# І П Булєєв - Вісник донецького університету - страница 10

В той же час необхідно відмітити, що інвестиційна діяльність комбінатом здійснюється в основному за рахунок власних коштів.

Таким чином, підприємства, які мають найкращі перспективи можуть забезпечити найбільш високу ефективність інвестицій. В свою чергу, інвестиції забезпечують підприємству можливість розвиватися, поступово переходячи від одного стабільного стану до якісно нового і підвищують його конкурентоспроможність.

Висновки: Інвестиції є забезпеченням нормального функціонування підприємства, його стабільного фінансового становища та максимізація прибутку.

Інвестиції у виробництво і нові технології допомагають вижити в жорсткій конкурентній боротьбі.

Направлення інвестицій на розширення та розвиток виробництва; недопущення надмірного морального й фізичного зносу основних фондів; підвищення рівня виробництва; підвищення якості, все це в першу чергу є забезпеченням конкурентоспроможності продукції і підприємства.

РЕЗЮМЕ

В статье определена связь инвестиционной привлекательности предприятия сповышением его производственного потенциала и конкурентоспособности. SUMMARY

The article deals with the connection of the investment attraction of the company and its production potential guarantee and competitiveness.

СПИСОК ИСТОЧНИКОВ:

1. Про інвестиційну діяльність. Закон України від 18 вересня 1991 р.

2. Інвестування: Навчальний посібник / Гриньова В.М., Коюда В.О., Лепейко Т.І., Коюда О.П./Під заг. ред. д-ра екон. наук, проф.. В.М. Гриньової. - 2-ге вид., доправ. і доп. - Х.: ВД «ІНЖЕК», 2004. - 404 с.

3. Денисенко М. П. Основи інвестиційної діяльності: Підручник для студентів вищих навчальних закладів. - К.: Алерта, 2003. - 338 с.

4. Майорова Т.В. Інвестиційна діяльність: Навчальний посібник. - Київ: «Центр навчальної літератури», 2004. - 376 с.

5. Федоренко В. Г. Інвестознавство: Підручник. - 3-тє вид., допов. - К.: МАУП, 2004. -

480 с.

Надійшла до редакції 02.12..2007року

УДК

COINTEGRATION ANALYSIS OF BULGARIAN IMPORTS AND EXPORTS

Petkov P., Head assistant-professor, PhD, Department of Mathematics and Statistics, Tsenov Academy of Economics, Svishtov, Bulgaria

l.Introduction

The demand for imports and exports in an economy and their relationships with other macroeconomic indicators are decisive factors for the design and conduct of economic policy. Determinants of trade flows have always attracted researchers in both academic area and policymaking institutions. Such an interest basically stems from the close linkage between the current account and exchange rate performances in any given economy. Closeness of an economy to some equilibrium is largely affected by its current account, balance of payments position and trade balance. An important indicator of a country's economic performance is the external account. Major external imbalances might predict future changes in a managed foreign exchange regime. External deficits or surpluses are caused by a number of factors - government spending, private consumption, income, fiscal and monetary policies etc. In a well-functioning economy, deficits are temporary phenomena that will be balanced by future surpluses. When a country conducts a "bad policy" there is distorted markets where no tendency towards balance of payments equilibrium and thus sustained external imbalances.

In this paper, we intend to determine whether there exist a long-run (static) and a short-run (dynamic) relationship between Bulgarian's aggregate import volume and export volume and their major determinants, on the basis of quarterly data for the period 2001-2007. The paper contributes in two ways. Firstly, we search for a statistical representation of real exports and imports using up to-date data so as to reflect the effects of the latest developments in the Bulgarian economy. In a simple statistical framework and using a fairly parsimonious set of explanatory variables, we demonstrate that the trade flows of Bulgaria can be explained

© Petkov P., 2007

adequately. It is important to note that parsimony of the regressors is crucial to have a clear-cut view of the trade balance. The hypothesis of existence of cointegrated relationships between aggregate import volume and its determinants, as well the hypothesis of existence of cointegrated relationships between aggregate export volume and its determinants are tested using cointegration techniques developed by Johansen (1988,1991) and Johansen and Juselius (1990,1992). If the hypotheses of no cointegration are rejected, stable long-run relationships between the aggregate import and export demand functions and their major determinants exist. Secondly, we attempt to estimate an error-correction model (ECM) to integrate the dynamics of short-run changes with long-run levels adjustment processes.

The remainder of the paper is organized as follows: Section 2 outlines the theoretical background on which our empirical analysis is based. Section 3 presents import demand and export supply functions for Bulgaria. After that, the empirical results are reported and discussed. The last section concludes the paper.

2. The theoretical background

In the literature, the investigations of the determinants of aggregate imports and exports are basically directed toward assessing the effects of a currency depreciation on the current account. There are two major approaches to investigate the effects of a real devaluation on the trade balance of a country, namely the 'elasticities' and the 'trade balance' approaches. From an econometric point of view, the elasticities approach is based on estimating the import and export demand functions. In most studies, export (import) volumes are regressed on effective exchange rates, relative export (import) price, and world (domestic) real income. Most empirical studies of international trade fall into one of two basic frameworks: "imperfect substitutes" model and the gravity model. While the imperfect substitutes framework focuses on the determinants of aggregate international trade with emphasis on structural parameters and their economic policy implications, the gravity modeling framework focuses on the determinants of bilateral trade flows, with an emphasis on location factors and their geo-political and geo-economic policy implications. Since the former approach is more relevant to our goal, we will examine in short its essence.

The key assumption of the imperfect substitutes model, presented by Goldstein and Khan (1985) is that imports and exports are not perfect substitutes for domestic goods. The model assumes that, in a simplified two-country world, each country produces a single tradable good that is an imperfect substitute for the good produced in other country. The simplest and most widely used procedure for estimating aggregate export supply and import demand functions in this framework is based on Marshallian demand function. Therefore, in order to see whether devaluation will help improving the trade balance, it is sufficient to estimate the import and export functions and to check whether the sum of the absolute price elasticity exceeds unity. The general aggregate import demand function is defined as:

M = f(Y,PM,PD)

t t t t /1 \

fi > 0,f2 < 0,f3 > 0

where Mt is volume of imports demanded, Yt is the nominal domestic income, PMt is the import price index in local currency, and PDt is the price index for domestically produced, import-competing goods; fi are the expected partial derivates of the import function with respect to the ith argument. Equation (1) suggests that demand for imports is determined by the level of expenditure or income (or another variable that captures domestic demand) and by the price of imports and domestic substitutes measured in the same currency. Under the assumption of homogeneity of degree zero in prices and nominal income, equation (1) is usually expressed with a relative price term as:

M = f(YRPM)

t t t (2) f>0,f<0

where Yt is real income, RPMt = PMt / PDt.

Correspondingly, the general function for aggregate exports has this theoretical form: X = f(YW,PX ,E*PW)

t t t t (3)

f > 0,f2 < 0,fs > 0

where Xt is volume of exports demanded by foreigners, YWt is world economic activity in constant prices, PXt is price of exports, PWt is foreign competitors' prices in the country's export markets, and E is nominal exchange rate in units of foreign currency per unit of home currency; fi are the expected partial derivates of the export function with respect to the ith argument. According to equation (3), the foreign country's demand for export is a function of its real income or expenditure, of the price of the domestically produced substitute goods and of the price of the foreign competing goods. As in equation (2), equation (3) can be rewritten with relative price term which can be viewed, alternatively, as the terms of trade or the real exchange rate:

Xt = f(YWt ,RPXt )

t t t (4)

f>0,f<0

12 where RPXt = PXt / E*PWt.

Functions (2) and (4) may be defined as the main framework that have most commonly used in empirical studies of import and export behavior. Demand for imports and supply of exports as a function of their major determinants were investigated by Khan (1974), Kouthakker and Magee (1969), Goldstein and Kahn (1985), Warner and Kreinin (1983), Carone (1996), Mayes (1981), Hooper and Marques (1995), Clarida (1991), Bahmani-Oskooee (1986). The estimation, testing, and identification methods are extremely varied -from conventional econometric methods to gravity equations, cointegration, error correction modeling and bounds testing approach. We will discuss a small subset of some fundamental studies.

Khan (1974) has estimated the world demand for several developing country aggregate exports and imports for the period 1951-1969 employing annual data. In the import demand equation are included the quantity of imports, relative price of import, and the real domestic GNP. In the export equation, the quantity of exports is regressed against relative price of export and real world income (proxied by OECD real GNP). Having estimated these functions using OLS, Khan reported that the prices did play important role in the determination of imports and exports of developing countries and Marshall-Lerner condition is satisfied.

Warner and Kreinin (1983) have also employed similar models. They have used three variants of import demand function and one export supply equation. First variant is with a relative price of import. The second one includes as separate regressors domestic prices and import prices in local currency. In the third variant independent variables are real GNP, domestic prices, import prices in foreign currencies and exchange rate. The export function includes as independent variables weighted GDP of major importing countries, the export unit value index of the country, export unit value index of the country, effective exchange rate index of local currency, expected rate of change in the exchange rate, and the weighted average export prices, expressed in foreign currency. Having estimated the demand for imports and exports using OLS techniques, Warner and Kreinin reported that the introduction of floating exchange rates appeared to have affected the volume of imports in several countries, but the direction of change varied between them. On the other side, the exchange

rate and the export price of competing countries are found to be powerful determinants of a country's export.

Bahmani-Oskooee (1986) have provided estimates of price and exchange rate response patterns by introducing a distributed lag structure on the relative prices and on effective exchange rate, applying Almon procedures (i.e. dynamics of the determination of the trade flows are involved). In the import demand model, quantity of import is regressed on relative import price (calculated as a ratio of import price level and domestic price level), real GNP and export weighted effective exchange rate. In the export supply model as regressors are included weighted average of real GNP of a country's trading partners, relative export price (calculated as a ratio of export price levels and weighted average of the export prices of a country's trading partners) and export-weighted effective exchange rate. Using OLS, Bahmani-Oskooee found that trade flows are more responsive to changes in relative prices than to changes in the exchange rates in the long-run.

Since the 1990s, cointegration and error correction modeling techniques have been use more and more in estimating the price and income elasticities of imports and exports. The some of the most recent studies using cointegration analysis in this area are Bahmani-Oskooee (1998), Bahmani-Oskooee and Niroomand (1998), Aydin et al. (2004), Bellesiotis and Carone (1997), Aurangzeb et al. (2005), Dutta and Ahmed (2004), Ahmed (2000).

Bahmani-Oskooee and Niroomand (1998) follow the previous literature without any modifications. They used a classical framework, introduced by Khan (1974), i.e. as major determinants of imports are employed relative import prices and real domestic income, and as major determinants of export are employed relative export prices and real world income. The authors concluded that for almost all countries1 devaluations could improve the trade balance.

Bahmani-Oskooee (1998) uses quarterly data with a slight modification of the classical models of export and import through the addition of nominal effective exchange rate (NEX) variable as a regressor. The expected sign of the coefficient attached to NEX variable in the import demand equation is positive and in the export supply equation, negative. He determined that the Marshall-Lerner condition is satisfied for almost all countries2.

Aydin et al. (2004) using quarterly data have estimated export supply and import demand models for the Turkish economy. Import quantity is regressed on real domestic GDP and real exchange rate, while in the export supply model as regressors are employed real domestic GDP, export prices, and the unit labor costs (ULC). The expected sign of the coefficient attached to ULC variable is negative. The authors concluded that the long-run relationship of exports with respect to real domestic income and export price is elastic, whereas the unit labor cost is inelastic. Meanwhile, the long-run relationship of imports with respect to domestic income is elastic, whereas the real exchange is inelastic. The short-run elasticities of export with respect to real income and export price are highly lower than long-run elasticity, but short-run elasticity of unit labor costs is a bit higher than the long-run elasticity.

Having provided the basic literature using the elasticities approach, we can emphasize the major common points of this part of studies. Firstly, all major studies regress import volumes on relative import prices and real domestic income; and export volumes on relative export prices and real world income. While doing this, the underlying framework is the imperfect substitutes model of the trade literature. Secondly, all elasticities approach models given above, focus on aggregate data for volume variables, such as import/export volumes and real incomes. Third, we may safely conclude that the satisfaction of Marshall-Lerner stability condition is dependent on the type of formulation employed, variables involved, and sample period. Finally, in the most studies in

1 Included countries are Australia, Austria, Belgium, Canada, Colombia, Cyprus, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, Korea, Mauritius, Morocco, Netherlands, Norway, New Zealand, the Philipines, South Africa, Spain, Sweden, Syria, Tunisia, the UK, the USA, and Venezuela.

2 Included countries are Greece, Korea, Pakistan, the Philipines, Singapore, and South Africa.

modeling aggregate import demand and export supply functions, the log-linear specification is preferable to the linear formulation.

3. The models and methodology

The standard specification of the import demand model is similar to any other demand model. It treats quantity of import demanded as dependent variable and import (or relative) price and income as independent variable. By analogy, the standard specification of the export supply model is similar to any other supply model. It regresses quantity of export supplied against world income and export (or relative) price. In modeling an aggregate import demand and export supply functions for Bulgaria, we follow the imperfect substitutes model, in which the key assumption is that neither imports nor exports are perfect substitutes for domestic goods. Since Bulgaria imports and exports only a relative small fraction of total world imports and exports, it may be quite realistic to assume that the world supply of imports to Bulgaria is perfectly elastic. This assumption seems to be realistic in the case of Bulgaria because the rest of the world may be able to increase its supply of exports to this country even without an increase in prices. To study the long-run equilibrium relation between volume of imports and its determinants in one relation and volume of exports and its determinants in another relation, following the literature and availability of the necessary data we assume that import and export demand equations take the following forms:

mt = 0O + 0yt + P2rPmt + Ppex + e (5)

and

xt = 0o + 0yw + 0rpx + 0ulc + є (6) where m = log M; y = log Y; rpm = log RPM = log (PM/PD); nex = log NEX; x = log X; yw

= logYW; rpx = log RPX = log (PX/PXW); ulc = log ULC. M is real imports; Y is domestic income; PM is import price; PD is domestic price level; NEX is nominal effective exchange rate; X is volume of exports; PX is export price; PXW is world export price level; YW is world income; and ULC are unit labor costs.

The data used for analysis are quarterly for the period 2001-2007. The data sources and definitions are cited in an Appendix.

The modeling strategy adopted in this study involves three steps:

1. Determining the order of integration of the variables by employing Dickey-Fuller

(DF), Augmented Dickey-Fuller (ADF) and Kwiatkowski-Phillips-Schmidt-Shin (KPSS)

unit-root tests;

2. If the variables are integrated of the same order, the Johansen-Juselius maximum likelihood method of cointegration is applied to obtain the number of cointegrating vector(s); and

3. If the variables are cointegrated, we can specify an error correction model and estimate it using standard methods and diagnostic tests.

4. Empirical analysis

Time-seriesproperties of the data

Before applying the cointegration and the error-correction methodology it is necessary to determine the order of integration of each variable, by testing whether they are stationary or they include a stochastic trend, i.e. how many times each variable needs to be differenced in order to achieve stationarity. To this end, we applied the DF (Dickey-Fuller, 1979), the ADF (augmented Dickey-Fuller, 1981) and the KPSS (Kwiatkowski-Phillips-Schmidt-Shin, 1992) tests. As far as the logarithms of variables are concerned, we tested the null hypothesis of non-stationarity against the alternative that the series are trend-stationary. In practice, the appropriate lag order is not known a priory, but has to determine by the researcher. Popular approaches to achieve this are information criteria, such as the Akaike Information Criterion

(АІС) and the Schwarz' Bayesian Information Criterion (BIC), and the general-to-specific approach of Ng and Perron (1995). In this study, we employ the latter approach, where one starts with a large value for lag-order (p) and sequentially eliminates the highest-order lag until it is significant at a pre-specified significance level1. The starting generous value ofp is defined using a formula, suggested by Schwert (1989) that allows the order of autoregression to grow with the sample size T: i.e., lag-length = int{4(T/100)1/4}. In applying ADF test, we use the Pantula principle (Dickey, Pantula, 1987) where we start with the null hypothesis that variable is I(2), and then reduce the order of differencing each time the null hypothesis is rejected until the null is not rejected.

Tables 1 and 2 report the results of the DF, the ADF and the KPSS tests for a second unit root and for one unit root, respectively. Comparing the DF and the ADF tests statistics in table 1 with their corresponding critical values, we conclude that the hypothesis of I(2) is rejected for all the series.

Table 1. The DF and the ADF" tests for a second unit root

Variables DF ADF

m

-17.408* b

-5.638 (3)c***

y

-4.754***

-3.732 (3)***

rpm

-11.253***

-7.792 (2)***

ulc

-5.149***

-21.112 (2)***

x

-6.615***

-8.266 (2)***

yw

-8.993***

-3.311 (3)***

rpx

-6.527***

-8.469 (2)***

nex

-6.764***

-7.913 (1)***

" The ADF test of second difference, where the null hypothesis is that there are two unit roots, is based

on the following regression: Ay = y/*Ay + ^ц/*Аy , + u .

t=i

b Three (two, one) asterisks indicate that two unit root hypothesis is rejected at the 1 percent (5 percent, 10 percent) significance level based on the model selected criteria.

c The number in parentheses is the chosen order of lags for the ADF test.

Unit-root tests are performed both in level and first difference forms using only an intercept. DF, ADF and KPSS statistics suggested that all variables include a unit root. In contrast, their first differences appear to be stationary. Therefore, each variable in our data set is integrated of order one.

_Table 2. The DF, the ADF and the KPSS tests for one unit root_

Variables

Level/First

Diff.

DF

ADF

KPSS"

Conclusion

m

Level

-1.555

0.811 (3) c

0.919***

First Diff.

-12.485***b

-8.240 (2)c***

0.056

1(1)

y

Level

-0.557

3.849 (3)

0.786***

First Diff.

-6.152***

-17.027 (2)***

0.102

1(1)

rpm

Level

-2.414

-2.674 (3)*

0.472**

First Diff.

-7.374***

-3.744 (2)***

0.176

1(1)

ulc

Level

-2.535

-0.090 (3)

0.908***

First Diff.

-4.713***

-14.629 (2)***

0.134

1(1)

x

Level

-0.849

-0.470 (3)

0.910***

First Diff.

-5.251***

-6.273 (2)***

0.052

1(1)

1 All calculations in this study were performed using Gretl 1.6.5 for Windows, where parameter p can be automatically determined.

ВІСНИК ДОНЕЦЬКОГО УНІВЕРСИТЕТУ, СЕР. В: ЕКОНОМІКА І ПРАВО, ВИП.2, 2007

Level

-0.557

1.719 (3)

0.822***

First Diff.

-6.152***

-4.355 (3)***

0.294

1(1)

rpx

Level

-1.685

-1.694 (3)

0.431*

First Diff.

-4.132***

-3.888 (2)***

0.084

1(1)

nex

Level

-1.839

-1.955 (3)

0.725**

First Diff.

-4.110***

-3.339 (1)***

0.301

1(1)

" The KPSS test is a unit root test in which the null hypothesis is opposite to that in the ADF test: under the null, the series in question is stationary; the alternative is that the series is Critical values are as follow: 10 % -0.347; 5 % - 0.463; 1 % - 0,739.

b Three (two, one) asterisks indicate the significance levels at 1 percent (5 percent, 10 percent) respectively. c The number in parentheses is the chosen order of lags for the ADF test.

Cointegration test

We are now in a position to apply Johansen (1988, 1991) and Johansen and Juselius (1990, 1992, and 1994) cointegration methodology. Their method is based on maximum likelihood estimation procedure amounts to calculating two test statistics known as X-max and trace. Before undertaking cointegration test, one first must decide about the order of the vector autoregression, i.e. specifying the relevant order of lags (p). When quarterly data are used, a common practice is to employ four lags in the procedure. Given the fact that the sample size is relatively small, we use a formula, suggested by Schwert (1989) and select 3 for the order of the vector autoregression (VAR).

The next issue raised in the process of formulation of the underlying VAR system is whether deterministic terms like a constant and a trend should enter the short and/or long-run models. To answer the question, we use the Pantula principle (see Harris, 1995, pp. 96-97.), i.e. a number of joint hypotheses tests testing simultaneously both the number of cointegrating relationships among the variables and the existence of deterministic components. For each function of aggregate import and export considered, three models are estimated. The most restrictive (named Model 2) assumes no linear trends in the levels of the data, i.e. an intercept that is restricted to the cointegration space. The second (named Model 3) assumes the existence of linear trends in the levels of the data, implying an intercept both in the long-run model as well as in the short run model. The two intercepts, when combined, leave only a constant in the short-run model. Finally, the least restrictive model (named Model 4), assumes the existence of some long-run linear growth which the model specification cannot account for, i.e. the existence of a trend term restricted to the cointegration space. The Pantula principle involves the estimation of all three models and the presentation of the results from the most restrictive hypothesis (i.e. r = number of cointegrating relations = 0 and Model 2) through the least restrictive hypothesis, i.e. r = number of variables entering the VAR -1 = n -1 and Model 4). The model selection procedure comprises of moving across the rows of the upper half of Table 3, from the most restrictive model towards the least restrictive one, and stopping when the null hypothesis is not rejected for the first time.

Table 3. Determination of cointegration rank and the case for the deterministic components"

## Похожие статьи

І П Булєєв - Вісник донецького університету