There has been considerable decline in the investment on road transport infrastructure in recent times, as a result of the dwindling economic investment owing to lowering gross domestic product (GDP) since 2009.

The objective of this study was to examine the relationship between road transport investment (ROTI) and economic development (ED) in South Africa. This article adopts the Harrod–Domar (HD) model of economic growth and development theory, endogenous growth theory and Solow–Swan neoclassical growth model.

Data were derived from the South African Reserve Bank, Quantec database and Statistics South Africa (StatsSA) between 1990 and 2014. It used time series, econometric models cointegration and vector error correction model (VECM) to analyse.

The results of the estimation demonstrate that the explanatory variables account for approximately 86.7% variation in ED in South Africa. Therefore, there exists a positive relationship between ROTI and ED.

This study established a long-run relationship between phenomena and demonstrates the role of road transport investment on economic development in South Africa.

The availability of road transport infrastructure investment (ROTI) in a country is of paramount importance, given that it enhances the economic chains and accelerates growth and development in the country. Various studies (Aschauer

It is imperative, therefore, to note that road transport infrastructural investment directly impacts ED, creates jobs and improves well-being of the people within the geographical enclave. It serves as input in the production process of various goods and services leading to better quality of life, reduced cost production and a greater value for money, while providing opportunities for its teeming populace. Directly or indirectly, ROTI raises productivity of workers, reduces cost of transportation, reduces turnover time and provides safety within a country. All these antecedents paved the way for investment of all sorts on the one hand, and positive external total output that exceeds the private returns. Considering South Africa’s curve in road transport infrastructural development since the dawn of democracy, one is tempted to note with caution that South Africa has one of the very best well-developed infrastructure investments, such as roads, among others, like health and educational facilities in the rural world in general and sub-Saharan Africa in particular. Yet, a bulk of infrastructure like roads, railways and ports require sizable investment on both maintenance and upgrade. This study is an attempt to establish the relationship between ROTI and ED in South Africa that is relatively not well exploited. A descriptive method based on empirical data and literature is pursued.

There are a number of theories to explain the linkage between ED and road transport infrastructural investment. ED theories are treated in this case as an extension of conventional economic growth theory and therefore development was merely equated to growth. Hence, gross domestic product (GDP) was a proxy for overall ED. ED, in this sense, includes national production, social equity aspect such as elimination of poverty, inequality and unemployment (Hall

Domar (

Where s is the marginal and average saving ratio.

Secondly, the economy is in equilibrium, that is, planned investments equal planned savings:

Thirdly, investment is determined by the expected increase in national income (∆Y) and a fixed technical coefficient known as incremental capital output ratio (ICOR):

By definition, ED (g) is the change in income per unit of income:

Substitution of the relationship in equation gives an alternative definition of growth as follows:

This model was developed by Solow and Swan in 1956. They related the assumptions of the fixed ICOR and the labour usage in the Harrod–Domar (HD) model. Solow (

According to the endogenous growth theory, the long-run growth rate and development depends on the provision of infrastructure services particularly road (Barro & Sala-i-Martin

Sturm, Kuper and De Haan (

Aschauer (

Montolio and Solé-Ollé (

Looney (

Kayode et al. (

Ashipala and Haimbodi (

South Africa has a better ROTI as compared with other emerging countries in the South. The ROTI recovered rising from 2.76% of GDP in 2004, 2.90% in 2005, 6.05% in 2008 and 7.64% in 2009 (Fedderke & Garlick

Perkins, Fedderke and Luiz (

According to Fedderke and Bogeti (

While for instance, Perkins et al. (

Though, having a better ROTI is paramount. Because several empirical studies (Aschauer

All research known to mankind are based on some underlying philosophical assumptions on what constitutes validity and reliability. The econometric models – cointegration and VECM is used to test the relationship between ROTI and ED in South Africa.

In examining the relationship between ROTI and ED, the study employed econometric technique. According to the endogenous growth theory, the long-run ED depends on the provision of infrastructure services particularly road (Barro & Sala-i-Martin

Based on the theoretical considerations discussed, the model is specified as follows:

Where, ED represents economic development, ROTI is the road transport investment, GENOT is government expenditure on road transport, EXCH is the exchange rate and I represents income.

The empirical model used in the study is consistent with that of Perkin et al. (

Cointegration is a statistical implication of the existence of a long-run relationship between the economic variables (Inder

The Johansen cointegration test is robust to various departures from normality in that it allows any of the four variables in the model used as the dependent variable while maintaining the same cointegration results. The cointegration and causality tests were carried out only on the stationary variables, I(1) or I(0). In the Johansen technique for cointegration, we test for r (the maximum number of cointegration relationships) which also includes testing procedures for linear restrictions on the cointegrating parameters, for any set of variables. In this technique, two statistics are used to identify the number of cointegrating vectors: the trace test statistic and the maximum-eigenvalue test statistic. The trace statistics evaluates the null hypothesis that there are at most r cointegrating vectors, whereas the maximum-eigenvalue test evaluates the null hypothesis that there are exactly r cointegrating vectors in _{t}

The endogenous growth theory argues that the provision of ROTI increases ED (Barro & Sala-i-Martin

The secondary data employed in the study are from 1990 to 2014 in quarterly time series. Data for the variables are obtained from the South African Reserve Bank, Statistics South Africa and Quantec database.

This study uses time series data in testing the stationarity properties. According to Brooks, stationary series can be defined as one with constant mean, constant variance and constant autocovariances for each lag (Brooks

The informal tests were carried out through graphical inspections and the results are presented in

Stationarity graphs at levels. (a) Economic development, (b), road transport investment (c) government expenditure on road transport, (d) income and (e) exchange rate.

In the ADF, the test adjusts the notion that error terms are independently and identically distributed. The ADF test is based on the hypothesis which states that:

_{0}):

_{1}):

The results of the ADF test consist of the t statistic and the critical values of a zero coefficient. If the t statistic is greater than the critical value, then the time series data are said to be non-stationary and unit root exists. In this case, the null hypothesis is not rejected. If the t statistic is less than the critical values, time series data are stationary and unit root does not exist therefore the null hypothesis is rejected.

The PP tests are performed to complement the ADF test because the former test can properly distinguish between stationary and non-stationary time series with high degree of autocorrelation and presence of structural break. The PP test is a more comprehensive theory of unit root testing. The PP tests are analogous to ADF tests, but they incorporate an automatic correction to the Dickey Fuller (DF) procedure to allow for auto-correlated residuals and usually give the same conclusions as the ADF tests (Gujarati

Results from

Unit root tests: Level series.

Variable | Augmented Dickey–Fuller test |
Phillips–Perron test |
||||
---|---|---|---|---|---|---|

Constant | Constant and trend | None | Constant | Constant and trend | None | |

ED | −1.634228 | −0.058578 | −1.144787 | −0.634228 | −0.058578 | −2.081128 |

ROTI | −0.385265 | −2.949198 | 2.529187 | 0.229008 | −2.926668 | 3.182100 |

GENOT | −1.398313 | −3.283108 | −1.321925 | −0.759928 | −2.308931 | −1.408210 |

EXCH | −1.950194 | −3.481184 | 11.39549 | −3.182252 | −3.362199 | 12.98635 |

I | 2.486681 | −0.358819 | −2.728705 | −8.218482 | −9.022105 | −2.275511 |

ED, economic development; ROTI, road transport investment; GENOT, government expenditure on road transport; EXCH, exchange rate; I, income.

Unit root tests: First difference series.

Variable | Augmented Dickey–Fuller test |
Phillips–Perron test |
||||
---|---|---|---|---|---|---|

Constant | Constant and trend | None | Constant | Constant and trend | None | |

ED | −9.476293 |
−9.435986 |
−9.520240 |
−46.72715 |
−47.07350 |
−46.95288 |

ROTI | −8.942108 |
−8.887047 |
−8.996896 |
−47.41466 |
−47.34633 |
−47.76597 |

GENOT | −8.990376 |
−8.936771 |
−9.043588 |
−15.24158 |
−15.01964 |
−15.39976 |

EXCH | −9.060199 |
−8.995857 |
−9.121416 |
−37.17517 |
−42.35735 |
−35.10184 |

I | −10.90503 |
−10.84363 |
−10.96445 |
−38.23886 |
−37.70277 |
−37.835058 |

ED, economic development; ROTI, road transport investment; GENOT, government expenditure on road transport; EXCH, exchange rate; I, income.

significance at 5%;

significance at 10%.

ROTI is non-stationary at level series using ADF but becomes stationary when using PP at 1% level of significance on the level series. GENOT is non-stationary at both ADF and PP at level series. At first difference, using ADF, GENOT is stationary at none (10% level of significance) and stationary at 1% level of significance on intercept, trend and intercept and none when using the PP test. EXCH is stationary at first difference series at intercept (10% level significance) and none (5% level of significance) when using ADF. Income (I) is stationary at 1% level of significance when using PP at intercept, trend and intercept and none.

The results suggest that all variables examined are not stationary at levels, which therefore implies that they are integrated of I (1) as the critical values are less than computed values and this require further differencing.

Cointegrating relations.

Hypothesised number of CE(s) | Trace statistic | 0.05 critical value | Maximum eigen statistics | 0.05 critical value |
---|---|---|---|---|

None |
64.73021 | 47.85613 | 29.13632 | 27.58438 |

At most 1 |
25.59389 | 29.79707 | 13.28200 | 21.13162 |

At most 2 | 12.31189 | 15.49471 | 9.669798 | 14.26460 |

At most 3 | 2.64209 | 3.841466 | 2.642090 | 3.841466 |

CE, competing endogenous.

Trace test indicates 1 cointegrating equations at the 0.05 level.

Denotes rejection of the hypothesis at the 0.05 level; MacKinnon, Haug and Michelis (

Vector error correction model results.

Variable | Coefficient | Standard error | t Statistic | Prob. value |
---|---|---|---|---|

C | −0.587486 | 5.086183 | 3.688633 | 0.0004 |

ROTI | 0.51927 | 0.221507 | 2.449209 | 0.0297 |

GENOTI | −0.010897 | 0.060236 | −0.182189 | 0.8559 |

EXCH | 0.509022 | 0.239206 | 2.477006 | 0.0021 |

I | 0.5925061 | 0.053237 | −1.352108 | 0.1800 |

^{2} |
0.866813 | - | - | - |

Adjusted ^{2} |
0.801525 | - | - | - |

ROTI, road transport investment; GENOTI, government expenditure on road transport; EXCH, exchange rate; I, income, ^{2}, explanatory variable; C, constant; ED, Dependant variable.

The variables were tested for the order of integration in first difference series and the results are presented in

Differential series. (a) Economic development, (b), road transport investment (c) government expenditure on road transport, (d) exchange rate and (e) income.

Brooks (

Other diagnostic tests.

Test | H_{0} |
Test statistic | Conclusion | |
---|---|---|---|---|

Jarque–Bera | Residuals are normally distributed. | 34.702 | 0.11 | Errors are normally distributed |

VEC residual serial correlation LM tests | There is no serial correlation in the residuals. | 78.473 | 0.11 | No second-order autocorrelation |

VEC residual heteroscedasticity tests | The residuals are homoscedastic. | 86.082 | 0.35 | No heteroscedasticity |

VEC, vector error correction; LM test, Lagrange multiplier test.

Because one cointegrating vector is found, estimation of VECM which adjusts to both short-run and long-run changes in variables and deviations from equilibrium follows:

The results of the estimation show that the explanatory variables account for approximately 86.7% variation in ED in South Africa. Therefore, there exists a positive relationship between ROTI and ED. Moreover, a positive relationship is expected between EXCH and ED as the theory asserts. I is income, the economic growth theory view on income provided that ED depends on income. It implies that an increase in income also increases ED. Therefore, a positive relationship is estimated. ^{2} is also highly equal to 86% and this signifies that variation in the regression explains the variation on dependent variable to 86%.

The VECM was subjected to rigorous diagnostic tests. Diagnostic checks are crucial in this analysis to establish if the assumptions which underlie the classical linear regression model is observed. The VECM was tested for AR Roots test and serial correlation and the results are indicated in

Inverse roots of AR characteristic polynomial.

The AR Roots Graph reports the inverse roots of the characteristic AR polynomial. The estimated VECM is stable that is stationary if all roots have modulus less than one and lie inside the unit circle. In our case, as illustrated in

The residuals were also examined for the normality, autocorrelation and heteroscedasticity and the results are reported in

This article focused on interpreting the results of models estimated. This article began with analysing the time series properties of the data using three methods of testing for unit root. All three methods confirmed that the variables are integrated of order one, I(1). Having determined the order of integration of the variables, the lag length used in the estimation for the Johansen cointegration was determined empirically. This, therefore, implied that there is a long-term relationship between ED and its determinants. The VECM was also estimated to analyse both the long-run and the short-run interaction between the variables. The long-run equation showed that all the variables employed in the model are significant and carried the correct signs. The results also observed all the assumptions which underlie the classical linear regression model.

The strength of this article lies in reaffirming the established relationships of road transport infrastructural investment with ED. Taking into cognisance the significance and the challenges confronting road transport infrastructure investment in South Africa, this article argued that ROTI is pivotal to ED. Stretching from an introduction to empirical review, where other studies demonstrated similar tendencies around the world, in Africa and in South Africa as well, outlining the studies of Perkins, Fedderke and Luiz, and Fedderke and Bogeti conducted in South Africa, it stretches the discussion even further by providing an analysis between 1990 and 2014 using the VECM AR Roots Johansen cointegration in assessing the relationship.

Based on both theoretical and empirical literature, the study explains the impact of ROTI and ED. The model explains ED as a function of road transport investment, GENOT, EXCH and income.

To empirically examine the run impact of ROTI on ED, the study employed Johansen’s cointegration approach, and the VECM was employed as to capture both the short-run and long-run dynamics of the estimated model.

As it is common that macroeconomic time series are trended, in most cases the variables are nonstationary and using a non-stationary data may lead to invalid results and conclusion. For this reason, before conducting the cointegration test, the study first conducted the stationarity test for all variables under investigation using both the informal and formal tests. For the informal test, graphical inspections were used, while PP and ADF tests were applied in order to formally test for stationarity. Having found that the variables are stationary after first differencing and are integrated of the same order, the study further conducted a cointegration test to check if there exists a long-run relationship between the two variables.

The results revealed that there is cointegration among the variables under examination. With the first model, the trace statistics suggested three cointegrating vectors, whereas the maximum eigenvalue suggested two cointegrated vectors. With the latter, the trace statistics suggest two cointegrating vectors, whereas the maximum eigenvalue indicates that there is no cointegration. The presence of at least one cointegrating vector allowed for estimation of the VECM, which was followed by diagnostic checks through autocorrelation (serial correlation), heteroscedasticity and normality of the residuals, the results from these tests are positive. These results are in accordance with most of the studies such as Peter et al. (2015) reviewed the literature in the sense that they confirm a long-run link between the two variables.

In general, the results confirm the existence of a link between ROTI and ED in both models. Also, the nature of the relationship is in accordance with the a-priori expectations as presented in this article. It is interesting to find that the ROTI, as sophisticated as it is, contributes this more towards growth.

Given the results above, the study makes the following policy recommendations:

The South African government should develop policies that encourage the incorporation of the ROTI into the economic system.

Policy makers (Fiscal and Monetary) should embark on economic activities that enhance the link between ROTI and ED, such as, stimulating savings which in turn improves the level of investment. Considering that literature proves investment as the main channel through which ROTI contributes towards ED.

Lastly, an environment that enables ROTI to directly impact ED should be created.

The use of data from mainly the South African Reserve Bank would limit the effect of the study to only the Reserved Bank analysis. Variables that were converted from annual to quarter might also result in the frequency disparities. This adjustment might have contributed to some of the challenges experienced in the study. Using the short vector error correction alone might portend another limitation on the contrary.

The authors declare that they have no financial or personal relationships that may have inappropriately influenced them in writing this article.

A.H. was responsible for data generating and E.A.N. was responsible for synergy, correction and literature.

Vector error correction estimates.

Cointegrating Eq: |
CointEq1 |
|||
---|---|---|---|---|

Date: 11/10/16 |
||||

ED(−1) | 1.000000 | |||

ROTI(−1) | 0.519275 | |||

(0.14846) | ||||

[3.49769] | ||||

GENOTI(−1) | −0.010897 | |||

(0.02035) | ||||

[-0.53541] | ||||

EXCH(−1) | 0.509022 | |||

(0.21532) | ||||

[2.36405] | ||||

C | −0.587486 | |||

CointEq1 | −0.113432 | −0.420689 | −0.241059 | −0.067092 |

(0.02500) | (0.44227) | (0.32633) | (0.04003) | |

[−4.53812] | [−0.95120] | [−0.73871] | [−1.67621] | |

D(ED(−1)) | 0.509464 | 0.814187 | −0.096950 | −0.230819 |

(0.12578) | (2.22554) | (1.64209) | (0.20141) | |

[4.05052] | [0.36584] | [−0.05904] | [−1.14600] | |

D(ED(−2)) | 0.259991 | −0.170035 | −0.049930 | 0.028214 |

(0.12926) | (2.28723) | (1.68761) | (0.20700) | |

[2.01132] | [−0.07434] | [−0.02959] | [0.13630] | |

D(ED(−3)) | 0.067026 | 0.363996 | 2.407533 | 0.069143 |

(0.07367) | (1.30359) | (0.96184) | (0.11797) | |

[0.90977] | [0.27923] | [2.50306] | [0.58608] | |

D(ED(−4)) | −0.642106 | −0.826820 | −0.995229 | −0.097624 |

(0.07736) | (1.36892) | (1.01004) | (0.12389) | |

[−8.29970] | [−0.60400] | [−0.98534] | [−0.78801] | |

D(ED(−5)) | 0.409964 | 0.384865 | −0.823270 | −0.008425 |

(0.11456) | (2.02711) | (1.49568) | (0.18345) | |

[3.57851] | [0.18986] | [−0.55043] | [−0.04592] | |

D(ED(−6)) | 0.213486 | 0.717173 | 0.253322 | 0.081986 |

(0.10968) | (1.94078) | (1.43198) | (0.17564) | |

[1.94637] | [0.36953] | [0.17690] | [0.46678] | |

D(ROTI(−1)) | 0.050538 | −0.641271 | 0.055571 | 0.069135 |

(0.01465) | (0.25921) | (0.19126) | (0.02346) | |

[3.44978] | [−2.47390] | [0.29055] | [2.94708] | |

D(ROTI(−2)) | 0.040283 | −0.353571 | 0.006260 | 0.074102 |

(0.01491) | (0.26387) | (0.19469) | (0.02388) | |

[2.70130] | [−1.33996] | [0.03215] | [3.10309] | |

D(ROTI(−3)) | 0.032206 | −0.317449 | −0.119951 | 0.060916 |

(0.01297) | (0.22943) | (0.16928) | (0.02076) | |

[2.48387] | [−1.38366] | [−0.70860] | [2.93384] | |

D(ROTI(−4)) | 0.008076 | 0.443273 | −0.177103 | 0.049247 |

(0.01142) | (0.20210) | (0.14912) | (0.01829) | |

[0.70703] | [2.19334] | [−1.18768] | [2.69257] | |

D(ROTI(−5)) | 0.002760 | 0.423052 | −0.232021 | 0.007910 |

(0.01014) | (0.17949) | (0.13243) | (0.01624) | |

[0.27209] | [2.35701] | [−1.75200] | [0.48699] | |

D(ROTI(−6)) | 0.002351 | 0.078661 | −0.091316 | −0.002703 |

(0.00764) | (0.13512) | (0.09969) | (0.01223) | |

[0.30791] | [0.58218] | [−0.91597] | [−0.22108] | |

D(GENOTI(−1)) | −0.017080 | 0.140487 | 0.002449 | 0.014481 |

(0.01092) | (0.19324) | (0.14258) | (0.01749) | |

[−1.56401] | [0.72701] | [0.01718] | [0.82807] | |

D(GENOTI(−2)) | −0.009873 | −0.084805 | 0.094848 | −0.021071 |

(0.01144) | (0.20246) | (0.14938) | (0.01832) | |

[−0.86287] | [−0.41887] | [0.63492] | [−1.14996] | |

D(GENOTI(−3)) | −0.023824 | −0.273758 | −0.084966 | −0.030515 |

(0.01130) | (0.19994) | (0.14753) | (0.01809) | |

[−2.10835] | [−1.36918] | [−0.57594] | [−1.68638] | |

D(GENOTI(−4)) | −0.030802 | −0.077042 | 0.288115 | −0.023120 |

(0.01244) | (0.22013) | (0.16242) | (0.01992) | |

[−2.47589] | [−0.34999] | [1.77389] | [−1.16052] | |

D(GENOTI(−5)) | −0.001345 | −0.219362 | 0.066641 | −0.021640 |

(0.01274) | (0.22547) | (0.16636) | (0.02040) | |

[−0.10558] | [−0.97293] | [0.40059] | [−1.06055] | |

D(GENOTI(−6)) | 0.008745 | 0.105475 | −0.113405 | 0.015815 |

(0.01263) | (0.22350) | (0.16491) | (0.02023) | |

[0.69236] | [0.47193] | [−0.68769] | [0.78187] | |

D(EXCH(−1)) | 0.024127 | 1.585671 | 0.477535 | 0.552556 |

(0.08685) | (1.53677) | (1.13389) | (0.13908) | |

[0.27780] | [1.03182] | [0.42115] | [3.97299] | |

D(EXCH(−2)) | 0.144663 | −0.172261 | −0.719269 | 0.166915 |

(0.09881) | (1.74837) | (1.29001) | (0.15823) | |

[1.46406] | [−0.09853] | [−0.55757] | [1.05491] | |

D(EXCH(−3)) | 0.069737 | 0.421250 | −0.476492 | 0.105299 |

(0.08573) | (1.51695) | (1.11926) | (0.13728) | |

[0.81344] | [0.27770] | [−0.42572] | [0.76702] | |

D(EXCH(−4)) | 0.461503 | 0.014257 | −0.512343 | −0.478010 |

(0.08251) | (1.45999) | (1.07724) | (0.13213) | |

[5.59317] | [0.00976] | [−0.47561] | [−3.61774] | |

D(EXCH(−5)) | −0.132806 | −2.577539 | 0.548396 | 0.366085 |

(0.11334) | (2.00548) | (1.47972) | (0.18150) | |

[−1.17174] | [−1.28525] | [0.37061] | [2.01704] | |

D(EXCH(−6)) | 0.033446 | −1.067723 | 0.848060 | 0.149730 |

(0.10812) | (1.91313) | (1.41158) | (0.17314) | |

[0.30934] | [−0.55810] | [0.60079] | [0.86480] | |

C | 0.010782 | −0.046330 | −0.041878 | −0.012128 |

(0.01210) | (0.21402) | (0.15791) | (0.01937) | |

[0.89145] | [−0.21648] | [−0.26520] | [−0.62618] | |

R-squared | 0.866813 | 0.862843 | 0.521001 | 0.639135 |

Adj. R-squared | 0.801525 | 0.795609 | 0.286197 | 0.462240 |

Sum sq. resids | 0.474880 | 148.6791 | 80.94164 | 1.217725 |

S.E. equation | 0.096495 | 1.707418 | 1.259798 | 0.154522 |

F-statistic | 13.27676 | 12.83342 | 2.218881 | 3.613084 |

Log likelihood | 86.64889 | −134.5907 | −111.1803 | 50.39434 |

Akaike AIC | −1.575296 | 4.171186 | 3.563125 | −0.633619 |

Schwarz SC | −0.783881 | 4.962601 | 4.354540 | 0.157796 |

Mean dependent | −0.009274 | −0.061039 | −0.037662 | −0.039247 |

S.D. dependent | 0.216598 | 3.776665 | 1.491117 | 0.210715 |

Determinant resid covariance (dof adj.) | 0.000940 | |||

Determinant resid covariance | 0.000181 | |||

Log likelihood | −105.2408 | |||

Akaike information criterion | 5.538721 | |||

Schwarz criterion | 8.826137 |