Trade Openness and Economic Growth in Nigeria: Further Evidence on the Causality Issue

The study specifically investigated the causality between the openness variable and economic growth, using data from the Nigerian economy. Previous studies in Nigeria have interpreted the regression results of output variables on the export trade variable as providing support for trade liberalization as an engine for growth with less emphasis on other measures like import. Such an interpretation is questionable since these regressions provided no means for determining the direction of causality. This paper performed causality tests with various forms of openness measures and economic growth. The results indicated a unidirectional relationship between openness and growth. This shows that an increasing level of openness will be beneficial, depending on the level of economic development in Nigeria. The result is robust across different measures of openness and analytical techniques.


INTRODUCTION
The idea that trade openness affects economic growth is not new again. However, the nature of the effect is being seriously debated in the literature. Some empirical studies have identified a positive linkage between a country's rate of economic growth and its openness to international trade, while others have failed to demonstrate such linkage, (Jin, 2002;Sinha & Sinha, 1996). The crux of the differences in these results has been the differences in methodology as well as the way the openness variables were defined (Baldwin, 2002 andAjayi, 2003).
Most striking fact about the existing volume of studies is that despite the fact that both exports and imports are equally important in promoting economic growth, most researchers have focused attention on the former (Sinha & Sinha 2003, Jonsson & Subramanian, 2001. A developing country like Nigeria is import dependent, therefore, its import effects on the growth process should not be ignored or assumed away without any empirical basis. Also, Nigeria has experimented with different exchange rate regimes, which might have implications for the trade-growth nexus. More importantly, with exception of Odusola and Akinlo (1995), no other studies on the Nigerian economy examined the causal relationship between trade (openness) variables and economic growth. The neglect of the causality nexus has implication on the correct modeling of the trade-growth equation. A causality test could provide insight on whether a single or simultaneous equation model is appropriate for trade-growth relationship. The fact that trade volume and economic growth are in tandem revealed nothing about the causal direction. Therefore, the issues of causality between trade and economic growth need to be investigated. This paper essentially contributes to existing studies in three ways: First, the paper included new time series from 1993 to 2000, which were excluded in the earlier studies on the Nigerian economy. The period of 1993 to 2000 coincided with the period when the external trade and exchange rate were extensively liberalized (see figures1 & 2) and thus could have implication on the empirical results. Second, rather than using the common narrow definition of openness, this paper used a variety of measures of trade openness and this helped to check the robustness of the causality results. Third, both VECM and standard granger causality tests were used. This was done to check whether omission of the error correction channel of influence in the granger causality test has any implication on the results. Therefore, this paper not only extended the existing literature but also improved the quality of the evidence.
The structure of the paper is as follows. Section 2 examined the trend of trade openness and economic growth over the years. Section 3 provided a review of the existing studies. Section 4 presented the methodology while section 5 analyzed the empirical results. Section 6 concluded the paper with policy implications of the findings.

TRADE OPENNESS AND ECONOMIC GROWTH IN NIGERIA
The analysis of the growth of exports and imports gives an indication as to the extent of the openness of an economy. However, trade flow analysis provides the basis of robust empirical investigation of the openness of an economy. Empirically, openness can be measured by the share of trade (import plus export) in total output, measured by the Gross Domestic Product (GDP). This is a broad concept of openness; in the narrow context, the ratio of imports or exports to GDP can represent the degree of openness of an economy.
A cursory look at Figure 1   While the economy has recorded remarkable progress in improving trade relations with other countries as reflected by the increasing rate of total trade to GDP, the rate of economic growth has remained sluggish. The low correlation depicted by the graph between the openness of the economy and the growth rate of the economy showed that openness has not contributed much to economic growth in Nigeria. In other words, Nigeria has not benefited as expected from the liberalization of the economy. Though, the above trend analysis is informative and indicative of the inverse relationship between economy growth and external trade in Nigeria, definite statements cannot be made without any empirical basis. The subsequent sections examined the empirical relationship between economic growth and openness in Nigeria.

REVIEW OF RELATED STUDIES
The relationship between openness and economic growth in developing economies has been fully analyzed by a large number of empirical papers. Primary attention has been given to the role of exports in economic performance with little attention paid to other growth promoting openness variables. In their paper, Cuadros, Orts and Alguacl (2001) employed a VAR model to examine the causal relationship between output level, inward Foreign Direct Investment (FDI) and trade in Argentina, Brazil and Mexico. They found that though FDI had significant impacts on growth, their results failed to find evidence in support of export led growth. They concluded that the fragility of their result may stem from the omission of other relevant mechanism through which openness can promote growth. Similarly, Goldberg and Klein (1999) had also opined that if capital flow is significant, focusing only on export as proxy for openness may be misleading. Using data from some Asian countries, Sinha and Sinha (2001) also reiterated that omitting import from trade openness measure creates a missing variable bias. Using data from imports and exports to capture openness, they found positive effects of openness on economic growth for some Asian countries.  Dreher (2002) using different measures of openness to capture economic globalization found that globalization promote growth but not to the extent that it can reduce poverty.
Studies carried out on Nigeria using time series data, has been scanty. Egwaikhide (1994) examined the link between the trade variable (export) and economic growth in Nigeria between 1959 and 1989, using cointegration and error correction methods. This approach is appealing as it could eliminate the problem of spurious estimates, which previous studies ignored. Ekpo (1995) also examined the openness and performance of the Nigerian economy for the period 1970 to 1992 using broad measures of openness. Using the aggregate production model, his analysis showed that capital stock and labour contributed positively to output growth during the period. However an increase in trade share measured by the black market premium rate and trade/GDP respectively, reduced output.
Oladipo (1998) extended the Ekpo and Egwaikhide (1994) model. He measured the degree of openness as the ratio of total trade (export + import) to GDP and as the ratio of export to GDP. Based on a sample period of 27 years (1970 to 1996), Nigerian quarterly data, the results showed that when the export/GDP ratio was used as a measure of openness it correlated positively with GDP growth. But, the conventional broad measure (import plus export) to GDP indicated a negative relationship. Olomola (1998) used the endogenous growth model to explore the long-run relationship between openness and economic growth. He adopted Dickey-Fuller and Augmented Dickey-Fuller test to examine the stationarity properties of the variables. Openness was proxied by export/GPD and total trade/GDP for sample period of 1960 to 1998, he found that total trade/GDP, has no significant relationship with long run growth in Nigeria.
In recent studies, Akinlo (2003a,b) examined the effects of FDI on economic growth in Sub-Saharan Africa. He found out that FDI had a positive impact on Sub Saharan African economic growth. In addition, both export and stock market development had positive effects. Though, this result is indicative of the possible role openness in the form of inward flows of capital goods could play in a developing sub-Saharan African country, the fact that the analysis was based on panel data makes country specific policy inferences difficult from the evidence. It is possible that individual country result may differ if time series data are employed. Apart from this shortcoming, one other important shortcoming of these studies is that they ignored the causal relationship between trade openness and growth. Their outcomes were monumental but the neglect of the nature and direction of influence between the trade variable and growth, made their results tentative.
Some studies have attempted addressing the issue of causality between the trade variable (especially export) and economic growth. Jung and Marshall (1985) used time series data to perform Granger (1969) causality tests between export and growth for 37 developing countries. The results provided evidence in favour of export led growth in only four instances. The result showed that the export promotion hypothesis is weaker than what previous statistical studies have indicated. Kunst and Mario (1989) also investigated the causality between productivity and exports using quarterly data for the period between 1965 and 1985. The result indicated no causal link from export to productivity.
Muhammed and Sampath (1999) also empirically examined the causality between exports (as a measure of trade openness) and economic growth for 97 countries using data for the period 1960 to 1992. While determining the stationarity of the two variables and their order of cointegration, they found that GDP and exports are integrated of different orders for 36 countries. Among the other 61 countries, for 17 countries, there were no long-run relationship between the two variables, 35 countries showed causality at least in one direction. Uni-directional causality from GDP to exports was reported for 10 countries, from export to GDP for 5 countries and bi-directional causality for 20 countries. Nine (9) countries did not show any causality between GDP and exports at all. One important issue of relevance to this study, is that Nigeria was among the 9 countries without causal relationship between exports and economic growth. The findings were contracted from existing studies. Anoruo and Ahmad (1999) used Johansen's (1991) cointegration technique instead of the Granger (1969) causality test to examine the causal link between trade openness and economic growth. In the five (5) Asian countries, selected covering 1960 to 1998, they found that both openness variables and economic growth are cointegrated and that there was a bi-directional causality between trade openness and economic growth.
Our perusal of literature on Nigeria has not indicated any detailed effort at investigating the causal link between trade variables and economic growth. The only exception was Odusola and Akinlo (1995). They used the traditional Granger (1969) causality test to examine the causal relationship between openness and GDP growth. The set of trade variables considered were export, import, and terms of trade and factor inputs, proxied by gross capital formation and labour force, using Nigerian data over 32 years from 1960 to 1992. The causality analysis using the Granger (1969)  First, the standard test adopted did not examine the basic cointegration properties of the variables. If the variables were cointegrated, their model which incorporated the difference variables, is mis-specified since the lagged error correction term is not included (Granger, 1988). Second, the noninclusion of the error correction term from the cointegration equation eliminated the long-run information embodied in the original form of the variables. The exclusion of the error correction term also foreclosed the detection of an additional channel of Granger causality through the lagged error correction terms. In view of this, Odusola and Akinlo's (1995) result could not be regarded as final and conclusive. It would, therefore, be interesting to reexamine the causality between trade openness and economic growth through the yet to be explored (error correction term) channel using data from Nigeria.
The fact that none of the existing studies was categorical about the nature and direction of the relationship between trade openness and economic growth makes further analysis imperative. Most importantly, the methodology adopted in these studies cannot capture the complex interrelationship between trade and economic growth. It cannot show the likely long run (feedback) effects that exist among trade variables and economic growth. This paper therefore addressed this empirical issue by employing Johansen (1991) cointegration and VECM procedure (in addition to the standard Granger causality test) with time series data for Nigeria and a larger set of trade variables. This method has been found to perform better than the Granger (1969Granger ( , 1980) causality approach (Anorou & Ahmed, 1999).

DATA AND METHODOLOGY
This study used annual data on gross domestic product (TGDP), industrial output (INDGDP) and openness measured by a set of trade openness variables for Nigeria. The data cover the period 1970 through 2000. The openness variable was proxied by a set of five (5) variables: export (EXPT), import (IMPT), and export plus import (TTRADE), exchanges rate premium (EXRT) and net capital inflow (NFCI Where, ∆ is the first-difference operator; t is the time trend, and l t is the stationary random error, n is maximum lag length. X t is total output (tgdp), industrial output (indgdp), and export (expt), import (impt), exchange rate premium (exrt), net capital inflow (nfci) and sum of import and export (ttrade),.
To ascertain the long run relationship between economic growth and openness, the Johansen cointegration procedure is utilized (Johansen, 1991) and Johnsen and Juselius (1990). The procedure involves the estimation of a vector errorcorrection model (VECM) in order to obtain the likelihood-ratios (LR). The VECM used in the study is as follows: Where, ∆ is the difference operator, ∆Y is total output (GDPt), industrial Output (indgdp), and export (expt), import (impt), exchange rate premium (exrt), net capital inflow (nfci) and sum of import and export (ttrade)], θ 0 represents the intercept, and l t represents the vector of white noise process. The matrix ∃ consists of r (r<n -I) co-integrating vectors. Similarly, the matrix ∀ contains the error correction parameters. In equation (2), the null hypothesis is that the matrix (Β = ∀∃') has a reduced rank of r< n -I. The alternative hypothesis, is that the matrix (Β = ∀∃') has full rank. Johansen procedure of cointegration provides two statistics. These include the value of the LR test based on the maximum eigenvalue of the stochastic matrix and the value of the LR test based on the trace of the stochastic matrix.
The causal relationship between economic growth-openness is examined with the help of Granger-causality procedure based on VECM. This procedure is particularly attractive over the standard VAR because it permits temporary causality to emerge from (1) the sum of the lagged coefficients of the explanatory differenced variable and (2) the coefficient of the error correction term. In addition, the VECM allows causality to emerge even if the coefficients of lagged differences of the explanatory variable are not jointly significant (Anoruo & Ahmad, 1999). It must be pointed out that the standard Grangercausality test used in earlier studies (e.g. Odusola & Akinlo, 1995) omitted the additional channel of influence (z t-1 ). In this study, the error correction model is based on: where z t-1 represents the error correction term lagged by one period, GDP t represents tgdp and indgdp, the growth variables, and trd t stands for the openness variables (i.e. expt; impt; ttrade; nfci and exrt). A, b, c, and d represent the optimal lag lengths obtained from the Final Prediction Error (FPE). All the variables are logarithmic and enter the co-integration analysis in levels.
In equation (3), the rejection of the null hypothesis that openness does notgranger cause economic growth, requires that (i) the θ j be conjointly statistically significant and/or (ii) the error-correction term (z t-1 ) be statistically significant. Similarly, in equation (4) the null hypothesis that economic growth does not granger cause openness is rejected on the condition(s) that the λ j s are jointly significant and/or the error correction term (z t-1 ) is significant.

EMPIRICAL RESULTS
Following the steps discussed in section 4, the result of the augmented Dickeyfuller (ADF) unit root tests are presented in Table 5.1. The non-hypothesis of non-stationarity of economic growth and trade openness variables is tested against the alternative hypothesis of stationarity. The results indicated that except for industrial output that was stationary at level, all the other series were not stationary in their levels. However, at first difference, the null hypothesis of no unit roots is not rejected as 5 per cent significance level. In all, the results indicated one order of integration for both economic growth and trade openness variables. The critical values at 5 per cent for levels and first difference are 2.967 and -2.971 respectively. The series are; ltgdp = aggregate gross domestic output; lindgdp=industrial output; lexpt=exports; limpt=imports; lttrade=sum of export and import; lexrt=exchange rate premium; lnfci= net capital inflow. All the variables are in log forms.
Having determined the stationarity of time series the next step of our study involved the application of Johansen procedure to ascertain whether trade openness variable are co-integrated with the economic growth variables for Nigerian data. In the bivariate system, consisting of economic growth and trade openness variables, the maximum number of co-integrating vectors is one so the null hypothesis is that there is no co-integrating vector and the alternative is that, there is one co-integrating vector The results of the procedure are presented in table 5.2. The null hypothesis of no co-integration between economic growth variables and trade openness (i.e. r=0) is rejected at the 5 percent significant level in all the cases. However, the null hypothesis that there is at most one cointegrating vector (r < 1) could not be rejected for the all the cases. The fact that the economic growth and openness variable are co-integrated for Nigerian economy suggests that (I) there is a long run equilibrium relationship between the time series and (ii) the existence of causality in at least one direction.
Given the results of the co-integration tests, the next step is the estimation of the VECM of equation 3 & 4 to determine the direction of causality between economic growth and trade openness variables. The results from the bivariate causality tests from the estimates of the VECM were presented in Panel A(i-ii) and B(i-v) of Table 5

.3. Panel A(i) and A(ii) contain the estimates of growth equation 3 for aggregate output (TGDP) and industrial output (INDPGDP)
respectively. The estimates of openness for each openness measure, LEXPT; IMPT; LEXRT; LTTRADE; and LNFCI, accordingly, were presented in Panel B(i-v).   * Represents rejection of hypothesis at 5 per cent of significance. The joint significance of the coefficient is ascertained by the standard F-test. The level of significance for the error correction terms is determined by standard tstatistics.
As could be observed from the Table 5.3, causality is only present in Panel A that is from openness to economic growth, not the other way round. This is evident from the significance of the error correction terms (ECMs) in Panel A and the insignificance of all the ECMs in Panel B. However, there was no causation through the statistical significance of the explanatory variables. As far as the joint significance test applied to the sum of the coefficients of each explanatory variable is concerned, there is no support for causality between trade openness and economic growth. Therefore, the principal source of causality is through the error correction terms, which validated the existence of a long-run equilibrium relationship between economic growth and openness. This result suggested that there is only uni-directional causality from trade openness to economic growth.
To check the robustness of the results and the implication of excluding ECM terms in modeling variables that were cointegrated, the standard Granger causality test was conducted. As shown in Table 5.4, except for the total trade measure of trade openness, which indicates bi-directional causality, there was no causality from economic growth to other openness variables. Unidirectional causality was indicated from trade measures to aggregate output but no causality was indicated between industrial output and the trade variables. The result from the standard Granger causality test suggested that the causal effects of economic growth on trade is weak in Nigeria, as reported by the VECM method. Thus, the estimates were robust and hence the omission of the ECM term might have resulted in specification bias, if the causality was presented only in the error correction term as observed in this paper. It is interesting to observe also that, the signs on the sum of the coefficients in most of the variables were negative. This implies ( Table 5.3) that there is an inverse relationship between the openness variables and economic growth. This also seems contradictory to the generally held view that openness should not only cause but also promote economic growth. The significant negative causal effects of increasing openness on real output growth is consistent with the argument of Aiken and Harrison (1999) and Jin (2002), that increased international competition due to openness may cause domestic investment to decline, and its decrease would be greater than an increase in capital inflows from abroad. If this net investment falls, aggregate demand will follow.
The results indicated further that trade openness or liberalization depressed the growth process in the Nigerian economy. The results are not surprising giving the nature of Nigerian external trade. Most often, the export volume is dominated by crude oil whose price and quantity is determined in the international market and has little or no relation with economic reality in the Nigerian economy. In the case of imports, the volume is skewed towards semiprocessed goods deceitfully packed as raw materials when being imported; this hindered the development of the local industries. There is a high incidence of exchange rate round tripping, a situation where some scrupulous people buy foreign currency at official rate in order to sell it on the black market. As a result, the productive sector was deprived access to foreign currency to procure locally unavailable raw materials. The premium therefore do not reflect the reality in the economy. This consequently depressed the production capacity of the economy rather than promoting it. In fact, the non-causal linkage from economic growth to trade openness indicated that the growth in Nigeria's total trade has nothing to do with Nigerian economic reality. This result thus suggested that increased trade openness might have contributed to the low economic growth in Nigeria in recent times.

SUMMARY AND CONCLUSION
For the past decade, the relationship between economic growth and openness has been a topic of sustained interest and controversy in economic development literature. This study improved on previous studies done on the Nigerian economy by suggesting a reasonable approach to re-investigating the economic growth openness nexus. Specifically, the study employed the Johansen cointegration technique for the long-run relationship between openness and economic growth. Finally, VECM was estimated to capture the permanent and transient movement between the variables contained in the co-integration equations.
The paper found that economic growth and openness were cointegrated for the Nigerian economy, thus there was a long-run relationship between them. Only unidirectional causality from economic growth to trade openness was present. This was corroborated by a standard Granger causality test. Alfaro and Chanda (2002) also argued that openness measured by FDI played an ambiguous role in contributing to economic growth. Countries with a well-developed financial market, benefit from openness while countries like Nigeria with a lessdeveloped financial market, lose more with increasing openness. The trade sector, therefore, could not be excused from the blame for overall low economic performance in Nigeria. In fact, the extensive liberalization of Nigerian external trade without taking cognizance of the economic reality in Nigeria, could have engineered the economic recession experienced in the country in recent times. The import dependent (industrial) sector, which is vulnerable to adverse development in the trade sector, seemed weak to absorb the shocks from disturbances from the international market.