Review of literature

Unequivocally, trade has played a remarkable role in the global economy, considering that the growth of real trade has outstripped that of world output. Global trade expansion during the 1995–2005 period was nearly 6% per year, almost twice as high as the growth in global GDP. The ratio of world exports of goods and services to GDP increased from 13.5% in 1970 to 32% in 2005 and many economies realised an excess of trade growth over output growth (Reference Jansen and LeeJansen and Lee, 2007). This fact is in line with the classical economists’ position that specialisation and division of labour enhance productivity and that a comparative advantage is attainable from trade because of the easy movement of scarce resources from a point of abundance to a region of scarcity.

Trade patterns in the global market have been changing since the adoption of liberalisation policies. Even following the economic crisis, trade has been growing in leaps and bounds. This has prompted many theoretical and empirical studies, focusing on both developed and developing economies (Reference Felbermayr, Prat and SchmererFelbermayr et al., 2011; Reference Moore and RanjanMoore and Ranjan, 2005; Reference Nanthakumar, Sukemi and KogidNanthakumar et al., 2011; Reference PortoPorto, 2008). For example, Reference Bernard, Redding and SchottBernard et al. (2007) see trade liberalisation as a welcome development because in the short term, it increases labour turnover in terms of the reallocation of workers from shrinking to expanding sectors. This development, they argue, has the ability to improve a national economy, by increasing productivity and generating income from world economies. Empirically, Reference Felbermayr, Prat and SchmererFelbermayr et al. (2011) show that over the long term, greater trade openness is associated with a lower structural rate of unemployment. This conclusion is based on the outcome of panel and cross-sectional data analysis of 20 Organisation for Economic Co-operation and Development (OECD)-member countries. It presupposes that trade openness has a way of encouraging production, which in turn repositions the utilisation of labour, thereby reducing the level of unemployment.

Regarding developing economies, the question of the relationship between trade openness and unemployment has been explored with mixed results. In their study of Malaysia, Reference Nanthakumar, Sukemi and KogidNanthakumar et al. (2011) find that an increase in the trade balance had negative Granger non-causality effects on the rigidity of unemployment dynamics. This implies that trade liberalisation is able to increase aggregate productivity in various sectors. Consequently, economic performance and efficiency raise the rate of labour utilisation. In Reference AlawinAlawin’s (2013) study of the trade balance and UNR in Jordan, using quarterly data for the 2000–2012 period, his major finding is the absence of a long-term relationship between the two factors. His results reveal that in the short-term, a trade balance deficit leads to unemployment and vice versa. Reference Kim and SunKim and Sun (2009) find that indicators of trade openness significantly play a role in labour market churning in most industries affected by the North American Free Trade Agreement (NAFTA) such as the automobile, chemicals and apparel sectors. This result buttresses the argument that trade openness promotes export and ushers in restructuring by some firms, often resulting in the decline of labour use in some sectors and its increase in others. The results of the study of India by Reference Hasan, Mitra and RanjanHasan et al. (2012) show no evidence of unemployment decrease due to trade reforms; on the other hand, urban unemployment falls with trade liberalisation in states with flexible labour markets and increases the employment share in net exporter industries. Their findings are similar to the cross-country estimation by Reference Dutt, Mitra and RanjanDutt et al. (2009) for 90 developing countries, even after excluding control variables. However, Reference Dutt, Mitra and RanjanDutt et al. (2009) also found a weak support for the Heckscher–Ohlin postulates that the relationship between trade openness and unemployment changes from negative to positive in labour-abundant and capital-abundant countries, respectively. A summary of the relevant empirical findings is presented in Table 1.

Table 1. Summary of relevant literature.

VECM: vector error correction model; OECD: Organisation for Economic Co-operation and Development; GDP: gross domestic product.

Similar studies relating to Nigeria are still latent. Reference Ogunkola, Bankole and AdewuyiOgunkola et al. (2006) evaluated the impact of Nigeria’s trade and investment policy reforms and reported the following findings. Trade and policy reforms do not have significant effects on aggregate output growth. The growth of investments, imports and foreign direct investments in the service sector significantly affect service output growth, while trade policy reform contributes significantly and positively to service export and import growth. However, with Nigeria’s large market for all categories of goods and services, trade openness (an important attribute of globalisation) is expected to help bring an influx of businesses that will assist in reducing the high UNR.

Overview of Nigerian economic performance

Nigeria has aspired in different ways to reposition her economy for development. This goal has compelled the country’s leaders at different periods to embark on reforms and policy formulation in various sectors of the economy. Globalisation and trade liberalisation have aimed at increasing capital inflow, technology transfer, manufacturing capacity utilisation and unemployment reduction. Domestic events have considerably affected the extent to which the positive intentions of various leaders of the country have been realised. Tables 2 –4 give an overview of how the Nigerian economy has fared in the recent past.

Table 2. Import, export and average manufacturing capacity utilisation and real GDP and total savings as a ratio of GDP (in million naira), 2000–2010.

Sources: Central Bank of Nigeria (CBN, 2009, 2012).

GDP: gross domestic product.

Table 3. Real output in activity sector of Nigerian economy, 2000–2010 (in million naira).

Sources: Central Bank of Nigeria (CBN, 2009, 2012).

Table 4. National unemployment rate, Nigeria, 1970–2011 (average).

Source: National Bureau of Statistics (NBS, 2011).

The total trade impact on GDP has shown a remarkable increase over the years, especially during the transition to civilian administration and its subsequent rule. Nigeria has also witnessed an increase in trade, resulting from globalisation and trade liberalisation of its economy (Table 2). This development can be inferred from the rising figures of the total trade of oil and non-oil, attributable to trade openness over the years. Agriculture has continuously contributed significantly to aggregate economic activity, as shown by the rising figures in the sector over the years (Table 3).

More attention and encouragement are required in the agricultural sector to optimise the utilisation of resources. The industrial sector’s real output rose from 2003 to 2006, fell slightly till 2007, then declined somewhat again. There have been efforts to expand appropriate industries in the rural sector in order to help raise its output. Moreover, the service sector has shown remarkable improvement, basically due to the government’s attention, while employment in this area is gradually rising. A positive change in real sector output was noted in 2010, resulting from government-intensified efforts. However, it is thought that the sectors mentioned would have achieved higher levels of performance those recorded, if not for the Boko Haram’s increased terrorist attacks in the northern part of the country and infrastructural deficiency. Nevertheless, Nigeria has experienced high UNRs over the years, as shown by the figures in Table 4. The steady rise in the UNR from 1970 to date is caused by the huge number of school dropouts and an increase in job seekers every year, with very few trained to opt for self-employment. This situation is also a result of the low savings rate that does not generate sufficient levels of investment and economic activity to hire unemployed workers. Furthermore, the 2007–2008 economic crisis further increased the UNR, already in double digits since 2001. This case is reflected by the adverse consequences of the global financial crisis in terms of massive job losses and labour supply exceeding demand. Similarly, a sharp spike in the UNR from 1990s to 2010 can be observed in Figure 1. The impacts of the policy shocks of 1986 and the crisis of 2007–2008 have had significant effects on the trend in this variable. The government’s efforts to tackle this macroeconomic problem are yet to achieve desirable results as unemployment continues to rise, reaching about 23.9% in 2011 (NBS, 2011).

Figure 1. Unemployment rate (1970–2010).

Source: plotted from data in National Bureau of Statistics (NBS, 2011).

UNR: unemployment rate.

The data

This study is based on an examination of annual data from 1970 to 2010, sourced from the Central Bank of Nigeria’s (CBN, 2009, 2012) Statistical Bulletin, NBS (2011) and the World Bank’s (2012) development indicators dataset. The vector of variables included percentage changes in the UNR, the commodity price index (CPI), real recurrent government expenditure on education (RGE) ^{Footnote 4} and the ratio of import and export to the RGDP as openness measures (OPEN). RGDP and per capita income (gross domestic product per capita (GDPC)) were included as alternative variable; hence, two different equations were estimated. The CPI and RGE were included as control variables. The underlying assumption was that the former would capture variability in primary commodity prices, given that the Nigerian economy is primary-sector driven ^{Footnote 5} (see Reference Dollar and KraayDollar and Kraay, 2004); hence, a variability in the foreign price shock would be transmitted more into the domestic economy, given the agrarian nature of the Nigerian economy. It was assumed that the latter variable would affirm that the increased public expenditure on education would improve the quality of human capital and, by extension, reduce unemployment. ^{Footnote 6} All variables were in their natural log forms, while a dummy variable was applied within the deterministic terms, in order to capture the impacts of the introduction of the SAP in Nigeria. JMulTi4 was the statistical software used in the estimation and analysis.

Graphs of the key variables reveal certain peculiarities. A keen observation of the fluctuating behaviour of the UNR in Figures 1 and 2 shows the trade openness variable. From 1980 till about 1986, trade openness shows a downward trend, reflecting the trade situation before the introduction of the SAP (Figure 2). An upward movement of the variable is also seen after the introduction of the trade liberalisation policy that allowed trade flows, devoid of restrictions on some goods and services. In Figure 3, changes in real output and per capita income show similar patterns over the sampled period, especially the gradual upward movement of these variables. However, we can equally confirm a downward trend in the variables during the crisis episodes. The World Bank’s (2013) development indicators show a recent growth in Nigeria’s real output to about 7.3% per year, surpassing those of South Africa and Turkey at 2% and 4%, respectively.

Figure 2. Trade openness measure (1970–2010).

Sources: plotted from data in Central Bank of Nigeria’s (CBN, 2012) Statistical Bulletin.

OPEN: trade openness measure, as defined in text.

Figure 3. Real gross domestic product and income per capita (1970–2010).

Sources: plotted from data in World Bank’s (2012) development indicators.

In this respect, an investigation of the dynamic relationship among these variables calls for the need in determining whether the variables are integrated or not, which naturally lead to capturing the long-run equilibrium if any.

Unit root tests

Based on Reference Nelson and PlosserNelson and Plosser’s (1982) traditional view that most macroeconomic time series variables have a basic unit root property, the initial step would have been to test whether the mean and variance of the variables are constant or change over time, employing the testing procedure of the Augmented Dickey–Fuller (ADF) and Fuller tests (Reference Dickey and FullerDickey and Fuller, 1979; Reference FullerFuller, 1976). However, in the presence of a shift in the level of the data-generating process that may arise due to structural breaks in the economy, an ADF test that ignores such shifts will have low power (Reference PerronPerron, 1989, Reference Perron1997). In order to account for this, a shift function may be added to the deterministic term while testing for a unit root. ^{Footnote 7} The shift function may be based on a simple shift dummy, such that d_tT_B captures the shift in the series by allowing d_tT_B = 0 for t ≤ T_B and d_tT_B = 1 for t > T_B . If the coefficient of the dummy variable is statistically significant at a given significance level, it means that there is an instantaneous shift to a new level after the break date, T_B . Alternatively, as Reference Lanne, Lütkepohl and SaikkonenLanne et al. (2002) and Reference Saikkonen and LütkepohlSaikkonen and Lütkepohl (2000) argue, the shift may be gradual, spreading over some periods of time, rather than being a simple shift: this then requires the estimation of a non-linear shift function. Hence, we applied the tests for the presence of unit roots, allowing for structural breaks in the series in the form of simple shifts or non-linear shifts. Should all variables be integrated in the same order, cointegration tests would also be conducted, allowing for these structural changes (Saikkonen and Lütkepohl (S&L) test) (Reference Saikkonen and LütkepohlSaikkonen and Lütkepohl, 2000).

Vector error correction model

The presence of cointegration involves estimating a vector error correction model (VECM). The VECM particularly captures both long-run and short-run dynamic relationships, using the restricted, generalised least squares (GLS) method. These restrictions are performed to allow for precision and a more robust estimate. The model is represented as

$\Delta y_t=\Pi y_{t-1}+{\Gamma }_1\Delta y_{t-1}+{\Gamma }_2\Delta y_{t-2}+\cdots +{\Gamma }_{\rho }\Delta y_{t-\rho +1}+C{d}_t+u_t$

where $\Pi =\alpha {\beta }^{\prime }$ ; $y_t=(UN{R}_t,CP{I}_t,OPE{N}_t,RGDP{C}_{T}orRGDP,RGE)$ , Δ being the differencing notation; $d_t$ represents the deterministic terms such as constant, a trend variable and shift or impulse dummy variables, and $u_t$ is the white noise stochastic disturbance term. Matrix $\alpha$ , including the speed of adjustment parameters, is also estimated, while matrix β is the cointegrating matrix and ${\Gamma }_i$ is the matrix of the short-run parameters and C is the parameter for deterministic terms. In the set-up of the model, the first term shows the long-run relationship among the variables, such that $y_t$ changes in response to one period lagged deviation from the long-run equilibrium or in response to stochastic shocks. Matrix $\alpha$ is the speed of adjustment parameters. The matrix for deterministic terms also includes the dummy variable $D_t=0$ for $T_B<1986$ and $D_t=1$ for T_B ≥1986, as well as a trend variable among the short-run parameters (not restricted to the long-run, cointegration equation). The variables real gross domestic product per capita (RGDPC) and RGDP are alternatively included in the equation.

Residual analysis of the model was tested for adequacy of the estimated model, especially for autocorrelation conditional heteroskedasticity (ARCH effects) and Reference LütkepohlLütkepohl (1991) and Chow tests for parameter stability (Reference HansenHansen, 2003).

Empirical analysis

A unit root test based on Reference Lanne, Lütkepohl and SaikkonenLanne et al. (2002) and Reference Saikkonen and LütkepohlSaikkonen and Lütkepohl (2002) was conducted, allowing for both shifts and impulse dummies, respectively, and accounting for any structural break in the series. The summary of the test results is presented in Table 5.

Table 5. Unit roots tests allowing for structural breaks.

Critical values based on Reference Lanne, Lütkepohl and SaikkonenLanne et al. (2002) and Δ is the differencing operator.

All variables have a unit root at the levels (before taking the first differences) but became stationary after taking the first differences. Thus, all variables are integrated into order 1, I(1). This implies that we need to check for the presence of cointegration relationships among the variables. The cointegration test results, allowing for structural breaks (Table 6), indicate one cointegrating relation in both equations, using GDPC and RGDP alternatively. Lag length was determined using the Hannan–Quinn and Schwarz information criteria.

Table 6. S&L cointegration tests for level shifts, $y_t=(UN{R}_t,CP{I}_t,GDPC,G{E}_t,OPE{N}_t)$ .

S&L: Saikkonen and Lütkepohl; GDPC: gross domestic product per capita; RGDP: real gross domestic product.

c: constant; t: trend; sd: shift dummy; based on Reference Saikkonen and LütkepohlSaikkonen and Lütkepohl (2000).

Estimation applying the VECM

Having established a single cointegration relationship among the variables, a VECM was estimated, based on rank r = 1 and three lagged differences, for the 1974–2010 sample period. All lag lengths were determined by minimising the information criteria of Hannan–Quinn criteria and Schwarz criteria. The estimated, long-run cointegration relationship is presented in Table 7.

Table 7. Estimated long-run cointegration vector.

GDPC: gross domestic product per capita; RGDP: real gross domestic product.

Coefficients indicate significance at the 1% level. Figures in ( ) show standard errors while those in { } are the p-values.

As observed, the coefficient of the first variable is normalised to unity. Estimates of the error correction model in both equations show that recurrent expenditure is not significant in reducing national unemployment, while GDPC and RGDP have large impacts on unemployment, with a negative sign, and are highly significant in reducing unemployment. Trade openness increases unemployment, given the highly significant and positive sign of the estimated coefficients. This result is consistent with the findings of Reference Helpman, Itskhoki and ReddingHelpman et al. (2010), Reference Menezes-Filho and MuendlerMenezes-Filho and Muendler (2011) and Reference Mitra and RanjanMitra and Ranjan (2007), where in frictional labour market conditions, trade openness leads to an economy-wide unemployment. Its significance is, however, minimal, at about 2.8% increase in unemployment. The dummy variables for 1986, capturing the structural change in the economy, also have significant negative impacts on the cointegration relationship. By implication, the SAP’s introduction in 1986 was effective in reducing unemployment in Nigeria as seen in both equations.

Considering the speed of adjustment parameters, the error correction term is said to be significant in the RGDP equation, except for CPI and UNR. This implies that for any discrepancy from the long-run equilibrium, these two variables are unresponsive in reverting the relationship to equilibrium. For the GDPC equation, only OPEN variable can react to return the system to equilibrium. The loading matrix estimates are reported in Table 8.

Table 8. Speed of adjustment coefficients for GDPC and RGDP (1974–2010).

GDPC: gross domestic product per capita; RGDP: real gross domestic product.

The estimates indicate that in the case of a discrepancy in one lagged period deviation from the long-run equilibrium, the most effective response to restore the system to equilibrium is trade openness, which will rise by about 12% and 1% in both equations, respectively. Also seen from the RGDP equation, RGDP and RGE will have to rise by about 2% and 6%, respectively, to bring the system back to equilibrium. It is also observed that the RGDP has not made a significant improvement in reducing unemployment, given the surging UNR in Nigeria, as previously reported in Table 4.

The short-run relationships further show that unemployment is significantly explained by its own first and second lags, while openness is significant in all lags, but with a negative relationship with unemployment in the first and second lags, then a positive relationship for the third and fourth lags, respectively, in both equations.

Diagnostic tests

The diagnostic tests were based on residual analysis and the Chow test for parameter stability. In checking for the presence of any possible autocorrelation ^{Footnote 8} in the residuals, we applied the Portmanteau test statistics under $H_0;E(u_t,u_{t-1})=0$ , where i = 1, …, h against the alternative that autocorrelation is not 0. To check for higher-order autocorrelation, the Lagrange multiplier (LM) test was also considered, as displayed in Table 9. The test results confirm no autocorrelation in the residuals. Besides, the univariate ARCH-LM tests indicate no ARCH effects in the residuals.

Table 9. Diagnostic test for VECM for GDPC and RGDP.

$Q_{10}$ : Portmanteau test statistic at lag 10; $Q_{10}^{\ast }$ : adjusted Portmanteau statistic for small samples; LM: LM-type test for autocorrelation with three lags; VECM: vector error correction model; GDPC: gross domestic product per capita; RGDP: real gross domestic product.

The Reference ChowChow (1960) tests were conducted for parameter stability under the null hypothesis of invariant coefficients and residual covariance matrices against the alternative (Reference Doornik and HendryDoornik and Hendry, 2001; Reference HansenHansen, 2003). However, Reference Candelon and LütkepohlCandelon and Lutkepohl (2001) proposed the use of bootstrapped values to accommodate small sample properties. The test results are presented in Table 10.

Table 10. Chow test for VECM for RGDP and GDPC.

VECM: vector error correction model; GDPC: gross domestic product per capita; RGDP: real gross domestic product.

Findings and recommendations

The empirical findings show that the cointegration relationship is significant where both the output and openness variables react to any disequilibrium in the long run. It is observed that recurrent expenditure on education in Nigeria could not properly explain changes in national unemployment. This could be due to the unstable nature of education funding in Nigeria, both at the formal and informal levels. Furthermore, an increase in both RGDP and per capita income is found to reduce unemployment at a minimal rate of about 1%. The magnitude of the coefficient appears insufficient and could be interpreted as representing the high UNR in Nigeria, despite the phenomenal growth rate in output over the years.

The foregoing results are not surprising, given the role of the Niger Delta crisis, a high crime wave such as kidnapping and the recent, incessant terrorist attacks by Boko Haram. This means that the government needs to employ more solid macroeconomic policies to prevent market failures, as well as undesired conditions such as terrorism. This solution necessitates the establishment of additional industries, especially in rural areas, with a view to harnessing idle resources. However, this remedy cannot work in isolation; the complement of other viable factors is also needed. The power sector must be tackled in all its ramifications.

Educational reform is also imperative to implement effective standards for admitting prospective students into areas of study in which they have high aptitudes and inclination. Skills development should be a major emphasis of education, not merely obtaining the certificate. Consequently, and over the long term, school graduates will be relevant in one sector or the other to minimise over-dependence on white-collar jobs that are not sufficiently available. The three tiers of the government have to plan job creation in line with graduations from schools.

Trade openness and commodity prices have raised the UNR instead of reducing it, contrary to expectations. Additionally, trade openness has increased the capital flight and brain drain from developing countries.

Furthermore, several factors arising from corruption, infrastructural deterioration and poor-functioning institutions could also pose a huge challenge to addressing the ever-increasing UNR. Improvement in the business environment through capital infusion will go a long way to encourage both domestic and foreign investors, thus optimising the utilisation of resources. Further efforts are also required to reduce the high degree of corruption in Nigeria. As Reference Uma and EbohUma and Eboh (2013) pointed out, corruption is one of the greatest obstacles to the development of the Nigerian economy, as it has brought about high-level greed and misallocation of scarce resources.

Obviously, there is a high cost of governance in Nigeria, as well as considerable income disparity between the political office holders and other lower ranking employees of the government. It is necessary to reduce the remuneration of political executives to have more funds available for industrial establishments. Moreover, high import duties and multiple taxes should be eliminated in the Nigerian economy. Reference OdusolaOdusola (2006) pointed out that in addition to federal income tax, companies are subjected to a wide range of taxes, levies and rates at the state and local levels. This situation can be an obstacle to foreign investment in Nigeria.