全球智库信息集成服务系统

We would like to thank Chris Udry and seminar participants at the Korea Institute for International Economic Policy, the ECAAR Congress, Southern California, Stanford, Georgetown, Hawaii, and the Australian National universities for helpful comments on an earlier draft.

Abstract

We construct the Korean Integration Model (KIM), a two-country computable general equilibrium (CGE) model linking the North and South Korean economies. Using KIM, we simulate the impact of a customs union and an exchange rate unification of the two economies both in the presence and absence of cross-border factor mobility, treating technological transfer in three ways. Factor mobility is of critical importance. If factor markets do not integrate, the macroeconomic impact on South Korea of economic integration with the North is relatively small, while the effects on North Korea are large. With a unified exchange rate and factor market integration, there is a significant impact on the South Korean income and wealth distribution. If investment flows from South to North and labor flows from North to South, there is a shift in the South Korean income distribution toward capital, and within labor toward urban high skill labor, suggesting growing income and wealth inequality in the South. Similarly, the specific form of technological transfer is important: if North Korea succeeds in adopting South Korean technology, it experiences a tremendous boost in productivity in the traded goods sector, which tends to increase the magnitude of change in macroeconomic aggregates and distributional shares exhibited by South Korea. If integration is accompanied by large capital inflows, there is a significant appreciation of the real exchange rate with deleterious implications for the South Korean traded-goods sector.

JEL codes: F1, O5, P2 Keywords: Korean unification, North Korea, economic integration

I. Introduction

The worsening economic plight of North Korea has been widely documented.1 In response, the regime has initiated some modest reforms that do not alter the fundamental centrally planned character of the economy, but the policy changes to date are probably inadequate to address the task at hand. In one poll of scholars, 38 percent of the respondents predicted that the current regime would not last a decade (Y.S. Lee, 1995). In a more recent poll, the respondents' mean subjective probability of collapse was 26 percent, while the mean estimate of significant reform was 40 percent (Noland, 1998, Table 1).

One obvious direction of reform would be to marketize the economy and open it to greater interaction to the outside world-including South Korea. Greater North-South economic integration, either in the context of a reform strategy initiated by the North, or in the context of collapse and absorption by the South, potentially could have profound effects on both economies, yet scant effort has been devoted to constructing economic models to analyze this possibility. In this paper we construct the Korean Integration Model (KIM), a two-country computable general equilibrium (CGE) model linking the North and South Korean economies, extending earlier modeling work on the North Korean economy by Noland, Robinson, and Scatasta ((NRS), 1997). That work developed an eight sector, four factor, constant returns to scale CGE model of the North Korean economy. The single-country NRS model was used to examine three issues associated with economic reform: the static gains to trade liberalization, the increase in total factor productivity induced by the importation of capital goods embodying new technologies, and the "obsolescence shock" reduction in the capital stock as a result of the introduction of new goods and significant changes in the structure of relative prices. The model was calibrated to 1990, the last year before North Korea entered a period of severe macroeconomic instability.

The main results obtained by NRS are that (1) the static gains from trade liberalization for North Korea are potentially huge-on the order of 25 to 35 percent of GDP depending on specific assumptions about factor market adjustment; (2) total factor productivity could increase by 18 percent, leading to a roughly 50 percent increase in national income in the complete liberalization scenario; and (3) that North Korea could absorb up to an approximately 50 percent "obsolescence shock" reduction in the capital stock before national income fell under successful economic reform.

Sectoral results reported in that paper indicated that there would be an enormous shift in the composition of output and employment towards the light industry sector (and to a lesser extent mining), while agriculture and the capital goods sector would tend to contract as factors were reallocated in a more economically rational way.

The NRS model was then used to calculate the "cost of unification" defined as the addition to the North Korean capital stock necessary to increase North Korean per capita incomes to 60 percent of those of the South, a target thought adequate to choke off incentives for mass migration.² In 1990 this "cost of unification" was $319 billion, rising to $754 billion in 1995, and $1,721 billion in 2000 as the gap between North and South Korean incomes grew with delay in the initiation of reform.³ In this paper we alter and extend the NRS model to a two-country setting by constructing a similar eight sector, four factor, constant returns to scale model of the South Korean economy and linking it to the North Korean model.⁴ The Korean Integration Model (KIM) is a member of a growing family of trade-focused, multi-country, computable general equilibrium models designed to analyze the impact of trade liberalization and the formation of free trade areas and customs unions. KIM consists of two linked country CGE models, one for North Korea and one for South Korea. The rest of the world is included by means of a simple representation of fixed world prices for North and South Korean exports and imports. The countries are linked by trade flows, and the model solves for all internal prices, including commodity and factor prices, and external prices of all goods traded between the two countries. Domestically produced and traded goods are specified as imperfect substitutes, which provides for a realistic continuum of "tradability" and allows for two-way intersectoral trade.

KIM has a standard neoclassical specification, except that the model incorporates severe quantity controls in exports and imports, with concomitant distortions in domestic product and factor markets. The markets for goods, factors, and foreign exchange are assumed to respond to changing demand and supply conditions, which, in turn, are affected by government policies, the external environment, and other exogenous influences. The model can be considered medium-to-long run in that all factors are assumed to be intersectorally mobile. It is Walrasian in that only relative prices matter. Sectoral product prices, factor prices, and the exchange rate are determined relative to an aggregate consumer price index, which defines the numeraire.⁵

The model allows us to examine three issues of major importance-the mechanism of technological convergence, the macroeconomic impact on the North and South Korean economies of different modes of economic integration, and the impact on the distribution of income in and across the two economies. Using KIM we simulate the impact of a North Korea-South Korea customs union, and an exchange rate unification linking the two economies both in the presence and in the absence of cross-border factor mobility. Our results indicate that this distinction is of critical importance. If factor markets do not integrate, the macroeconomic impact on South Korea of economic integration with the North is relatively small. Far larger macroeconomic results are obtained when we unify the real exchange rates and begin allowing factor market integration. Indeed, factor market integration has a significant impact on the South Korean income and wealth distribution. If investment flows from South to North and labor flows from North to South, there is a shift in the South Korean income distribution toward capital, and within labor toward urban high skill labor, suggesting increased income and wealth inequality in the South. The magnitude of these effects is increased if North Korea adopts South Korean technology. If integration is accompanied by large capital inflows, there is a significant appreciation of the real exchange rate and output falls in the South Korean traded-goods sector.

Given the model's medium-to-long-run orientation, our focus in this paper is primarily of sectoral adjustment issues in the context of a simple macroeconomic framework. For two principal reasons we do not address a number of interesting macroeconomic issues such as exchange rate overshooting which have been prominent in the literature on German unification.⁶ First, the disimilarity of factor endowments is far more pronounced in the Korean case than in the German case (Noland, 1997). As a consequence, integration may have more dramatic sectoral implications in the Korean case compared to the German case. This fact, combined with the far larger differences in economic size between the two Koreas compared to pre-unification Germany, suggests that in certain respects NAFTA may be a closer analogue to the prospective Korean situation than the German experience with unification. KIM is well-suited for examining these integration issues.

Second, history does not operate by analogy: there is no particular reason to believe that adjustment issues that have arisen in the German case which have been at least partly due to avoidable policy mistakes (such as the wage equalization policy) will occur in the prospective Korean case. In point of fact, the Koreans can learn from the German experience and avoid some of the German errors.⁷

To cite a specific example, in contrast to the German wage equalization policy, mainstream Korean analysts expect the maintenance of the existing demilitarized zone to control population movements post-economic integration, and the perpetuation of greatly differing wage structures in the two halves of the peninsula for some extended period of time (cf. Young, Lee, and Zang, 1998). Finally, we should observe that what is modeled in this paper are a customs union and exchange rate unification; these do not require political unification, just economic integration, and on the political issue our paper is agnostic.

II. The Korean Integration Model (KIM)

KIM has eight sectors: agriculture/forest/fisheries, mining, light manufacturing, industrial intermediates, capital goods, construction, public administration, and services. There are three "demanders a single aggregate household which buys consumer goods government spends on and public administration an capital account purchases investment goods. Primary factors of production are agricultural laborhigh-skill urban low-skill labor. Land is not explicitly modeled as separate factor can be considered subsumed in capital. Sectoral production technology is represented by a set of Cobb-Douglas functions of the primary factors, while intermediate inputs are demanded according to Leontief, fixed input-output coefficients.⁸ On the demand side, import demand functions are specified as AIDS (Almost Ideal Demand System)-translog-which allows substitution elasticities to differ between domestic-, Korean partner-, and rest-of-the-world-produced goods.

KIM focuses on real trade flows, relative prices, and the real exchange rate. The aggregate price level in each country is taken as exogenous, and the model does not include money or other assets. The model includes the basic macro aggregates for each country, including the government deficit, the balance of trade, and the savings-investment balance. The balance of trade for each country is fixed exogenously (except when modeling full integration), so the model does not consider any possible macro feedbacks from trade liberalization to changes in international capital flows. The macro "closure" for each country is simple. Government revenue is determined endogenously, given a variety of fixed tax rates, while government expenditure is fixed endogenously. The government deficit is endogenous. Aggregate investment in each country is assumed to be a fixed share of GDP, and aggregate savings is assumed to adjust to equate total savings and investment.

KIM includes quantity rationing of both exports and imports. North Korea is assumed to have levels of "desired" exports and imports that would be typical for a country of its size and per capita income, but that exports and imports are rationed physically, yielding the low levels observed in the base data.⁹ South Korean trade with North Korea is similarly assumed to be rationed in physical terms, and "desired" trade between the two countries is assumed to equal levels that would be predicted from a gravity model. Trade liberalization and integration in the form of a customs union is modeled by removing all quantity rationing and imposing a common external tariff equal to South Korean tariffs.

KIM also includes a facility for modeling exchange rate unification by specifying: (1) a fixed exchange rate between North and South Korea, and (2) a unified, fixed, balance of trade for the two countries together. The result is that, in the various experiments done with this specification, the separate North and South Korean trade balances can vary, but their sum is fixed.

Modeling Quantity Controls in Trade

In the case of North Korea, the major distortion in the economy is assumed to be quantitative controls on both imports and exports. Because of data problems, discussed below, we assume no other sources of price distortions such as sectorally differentiated taxes and subsidies, which we treat explicitly in the case of South Korea. Such sectoral distortions undoubtedly exist in North Korea, but due to the organization of the North Korean economy are effectively impossible to conceptualize much less measure, so we focus only on trade liberalization.¹⁰ Demanders are assumed to treat imports and domestically produced goods as imperfect substitutes (the Armington assumption), and have an AIDS-translog sectoral import demand function that depends on the relative prices of imports and domestically produced goods on the domestic market. These demand functions are parameterized according to the "normal" levels of sectoral imports that one would expect North Korea to have without any rationing, given the results from the gravity model. Then, we assume the difference between desired imports and observed imports is due to the imposition of quantity rationing by the government. That is:

where M is imports, D is domestic supply, qr is the quantity rationing rate, and the subscript I refers to the sector.¹¹

The model also specifies sectoral export supply functions, where the export supply ratio depends on the ratio of the export price to the price on the domestic market.12 The supply functions are parameterized so that the desired ratio is consistent with the results from the gravity model. Symmetrically with the treatment of imports, quantity controls are specified so that actual exports are less than desired.

The result is that demanders are forced off their import demand curves and producers are forced off their export supply curves.¹³ The distortions are quite large, indicating large potential gains from liberalizing trade and allowing markets to clear. The trade rationing leads to major distortions in the domestic price system as well.

Data

The model utilizes four main databases, macroeconomic and microeconomic Social Accounting Matrices (SAMs) of North and South Korea for 1990, the most recent year for which data were relatively unaffected by the severe macroeconomic shocks that North Korea began to suffer in 1989 (see Appendix 1). In the case of South Korea, construction of the SAMs was straightforward. However in the case of North Korea, the approach we adopted was to draw on a variety of data sources and use a new matrix balancing technique to ensure consistency that uses an approach in the area of maximum entropy econometrics that is essentially Bayesian in that it stays "close" to known controls (or Bayesian prior) while imposing all the consistency requirements of the balanced accounts.¹⁴

Data for the North Korean macroeconomic SAM were primarily derived from North Korean government budget data as reported in Hwang (1993). One assumption made to build the macro SAM is that the North Korean government makes all investments. Government revenues are treated as being derived solely from direct household and enterprise taxes. Indirect taxes, import tariffs, and export tax rates are set to zero. In reality, revenues are raised from a transaction tax which varies depending on the legal status (state-owned, co-op, etc.) of the transacting parties, thus obviating the whole notion of a sectoral tax rate. In the absence of precise information about tax incidence, this was computed on the basis of a number of assumptions: (i) households' marginal propensity to save is between 30 percent and 40 percent; (ii) private savings are seized by the government via a number of instruments which are here summarized as a direct income tax; (iii) data about government current expenditure and investment are assumed to be reliable; (iv) part of capital/land returns are distributed to households, but capital/land income from public enterprises is appropriated by the government in the form of a enterprise tax.

The input-output coefficients are contained in a microeconomic SAM which was derived from a pre-reform (1979) Chinese input-output table compiled by the World Bank. This table was constructed to SNA standards, expanding on the material product accounts (World Bank, 1985). The assumption is that a good starting point (or prior) for inter-industry input-output relations in North Korea is pre-reform China, reflecting their common links to 1970s vintage Soviet manufacturing technology. (This assumption is subsequently relaxed, as explained below.)

Urban workers are divided into high skilled (professional, technical, and managerial) and low skilled (the remainder). The initial starting point for industry employment structure was taken from the Chinese data. The wage premium was calculated on the basis of South Korean data. While one might expect a priori that wage dispersion in the North would be less than in the South, at this level of sectoral aggregation, the skilled wage premium obtained from the South Korean data was within the dispersion observed in fragmentary data on North Korean wages. Sectoral outputs are derived from estimates of North Korean GDP (Noland, 1996) and output shares reported by the Korea Development Bank (1994). When these output shares were applied to the labor data they yielded a rural wage that was too high relative to urban wages. The agricultural sector's share was reduced to about 21 percent of value-added which reduced agricultural wages to a level more consistent with the fragmentary North Korean wage data.¹⁵ A real exchange rate was constructed from the GDP estimates reported in Noland (1996). The real (PPP adjusted) North Korean won-US dollar exchange rate was used to convert export and import data from dollars into won to obtain the domestic resource equivalent of external trade. (The model equations and further description are presented in Appendix 2.)

III. Policy Experiments

Integration is studied under two main scenarios. The formation of a customs union which involves: (a) the elimination of North Korean quantity rationing of trade, (b) the elimination of intra-Korean barriers to trade, and c) the adoption of South Korea tariffs as the common external barrier. In this scenario, there is product, but not factor, market integration between North and South Korea. The second main scenario involves exchange rate unification, fixing the real exchange rate between North and South Korea. Three variants are examined. In the first, capital moves from South to North Korea until North Korean per capita income rises to 60 percent that of the South's. In the second variant, this is achieved by allowing labor to migrate from North to South Korea. In the third variant, the per capita income target is achieved through the movement of both labor and capital. This formulation not only allows us to calculate the macroeconomic impacts of product and factor market integration, but also permits us to calculate income and its distribution with respect to the both the original populations of North and South Korea, and the post factor market integration distribution of population on the Korean peninsula.

A final issue involves the specification of the North Korean economy. As argued in NRS, liberalization of the North Korean economy is likely to involve at least three identifiable effects: static reallocation of factors according to comparative advantage; an increase in total factor productivity (TFP) associated with importation of capital equipment embodying new, superior, technology developed abroad; and an "obsolescence shock" reduction in the value of the existing capital stock.

The Static Reallocation Effect

This is illustrated graphically in Figure 1, which presents a simplified model with an imported good (M), an export (E), and a domestic non-traded good (D). The country produces two goods, D and E, and consumes two goods, D and M. The production possibility frontier is given in quadrant IV (lower right) and the balance-of-trade constraint is given in quadrant I (upper right). The consumption possibility frontier is given in quadrant II (upper left), which indicates supplies to the domestic market of domestically produced goods (D) and imports (M), with M being purchased from export receipts (quadrant I).

Figure 1 shows the movement from rationing equilibrium A to free trade equilibrium B in quadrants II and IV.¹⁶The movement from point A to point B in the fourth (lower right) quadrant changes the structure of production and will yield an increase in real output (GDP) measured in base prices, even though it represents a movement along the same production possibility frontier. In quadrant II, real expenditure (absorption) measured at base prices also increases, as does welfare (measured by the difference between two indifference curves). These three measures all reflect the increase in efficiency arising from the removal of rationing. ¹⁷

Technological Transfer

Recent research suggests that the world is characterized by international technological spillovers. These are quite important in the case of developing countries which benefit from technological innovations abroad primarily transmitted through international trade in capital goods embodying these innovations. In the case of North Korea, the parameters estimated by Coe, Helpman, and Hoffmaister (1996) indicate that complete liberalization would result in a total factor productivity gain of approximately 18 percent.¹⁸ This is depicted in Figure 1 as the movement from the rationing equilibrium A to the free trade equilibrium C on the new, larger, production and consumption possibility frontiers.

However, the specific case in hand may differ fundamentally from the generic phenomenon analyzed by Coe, Helpman, and Hoffmaister. For the purposes of their regression model they classify South Korea as a developing country. Thus no technological spillovers would be attributed to North Korea importing capital goods from the South. Moreover, in the monetary union simulations, we allow cross-border factor flows; in particular, we allow capital to move from South to North Korea. In this case it would be plausible to expect that the North would adopt South Korean technology embodied in the capital. We model this channel of technology transfer in two ways.

In the first approach, the North Korean average level of productivity is assumed to converge to the South's. Operationally, the North's production function shift parameter (its productivity level) increases to the level of the South's. In the second, more sophisticated version, not only does the North attain the South's level of productivity, but it also adopts its technology in the form of the South's input-output coefficients. The rationale behind this approach is that as new plants are built using South Korean capital, and new production technologies are adopted in North Korea, this will be reflected in the allocation of basic inputs and produced intermediates. Thus as South Korean techniques become the norm, the input-output coefficients in the North converge to those of the South. These are presumably optimal given the existing factor prices and distortions in South Korea, so their adoption by North Korea would imply the elimination of non-trade related distortions which we are unable to model explicitly.¹⁹

In the model, the sectoral production technologyIn the model, the sectoral production technology is described by a CES real value-added function of primary inputs (labor and capital) and fixed input-output coefficients, which give the demand for intermediate inputs. In the first step, we assume that North Korea achieves the South's value-added technology (sector by sector). In the second step, we assume that the North also adopts the South Korean input-output coefficients. To allow North and South value-added productivity levels to converge, we first measure the differences in the level of value-added total factor productivity (by sector) in the two countries. The value-added productivity gap is defined as:

where PRODGAPi is the productivity gap in sector I; MRP is the value-added marginal revenue product, which equals

where Xi is the output; FCTR is a primary factor (labor or capital); and PVi is the value-added price of output.

In the first approach, we adjust the sectoral shift parameters in North Korea to reflect capital transfers from the South, with the Northern shift parameter attaining the South Korean value in the final experiment.

While the technology transfer mechanism described above is arguably an improvement over the simple uniform increase in TFP in the particular application at hand, it has some unrealistic features. Although North Korea now possesses the same value-added productivity level as does South Korea, Northern production still embodies an inefficient set of production processes as embodied in its input-output coefficients. In the second step, we also allow the input-output coefficients to converge gradually. This is implemented by computing a linear combination of the North and South Korean input-output coefficients. The only difference here is that the weight or the share of the linear parameter changes in each experiment as capital investment starts to move from the South to the North. So at the last experiment, the weight attached to the South Korean input-output coefficients is one and the weight attached to the North Korean coefficients is zero. Thus, at the end of the simulation, North Korea fully adopts the South Korean input mix. This feature, combined with the North Korea productivity convergence, fully captures the technology spillover effect associated with capital migration. Indeed, in the final experiment, North Korea can be said to have adopted South Korean technology.

These two types of technology transfer are illustrated graphically in Figure 2. Before the cross-border factor flows take place, North and South Korean outputs and technologies are represented by their respective isoquants, Q(NK1) and Q(SK1), and their respective rates of factor substitution, RTSnk1 and RTSsk1. The tangent points, A1 and B1, represent the equilibrium (and optimal) factor usage, given technology and factor prices. As factor flows and technology transfers occur, the North Korean isoquant moves outward to A2 due to the increased capital investment from the South, and the South Korean isoquant moves inward to B2 due to the capital outflow to the North and labor inflow from the North. Although South Korean output declines slightly, the combined output of North and South Korea increases.

Technology transfer is indicated by the convergence of the technical rates of substitution between capital and labor in North Korea (RTSnk2) and South Korea (RTSsk2). The parallel slopes indicates that the North and South are using the same input factor proportions (though at different levels of aggregate factor supplies). If the figure were redrawn, using unit isoquants, the unit isoquants for North and South Korea would converge.

As expected, the overall level of North Korean TFP is considerably below that of the South. Sectoral differences range from no difference in TFP in the case of agriculture (remember the model is calibrated for 1990-probably the peak year of North Korean agricultural output) to more than 300 percent higher TFP in South Korean construction, with TFP differences in the non-agricultural traded goods sectors ranging between 49 percent (industrial intermediates) to 110 percent (light manufacturing). Patterns of intermediate input usage vary significantly across the two economies as well, with North Korea's use of intermediate inputs in its traded-goods sectors higher than South Korea's in most cases, and in some cases significantly so. Wastage of intermediate inputs is typical of centrally planned economies. In summary we have three ways of modeling technological transfer. In the first, North Korea experiences a uniform increase in TFP across sectors. In the second, it experiences a sectorally non-uniform convergence to South Korean productivity levels. In the third it adopts the South Korean pattern of input usage as well as achieving the South Korean level of productivity.

Capital Obsolescence

Finally, an important question involves the value after liberalization of the pre-existing capital stock. There are two points to consider. First, due to the putty-clay nature of technology, the capital stock accumulated under one set of output and factor prices is likely to be sub-optimal for different relative prices. While this is true for all economies, the impact is particularly acute for transition economies, where the relative prices under central planning were wildly at variance with those observed in world markets (and the notion of optimizing choice of technique with respect to factor prices was of questionable relevance). Second, economies sheltered from international trade may manufacture products that are essentially worthless in world markets. (Think of televisions or radios without tuners-both of which are produced in North Korea.) To the extent that capital is product-specific, this capital will be effectively worthless when the economy is opened up to trade.²⁰ Sinn and Sinn (1992) report that one-half to two-thirds of East Germany's capital stock was worthless after unification.²¹ If lack of exposure to international trade is taken as a proxy for internal distortion, the No