formal US trade actions against Korea. At the same time, the growing dependence of the region on the US market will increase the likelihood of disputes being resolved in ways satisfactory to the US.

I. Introduction

The financial market turbulence that began in 1997 has led to large nominal exchange rates movements throughout Asia. These changes have significantly altered the pattern of relative competitiveness through out the region, and can be expected to significantly affect the volume and composition of international trade flows.

Observers have identified a number of possibly contributing factors to the financial turbulence of 1997-8, including the large capital outflows from Northeast Asia starting in the mid-1980s, the weak financial sectors throughout the region, and moral hazard issues associated with previous international financial rescues, most notably the Mexican bailout of 1995. Some claim that the roots of the 1997 crisis lie in the Chinese currency devaluation of January 1994.¹ The fall of the reminbi implied a real exchange rate appreciation for the dollar-pegged currencies in Southeast Asia which their fragile financial systems were unable to absorb. The crisis in Southeast Asia put downward pressure on the Japanese yen and Australian dollar which float, and the Korean won which also was pegged to the dollar. Under the pressure of emerging banking crises, the exchange rate pegs eventually were abandoned leading to a period of exchange rate instability that has yet to fully subside (Table 1).

Whatever the ultimate source of these exchange rate changes, they will significantly alter the international pattern of trade competitiveness. With domestic demand in these economies flat or declining, it is likely that recovery will at least partly come through increases in net exports. Given existing trade patterns and the exchange rate changes that have occurred to date, the most obvious market for these exports is the United States. Yet research has shown that bilateral trade balances are a key indicator of the extent of trade conflict between the US and its partners. This suggests that the most likely path of recovery will bring these countries into conflict with the US. In this paper we examine three issues. The first is the path of China's nominal and real exchange rates since 1990. As it turns out, this is more complicated than is commonly assumed, with basic results exhibiting sensitivity to the exchange rate measure used. We conclude that while China did experience a large nominal depreciation, its much higher relative inflation eroded this devaluation, and in real terms the reminbi has actually appreciated during the 1990s. The Chinese devaluation was at best a contributing factor to the Asian financial crises, not their primary cause.

The second issue we examine is the impact of exchange rate changes to date on world trade. Using a computable general equilibrium (CGE) model we simulate the series of devaluations which have occurred throughout Asia and generate estimates of the impact of these exchange rate changes on the volume and partner country and sectoral composition of world trade. In the case of the US, for example, we find that the exchange rate changes to date could add more than $50 billion to the US trade deficit. The impact would be felt throughout the traded goods sector, with the light manufacturing and machinery sectors hit particularly hard.

With respect to China, we find that a modest 5-10 percent real effective devaluation of the reminbi would be sufficient to re-establish the status quo ante in light of the recent devaluations elsewhere in the region. At the same time, given the relatively minor impact on China of the events observed thus far, and the likelihood that a devaluation in China could spark another round of financial turbulence in the region, this result surely argues for restraint on the part of the Chinese.

Finally, we apply the results of these simulations to econometric models of trade conflict. We find that the projected trade flow changes will increase the likelihood of bilateral friction between the US and its Asian trade partners. The impact may be felt acutely in Korea, where we estimate that the results in train significantly increase the likelihood of formal trade actions against Korea. At the same time, the growing dependence of the region on the US market will increase the likelihood of disputes being resolved in ways favorable to US trade interests.

II. China's Exchange Rate

As the examples cited in the previous section demonstrate, it has been argued that China's exchange rate management amounts to a "cheap currency policy," which in turn has caused the financial crises in Asia. Reality is a bit more complicated for at least three reasons. First, while it is true that the reminbi depreciated in nominal terms, China has experienced higher inflation than its trade partners, eroding its real depreciation.² Second, these calculations are normally done with respect to the reminbi's value against the US dollar. But for the most part, Chinese exports do not compete against US exports in third markets nor even with US domestic production in the US market.³ The relevant measure of export competitiveness is the real exchange rate defined relative to China's export competitors. Lastly, most analysts implicitly assume that all transactions occurred at the official rate. Yet there is evidence that prior to the unification of the exchange rate in 1994, a considerable share of transactions occurred at the swap rate.⁴ Depending on the relative weights one gives official and swap rate transactions, one can obtain considerably different indications of the path of the exchange rate.

Table 2 reports export share weighted nominal and real effective exchange rates constructed using export data of China's major trading partners. Two things are immediately apparent: there are significant differences in the paths of the nominal and real exchange rate, and the relative weights assigned to the official and swap rates matter a great deal in assessing cumulative exchange rate movements since the early 1990s. If one assumes that all transactions occurred at the official rate, then both the nominal and real effective exchange rates have depreciated since the early 1990s, though the nominal rate has depreciated far more, and since 1995 much of the real depreciation has been eroded by China's relatively high rate of inflation. As one increases the importance accorded to transactions at the swap rate prior to 1994, a very different story emerges, suggesting that there has been cumulative appreciation of the real exchange rate since the early 1990s. (Indeed, there may have even been cumulative effective nominal appreciation over this period, as well.) This fact, together with tremendous uncertainty about the extent of relative productivity increases in China, makes it very difficult to ascertain how close China's current exchange rate is to some notion of a fundamental equilibrium rate.

What can be said with more certainty is that there has been considerable divergence between the effective nominal and real exchange rates since 1994, with the real effective rate appreciating far more than the nominal rate. In the succeeding sections below we construct a CGE model calibrated for 1992, the most recent year for which a complete social accounting matrix for China can be constructed, and simulate a series of real exchange rate depreciations involving China and its partners, implicitly assuming that this plausibly could be defended as an equilibrium situation.⁵

III. Structure of the Model and Major Assumptions

The model is neoclassical in spirit and is part of a now long tradition of multi-country, multi-sector, computable general equilibrium (CGE) models that started with models developed by Deardorff and Stern (1990) and Whalley (1985) to examine the impact of the Tokyo Round of GATT negotiations. The lineage of our model starts from the Walras world model developed at the OECD (Burniaux et al., 1990) and is very close to models of ASEAN and APEC developed by Lewis, Robinson, and Wang (1995) and Lewis and Robinson (1996). Our particular China-focused model was developed by Wang (1994, 1997). It has been adapted for this application to reflect the crisis aspect of the current situation as described below.

The model includes twenty-one regional models, each with twenty-one sectors and four primary factors: agricultural land, capital, unskilled labor, and skilled labor.⁶ The models are linked by commodity trade, with world commodity prices determined endogenously to clear world markets. Primary factors are mobile across sectors within regions, but not internationally. Prices and wages (including the rental rate of capital) are determined to clear regional commodity and factor markets. However, the financial crises could be expected to cause a variety of disruptions in the economy that would push it away from a full employment equilibrium. We model this by introducing negative supply-side shocks as described in the next section.

Economic Agents and Factor Endowments. Three demand-side agents are assumed for each region: a single aggregate private household which buys consumer goods, the government which buys goods and public administration, and an aggregate capital account or an aggregated investor which purchases capital goods. Factor endowments are owned by households and set exogenously. Households sell the two categories of labor and rent capital to firms, and allocate their income from factor returns to savings and expenditures. The capital account collects savings from households, government, and firms, also accounting for foreign capital inflows or outflows. Total regional savings are available to the investor to purchase capital goods for gross investment.

Production. There is one competitive firm in each sector for every region, which produces only one product. The production is characterized by two-level nesting of constant elasticity of substitution (CES) functions. At the first level, firms optimize over two inputs: a composite primary factor and an aggregate intermediate input, according to a CES cost function. At the second level, firms can substitute among the four primary factors according to a CES cost function, but demand for produced intermediate inputs follows a Leontief specification foreclosing substitution among produced intermediates. The degree of substitutability between the composite primary factor and the aggregate intermediate, as well as among the four primary factors depends on their base year share in production and on the elasticity of substitution which is assumed to be constant. Technologies in all sectors exhibit constant return to scale implying constant average and marginal cost. Firms' output is sold on the domestic market or exported to other regions through a constant elasticity of transformation (CET) function. The structure of production is illustrated in Figure 1. The CET function can be partially or entirely turned off in the model. In this case, exports and domestic sales become perfect substitutes.

Demands. Agents in each region value products from different regions as imperfect substitutes (the Armington assumption). The private household in each region maximizes a Stone-Geary utility function over the eight composite goods, subject to their budget constraints, which leads to the Extended Linear Expenditure System (ELES) of household demand functions. Household savings are treated as demand for future consumption goods with zero subsistence quantity (Howe, 1975). An economy-wide consumer price index is specified as the price of savings. It represents the opportunity cost of giving up current consumption in exchange for future consumption (Wang and Kinsey, 1994). Government spending and investment decisions in each region are based on Cobb-Douglas utility functions, which generate constant expenditure shares for each composite commodity. In each region, firm intermediate inputs, household consumption, government spending and investment demand constitute total demand for the same Armington composite of domestic products and imported goods from different sources. A two-level nested CES aggregation function is specified for each composite commodity in each region. The total demand is first divided between domestic produced and imported goods, then the expenditure on imports is further divided according to geographical origin under the assumption of cost minimization. Sectoral import demand functions for each region are derived from the corresponding cost function according to Shephard's lemma. Complete trade flow matrices for all trade partners are part of the model solution. The structure of demand is depicted in Figure 2.

International Shipping. There is an international shipping industry in the model to transport products from one region to another. Each region is assumed to allocate a fraction of the output of its transportation and service sector to satisfy the demand for shipping that is generated by interregional trade. The global shipping industry is assumed to have an unity elasticity of substitution among supplier sources. The margins associated with this activity are commodity/route specific. At the equilibrium, total value of international transportation service at world price equals to the sum of the export proportion of the service sector's output from each region.

Trade Distorting Policy. Each region's government imposes import tariffs, export subsidies and indirect taxes, all in ad valorem terms. Tariff and tax (subsidy) rates vary by sector and by destination. Nontariff barriers are applied in terms of their tariff-equivalents, with the usual caveat that this treatment is inadequate with respect to non-tariff barriers which take the form of quantitative restrictions.

Price System. There are 10 types of prices for each good with the same sector classification in each region. They are value-added prices, aggregate intermediate prices, average output prices, composite good prices, consumer prices, producer prices, export prices, import prices, f.o.b. prices, and c.i.f. prices. The value-added price equals the unit cost of primary factor inputs. The aggregate intermediate price is a fixed proportion (IO coefficients) weighted average of composite good prices. A CES aggregation of the two equals the average output prices. Adding to it the production taxes yields the producer prices which are tax inclusive CET aggregation of domestic and export prices. Sellers receive this price. The composite good price is a tax inclusive CES aggregation of domestic and import price, which in turn is an aggregation of tariff inclusive import prices from different sources. The consumer price is the composite good price plus sales tax. Buyers pay this price. The f.o.b. price of each Armington good is the firm's export price plus the export taxes or minus export subsidies. Adding to it the international transportation margins yields the c.i.f. price. The relation among the ten categories of prices in the model is illustrated in Figure 3. An exchange rate, as a conversion factor, translates world market prices into domestic prices.

It is important to emphasize that the "nominal" exchange rate variable in the model is not a financial exchange rate, but measures the "parity" that translates international prices, which are measured in $US, into domestic prices (de Melo and Robinson, 1989; de Melo and Tarr, 1992; Devarajan, Lewis, and Robinson, 1990). In the model, all prices are relative prices, including the exchange rate, since the model includes no money or assets and thus cannot determine the aggregate price level endogenously. "Relative to what" is determined by choice of numeraire. Devarajan, Lewis, and Robinson (1993) sort out the relationship between the exchange rate variable in a CGE model and various measures of the "real" and/or purchasing power parity exchange rate. De Melo and Tarr (1992) and de Melo and Robinson (1989) show that, in a single-country model, when the price index of "home" goods is chosen as numeraire, the "nominal" exchange rate in the model is equivalent to the real exchange rate of trade theory--that is, it measures the relative price of traded goods sold on world markets to the prices of "semi-tradable" domestically produced goods sold on domestic markets.⁷ In a multi-regional model, where world prices are endogenous, measuring the appropriate real exchange rate requires considering both international terms of trade effects (the relative price of exports and imports on world markets) as well as domestic terms of trade effects (relative price of traded and domestically produced goods sold on domestic markets).

Equilibrium In our multi-country model, equilibrium in product and factor markets in each country determines a set of relative product and factor prices such that supply and demand is equated for all goods and factors, subject to the negative supply-side shocks described below. Equilibrium in world markets determines a set of relative world prices so that desired sectoral supply of goods (exports) equals desired demand (imports) across all countries. Relative world prices are all measured in terms of exports and imports of the United States, whose exchange rate is set to one.

Choice of Numeraire In this model, we specify the aggregate consumer price index (CPI) in each region as the numeraire price index. This choice implies that, in model solutions, factor returns and household income measure real returns in terms of welfare. For this choice, the exchange rate variables by region are defined as price-level-deflated (PLD) real exchange rates, deflated by CPI. In addition, a numeraire is required for aggregate world prices, which is given by setting the exchange rate for the United States at one.

Choice of Macro Closure For each region, the model includes the three macro balances: savings-investment, balance of trade (in goods and non-factor services), and government expenditure-receipts (government deficit). Since each regional model satisfies Walras' Law, the three balances are not independent and only two need to be determined by means of a specified macro adjustment mechanism, or macro "closure" of the model.

Since this model is being used to explore the impact of changes in real exchange rates on the real economy in a full-employment environment, we specify a macro closure with the major macro aggregates set by simple macro relationships. Aggregate investment and government expenditure are specified as fixed shares of GDP in each region. The government deficit is endogenous--direct and indirect tax rates (including tariffs) are fixed, so total tax receipts are a function of the level and distribution of economic activity. Aggregate private savings are assumed to adjust endogenously to match aggregate investment through changes in the rate of retained earnings by producers, thus achieving savings-investment equilibrium even with endogenous changes in the government deficit and the balance of trade.

The model specifies sectoral export supply and import demand functions which depend on sectoral relative prices of goods sold on international and domestic markets. With imperfect substitutability between imports and domestically produced goods in the same sector and imperfect "transformability" between sectoral supply to world markets (exports) and domestic markets, each sector has distinct prices for exports, imports, and domestic sales. In the aggregate, there is a functional relationship between the balance of trade (in goods and non-factor services, or the current account balance) in each region and the real exchange rate. If the real exchange rate depreciates, the price of traded goods increases relative to the price of domestically produced goods sold on the domestic market. Exports increase, imports decrease, and the trade balance will improve. Given our macro closure, note that changes in the trade balance are assumed not to have any effect on aggregate investment in the region. They lead instead to an equilibrium adjustment in the domestic savings rate.

In the model, we specify a fixed real exchange rate for each region. In the base solution, the initial trade balance and exchange rate are assumed to be in equilibrium--that is, the initial trade balance is assumed to be "sustainable" and consistent with the initial real exchange rate. In simulation experiments, we change that rate for a particular region, which changes the equilibrium trade balance both in aggregate and bilaterally. In the multi-region model, there is a ripple effect, since a depreciation in one region's real exchange rate implies a relative appreciation for its trading partners, leading to changes in their equilibrium trade balances as well. The world model is closed in the sense that the sum of all regional trade balances must be zero. Thus, we can use the model to see how real exchange rate shocks lead to adjustments in trade balances worldwide in a consistent framework.

In the simulations, we change the real exchange rate of one region at a time, keeping all other regional real exchange rates fixed. We use the model as a simulation laboratory to isolate the exchange rate transmission effect. We do not attempt to model a mix of exchange rate and trade balance adjustments or to forecast what might actually happen. This real model, with no money or asset markets, simply cannot be used for such forecasting.⁸ The model does allow us to explore structural adjustment effects such as changes in sectoral production, exports, and imports. In principle, the model could be linked to a macro model that included asset flows and could determine the set of real exchange rates resulting from some macro shock. Our model could then be used to determine the resulting real trade flows and regional sectoral structural adjustments.

IV. Results

To simulate the model we need to specify a series of simulation experiments. With respect to the exchange rate realignments occurring in Asia, there are two sources of uncertainty. First, as of this writing, considerable nominal exchange rate volatility continues in Asia. Second, we do not know how much of the nominal depreciations will ultimately be reduced by higher inflation. So what we do here is to simply choose what we believe are a defensible set of real exchange rate changes based on the nominal changes that have already been observed (Table 1). We assign the Asian economies to three groups: those experiencing a 30 percent real devaluation (Thailand, Indonesia, and Korea), those experiencing a 20 percent real depreciation (Malaysia and the Philippines), and those experiencing a 10 percent real depreciation (Singapore, Taiwan, and Japan). In our experiments, we attempt to follow the historical evolution of these changes, beginning in Southeast Asia, and moving northward to Northeast Asia in experiments 1 through 8.

This approach, while adequate for normal situations, may inadequately capture the crisis aspects of the current situation. The precipitous nominal exchange rate depreciations have been accompanied by significant disruptions of domestic financial markets and institutions. As a result, firms may suddenly find themselves liquidity-constrained, unable for example to finance purchases of intermediate inputs. The economy may not respond to the real exchange rate changes as expected under more normal circumstances.

We model this financial disruption to the economy as a series of negative supply-side shocks to total factor productivity, in effect shifting the production possibility frontier inward (or alternatively allowing for less-than full employment of resources). For countries experiencing a real exchange rate depreciation of 30 percent this is accompanied by a 5 percent negative supply-side shock; countries experiencing 20 percent depreciations are also experience a 3 percent supply shock; and countries experiencing 10 percent depreciations are hit with a 1 percent supply shock.

This leaves the shoe that has not dropped, China. In the final experiment (EXP9), we subject China to a real depreciation of 10 percent, roughly the amount needed to re-establish its baseline trade balance (assuming it adjusts as would a competitive economy) .

Figures 4, 5, and 6 display the paths of Japanese, Chinese, and US aggregate imports and exports through the nine experiments. (Similar graphs could be constructed for other countries but are omitted here for the sake of brevity.) As can be seen in Figure 4, Japan is initially hurt by the devaluations in the rest of Asia, but gets an enormous boost from its own depreciation in experiment 8.⁹ This boost is only partly offset by the Chinese devaluation in experiment 9, increasing its global trade surplus by more than than $45 billion from the base.¹⁰ The bulk of this increase comes at the expense of the US and Western Europe, as seen in Table 3.

In the case of China (Figure 5), its exports shrink and imports grow steadily through the first eight experiments until experiment 9, where it experiences a 10 percent real depreciation. The result is a big increase in exports, and a significant reduction in imports, bringing the Chinese into surplus.

Since the destination of much of Asian exports is the United States (in David Hale's words, the world's "consumer of last resort") the impact of Asian competitive devaluation on the US trade balance is huge. As shown in Figure 6, US exports decline and imports rise steadily through the nine experiments, with its trade deficit growing more than $50 billion from the baseline in experiments 8 and 9.¹¹ The biggest increases in the deficit generated by trade with Japan and Korea. Trade with China accounts for a relatively small share of the increase.¹² Sectorally, the deficit increase is generated largely by US decreased exports to the region of agriculture products, intermediate capital goods, machinery, transportation equipment, and services, and by increased imports of textile, apparel, light manufacturing products, intermediate capital goods, machinery, and transportation equipment from the region (Table 4).¹³ Production decreases throughout the traded goods sector (with the exception of the energy sector), with the apparel and machinery sectors hit particularly hard. In terms of macroeconomic aggregates, the terms of trade effect generates an increase in consumption, and with government expenditure and investment relatively stable, this is manifested offset by a fall in saving. In these respects, this outcome resembles the US experience of the first half of the 1980s, though relative to national income, the impact is only about half as much.

Perhaps the most striking result in this whole exercise is the enormous increase in the Korean surplus, which increases more than $45 billion relative to the baseline. This is mainly due to Korea's large 30 percent real depreciation (30 percent) in experiment 6. Obviously, if the eventual real depreciation is smaller than that in experiment 6, the impact on Korea would be commensurately smaller as well.

V. Implications for Trade Conflict

The results reported in the previous section indicate that the changes in trade flows in train is expected to substantially increase US bilateral trade deficits with a number of its Asian trade partners. The pattern of bilateral trade conflict (and its resolution) between the US and its trade partners has been a topic of recent research (Noland 1996, 1997b). Significantly, the most striking result is the one that strongly links trade conflict to the magnitude of bilateral imbalances.

A fundamental problem faced by analysts is that the US takes relatively few formal actions against its trade partners--most trade disputes are resolved before formal actions or sanctions are applied. To get a handle on the degree of interaction prior to the formal dispute stage, the number of pages devoted to the analysis of trade issues with partner countries in the United States Trade Representative's National Trade Barriers annual has been taken as an indicator of attention while the initiation of legal processes that could lead to sanctions has been used as a proxy for action. As can be seen in Table 5, Asian countries appear to get more than their share of attention from USTR. Some simple correlations between these trade conflict variables and a number of other variables are reported in Table 6.¹⁴ Note the strong correlation between trade conflict to the magnitude of bilateral imbalances.

As revealed in Table 7, the simple correlation is born out in multivariate regressions -- the degree of trade conflict between the US and its partners is highly correlated with the magnitude of bilateral trade imbalances and other variables.¹⁵ The question immediately arises as to whether this trade pressure is successful -- that is, has it had the effect of eliminating access barriers faced by US firms? On this point the evidence is weak. There is evidence, however, that the US has tended to be the most successful in resolving issues such as export subsidies or countertrade where well specified intern