The Pricing Behavior of Korean Manufactured Goods During Trade Liberalization

1. Introduction

The idea that competition from imports would force oligopolistic firms in protected domestic markets to behave more competitively in setting their price is both old and widely accepted among economists and policy makers alike. There are, however, only a few empirical studies directly examining the response of domestic prices to changes in corresponding import prices. Some of these studies, conducted in the context of developed countries by Ceglowsky(1992), Levinsohn(1993), and Swagel(1996), are usually known as "competing goods effect" literature and they in general found that while the competing goods effect is significant, it is typically small. In the context of developing countries, however, we found only one empirical study on the topic, the 1989 paper by Corbo and McNelis. They conducted a comparative analysis of the response of domestic prices of manufactured goods to changes in external prices in Chile, Israel, and Korea during the period from the mid-1970s to mid-1980s. They found that domestic prices of manufactures in these countries responded substantially to changes in external prices, and that the coefficients of external prices and unit labor costs change systematically with the degree of openness of the economy.

However, while they intended a comparative analysis of pricing behavior for the three countries, the lack of uniform data prevents it from being a serious and genuine comparative analysis.¹ At the same time, their findings of high external price coefficients for these three countries appear to be overstated.² Taken literally, their high external price coefficients indicate a strong linkage between domestic and external prices which then provides the basis for a substantial competitive effect of import prices on domestic firms in setting the domestic prices of manufactured goods. However, the existence of such large competitive pressures of foreign imports on domestic prices of these industrializing countries, indicated by their results, is doubtful. In fact, domestic prices of manufactured goods in one of these countries, namely Korea, did not appear to respond to changes in external prices during a similar period, as reported by Yang and Hwang(1994). We maintain that their external price coefficients are highly overstated due largely to misspecification of their model (which will be discussed later).

The size and speed of the response of domestic prices to changes in external prices are important for the purpose of stabilization policy. For instance, if the response of domestic price to changes in external prices is quick and substantial, then exchange rate stabilization is a more effective instrument for reducing inflation in manufacturing prices. But if domestic (manufacturing) prices do not follow corresponding external prices, then a macroeconomic policy of reducing excess demand and wage policy is more effective. At the same time, the coefficients of external prices also have important implications for competition policy. If the linkage between the two prices is strong, then open international trade through trade liberalization should be considered as a viable alternative, or at least a complementary policy, to promote domestic competition.

Given the presumed importance of the size and speed of the impact of external prices on domestic prices in the formulation of these policies, it is necessary for economists and policy makers to accurately measure of the size and speed of the impact of import prices on domestic prices, particularly on prices of manufactured goods. The purpose of this paper is to estimate the size and trajectory of the response of domestic prices of Korean manufactured goods to changes in corresponding import prices and other determinants by extending the work of Corbo and McNelis.³ We use monthly aggregate time series data for the period from 1975.1 to 1996.6, which includes recent years from the mid-1980s to mid-1990s when import liberalization was pursued aggressively in Korea. In addition, we use both ordinary least squares and a Kalman-filter time-varying parameter regression, which allows the capture of the time-varying price response of domestic manufactured goods to changes in corresponding import prices during the period of import liberalization.

P^c_t = a₀ + a₁ ULC_t + a₂UMC_t + a₃ED_t, (1)

where P^c represents the rate of change in domestic manufactured goods at time t in a closed economy; and ULC represents the rate of change in unit labor cost at time t; and UMC represents the rate of change in unit material cost for all manufacturing, and ED_it represents a measure of excess demand at time t.

When all manufactured goods are homogeneous tradables and the economy is completely open with no import restrictions, then domestic prices of manufactured goods follow the law of one price, as specified in Aizenman (1986)

external prices of manufactured goods in the local currency at time t.

When the country is semi-open, it is reasonable to consider pricing behavior as a linear combination of the two pricing rules as below:

P^so_it = dP^o_it + (1-d)P^c_it (3)

where d measures the extent that an open-economy pricing rule is in force and (1-d) measures the extent that the closed-economy pricing rule is in force. Putting (1) and (2) into (3) yields the price equation for a semi-open economy:

P^so_it =d PE_it + (1-d) [a₀ + a₁ULC_it +a₂UMC_t + a₃ED_t] (4)

Corbo and McNelis at this point assume that the rates of change in UMC_t and PE_t are equal in deriving their final estimating equation:

P^so_it = (1-d)a₀ + [d + (1-d)a₂] PE_t + (1-d)a₁ ULC_t + (1-d)a₃ Ed_t (5)

While their assumption is made on statistical grounds, it is difficult to justify why the rates of change in external prices of final goods will be the same as that of unit material costs of imported raw materials. Once this assumption is made, however, the two variables (PE and UMC) are now collapsed into one variable. Consequently, the separate roles of external price and unit material cost of imported materials in the determination of prices for manufactured goods cannot be ascertained. Not being able ascertain separately the role of imported raw and intermediate materials in the determination of domestic price is an important omission, particularly for a country like Korea, which relies heavily on imported raw and intermediate materials.

In the present paper, we drop this restrictive assumption. By expanding Equation (4), we obtain the price equation for a semi-open economy:

P^so_t = a₀ (1-d) + dPE_t + a₁(1-d) ULC_t + a₂(1-d)UMC_t + a₃(1-d)ED_t (6)

4. Empirical Analysis

Since the data for unit labor costs in each sector (ULC) are not available, they are constructed following a method originally used by Hooper and Mann(1989), utilizing data on earnings, man days, employment, and an industrial production index in each sector.⁸ The data on unit material cost (UMC) is the ‘raw materials and intermediate goods’ component of import price indexes on won basis, also available from Price Statistics Summary 1993 and updated using various issues of Monthly Prices. Finally, excess demand (ED) is measured as the rate of change of new orders less the rate of change of inventories in each manufacturing sector, the same measure as used in Corbo and McNelis. Data on producer’s shipment indexes and producer’s inventory indexes in each sector are obtained from Economics Statistics Years Book and Monthly Statistical Bulletin, both published by the Bank of Korea.

OLS Estimation Results:

Table 1 reports the results of applying ordinary least squares to Equations (5) and (6) for total manufacturing together with all relevant regression statistics. We use the maximum likelihood method of Beach and MacKinnon(1978) to deal with possible autocorrelation. One of our tasks is to show, using the Korean case, how misspecifcation in the Corbo and McNelis’study overstates the magnitude of the impact of changes in the prices of imported manufactured goods on corresponding domestic prices.

The first row in Table 1 presents regression results based on the Corbo and McNelis specification of combining the influences of changes in external price and unit material costs. We note that the estimated coefficient for PE, 0.376, is fairly high and significant. Since the two price variables are measured as rates of change, this estimated coefficient is the external price elasticity and it indicates a high degree of openness of the economy. Estimated coefficients for the unit labor cost and excess demand are found to be very small and insignificant.

The second row in Table 1 presents regression results based on Equation (6) by allowing for the separate roles of external price and unit material cost in the determination of domestic price. We note that the estimated coefficients for external price and unit material cost are 0.221 and 0.200, respectively, and are both substantial and highly significant. When the second variable UMC is combined into one variable, as in the first row based on the Corbo and McNelis specification, then the estimated coefficient for the first variable PE (0.376) measures the gross effect of PE on PD. This is approximately equal to the sum of the direct effect of PE on PD (0.221) and the indirect effect of UMC on PD (0.200 *0.860), where 0.860 is the slope coefficient in the regression of UMC on PE. This is a classic case of specification bias due to the omission of a variable.⁹ Reviewing the regression results in the first two rows in Table 1, it is clear that the estimated coefficient for external price of 0.376 captures the sum of direct and indirect effects on domestic price. The extent of the upward bias in the estimated external price coefficient in the Corbo and McNelis study is equal to the size of indirect effect of UMC on PE.

While we are successful in separating the role of changes in external price and imported material cost in the determination of domestic price, the estimated external price coefficient in the first two rows may still not give a true measure of the impact of import prices on domestic price during trade liberalization. Indeed, we argue that it is substantially affected by the high rate of inflation during the period of 1980-81. There are two reasons why we believe this is the case. First, we note that the estimated external price coefficients in the Corbo and McNelis study rose sharply in the two periods of rapid inflation during 1974-75 and late 1979-1981 for both Israel and Korea. However, no serious import liberalization took place either in Israel or Korea during these periods. Second, examination of the actual data for the rates of changes in the dependent variable and external prices for Korea reveal sharp increases for the two variables during the 1980-81 period.

The effect of high inflation during the 1980-81 period on the estimated coefficient for the external price is first tested in two ways. First, we re-estimate the final price equation with data for the two sub-periods: 1975.1-1982.12 and 1983.1-1996.6. The estimated external price coefficient obtained from the first sub-period which includes the high inflation period of 1980-81 turns out to be 0.393 (in the third row), substantially higher than 0.221 in the second row. In contrast, the estimated external price coefficient from the second sub-period is very small and insignificant, as shown in the fourth row of Table 1. Second, the fifth row in Table 1 presents regression results with the final price equation after introducing the slope dummy for external price. The slope dummy variable takes the values of "1" for the 1980-1981 and "0" for the remaining years. Once we introduce the slope dummy variable for external price, the magnitude of the estimated coefficient for the external price becomes very small and insignificant, results similar to those displayed in the fourth row. In contrast, the coefficient for the interaction term for the external price, DV*PE, is 0.697 and highly significant. Examining this regression, it is interesting to note that the estimated coefficient for the imported unit material cost, UMC, is 0.108 and highly significant. In addition, the effect of inflation is also tested by omitting the years 1980 and 1981 from the regression and the results are presented in the last row of Table 1. As expected, the magnitude and the sign of estimated coefficients are very similar to those with the slope dummy variable in the above regression results in the fifth row.

Therefore, once we account for the effect of high inflation during the 1980-81 period, changes in external prices do not seem to affect the pricing decisions of manufactured goods in Korea. Does this mean that domestic prices do not respond to changes in external price at all? Not necessarily. It simply says that the rate of change in domestic prices does not respond on the average to the rate of change in import prices. As is well known, the OLS estimates are based on a fixed-coefficient model and thus provide only averages for the variable coefficients for the whole sample period. In fact, this evidence should caution us about the limitations of OLS estimates, particularly when the period of estimation includes a unusual subperiod (like 1980-81), which may be structurally different from the rest of the period. It is therefore desirable to examine how the estimated coefficients vary over time as the economy gradually opens up over the course of the period. We now turn to a time-varying parameter regression model.