The Role of Urbanisation in the Economic Development Process: Evidence from Nigeria

Abstract

The aim of this article is to investigate the relationship between urbanisation and economic growth, while controlling for the agricultural sector, industrial development and government expenditure in Nigeria. The autoregressive distributed lag (ARDL) approach to cointegration is applied to examine the long-run relationship between the variables over the period 1961–2012. In the process of estimating the long-run coefficients, the ARDL method is augmented with a fully modified ordinary least squares (FMOLS) estimator and a dynamic ordinary least squares (DOLS) estimator. The direction of causality between the variables is examined through the vector error correction method (VECM) Granger causality test. The results establish the existence of a long-run relationship in the variables. The results of the long-run regressions indicate the presence of long-run causality from urbanisation, agriculture and industrialisation to economic growth. Due to the deficiencies associated with the single-equation methods (including the ARDL model), we also use the structural vector error correction model (SVECM) to analyse the relationship between the variables. The impulse response and variance decomposition analyses derived from the SVECM method suggest that urbanisation, agriculture and industrialisation are important determinants of economic growth. The implications of the results are discussed.

JEL Classification: Q43, O55, O18

Keywords

Urbanisation Economic Growth Nigeria

1. Introduction

The world has witnessed a general and continuous increase in the rate of urbanisation (share of urban population in the total population) and the trend is expected to continue into the future. However, it has been documented that migration from the rural areas into urban centres may have a double-edge role in the economy.

On a positive note, the exodus of people from the rural areas to the urban leads to a high population density in the cities, which minimises the effect of man on local ecosystems. Cities generate big markets for products and can entice tourism and international investment from around the globe. It is believed that innovations and economies of scale are urban phenomena, and result from the dense networks of people. The structural transformation path usually comprises the movement of economic activities and people from the agricultural sector and rural areas into manufacturing and services of the urban centres (Cohen, 2006; Sackey, Liverpool-Tasie, Salau, & Awoyemi, 2012; Solarin, 2011, 2014). The movement to cities offers more education and employment opportunities for women and provides city-based activities the necessary resources such as human capital input for production and manufacturing. Returns to education are greater in the cities, as are the average educational accomplishment levels and literacy rates, thereby generating new ideas, fostering creative and innovative civic cultures (Bertinelli & Black, 2004).

Beyond the role of generating well-paid jobs, urban centres ensure, through their density, greater accessibility to public services. They ensure more efficient channelling of social services including healthcare and government assistance. Even in the urban centres of the poorest countries, there is better accessibility to basic services including those associated with the attainment of the Millennium Development Goals (MDGs), such as access to sanitation facilities and safe water. In the same vein, urban infrastructure developments ensure larger positive externalities in major cities than in rural areas as they boost the economic capability of households (via improved life expectancy and health) and human capital, which stimulate higher economic growth. Urbanisation is usually regarded as a proxy for accessibility to a country’s national electricity grid—therefore, more and wealthier individuals with such privilege would stimulate more electricity use (Liddle & Lung, 2014; Turok & McGranahan, 2013; World Bank, 2013).

Considering the negative implications, urbanisation causes an increase in the activities of the informal economy in cities, which limits tax revenues for the authorities. In turn, governments are forced to increase expenditure on the provision of basic services and, in some cases, to seek the support of domestic and international donor agencies for assistance. It is worth noting that urbanisation robs the agriculture sector of resourceful youths and, with the lack of modern farming equipment, weakens the chances of the sector feeding the country (Sackey et al., 2012).

Urbanisation is universally acclaimed as a burning issue in several countries, but it tends to portend more challenges to third-world countries because they experience a relative higher growth rate of urbanisation. According to World Bank (2014), OECD countries recorded an urbanisation rate of 76.98 per cent in 2012, while the rate in Sub-Saharan African countries was 36.78 per cent in the same year. However, the growth rates of urbanisation in Sub-Saharan Africa and low-income economies countries are higher than the figures for the European Union and the OECD (Table 1). Moreover, there are more people residing in cities with the least population of 100,000 in underdeveloped nations than in developed economies. Over the past 20 years, developing economies have urbanised rapidly, with urban citizens increasing from about 1.5 billion in 1990 to 3.6 billion in 2011 (World Bank, 2013). The ratio of the population living in urban centres in poor countries is expected to reach 52 per cent by 2020 (Ravallion, 2002).

Table 1
Urbanisation Rate in Selected World Regions, 1961–2012

Region 1961 1970 1980 1990 2000 2010 2011 2012

European Union 61.41 65.44 68.70 70.23 71.40 73.63 73.85 74.07

Low income 11.42 14.83 18.50 21.40 24.15 27.42 27.80 28.19

OECD members 62.83 67.33 70.32 73.14 75.65 79.38 79.68 79.98

Sub-Saharan Africa 15.24 19.33 23.62 27.91 32.00 35.93 36.36 36.78

Nigeria 16.95 22.71 28.58 35.28 42.35 49.00 49.62 50.23

Region	1961	1970	1980	1990	2000	2010	2011	2012
European Union	61.41	65.44	68.70	70.23	71.40	73.63	73.85	74.07
Low income	11.42	14.83	18.50	21.40	24.15	27.42	27.80	28.19
OECD members	62.83	67.33	70.32	73.14	75.65	79.38	79.68	79.98
Sub-Saharan Africa	15.24	19.33	23.62	27.91	32.00	35.93	36.36	36.78
Nigeria	16.95	22.71	28.58	35.28	42.35	49.00	49.62	50.23

Source: World Bank (2014).

Note: The figures are expressed in percentages.

Nigeria is one of the most urbanised countries in the African continent and the developing countries. The emergence of several urban centres in the country typically represents its rapid pace of urbanisation. With the urban population doubling almost every 15 years, the share of the population residing in Nigerian cities with more than 20,000 inhabitants increased from less than 15 per cent in 1950 to 23.4 per cent in 1975 and 43.3 per cent in 2000 (UN-HABITAT, 2006). The country currently has eight cities with more than a million population, and with a projection that the number of such cities will increase to 14 by 2050 (UN-HABITAT, 2014). Lagos, with an estimated population of 21 million inhabitants, is the largest city in Nigeria and in Africa (The Atlantic, 10 July 2012; The New York Times, 14 April 2012). The urbanisation rate in the country increased from less than 17 per cent in 1961 to 42.35 per cent in 2000 and further to 50.23 per cent in 2012 (World Bank, 2014).

The rapid pace of urbanisation in Nigeria may be explained by several factors. The emergence of petroleum and gas led to increased urban wages, education investments and food imports, and a sharp decline in the profitability of tradable agriculture, including importable cereals, and in the export crop sector (particularly cocoa and oil palm). Besides, the civil war (1967–70) forced the government to create new states, which fostered the growth of intermediate and small urban centres as their capitals (Sackey et al., 2012). The prevalence of insufficient social amenities, infrastructures and facilities including poor health services, erratic power supply, appalling road conditions, poor educational facilities, lack of housing conditions and social life have pushed people to move from the rural to urban centres (Oyeleye, 2013). For instance, the access to electricity in the rural areas is 34.9 per cent, while in the urban areas it is 79.8 per cent (World Bank, 2014). It is suggested that urbanisation has been a reaction to ‘push’ factors such as rural droughts, dwindling prices of agricultural produce and tribal skirmishes instead of the pull of economic opportunities (Turok & McGranahan, 2013).

Despite the growth of urban centres in the country, the interaction between urbanisation and economic growth remains imprecise and ambiguous. There are arguments that urbanisation is not a catalyst for economic growth as it only accompanies periods of sustained economic growth. As such, it has been argued that urbanisation is more an indicator than a tool of economic growth (Liddle, 2013). The authorities who intend to boost long-term economic growth by increasing or deterring the growth of the population in cities are expected to fail in their mission (Bloom, Canning, & Fink, 2008). Some experts believed that urbanisation negatively influences economic activities, while others suggested that urbanisation has a positive effect on the economy (Sackey et al., 2012; Turok & McGranahan, 2013).

The lack of knowledge on the exact relationship between urbanisation and economic growth is a disadvantage to policies makers, who might want to pursue certain policies in the area of urban development. Unfortunately, we are not aware of any time-series study which has attempted to investigate the link between urbanisation and the Nigerian economy. The present study examines the relationship between urbanisation and economic development, while incorporating proxies for the agricultural sector, industrial development and government expenditure. This research is distinct from the existing literature in many respects. According to our knowledge, this is the first time-series study on the relationship between urbanisation and economic growth in Nigeria, which is the largest country in Africa and has the most urban centres in the continent. Moreover, we utilise a structural vector error correction model (SVECM) to probe the relationship between the variables. SVECM clearly separates structural exogenous shocks from movements in the variables due to endogenous responses to current development of the economy (Massidda & Mattana, 2013). This method is useful because structural shocks can affect many of the variables used in this study. The outcomes of this study may assist authorities to have a better understanding of the complexities surrounding the relationship between urbanisation and economic growth. The rest of this article is organised as follows. The second section presents a literature review. The third section introduces the methodology employed in this article. The fourth section deals with the empirical findings, and the fifth section presents the conclusion of this study.

2. Literature Review

The empirical literature on the relationship between urbanisation and economic growth can be divided into two strands. The first strand examines the nexus from a bivariate perspective, whereby urbanisation and economic development are considered without the incorporation of other variables. The second strand utilises the multivariate approach, whereby other determinants of economic growth are included in the model. The studies with the bivariate approach include Laidlaw and Stockwell (1979) who showed a weak association between urbanisation and economic growth for Asian and African countries but a moderate positive association for Latin America. Henderson (2003) noted a positive correlation coefficient of around 0.85 between urbanisation and GDP and thereby opined that urbanisation and economic development are interconnected. Using a bivariate Granger causality test on 28 countries (developing and developed countries) for the period 1950–2000, Lo (2010) confirmed a long-run connection between urbanisation and economic growth, provided evidence for causality running from urbanisation to economic activity for less-developed countries, and observed the opposite for advanced countries.

However, Lütkepohl (1982) argued that such an approach may create omitted variable bias, which is resolved with adding more variables. Among the research works which attempted to resolve biasness caused by the omission of variables is McCoskey and Kao (1999) who utilised the Cobb–Douglas production function to examine the nexus, with urbanisation serving as a shift factor for 30 developing countries and 22 developed countries for the period covering 1965–89. The study observed that urbanisation has a long-run effect on output per worker. Bertinelli and Strobl (2003) assessed the link among urbanisation, urban concentration and economic growth for 39 countries for 1960–90, but could not establish a connection between urbanisation and growth.

Bloom et al. (2008) examined the relationship between urbanisation and economic growth in 180 countries. The authors observed no causality from urbanisation to GDP per capita. Brülhart and Sbergami (2009) conducted another cross-sectional study on the nexus by looking at several determinants of economic growth in 105 countries. Using panel techniques, Brülhart and Sbergami (2009) observed that urbanisation was positively related to economic growth up to a threshold (10,000 US dollars in 2006 prices), while government expenditure was negatively related to economic growth. Bradshaw and Fraser (1989) conducted cross-sectional research in which the influence of urbanisation and industry (proxied by industrial employment) on economic growth was assessed in various districts of China. Using the ordinary least square (OLS) techniques, the findings suggested that industrial employment and urbanisation are positively associated with economic development. Brückner (2012) examined the urbanisation rate and economic growth link with an indicator of the agricultural sector as a conditioning variable for a panel of 41 African countries in the period, 1960–2007. The results demonstrate that increases in the urbanisation rate yield an inverse shift in growth in the agricultural sector. Brückner (2012) further provided evidence for feedback from economic growth to urbanisation, which was found to be significant when conditional changes in agriculture were ignored.

Liddle (2013) examined the relationship between urbanisation and economic growth, while adding energy consumption within the neoclassical production function for 79 countries. The results revealed that urbanisation has a positive impact on economic growth in high- and upper-middle-income countries but has a negative impact for low-income countries. Thus, Liddle (2013) posited that urbanisation has a ‘ladder’ impact on economic growth, which implies that the positive influence of urbanisation on economic growth tends to be more pronounced in developed nations rather than in developing ones.

Generally, panel data approaches suffer from heterogeneity problems. More importantly, the effect of urbanisation differs across countries and largely depends on the stage of economic development of each country. As such, the results obtained from the panel data techniques may not be relevant for all the sampled countries. Using cross-section techniques to determine the long-run relationship between urbanisation and growth effects may yield inconsistent and biased results (McCoskey & Kao, 1999). Conversely, time-series analysis can substantially aid our knowledge of the interconnection between urbanisation and economic activity and can capture the dynamic and temporal nature of the nexus (Thanh, 2007). There are few papers that have employed time-series procedures in the urbanisation–economic growth nexus. Abdel-Rahman, Safarzadeh and Bottomley (2006) focused on 35 developing countries and established a negative link between urbanisation and economic growth. Alam, Fatima and Butt (2007) used data for Pakistan for the period 1971–2005 to reach the conclusion that urbanisation has negatively influenced economic growth. Shahbaz and Lean (2012) considered the interaction among economic growth, industrialisation and urbanisation with other variables including energy consumption and financial development in Tunisia from 1971 to 2008. Using the autoregressive distributed lag (ARDL) cointegration approach and Granger causality test, the authors observed the existence of bidirectional causation and a long-run relationship among the variables. In a related work, Liu (2009) could only observe a short-run causality flowing from growth to urbanisation in China, for the period, 1978–2008. Solarin and Shahbaz (2013) considered the nexus between urbanisation and economic growth in Angola, while providing for electricity consumption. The authors observed a long-run bidirectional causality between the variables.

3. Methodology

3.1 Model

The direction of the link between urbanisation and economic development remains ambiguous, especially for developing countries, where the intensity of growth of urbanisation is highest. The exclusion of other variables that appears to moderate the relationship between the two variables tends to aggravate the ambiguity of the nexus, as the preceding reviewed literature has highlighted. In this section, we start the process of attempting to unravel the relationship in Nigeria by providing a simple linear-log model as follows:

G D P_{t} = f (U R B_{t}, A G R_{t}, I N D_{t}, G O V_{t})

(1)

Here, GDP is real gross domestic product per capita; URB is the share of urban population in the total population in Nigeria, AGR is the share of agriculture (inclusive of crop production, livestock, forestry and fishing) in the GDP, IND is the share of industry (inclusive of manufacturing) in the GDP and GOV is the share of government expenditure (recurrent plus capital expenditure) in the GDP. We combed the World Development Indicators of the World Bank (2014) to generate GDP and URB data; and the Statistical Bulletin of the Central Bank of Nigeria (CBN, 2014) to obtain data for AGR, IND and GOV for the period, 1961–2012.

There are theoretical justifications for the inclusion of the control variables in this study. Key variables in the economy—such as agricultural development, industrialisation and government expenditure—change as urbanisation interacts with economic development. Although the petroleum and gas sector dominates foreign exchange earnings and government revenues, it is the agricultural sector that usually drives domestic consumption (Sackey et al., 2012). Agriculture is still central to the economic development of the country, and it remains the backbone of most rural economies, given the scale of rural nonfarm earnings. In many developing countries, agricultural value added per worker (which is a measure of agricultural productivity) has been increasing over the years, which tends to perpetuate the significance of this sector in the economy. Consequently, agriculture is still expected to positively contribute to the economic development process.

Urban economies would involve industrialisation that stimulates other forms of undertakings such as commerce and services through forward and backward linkages. Specifically, as urbanisation gains momentum, agricultural productivity tends to decrease, while industrialisation is bolstered because of the shifting of production factors from the former to the latter (Davis & Henderson, 2003; Lu, Liang, Bi, Duffy, & Zhao, 2011; Thanh, 2007). A crucial reason why urbanisation tends to go along with economic development is the process of industrialisation, which instigates rural labour to migrate to urban centres to work in manufacturing plants (Liddle & Lung, 2014). As the economy develops and continues to imbibe the use of modern machinery, the industrial value added per worker is expected to continue to increase. Consequently, industrialisation is projected to have a positive impact on economic growth. Equitable and sustained economic growth is obviously a principal aim of government expenditure policy, and many public programmes are geared towards ensuring equitable and sustained economic growth. Public expenditure performs a vital part in human and physical capital formation in a country. Optimal government expenditures can boost economic growth, even in the short run, when constraints in skilled manpower or infrastructural facilities become a substantial limit to a rise in output. Moreover, the urbanisation generates government expenditure on social amenities, which includes government spending on transportation, communications and energy infrastructures (Moomaw & Shatter, 1996). However, the relationship between government expenditures and economic growth is not automatically unidirectional because, according to the Wagner’s law, economic growth can generate variations in either aggregate public expenditure or its components (Chu, Gupta, & Clements, 1995).

3.2 Unit Root Tests

Time-series literature provides several unit root tests including the routinely used augmented Dickey–Fuller (ADF) test, whose power is weakened with the occurrence of structural breaks. In the presence of structural breaks, the Perron (1989) and Zivot and Andrews (1992) tests are superior to ADF. However, the Perron approach of determining breaks exogenously has been criticised on the basis that it provides an arbitrary approach to choosing the break date. Although the Zivot and Andrews test advanced an endogenous single structural break unit root method, its assumption of no breaks under the null may lead to size distortion. The method does not provide for more than a single structural break. Lee and Strazicich (2001) also noted that the Zivot and Andrews test is likely to incorrectly compute the break point. Lee and Strazicich (2003, 2004) suggested unit roots tests with structural break(s), which are free of the problems associated with the ADF and Zivot and Andrews tests. In order to test for unit roots in the series, we adopt the Lee and Strazicich (2003, 2004) tests, which are specified follows:

Δ y_{t} = δ' Δ Z_{t} + ϕ {\bar{S}}_{t - 1} + \sum_{i = 1}^{p} γ Δ {\bar{S}}_{t - i} + μ_{t}

(2)

Here, ${\bar{S}}_{t} = y_{t} - {\hat{ψ}}_{x} - Z_{t} {\hat{δ}}_{t,} t = 2.............., T, \hat{δ}$ denotes the coefficients in the regression of Δy_t on ΔZ_t and ${\hat{ψ}}_{x} = y_{t} - Z_{t} \hat{\partial},$ in which y₁ and Z₁ represent the first observations of y_t and Z_t, respectively. μ_t is contemporaneous error term that is independent and identically distributed with zero mean and finite variance. For a one-time change in both level and trend, Z_t = [1, t, D_1t, DT_1t]' where DT_jt = t if t ≥ T_B + 1, and 0 otherwise. The location of a one-time change is given by λ_j = T_Bj/T, j = 1. Two-time changes in level and trend are specified as Z_t = [1, t, D_1t, D_2t, DT_1t, DT_2t]' where DT_jt = t if t ≥ T_Bj + 1, j = 1, 2, and 0, otherwise, while T_B is the break date. The location of a two-time change is given by λ_j = T_Bj/T, j = 1, 2. Generally, break date(s) are determined with the location of the value(s), which minimises the t-test statistic in an LM test on the null hypothesis of unit root (ϕ = 0). In our study, the augmented terms of $ε {\bar{S}}_{t}$ are incorporated to accommodate the incidence of serial correlation in errors.

3.3 SVECM Analysis

Under the assumption that the unit root test has been performed, we start the analysis by considering a reduced finite-order VAR of the following specification:

Y_{t} = \sum_{k = 1}^{m} Ψ_{k} Y_{t - k} + Ξ D_{t} + e_{t}

(3)

where m denotes the relevant order lag polynomial, Ψ_k denotes the matrix housing the relevant parameters of the variables under consideration, e_t is a vector containing the observable residuals and Ξ is the coefficient matrix connected to the deterministic terms, including a constant, trend and two structural breaks in line with Johansen, Mosconi and Nielsen (2000). In the foregoing specification, Y_t = [GDP_t, URB_t, AGR_t, IND_t, GOV_t]'. When the variables are stationary but cointegrated, model (3) can be specified in a vector error correction method (VECM) specification by removing Y_t_–1 from the two sides of the equation. The equation can be rewritten in the following form:

Δ Y_{t} = α β' Y_{t - 1} + \sum_{k = 1}^{m} Γ_{k} Δ Y_{t - k} + e_{t}

(4)

Γ_k comprises the coefficients of the series in difference, while k is depicting the lag order and m is the maximum lag order. β' is a (5 × r) matrix of long-run coefficients, with r the cointegration rank. For simplicity, Equation (3) excludes the deterministic components.

There are three potential outcomes in the pattern of the αβ' impact matrix. The relevant one for this study is when the variables are not stationary but cointegrated. There are r cointegrating vectors, which link the series in the long run, when the impact matrix is known to have an intermediate rank, r. Since these vectors are not uniquely determined, it is ideal to impose few restrictions in a bid to identify the cointegrating space. This entails one normalisation and (r − 1) restrictions on each of the cointegrating vectors. Causality analysis associated with the series can be conducted, once an identification system is selected. However, when structural exogenous innovations have to be detached from movements in the series due to endogenous responses to contemporary developments in the economy, it is hard to provide an economic explanation to the reduced-form VAR model. Adopting a SVECM method can address this issue. The initial step to having a structural depiction of Equation (3) requires identification of the fundamental innovations that induce informative responses of the system variables. Specifically, the impacts of the structural shocks e_t on the series under consideration, Y_t can be explained by the following expression:

A Δ Y_{t} = A α β' Y_{t - 1} + A \sum_{k = 1}^{m} Γ_{k} Δ Y_{t - k} + ε_{t}

(5)

The impacts of the structural shocks on the series Y_t are specified as:

e_{t} = ε_{t} A^{- 1} = B ε_{t}

(6)

where the (K × 1) is the vector of unobservable structural disturbances ε_t fulfils the classical assumption. In order to estimate the responses to structural shocks, ε_t, there is a need to recover the K² elements of matrix B (Amisano & Giannini, 1997). For this, there is need for a set of identifying restrictions. However, the mechanism for selecting contemporaneous restrictions is always rather subjective. With the availability of relevant theories on the relationship between the series under consideration, some direction can be offered by economic rationale. Alternatively, a more empirically based method can be utilised. For example, feasible restrictions can be investigated through a Granger causality framework.

Contemporaneous relationships can also be determined by statistical tests for the coefficients’ significance. Under the assumption that structural shocks are uncorrelated and have unit variance Σ_ε = I_k, using Equation (6), the following is obtained:

Σ_{e} = E [e_{t} {e^{'}}_{t}] = B E [ε_{t} {ε^{'}}_{t} ​] B^{'} = B Σ_{t} B^{'} = B B^{'}

(7)

Hence, in a bid to uniquely identify the K² matrix B, the K(K + 1)/2 independent linear restrictions from the covariance matrix Σ_e = BB' are not sufficient. At this point, there is a need to introduce extra K(K – 1)/2 linearly independent restrictions. Impulse response and variance decomposition analyses can be adopted once the B matrix is identified. These analyses demonstrate the time-profile of the impact of structural shocks on series of the system, and turn out to be very valuable for policy simulation purposes.

4. Empirical Findings

The results of the stationarity tests are presented in Table 2. The unit root tests on the level form of the series are presented in the upper panel. With the ADF test, we cannot reject the null of a unit root for any of the variables. However, the ADF test loses its power in the presence of structural breaks and therefore needs to be supported with other unit root tests. Using the Lee and Strazicich tests for one and two structural breaks on the level form of the series, we cannot reject the null of unit root of all the variables. We report the unit root tests of the first difference variant of the series in the lower panel of Table 2. With the use of the ADF test, we can reject the null hypothesis at the 10 per cent level or higher. Similar outcomes were observed when the Lee and Strazicich tests are applied, which is an indication that the series are I(1). More than 50 per cent of the structural breaks lie within the late 1960s to early 1970s, which is a period coinciding with the Nigerian civil war (6 July 1967–15 January 1970). During the civil war, there were unusual changes in macroeconomic variables, mass displacement of the population and the deaths of more than two million people (Nafziger, 1972).

Table 2
Unit Root Tests

Panel A: Series at Level

Variable ADF
Lee and Strazicich Test (2004)
Lee and Strazicich Test (2003)

T-stat T-stat Break T-stat Break (1) Break (2)

GDP_t -1.450[1] -2.523[1] 1985 (0.4) -4.126[1] 1973 1995 (0.4, 0.6)

URB_t -0.494[1] -2.080[1] 1968 (0.2) -2.786[1] 1972 1988 (0.4, 0.6)

AGR_t -1.819[0] -3.462[0] 1970 (0.2) -3.305[0] 1972 1981 (0.2, 0.4)

IND_t -1.848[4] -2.711[0] 1969 (0.2) -4.452[0] 1978 1999 (0.4, 0.8)

GOV_t -2.167[1] -3.938[3] 1970 (0.2) -2.768[1] 1971 2001 (0.2, 0.8)

Panel B: Series at First Difference

Variable ADF
Lee and Strazicich Test (2004)
Lee and Strazicich Test (2003)

T-stat
T-stat
Break

T-stat
Break (1)
Break (2)

ΔGDP_t -5.057[0] -5.339[1] 1982 (0.4) -6.013**[1] 1974 1981 (0.2, 0.4)

ΔURB_t -4.092[0] -4.865[0] 1969 (0.2) -5.852**[4] 1974 1981 (0.2, 0.4)

ΔAGR_t -7.806[0] -9.116[0] 1979 (0.4) -9.698***[0] 1979 1990 (0.4, 0.6)

ΔIND_t -7.354[1] -9.218[0] 1972 (0.3) -8.883***[0] 1984 2000 (0.4, 0.8)

ΔGOV_t -9.707[0] -9.598[0] 1995 (0.6) -9.807***[0] 1983 1995 (0.4, 0.6)

Panel A: Series at Level
GDP_t	-1.450[1]	-2.523[1]	1985	(0.4)	-4.126[1]	1973	1995	(0.4, 0.6)
URB_t	-0.494[1]	-2.080[1]	1968	(0.2)	-2.786[1]	1972	1988	(0.4, 0.6)
AGR_t	-1.819[0]	-3.462[0]	1970	(0.2)	-3.305[0]	1972	1981	(0.2, 0.4)
IND_t	-1.848[4]	-2.711[0]	1969	(0.2)	-4.452[0]	1978	1999	(0.4, 0.8)
GOV_t	-2.167[1]	-3.938[3]	1970	(0.2)	-2.768[1]	1971	2001	(0.2, 0.8)
Panel B: Series at First Difference
Variable	ADF	Lee and Strazicich Test (2004)	Lee and Strazicich Test (2003)
T-stat	T-stat	Break		T-stat	Break (1)	Break (2)
ΔGDP_t	-5.057***[0]	-5.339***[1]	1982	(0.4)	-6.013**[1]	1974	1981	(0.2, 0.4)
ΔURB_t	-4.092***[0]	-4.865***[0]	1969	(0.2)	-5.852**[4]	1974	1981	(0.2, 0.4)
ΔAGR_t	-7.806***[0]	-9.116***[0]	1979	(0.4)	-9.698***[0]	1979	1990	(0.4, 0.6)
ΔIND_t	-7.354***[1]	-9.218***[0]	1972	(0.3)	-8.883***[0]	1984	2000	(0.4, 0.8)
ΔGOV_t	-9.707***[0]	-9.598***[0]	1995	(0.6)	-9.807***[0]	1983	1995	(0.4, 0.6)

Source: Author’s computation.

Notes: ***, ** and * imply 1%, 5% and 10% levels of significance, respectively. TB is the estimated break points. Critical values for tests with structural break(s) are from Lee and Strazicich (2003, 2004).

Before conducting the SVECM analysis, we report the outputs from the ARDL test and the resulting causality analysis. Having established that the variables are I(1), we present the findings of the ARDL approach to cointegration, while utilising the critical values of Pesaran, Shin and Smith (2001) and Narayan (2005) in Table 3. The results demonstrate that cointegration is present, when the GDP is the dependent variable. In this case, the F-statistic (6.822) is higher than the upper-bound critical values (4.781 for Pesaran et al., 2001; 5.583 for Narayan, 2005) at the 1 per cent level. However, the ARDL test indicates that when urbanisation, agriculture, industrialisation and government expenditure are specified as the dependent variables, the F-statistics (2.577, 3.206, 2.480 and 2.719, respectively) either lie within the bound critical values or are lower than the lower-bound critical value, which suggests that we cannot reach the conclusion of cointegration, when these variables are the dependent variables. The validity of the ARDL bounds testing becomes questionable once there is evidence of structural breaks in the series. Thus, we apply the Gregory and Hansen (1996) structural break cointegration method, which provides for one structural break. The results, as shown in Table 4, suggest that we cannot reject the null of no cointegration when GDP is the regressand. The findings further indicate that the structural breaks occurred in the early 1970s, which coincided with the civil war in the country. In summary, these findings imply the existence of cointegration with structural breaks, which corroborate the prior results.

Table 3

ARDL Cointegration Test

Dependent Variable	GDP_t	URB_t	AGR_t	IND_t	GOV_t
F-statistics	6.822***	2.577	3.206	2.480	2.719
Critical values	10%i(0)	10%i(1)	5%i(0)	5%i(1)	1%i(0)	1%i(1)
Pesaran et al. (2001)	2.425	3.574	2.850	4.049	3.817	5.122
Narayan (2005)	2.435	3.600	2.900	4.218	3.955	5.583

Source: Author’s computation.

Notes: ***, ** and * imply 1%, 5% and 10% levels of significance, respectively. The optimal lag length is determined by AIC. Critical values are for the model with unrestricted intercept and no trend. The null is no cointegration.

The evidence for a single cointegration link in the previous section signifies Granger causality in at least one path, but does not specify the direction of causality in the series. In Table 5, the article presents Granger causality tests within the framework of VECM. The F-statistic on the lagged explanatory variables is examined for short-run causality, while the t-statistic on the lagged ECT is examined for long-run causality. Beginning with the short-run causality, the findings suggest the flow of causation from economic growth to urbanisation at the 5 per cent level, with feedback from urbanisation towards economic growth at the 5 per cent level, thus demonstrating a two-way causation from economic growth to urbanisation in the short run. Agriculture and industry Granger cause economic activities in the short run at the 10 per cent and 1 per cent levels, respectively, without any feedback from economic growth, thereby illustrating a one-way causation from agriculture and industrialisation to economic growth in the short run. Two-way causality is observed for urbanisation and industry at a maximum of the 5 per cent level, while one-way causality is observed from agriculture to urbanisation at the 5 per cent level. No causality is reported for urbanisation and government expenditure in the short run. This implies that urbanisation is associated with agriculture and industrialisation, with no connection between urbanisation and government expenditure in the short run.

Table 4

Gregory and Hansen (1996) Structural Break Cointegration Test

Dependent Variable	ADF test	TB
GDP_t	-6.56***[3]	1966
URB_t	-4.29[0]	1971
AGR_t	-4.12[1]	1966
IND_t	-4.06[0]	1970
GOV_t	-4.69[0]	1968

Source: Author’s computation.

Notes: *** implies the 1% level of significance. The ADF statistics show the Gregory–Hansen tests of cointegration with an endogenous break in the intercept and trend. Critical values for the ADF test (for four regressors) at the 1%, 5% and 10% levels are −6.36, −5.83 and −5.59, respectively. The AIC is used to determine the optimal lag length (out of a maximum of 4), while the trimming region is set at 0.15. TB is the estimated break points.

Table 5

Granger Causality Test

Dependent Variable	Short-run Causality					Long-run Causality
Dependent Variable	ΣΔGDP_t–i	ΣΔURB_t–i	ΣΔAGR_t–i	ΣΔIND_t–i	ΣΔGOV_t–i	ECT_t–1
ΔGDP_t	–	4.166**	2.527*	8.035***	0.712	-3.470***
ΔURB_t	3.178**	–	3.619**	4.004**	0.259	–
ΔAGR_t	0.614	0.077	–	0.427	1.764	–
ΔIND_t	0.524	4.447***	0.512	–	1.039	–
ΔGOV_t	0.785	1.449	1.596	0.748	–	–

Source: Author’s computation.

Notes: ***, ** and * imply 1%, 5% and 10% levels of significance, respectively.

The causality test does not indicate whether changes in each variable will positively or negatively affect changes in economic activity, especially in the long run. In Table 6 we report the long-run and short-run elasticities, with economic growth as the regressand. Due to robustness issues, the ARDL estimates are supported with estimates of the fully modified ordinary least squares (FMOLS) estimator of Phillips and Hansen (1990) and the dynamic ordinary least squares (DOLS) estimator of Stock and Watson (1993). The coefficient of urbanisation turns out positive, in contrast to the findings of Abdel-Rahman et al. (2006), Brückner (2012) and Liddle (2013). Based on the other results, we note that agriculture and industrialisation benefit economic growth. Government expenditure is found to be insignificantly related to economic growth. Due to the fact that the civil war (1967–70) ushered in the creation of many states and several structural breaks observed in the earlier findings are within this period, we further check the long-run relationship for the period covering the onset of the war and thereafter (1967–2012). The results are not materially different from the findings on the entire sample, but the subsample estimates appear to produce more robust outputs. Despite the similarity of the signs, we note that the coefficients reported under the DOLS estimator are much larger than under the ARDL and FMOLS, in line with the results of Solarin and Shahbaz (2013). In the short run, it is observed that urbanisation has a significant effect on the economy. Diagnostics tests suggest that the model is free from serial correlation, heteroscedasticity, the functional form problem and normality problems.¹ Moreover, the CUSUM and CUSUMSQ tests in Figures 1 and 2 largely support the stability of the coefficients of the regression equations.

Explaining the results, the long-run causality from the positive impact of urbanisation on economic growth connotes a favourable impact of urbanisation on the economy. The fact that the urban centres in Nigeria are the financial keys to the country’s prosperity, makes this regression result not too surprising. Aba, Abuja, Lagos, Kano and Port Harcourt are the commercial nerves of Nigeria and, at the same time, are the most urbanised cities in the country. Thus, the role of urbanisation cannot be overemphasised. This result may not be peculiar to Nigeria, as global cities and towns are the centres of affluence—more than 80 per cent of global economic activities are produced by urban citizens, who constitute just over 50 per cent of the global population. Economic agglomeration increases efficiency, which in turn entices more companies and generates well-paid jobs. Urbanisation offers higher wages for employees, earlier working on farms, and it creates additional avenues to move up the income ladder (World Bank, 2013).

Figure 1

Source: Author’s computation.

Figure 2

Source: Author’s computation.

Table 6

Long-run and Short-run Elasticities

Panel A: Long-run Elasticities
Dependent Variable	GDP_t
	ARDL		FMOLS		DOLS
Period	1961–2012	1967–2012	1961–2012	1967–2012	1961–2012	1967–2012
URB_t	2.681***	2.692***	0.627**	0.553**	1.177***	2.487***
AGR_t	3.186**	5.081***	1.178***	4.194***	5.071***	5.859***
IND_t	1.293**	1.298**	0.827*	2.826***	5.668*	-2.158
GOV_t	0.604*	0.622	0.029	-0.032	-0.104	-0.086
Panel B: Short-run Elasticities
Dependent Variable	ΔGDP_t
Period	1961–2012			1967–2012
ΔURB_t	2.115*			0.276*
ΔAGR_t	0.352**			0.332***
ΔIND_t	0.203*			0.252**
ΔGOV_t	0.064			0.074

Source: Author’s computation.

Notes: The optimal lag for ARDL is based on the Akaike information criterion, while the optimal lag for the FMOLS is based on Bartlett weights. As required, we apply the Newey-West on the estimates of DOLS. ***, ** and * imply 1%, 5% and 10% levels of significance, respectively.

In spite of the declining relevance of agriculture in Nigeria, the positive impact of the sector on economic growth has some factual evidences. The sector remains a significant component of the total production in the economy. In addition to providing some of the necessary inputs for the industrial sector, agriculture has generally expanded over the years. Since 1985, the valued added of the agricultural sector has generated positive growth rates. Agriculture value added per worker has been rising since 1987. It rose from US$1,489 in 2000 to US$4,452 in 2012 (World Bank, 2014). The positive impact of industrialisation on economic growth can also be explained by the continuous importance of the sector to economic growth. For instance, the growth rate of the value added of the industrial sector was 2.42 per cent in 2012 and the share of its value added in total GDP was more than 23 per cent in the same period (World Bank, 2014). The results of the negative impact of the government on economic growth are not surprising since the recurrent expenditure accounts for the lion share of total government expenditure. Recurrent expenditures are usually inefficient and unproductive relative to capital expenditure, especially in the developing countries.

Table 7

Cointegration Test Results

Null Hypothesis	Trace-statistics	Critical Values
Null Hypothesis	Trace-statistics	99%	95%	90%
r = 0	139.407***	134.93	123.21	117.24
r = 1	111.192**	112.53	101.99	96.64
r = 2	66.472	89.99	80.51	75.73
r = 3	36.162	69.45	60.88	56.61
r = 4	10.467	52.27	44.45	40.60

Source: Author’s computation.

Notes: *, ** and *** denote significance at the 10%, 5% and 1% levels, respectively. The model estimated includes linear trends, which correspond to the H_l version of Godwin and Giles (2012). The AIC is used to determine the optimal lag length (out of a maximum of 4).

The foregoing analysis has relied on a single equation, the ARDL approach, to motivate the direction of causality among a vector of non-stationary variables. Besides, the analysis does not separate structural exogenous shocks from movements in the variables due to endogenous responses to current development of the economy (Massidda & Mattana, 2013). Hence, we report the SVECM analysis to support the foregoing analysis. The cointegration test of Johansen et al. (2000) is the starting point and the results are reported in Table 7. We have introduced two structural breaks (1973 and 1995) into the cointegration analysis and the selection of the dates is based on the unit root analysis of economic growth in level form. The critical values are premised on the procedures of Giles and Godwin (2012). The results provide evidence for two cointegrating vectors. After deciding the dimension of the cointegration space, we examine the error correction analysis. For brevity, we report a long-run regression wherein all the variables are normalised on economic growth in Table 8. The results suggest that with the exception of government expenditure, all the variables are positively related to economic growth. The results are not different from the ARDL output, but the results are not as significant as the ARDL output. Generally, the model is valid because it does not suffer from non-normality and autocorrelation of the residuals (statistics not reported here).

Table 8

Estimation Results

Long-run Analysis
Independent Variable			1961–2012
URB_t			0.111***
AGR_t			1.976*
IND_t			0.105*
GOV_t			-0.114
Granger Causality Test
Dependent Variable	Short-run Causality					Long-run Causality
Dependent Variable	ΣΔGDP_t–i	ΣΔURB_t–i	ΣΔAGR_t–i	ΣΔIND_t–i	ΣΔGOV_t–i	ECT_t–1
ΔGDP_t	–	1.938	3.724	6.770*	0.836	-0.118**
ΔURB_t	8.585**	–	4.162	2.923	0.803	-0.003**
ΔAGR_t	4.644	13.462***	–	3.800	1.922	-0.417**
ΔIND_t	2.543	8.600**	2.129	–	6.649*	-0.289*
ΔGOV_t	2.737	3.921	2.193	10.978***	–	-0.668***

Source: Author’s computation.

Notes: ***, ** and * imply 1%, 5% and 10% levels of significance, respectively. We have not created a subsample of 1967–2012 in this case because the Johansen method is not robust when dealing with a small sample size.

We proceed with the causal analysis of the variables within a VECM framework. The Wald test provides evidence for limited causality in the short run. There is long-run bidirectional causality in the variables. The inference of the causality test is that none of the variables can be regarded as being weakly exogenous in the long run. However, the findings are contrary to the results of Bloom et al. (2008), who observed the absence of causality between urbanisation and economic growth for 180 countries. One of the reasons for the differences between the results can be attributed to the difference in datasets. Although Nigeria is among the 180 countries used in the work of Bloom et al. (2008), the causality results generated from their article does not represent all the countries involved in the sample. This is one of the setbacks of using a panel data causality approach.

After the causality test, we turn to the estimation of contemporaneous coefficients of the structural model. In doing this, a set of K(K – 1)/2 linearly independent overidentifying restrictions on the coefficients of matrix B is needed. Setting zero to the entries of matrix B can follow different processes. First, the process can be based on existing economic theories. For example, the process can be based on the identification that relies on the modernisation theory or dependency theory of urbanisation, Wagner’s law or Keynes’ aggregate expenditure model. These theories provide contradictory arguments and hence breed uncertainty about the links in the variables. Hence, we follow the work of Massidda and Mattana (2013) by using a procedure that is largely based on a statistical criterion by searching for significant coefficients. Hence, zeros are allocated in matrix B where the null hypothesis of no contemporaneous impact is rejected. The following matrix B is derived through the process:

B = [\begin{array}{l} \begin{matrix} 0.075 0.005 - 0.029 0.014 0.000 \end{matrix} \\ \begin{matrix} 0.000 0.003 0.003 0.000 0.000 \end{matrix} \\ \begin{matrix} 0.000 0.000 0.131 - 0.068 0.000 \end{matrix} \\ \begin{matrix} 0.000 0.000 - 0.100 0.144 0.000 \end{matrix} \\ \begin{matrix} 0.020 0.000 0.000 0.009 0.242 \end{matrix} \end{array}]

(8)

The ordering corresponds to the following system of equation:

\begin{matrix} e_{G D P} = b_{11} ε_{G D P} + b_{12} ε_{U R B} + b_{13} ε_{A G R} + b_{14} ε_{I N D} = 0.075 ε_{G D P} + 0.005 ε_{U R B} \\ - 0.029 ε_{A G R} + 0.014 ε_{I N D} \end{matrix}

(9)

e_{U R B} = b_{22} ε_{U R B} + b_{23} ε_{A G R} = 0.003 ε_{U R B} + 0.003 ε_{A G R}

(10)

e_{A G R} = b_{33} ε_{A G R} + b_{34} ε_{I N D} = 0.131 ε_{A G R} - 0.068 ε_{I N D}

(11)

e_{I N D} = b_{43} ε_{A G R} + b_{44} ε_{I N D} = - 0.100 ε_{A G R} + 0.144 ε_{I N D}

(12)

\begin{matrix} e_{G O V} = b_{51} ε_{G D P} + b_{54} ε_{I N D} + b_{55} ε_{G O V} = 0.020 ε_{G D P} \\ + 0.009 ε_{I N D} + 0.242 ε_{G O V} \end{matrix}

(13)

From the foregoing results, it is evident that GDP is contemporaneously endogenous in the system and responds to all innovations with the exceptions of the government expenditure innovations. Moreover, urbanisation responds to its own innovations and the one from agriculture. Industrialisation negatively responds to shocks in the agriculture sector, which is not surprising given the substitutability of the two sectors. GDP and industrialisation have a positive impact on government expenditure. In comparison with the causality test, the contemporaneous analysis seems to provide evidence for a less strong relationship between the variables.

Having examined the contemporaneous relationship, we proceed with the impulse response and variance decomposition analyses to shed more light on the relationship between the variables. The impulse response analysis provides a dynamic description of the effects of structural shocks to the economy, while the variance decomposition analysis indicates the magnitude of the predicted error variance for a series accounted for by innovations from each of the independent variables over different time horizons.

The results of the impulse response analysis are reported in Figures 3–7, which show a one standard deviation shock to one variable on other system variables over time. Commenting on the shocks of GDP hitting the other variables, it is observed that positive shocks of GDP started hitting urbanisation in the 16th period. Shocks in GDP started to have a negative effect on industrialisation and government expenditure after the fifth and seventh period, respectively. We also observe that the shocks in urbanisation initially have a negative impact on GDP. However, starting from the 16th period, the shocks from urbanisation start to have positive impact on GDP. The shocks from urbanisation do not have a stable impact on industrialisation and agriculture. The shocks in agriculture have a positive impact on the RGDP throughout the period under review, while the shocks in government expenditure initially have a positive impact on the RGDP, but a negative impact on it after the 12th period. Table 9 shows the results of the variance decomposition analysis, which indicates the share of fluctuations in the variables caused by different shocks. The results indicate that after the 1st, 10th and 20th periods, urbanisation contributes 7 per cent, 19 per cent and 20 per cent, respectively, to the variance in RGDP. Both agriculture and industrialisation continue to account for an increasing share of the fluctuations in RGDP over time. On the other hand, government expenditure accounts for a decreasing share of the fluctuations in the RGDP. Therefore, these results suggest that urbanisation, agriculture and industrialisation are important determinants of economic growth in the country.

Figure 3

Source: Author’s computation.

Figure 4

Source: Author’s computation.

5. Conclusion

Although most developed countries are more urbanised than developing countries, the growth rate of migration into cities is higher in developing countries, but in spite of this trend, the connection between urbanisation and economic growth remains vague. Unfortunately, time-series studies on this subject are virtually non-existent. Hence, the present exercise assesses the link between urbanisation and economic development, while considering a few variables such as agriculture, industrial development and government expenditure. Recent time-series procedures are employed to look at the relationship between the series.

Figure 5

Source: Author’s computation.

Figure 6

Source: Author’s computation.

Figure 7

Source: Author’s computation.

Table 9

Variance Decomposition Analysis

Period	Proportion of Variance in ‘RGDP’ Accounted for by:
Period	RGDP	URB	AGR	IND	GOV
1	100	0	0	0	0
5	66.640	7.148	16.457	3.848	5.907
10	44.914	19.178	22.699	9.385	3.823
15	40.627	19.443	25.823	10.846	3.261
20	39.362	19.758	26.551	11.182	3.148
	Proportion of Variance in ‘URB’ Accounted for by:
1	5.549	94.451	0	0	0
5	29.407	66.487	1.471	0.593	2.042
10	42.682	42.440	11.552	0.302	3.024
15	38.325	39.695	17.226	0.323	4.431
20	38.098	39.661	17.322	0.447	4.472
	Proportion of Variance in ‘AGR’ Accounted for by:
1	18.849	7.306	73.845	0	0
5	38.231	16.200	31.786	4.053	9.729
10	36.897	17.746	30.490	5.869	8.999
15	36.066	22.378	27.311	5.960	8.285
20	36.281	22.344	26.539	6.512	8.325
	Proportion of Variance in ‘IND’ Accounted for by:
1	25.219	0.382	37.142	37.257	0
5	28.838	5.468	16.585	26.950	22.158
10	28.232	5.072	15.621	29.165	21.910
15	25.472	5.273	19.696	27.564	21.995
20	24.874	6.637	20.896	25.303	22.290
	Proportion of Variance in ‘GOV’ Accounted for by:
1	1.551	0.013	0.268	12.346	85.823
5	2.057	14.783	16.022	17.621	49.516
10	7.451	12.605	14.725	19.855	45.364
15	9.114	10.881	18.252	19.994	41.759
20	9.485	10.410	20.779	19.243	40.083

Source: Author’s computation.

The analysis reveals that there is a long-run connection among the series, though they attain stationarity at the first difference. The causality flows from urbanisation, agriculture, industrialisation and government expenditure to economic growth in the long run. The analyses from the long-run estimates, which were generated from the ARDL method, suggest urbanisation has a positive effect on economic activity, agriculture and industrialisation positively influence national output, and the impact of government expenditure is insignificant on national output. Due to the deficiencies associated with the single-equation method (including the ARDL model), we have also used SVECM to analyse the relationship between the variables. The impulse response and variance decomposition analyses derived from the SVECM method suggest that urbanisation, agriculture and industrialisation are important determinants of economic growth.

The fact that the urban centres in the country account for the largest share of economic activity supports the results. Urbanisation is an instrument rather than an indicator of economic development in the country. It also implies that the nation is not over-urbanised. Therefore, the promotion of urbanisation cannot be ignored in the process of development (Shahbaz & Lean, 2012). However, one of the negative aspects of urban development in the country is the lack of basic amenities. Thus, for the effective and sustainable contribution of urbanisation to economic development, the authorities should invest in the social and physical infrastructures of the urban centres, which include roads, water, energy, waste management, street lighting, health care and policing. Once established, the physical structures of a city may last for more than 150 years (World Bank, 2013). Such activities will also boost the industrialisation process, since most industries in the countries are located in the urban centres.

Complementary policies for the urban centres must be consolidated with corresponding programmes to improve agricultural activities, such as the provisions of low-cost agricultural tools, adequate markets for agricultural outputs and other incentives. As agriculture is a rural phenomenon, the development of urban centres should not jeopardise government efforts to resuscitate the rural areas. The government should redirect some of its expenditure towards the long-term development of urban centres and rural areas, to make these expenditures more efficient and productive.

Footnotes

Appendix

Notes

References

Abdel-Rahman

A.N.

Safarzadeh

M.R.

& Bottomley

M.B.

(2006). Economic growth and urbanization: A cross-section and time-series analysis of thirty-five developing countries. Internal Review of Economics, 53(3), 334–48.

Alam

Fatima

& Butt

M.S.

(2007). Sustainable development in Pakistan in the context of energy consumption demand and environmental degradation. Journal of Asian Economics, 18(5), 825–37.

Amisano

, & Giannini, C. (1997). Topics in structural VAR econometrics (2nd ed.). Berlin: Springer-Verlag.

Bertinelli

& Strobl

(2003). Urbanisation, urban concentration and economic growth in developing countries. CREDIT Research Paper No. 03/14. Nottingham, UK: University of Nottingham

Bertinelli

& Black

(2004). Urbanization and growth. Journal of Urban Economics, 56(1), 80–96.

Bloom

D.E.

Canning

& Fink

(2008). Urbanization and the wealth of nations. Science, 319(5864), 772–75.

Bradshaw

Y.W.

& Fraser

(1989). City size, economic development, and quality of life in China: New empirical evidence. American Sociological Review, 54(6), 986–1003.

Brückner

(2012). Economic growth, size of the agricultural sector, and urbanization in Africa. Journal of Urban Economics, 71(1), 26–36.

Brülhart

& Sbergami

(2009). Agglomeration and growth: Cross-country evidence. Journal of Urban Economics, 65(1), 48–63.

10.

CBN (2014). Central Bank of Nigeria statistical bulletin. Retrieved 12 December 2014, from http://www.cenbank.org/OUT/2011/PUBLICATIONS/STATISTICS/2010/INDEX.HTML

11.

Chu

K.Y.

Gupta

& Clements

(1995). Unproductive public expenditures: A pragmatic approach to policy analysis. IMF Pamphlet Series, 48, 1–45.

12.

Cohen

(2006). Urbanization in developing countries: Current trends, future projections, and key challenges for sustainability. Technology in Society, 28(1), 63–80.

13.

Davis

J.C.

& Henderson

J.V.

(2003). Evidence on the political economy of the urbanization process. Journal of Urban Economics, 53(1), 98–125.

14.

Engle

R.F.

& Granger

C.W.

(1987). Co-integration and error correction: Representation, estimation, and testing. Econometrica, 55, 251–76.

15.

Giles

D.E.

& Godwin

R.T.

(2012). Testing for multivariate cointegration in the presence of structural breaks: p-values and critical values. Applied Economics Letters, 19(16), 1561–65.

16.

Gregory

A.W.

& Hansen

B.E.

(1996). Residual-based tests for cointegration in models with regime shifts. Journal of Econometrics, 70(1), 99–126.

17.

Henderson

(2003). The urbanization process and economic growth: The so-what question. Journal of Economic Growth, 8(1), 47–71.

18.

Johansen

Mosconi

& Nielsen

(2000). Cointegration analysis in the presence of structural breaks in the deterministic trend. The Econometrics Journal, 3(2), 216–49.

19.

Laidlaw

K.A.

& Stockwell

E.G.

(1979). Trends in the relationship between urbanisation and development in Africa. The Journal of Modern African Studies, 17(04), 687–94.

20.

Lee

& Strazicich

M.C.

(2001). Break point estimation and spurious rejections with endogenous unit root tests. Oxford Bulletin of Economics and Statistics, 63(5), 535–58.

21.

Lee

& Strazicich

M.C.

(2003). Minimum LM unit root test with two structural breaks. The Review of Economics and Statistics, 63, 1082–89.

22.

Lee

& Strazicich

M.C.

(2004). Minimum LM unit root test. Working Paper, Department of Economics, Boone, North Carolina: Appalachian State University.

23.

Liddle

(2013). The energy, economic growth, urbanization nexus across development: Evidence from heterogeneous panel estimates robust to cross-sectional dependence. The Energy Journal, 34(2), 223–44.

24.

Liddle

& Lung

(2014). Might electricity consumption cause urbanization instead? Evidence from heterogeneous panel long-run causality tests. Global Environmental Change, 24, 42–51.

25.

Liu

(2009). Exploring the relationship between urbanization and energy consumption in China using ARDL (autoregressive distributed lag) and FDM (factor decomposition model). Energy, 34(11), 1846–54.

26.

(2010). Urbanisation and economic growth: Testing for causality. The 16th annual Pacific Rim Real Estate Conference, 24th–27th January 2010, Wellington.

27.

Liang

Duffy

& Zhao

(2011). Effects of urbanization and industrialization on agricultural land use in Shandong Peninsula of China. Ecological Indicators, 11(6), 1710–14.

28.

Lütkepohl

(1982). Non-causality due to omitted variables. Journal of Econometrics, 19(2), 367–78.

29.

Massidda

& Mattana

(2013). A SVECM analysis of the relationship between international tourism arrivals, GDP and trade in Italy. Journal of Travel Research. 52(1), 93–105.

30.

McCoskey

& Kao

(1999). Testing the stability of a production function with urbanization as a shift factor. Oxford Bulletin of Economics and Statistics, 61(S1), 671–90.

31.

Moomaw

R.L.

& Shatter

A.M.

(1996). Urbanization and economic development: A bias toward large cities? Journal of Urban Economics, 40(1), 13–37.

32.

Nafziger

E.W.

(1972). The economic impact of the Nigerian civil war. The Journal of Modern African Studies, 10(02), 223–45.

33.

Narayan

P.K.

(2005). The saving and investment nexus for China: Evidence from cointegration tests. Applied Economics, 37(17), 1979–90.

34.

The New York Times (2012, April 14). Nigeria tested by rapid rise in population. Retrieved 26 June 2016, from http://www.theatlantic.com/international/archive/2012/07/this-is-africas-new-biggest-city-lagos-nigeria-population-21-million/259611/

35.

Oyeleye

O.I.

(2013). Challenges of urbanization and urban growth in Nigeria. American Journal of Sustainable Cities and Society, 2(1), 79–95.

36.

Perron

(1989). The great crash, the oil price shock, and the unit root hypothesis. Econometrica, 57, 1361–401.

37.

Pesaran

M.H.

Shin

& Smith

R.J.

(2001). Bounds testing approaches to the analysis of level relationships. Journal of Applied Econometrics, 16(3), 289–326.

38.

Phillips

P.C.

& Hansen

B.E.

(1990). Statistical inference in instrumental variables regression with I (1) processes. The Review of Economic Studies, 57(1), 99–125.

39.

Ravallion

(2002). On the urbanization of poverty. Journal of Development Economics, 68(2), 435–42.

40.

Sackey

Liverpool-Tasie

Salau

& Awoyemi

(2012). Rural-urban transformation in Nigeria. Journal of African Development, 14(2), 131–68.

41.

Shahbaz

& Lean

H.H.

(2012). Does financial development increase energy consumption? The role of industrialization and urbanization in Tunisia. Energy Policy, 40, 473–79.

42.

Solarin

S.A.

(2011). Electricity consumption and economic growth: Trivariate investigation in Botswana with capital formation. International Journal of Energy Economics and Policy, 1(2), 32.

43.

Solarin

S.A.

(2014). Multivariate causality test of electricity consumption, capital formation, export, urbanisation and economic growth for Togo. Energy Studies Review, 21(1), 109–32

44.

Solarin

S.A.

& Shahbaz

(2013). Trivariate causality between economic growth, urbanisation and electricity consumption in Angola: Cointegration and causality analysis. Energy Policy, 60, 876–84.

45.

Stock

J.H.

& Watson

M.W.

(1993). A simple estimator of cointegrating vectors in higher order integrated systems. Econometrica, 64(4), 783–820.

46.

Thanh

(2007). Economic development, urbanisation and environmental changes in Ho Chi Minh City, Vietnam: Relations and policies. Paper presented at the workshop, Urban Population, Development and Environment Dynamics in Developing Countries, 11–13 June, Nairobi.

47.

The Atlantic (2012, July 10). This is Africa’s new biggest city: Lagos, Nigeria, population 21 million. Retrieved 26 June 2016, from http://www.theatlantic.com/international/archive/2012/07/this-is-africas-new-biggest-city-lagos-nigeria-population-21-million/259611

48.

Turner

(2009). Testing for cointegration using the Johansen approach: Are we using the correct critical values? Journal of Applied Econometrics, 24(5), 825–31.

49.

Turok

& McGranahan

(2013). Urbanization and economic growth: The arguments and evidence for Africa and Asia. Environment and Urbanization, 25(2), 465–82.

50.

UN-HABITAT (2006). Documentation of lessons learnt in the Sustainable Cities Programme in Nigeria. Retrieved 26 June 2016, from http://www.unhabitat.org/downloads/docs/5431_36914_7550_NIG_NATIONAL_REPORT-FINAL.doc.

51.

UN-HABITAT (2014). The state of African cities 2014q. Retrieved 26 June 2016, from http://unhabitat.org/the-state-of-planning-in-africa/

52.

World Bank (2013). Rural-urban dynamics and the Millennium Development Goals. Retrieved 26 June 2016, from http://siteresources.worldbank.org/INTPROSPECTS/Resources/334934-1327948020811/8401693-1355753354515/8980448-1366123749799/GMR_2013_Full_Report.pdf

53.

World Bank (2014). World development indicators on CD-ROM , Washington DC: World Bank.

54.

Zivot

& Andrews

D.W.

(1992). Further evidence on the great crash, the oil-price shock, and the unit-root hypothesis. Journal of Business & Economic Statistics, 10, 251–70.

Panel A: Series at Level
Variable	ADF	Lee and Strazicich Test (2004)			Lee and Strazicich Test (2003)
Variable	T-stat	T-stat	Break		T-stat	Break (1)	Break (2)
GDP_t	-1.450[1]	-2.523[1]	1985	(0.4)	-4.126[1]	1973	1995	(0.4, 0.6)
URB_t	-0.494[1]	-2.080[1]	1968	(0.2)	-2.786[1]	1972	1988	(0.4, 0.6)
AGR_t	-1.819[0]	-3.462[0]	1970	(0.2)	-3.305[0]	1972	1981	(0.2, 0.4)
IND_t	-1.848[4]	-2.711[0]	1969	(0.2)	-4.452[0]	1978	1999	(0.4, 0.8)
GOV_t	-2.167[1]	-3.938[3]	1970	(0.2)	-2.768[1]	1971	2001	(0.2, 0.8)
Panel B: Series at First Difference
Variable	ADF	Lee and Strazicich Test (2004)			Lee and Strazicich Test (2003)
Variable	T-stat	T-stat	Break		T-stat	Break (1)	Break (2)
ΔGDP_t	-5.057***[0]	-5.339***[1]	1982	(0.4)	-6.013**[1]	1974	1981	(0.2, 0.4)
ΔURB_t	-4.092***[0]	-4.865***[0]	1969	(0.2)	-5.852**[4]	1974	1981	(0.2, 0.4)
ΔAGR_t	-7.806***[0]	-9.116***[0]	1979	(0.4)	-9.698***[0]	1979	1990	(0.4, 0.6)
ΔIND_t	-7.354***[1]	-9.218***[0]	1972	(0.3)	-8.883***[0]	1984	2000	(0.4, 0.8)
ΔGOV_t	-9.707***[0]	-9.598***[0]	1995	(0.6)	-9.807***[0]	1983	1995	(0.4, 0.6)