ISSN: E 2347-226X, P 2319-9857

All submissions of the EM system will be redirected to Online Manuscript Submission System. Authors are requested to submit articles directly to Online Manuscript Submission System of respective journal.

The Role of Agriculture Sector in Shaping Nigerian Economic Development in the Light of Slow Global Economic Recovery and Global Financial Developments

Sadiq MS1,2* and Singh IP1

1Department of Agricultural Economics, SKRAU, Bikaner, India

2Department of Agricultural Economics and Extension Technology, FUT, Minna, Nigeria

*Corresponding Author:
Sadiq M Sanusi
Department of Agricultural Economics
SKRAU, Bikaner, India

Received Date: 12/11/2017; Accepted Date: 22/12/2017; Published Date: 27/12/2017

Visit for more related articles at Research & Reviews: Journal of Agriculture and Allied Sciences


This empirical research investigated the future outlook of Nigeria economic development in the light of slow global economic recovery and global financial developments using long-run and short-run [Error correction model (ECM)] regression models. The study made use of annual time series data covering Gross domestic product (GDP), agriculture, industry and services sectors which spanned from 1990-2012. The findings showed that these variables have long-run association and the GDP was found to establish long-run equilibrium with these economy drivers, though the convergence rate was very slow as indicated by the attractor coefficient. Furthermore, findings showed that all the economic indicators exert positive influence on GDP with agriculture sector having a lead influence when compared to the other economic indicators. Therefore, the study recommends that government should adopt adjustment strategies that hinges on shoring-up non-oil revenues to compensate for the dwindling oil revenues given that the prospect for the country economy depend on the policies articulated for the medium-to-long term and the seriousness with which they are implemented.


Economic indicators, development, economy, ECM, Nigeria


The concept of economic development has been in existence in the west for centuries and has been frequently used by the economists, politicians, and others since 20th century. The term refers to economic growth accompanied by changes in output distribution and economic structure-concerned with quality improvements, the introduction of new goods and services, risk mitigation and the dynamics of innovation and entrepreneurship. Whereas economic development is a policy intervention endeavour with aims of economic and social well-being of people, economic growth is a phenomenon of market productivity and rise in (GDP). Consequently, as an economist Amartya Sen points out, “economic growth is one aspect of the process of economic development”.

The economic aspirations of Nigeria have remained that of altering the structure of production and consumption patterns, diversifying the economic base and reducing reliance on oil, with the aim of putting the economy on the part of sustainable, all-inclusive and non-inflationary growth. The implication is that while rapid growth in output, as measured by the real gross domestic product (GDP), is important, the transformation of the various sectors of the economy is even more critical. This is consistent with the growth aspirations of most developing countries, as the economy structure is expected to change as growth progresses.

Since independence, successive governments in Nigeria have pursued the goal of structural changes without much success. The growth dynamics have been propelled by the existence and exploitation of natural resources and primary products. Initially, the agricultural sector, driven by the demand for food and cash crops production was at the centre of the growth process, contributing 54.7% to the GDP during the 1960s. The second decade of independence saw the emergence of the oil industry as the main driver of growth. Since then, the economy has mainly gyrated with the boom burst cycles of the oil industry. Government expenditure outlays that are dependent on oil revenues have more or less dictated the pace of growth of the economy. Looking back, it is clear that the economy has not actually performed to its full potential, particularly in the face of its rising population.

The economy of the country has grossly underperformed relative to her enormous resource endowment than her peer nations despite its natural and human resource potentials: its economic performance has been rather weak and does not reflect these endowments. When compared with the emerging Asian countries, notably, China, India, Malaysia, Indonesia and Thailand that were far behind Nigeria in terms of GDP per capita in 1970, these countries have transformed their economies and are not only miles ahead of Nigeria, but are also major players on the global economic arena.

Research Methodology

The present study used deflated price annual time series data covering GDP, agriculture, industry and services sectors spanning from 1990-2012, sourced from CBN database. The data were synthesized using simple regression model, ADF unit root, Elliot-Rothenberg-Stock (ERS) ADF-GLS test, and linear and symmetric price mechanism model (ECM). The empirical models used are given below:

Simple regression and Engel-Granger ECM model

Consider a multivariate co-integration model as follows:

image (1)


GDP=Gross Domestic Product



εt=pure random walk

Co-integration of the multiple variables can be tested if all the variables display the same order of integration. The revenue adjustment mechanism between these variables, measured by Equation (1), was estimated through the Ordinary Least Squares (OLS) approach. ADF unit root test was applied to the residual of the estimation. These variables are said to be co-integrated if their residual is stationary, suggesting that there is a revenue adjustment mechanism between these multivariate series, which makes them converge to their long-term equilibrium relationship. In addition, short-term integration tests enable checking whether revenue responses on the variables are instantaneous. The short-term relationship is derived from the Granger [1] representation theorem in the form of an Error-Correction Model (ECM) and is presented as follow:

ΔGDP = α0 + α1ΔAgriculture + α2Δindustry + α3ΔServices + α4εt-1 + Vt (2)


Δ=First difference


α1-4=coefficient of variable

εt-1=lagged value of the residual derived from Equation (1); and,

Vt=white noise

Augmented dickey fuller test

The Augmented Dickey-Fuller test (ADF) is the test for the unit root in a time series sample [2-4]. The autoregressive formulation of the ADF test with a trend term as cited by [5,6] is given below:

image (1)

Where, pit is the price in market i at the time t, Δpit (pit – pt-1) and α is the intercept or trend term.

ARIMA (Autoregressive integration moving average) model

A generalization of the ARMA models which incorporates a wide class of non-stationary time-series is obtained by introducing the differencing into the model. The simplest example of a non-stationary process which reduces to a stationary one after differencing is Random Walk. A process {yt} is said to follow an integrated ARMA model, denoted by ARIMA (p, d, q), if ∇dyt = (1-β) dεt is ARMA (p, q), and the model is written below [7,8]:

image (22)

Where, εt ~ WN(0, σ2), and WN indicates white noise. The integration parameter d is a non-negative integer. When d = 0, ARIMA (p, d, q) = ARMA (p, q).

Forecasting accuracy

For measuring the accuracy in fitted time series model, mean absolute prediction error (MAPE), relative mean square prediction error (RMSPE), relative mean absolute prediction error (RMAPE) [9] and R2 were computed using the following formulae:

MAPE=1/T Σ {At – Ft}                               (23)

RMPSE=1/T Σ {(At – Ft)2 / At}                     (24)

RMAPE=1/T Σ {(At – Ft) / At} X 100             (25)

image (9)                 (26)

Where, R2 = coefficient of multiple determination, At = Actual value; Ft = Future value, and T = time period(s)

Results and Discussion

Estimates of long-run effects using simple regression model (Equation 1)

The regression of a non-stationary time series on another may cause spurious/nonsense regression as evidenced by the coefficient of multiple determination R2 (0.9987) which was greater than Durbin-Watson statistics (0.9909) as shown in Table 1a. A spurious/nonsense model is not desireable given that it is not ideal for policy making and cannot be used for longrun prediction. Furthermore, a unit root test on the residual variable at level generated from the simple regression (Equation 1) was found to be stationary as indicated by the ADF τ-statistics which was greater than the Engel-Granger critical value at 5% significance level (Table 1b), implying that the double logarithm simple regression model is not a spurious regression, thus a long-run model which can be use for ideal policy making and long-run prediction. Also, the stationarity of the residual variable implies that all the variables included in the model were cointegrated i.e these variables have long-run association or they move together in the long-run-these variables shared the same stochastic trend. Furthermore, a perusal of Table 1a shows that all the predictor variables included in the model viz. agricultural, industry and services sectors significantly influence the country GDP positively in the long-run. For ideal policy making it implies that 100% increase in the revenues that accrue to each of this economy driver’s viz. agriculture, industry and services sectors will increase the country GDP by 47.4, 35.6 and 13.3%, respectively.

Variable Coefficient Standard error t-value
Intercept 0.278 0.061 4.56***
Ln(Agriculture) 0.474 0.047 10.15***
Ln(Industry) 0.356 0.036 9.86***
Ln(Services) 0.133 0.042 3.17***
R2 0.9989
R2-Adjusted 0.9987
D-W stat 0.9909
F-statistic 5657.47***
AIC -3.542
SIC -3.345
HIC -3.492

Table 1a. Results of simple regression showing long-run effects.

Variable t-statistic Engel-Granger critical value Decision
5% 10%
Residual (U) -3.64 -3.34 -3.04 Stationary at level I(0)

Table 1b. ADF unit root test on residual.

Estimates of short-run effects using error correction model (ECM)

If variables are found to be cointegrated we can specify an error correction model (ECM) and estimate it using standard methods and diagnostic test. Also, since these variables were found to be integrated in the long-run i.e there is co-movement or long-run association between these variables, they are likely to establish long-run equilibrium, thus, the need to model a short-run equation which will capture both the long-run and short-run equilibriums. The results of the unit root test applied to the transformed variables at level using ERS (Elliot-Rothenberg-Stock) rejected the alternative hypothesis of stationarity in favour of the null hypothesis (non-stationarity) as indicated by the estimated t-statistics which were higher than the t-critical value at 10% significance level. At first difference the unit root test performed on all the transformed variable series rejected the null hypothesis of non-stationarity in favour of the alternative hypothesis (non-presence of unit root) as shown by the estimated t-statistics which were lower than their respective t-critical values at 5% significance (Table 2a). In summary, the ERS unit root test indicated that all the variable series at level where non-stationary i.e have unit root, but at first difference all the variable series became stationary i.e no unit root present, thus, meaning that these variables were integrated of order one. The reason for the application of the ERS (Elliot-Rothenberg-Stock) test also termed ADF-GLS test against the ADF test which is the most widely and commonly used unit root test was due to Type II error inherent in the estimated ADF test results (though not reported in the Table) which clearly proved that the ADF test had lost its power to test for stationarity due to the presence of structural points (SAP and post-SAP periods) in the captured data. It is worthy to note that [10] as cited by [11,12] advocated that all the traditional unit root test models (DF, ADF and PP tests) should be discarded because of their weaknesses which cause Type I and II errors. The results of the short-run model which capture both the long-run and short run equilibrium is shown in Table 2b. The attractor coefficient termed the error correction term (ECT) of the GDP against the economy driver sectors was found to be negative sign and significant, indicating that the GDP established a long-run equilibrium with all the economy driver sectors. The estimated attractor coefficient of the GDP was -0.278, indicating that the GDP absorbed 27.8% of the shocks in order to maintain a long-run equilibrium per annum i.e it corrects its previous error from the long-run equilibrium due to any shock from short-run equilibrium at the speed of 27.8% annually, which will take approximately 8 months and 10 days for GDP to re-establish long-run equilibrium. Therefore, it can be inferred that there was long-run causality running from the economy driver sectors jointly to the GDP. However, the speed of convergence of the GDP towards the long-run equilibrium level with the economy driver sectors was very slow, and this can be attributed to inefficiency of the reform programmes in the country. The results of the short-run equilibrium show that each of the sectors viz. agriculture, industry and services sectors has short-run causal effect on the GDP. In other words, the agriculture, industry and services sectors positively influence the GDP in the short-run. The implication is that a 100% increase in the sectoral revenue of agriculture, industry and services will increase the GDP by 36.5, 44.3 and 27.4%, respectively (Table 2a). Except for the industrial sector, all the long-run elasticities for agriculture and services were found to be higher than their respective short-run elasticities. Therefore, it can be inferred that the impact of holistic reform programmes targeting agriculture and services sectors-non-oil sector, if well planned, executed and sustained will be more visibly in the long-run while impact of reform programmes targeting industrial sector will be more visible in the short-run.

Variables t-statistics t-critical value (5%) Decision
Ln(GDP) -0.802 -1.608 Non-stationary I(0)
Δ(GDP) -3.517 Stationary I(1)
Ln(Agriculture) -1.017 -1.608 Non-stationary I(0)
Δ(Agriculture) -1.892 Stationary I(1)
Ln(Industry) -0.570 -1.608 Non-stationary I(0)
Δ(Industry) -5.215 Stationary I(1)
Ln(Services) -1.288 -1.608 Non-stationary I(0)
Δ(Services) -1.763 Stationary I(1)

Table 2a. Unit root tests.

Variable Coefficient Standard error t-value
Intercept -0.018 0.0069 -2.583**
Δ(Agriculture) 0.365 0.0376 9.718***
Δ(Industry) 0.443 0.0166 26.67***
Δ(Services) 0.274 0.0495 5.533***
ECTt-1 -0.278 0.1494 -1.858*
R2 0.99
R2-Adjusted 0.98
D-W stat 2.380
F-statistic 464.58***
AIC -4.913
SIC -4.665
HIC -4.855

Table 2b. Results of ECM showing short-run effects.

Diagnostic testing

Table 2c shows the diagnostic statistics results of the ECM model. The test of autocorrelation showed that the residuals were not serially correlated as indicated by the Breusch-Godfrey Lagrange Multiplier (LM) statistic which was not different from zero at 10% probability level (p>0.10), thus the acceptance of the null hypothesis of no autocorrelation. The Arch test results revealed that the variance of the current residuals and that of the lagged residuals do not correlate as indicated by the Q-statistic which was not different from zero at 10% probability level (p>0.10), thus the acceptance of the null hypothesis of no Arch effect. The stability test depicted in Figure 1 show that the Cusum line was within the boundary of 5% probability level, implying that there was no structural break in the equation: the model was not misspecified. Furthermore, the result of normality test showed that the residuals were normally distributed as indicated by Jarque-Bera statistic which was not different from zero at 10% probability level (p<0.10). Therefore, based on the outcome of the diagnostic statistics, it can be inferred that the ECM model used was the best fit and valid for prediction.

Test Statistic P-value
Autocorrelation Breusch-Godfrey LM test (F-stat) 0.959 0.342
Breusch-Godfrey LM test (Obs. R2) 1.244 0.265
Arch effect Q-stat 0.0347 0.852
Normality Jarque-Bera 0.662 0.718

Table 2c. ECM Diagnostic checking.


Figure 1. Stability test.

Forecasting using ARIMA

Various combinations of the ARIMA models were tried after the first differencing of each variable series, and based on the smallest AIC value the best ARIMA model was selected. Of all the ARIMA models tested, ARIMA (0,1,1) model proved to be the best for almost all the variables except the Agriculture sector which proved that ARIMA (1,1,0) was the best given that it has the lowest AIC value (Table 3a).

Variable 1,1,1 1,1,0 0,1,1
GDP AIC 307.48 305.75 305.50**
SBC 310.76 307.93 307.68
Log likelihood -150.74 -150.88 -150.75
Agriculture   AIC 270.17 268.25** 268.34
SBC 273.44 270.43 270.52
Log likelihood -132.08 -132.12 -132.17
Industry AIC 370.14 368.46 368.24**
SBC 373.41 370.64 370.42
Log likelihood -182.07 -182.23 -182.12
Services AIC 266.15 264.58 262.76**
SBC 269.43 266.76 264.94
Log likelihood -130.08 -130.29 -129.38

Table 3a. AIC and BIC values of different ARIMA models.

Out of the total 22 data points (1990 to 2012), the first 17 data points (from 1990 to 2007) were used for model building, while the remaining 5 data points (from 2008 to 2012) were used for model validation. One-step ahead forecasts of the revenue for each variable along with their corresponding standard errors using naïve approach for the period 2008 to 2012 with respect to the fitted models were computed (Table 3b).

Date GDP Agriculture Industry Services
Actual Forecast Actual Forecast Actual Forecast Actual Forecast
2008 3614.44 3449.08 2818.53 2697.14 6633.59 5674.29 2461.57 2456.86
2009 3448.54 3771.57 3063.90 3040.26 5399.20 6738.90 2477.29 2568.97
2010 4377.60 3747.57 3249.69 3269.07 9605.23 6317.14 2448.61 2493.13
2011 4485.59 4399.72 3458.87 3428.33 9950.64 8887.38 2481.50 2509.15
2012 4561.20 4665.81 3849.10 3647.93 9709.80 10019.20 2477.10 2558.05

Table 3b. One step ahead forecasts.

The forecasting ability of the selected ARIMA models of revenue series for the economic variables were judged on the basis of R2, the mean absolute prediction error (MAPE), root mean square error (RMSE) and relative mean absolute prediction error (RMAPE) values. A perusal of Table 3c shows that the RMAPE of each variable is less than 10%, indicating the accuracy of the models used.

Variables ARIMA model R2 MAPE RMSPE RMAPE (%)
GDP 0,1,1 0.994 90.72 26.51 1.8
Agriculture 1,1,0 0.99 71.47 3.26 2.1
Industry 0,1,1 0.96 732.31 344.04 6.2
Services 0,1,1 0.99 48.02 1.43 -1.94

Table 3c. Validation of models.

One step ahead out of sample forecast of the revenue for the economic variables selected during the period of the year 2013 to 2024 have been computed. The absolute data points are shown in Table 3d and also depicted in Figure 2a-d to visualize the performance of the fitted model. A cursory review shows that the revenue of each variable would be marked by an increase, which is an indication of good prospect for the country economy. Under normal growth the forecasted trend will prevail; while under high and low growth the future trend of revenues for each variable will not exceed the upper and lower confidence limits, respectively. It can be observed from the forecasted results that agriculture will have the highest contribution to the country GDP with industry trailing behind. Under normal growth, the forecasted annual revenue growth rate for GDP, agriculture, industry and services will be 3.5, 3.6, 3.5 and 3.4%, respectively. In the case of high growth, the annual revenue growth rate for GDP, agriculture, industry and services are estimated to be 4.1, 4.7, 4.4 and 5.1%, respectively; and while under low growth, the estimated revenue growth rate for GDP, agriculture, industry and services sectors will be 2.8, 2.1, 2.3 and 1.0%, respectively. Therefore, since the country anticipated growth exceed the future outlook, onus lies on the stakeholders to implement sound policies to make the non-oil sectors main driver of the economy due to the low oil prices which leads to decline in fiscal revenues, vulnerability to slow global economic recovery and global financial development. Also, the overall impact of non-oil sector-GDP is relatively muted [13].

Date GDP Agriculture 
LCL Forecast UCL LCL Forecast UCL
2013 4327.2270 4796.8058 5266.3845 3917.9934 4118.7760 4319.5585
2014 4426.2598 5002.0050 5577.7501 3981.8701 4334.7613 4687.6524
2015 4542.0270 5207.2042 5872.3814 4043.6660 4526.8345 5010.0030
2016 4668.4686 5412.4034 6156.3382 4112.9498 4708.2582 5303.5677
2017 4802.4846 5617.6026 6432.7205 4191.4752 4884.9389 5578.4025
2018 4942.2363 5822.8081 6703.3672 4278.5429 5059.5072 5840.4714
2019 5086.5267 6028.0010 6969.4752 4372.8752 5233.1347 6093.3943
2020 5234.5230 6233.2002 7231.8753 4473.2346 5406.3433 6339.4519
2021 5385.6267 6438.3994 7491.1720 4572.8752 5579.3652 6580.1458
2022 5539.2755 6643.5986 7747.8216 4688.0991 5752.3040 6816.5090
2023 5695.4171 6848.7978 8002.1784 4801.1298 5925.2059 7049.2819
2024 5853.4700 7053.9970 8254.5240 4917.1671 6098.0912 7279.0153
GR% 2.8 3.5 4.1 2.1 3.6 3.5
Industry Services
LCL Forecast UCL LCL Forecast UCL
2013 8311.2137 10263.5476 12215.8816 2334.9859 2500.8658 2666.7457
2014 8383.1307 10707.9411 13032.7516 2232.8783 2603.7971 2974.7158
2015 8506.9841 11152.3346 13797.6851 2209.0886 2706.7283 3204.3681
2016 8665.6848 11596.7281 14527.7714 2211.5711 2809.6596 3407.7481
2017 8894.8600 12041.1216 15232.3832 2228.6505 2912.5909 3596.5313
2018 9053.7099 12485.5151 15917.3202 2255.3649 3015.5222 3775.6794
2019 9273.3498 12929.9086 16586.4674 2289.0539 3118.4534 3947.8530
2020 9506.0262 13374.3021 17242.5779 2328.0941 3221.3847 4114.6754
2021 9749.7038 13818.6955 17887.6873 2371.4085 3324.3160 4277.2235
2022 10002.8274 14263.0890 18523.3507 2418.2392 3427.2473 4436.2554
2023 10264.1769 14707.4825 199150.7882 2468.0289 3530.1786 4592.3282
2024 10532.7743 15151.8760 19770.9777 2520.3536 3633.1098 4745.8661
CG% 2.3 3.5 4.4 1.0 3.4 5.1

Table 3d. Out of sample forecasts (Billion N).


Figure 2a. Forecast of GDP.


Figure 2b. Forecast of Agiculture.


Figure 2c. Forecast of Industry.


Figure 2d. Forecast of Services.

ARIMA diagnostic Checking

The model verification is concerned with checking the residuals of the model to see if they contained any systematic pattern which still could be removed to improve the chosen ARIMA. The results of the autocorrelation tests for each variable showed the residuals to be purely random as indicated by the Ljung-Box Q-statistics tests which were not significantly different from zero at 10% probability level. Also, the Arch effect tests showed no arch effects in the residuals as evidence by Arch-Lagrange multiplier (LM) test statistics which were not different from zero at 10% probability level. The normality tests for each variable showed that only the residuals of GDP and agriculture were found to be normally distributed as evidence by Jarque-Bera test statistics which were not significantly different from zero at 10% probability level (Table 3e), while that of industry and services sectors were not normally distributed. However, normality test is not considered a serious matter because in most cases data are not normally distributed. Therefore, these proved the selected model to be the best fit and appropriate model for forecasting.

Variables ARIMA model Autocorrelation test (Ljung-Box Q) Arch test (LM) Jarque-Bera Normality test (Chi2)
GDP 0,1,1 0.038 (0.981)NS 8.719 (0.3665)NS 3.2940 (0.1926)NS
Agriculture 1,1,0 0.254 (0.614)NS 7.780 (0.455)NS 1.641 (0.4402)NS
Industry 0,1,1 0.056 (0.813)NS 4.296 (0.830)NS 26.95 (0.000)***
Services 0,1,1 1.814 (0.178)NS 3.574 (0.893)NS 4.632 (0.099)*

Table 3e. Diagnostic checking for best ARIMA models.

Conclusion and Recommendation

This empirical research investigated the future outlook of Nigeria economic development in the light of slow global economic recovery and global financial developments using linear and symmetric price transmission mechanism model (ECM). The results showed that the residual of the OLS regression was stationary at level, indicating that GDP and the economy drivers are co-integrated. Furthermore, it was observed that the GDP established a long-run equilibrium with the all economy drivers, and corrects any deviation from the equilibrium that originated from any of the short-run equilibrium at a very slow rate per annum. Also, the economic parameters were found to exert positive influence on the GDP formation. It was observed that the future outlook of the GDP and the economy drivers will exhibit an increasing trend with the forecasted growth rate ranging from 3.4-3.6%. Therefore, in the light of the recent macroeconomic challenges, study recommends that government should adopt adjustment strategies that hinges on increasing non-oil revenues to compensate for the dwindling oil revenues because revenues from non-oil sectors can stand as the main driver of economy growth over the medium term.