Poverty and Agriculture in Southern Africa Revisited: A Panel Causality Perspective

Abstract

Agricultural advancement is considered a panacea for poverty reduction, particularly, in developing countries. This study empirically investigates the dynamic linkage between agricultural value added (AVA) and poverty reduction for a panel of nine countries in Southern Africa using a second-generation panel approach for the period 1990 to 2015. Empirical results show that agricultural development is necessary but not a sufficient policy to combat poverty as it is only viable in the short run. Thus, we suggest long-run economic programs and/or strategies that will complement agricultural development toward poverty alleviation to spur economic growth in the sampled region.

Keywords

poverty reduction agricultural development panel cointegration panel Granger causality Southern Africa

Introduction

Over the years, agriculture has attracted many discourses, and there has been much research on the role it plays in enhancing poverty reduction (Anríquez & Stamoulis, 2007; Christiaensen, Demery, & Kuhl, 2006; Dandekar, 1986; De Janvry & Sadoulet, 2010; Machethe, 2004; Pauw, 2007; Peskett, Slater, Stevens, & Dufey, 2007; Schneider & Gugerty, 2011; Thompson, Rusastra, & Baldwin, 2008) as well as its multiplier effects on other nonagricultural sectors in both developed and developing economies. This study revisits the question raised by Machethe (2004) on the role of agriculture in economic growth to ascertain whether agricultural improvements can lead to poverty reduction in the Southern Africa. The empirical findings from this study have policy implications for this region and Africa countries at large, though the study investigates the dynamic causality relationship between agriculture and poverty reduction to be precise.

Other scholars have also examined the relationship between poverty and agriculture for different countries and regions, such as Zimbabwe (Mbiba, 1995), Malawi (Orr, 2000), Kenya (Kimenyi, 2002), Zambia (Siegel & Alwang, 2005), Southern Africa (Machethe, 2004; Munthali, 2007; Pauw, 2007), Africa as a whole (Fan, Johnson, Saurkar, & Makombe, 2008), and other selected African and non-African countries (Ellis & Freeman, 2004; Thompson et al., 2008), using qualitative, time series, and cross-country (panel) analyses. This study reconsiders the dynamic causality relationship between agriculture and poverty using a second-generation panel for countries in Southern Africa. A gap arises from previous studies created by the failure to unravel the possible existence of short-run or long-run agricultural development influences on poverty reduction. It is unproductive to investigate the growth impact of agriculture and how its development will influence poverty reduction without putting the direction and duration of causality in perspective. On that premise, this study seeks to address this gap in literature for countries in Southern Africa.

It is imperative to provide a brief background of the existing relationships between agricultural development and poverty reduction. By all indications, poverty is prevalent across the African continent. This is evident in the absence of fundamental privileges for human growth and development, which serve to promote creative, decent, and healthy living. According to reports presented by the United Nations Environment Programme (UNEP), about 48.5% of the inhabitant in Africa is living on less than US$1.25 per day, whereas 69.9% on less than US$2 per day, and there is a projected increase of 9% at the end of 2015. This figure conflicts with the Millennium Development Goals (MDGs), which set to halve poverty rates by the end of the same period. However, the expected rise in poverty rates in most of sub-Saharan Africa, especially, for those in the Southern-most region, is widely attributed to a slow down of growth and/or economic underperformance in the region. These economic scenarios can be divided into internal and external factors. According to the International Council on Social Welfare (2017), some of the internal factors responsible for poverty in the region are high morbidity, HIV/AIDS, gender inequality, and marginalization. The external factor includes economic competition, which originated from trade liberation occasioned by globalization. In other words, the rise in agricultural subsidies in Western countries tends to have attendant effects in this region, thereby jeopardizing agricultural economic activities (Munthali, 2007).

In addition to being an important factor in infant development, the contributions agriculture offers to the growth of other sectors of the economy are enormous (Johnston & Mellor, 1961; Schultz, 1964). This implies that policy reform in agriculture and investment will speedily enhance overall economic growth. The reality of the enormous multiplier effects of agriculture on other sectors in sub-Saharan Africa is clearly documented by various scholars (see Haggblade, Hazell, & Reardon, 2007).

Looking back at developmental programs across Africa, the advocacy of the MDGs shifted attention from promoting economic growth to fostering poverty reduction, knowing fully well that, poverty reduction does not solely rely on overall economic growth but rather on the ability of the poor to take part in the growth process by renewing their interests in participating in agriculture and, by extension, the growth and development process. It is well known that a large proportion of the poor population rely on agriculture as their means of livelihood; therefore, compared with the rich, the benefits accrued from economic growth emanating from agriculture are more poor centric than when growth occurs in nonagricultural sectors and/or subsectors. Therefore, maximizing the benefits of poverty reduction means ensuring investments in forms of physical and human capital, leading to sound economic policies to promote the desired development in the agriculture sector.

The relevance of the agricultural sector in Southern Africa cannot be overemphasized. Between 4% and 27% of real gross domestic product (GDP), that is, the economic growth of member states, and roughly 13% of gross export earnings are based in agriculture. In addition, more than 70% of Southern Africa relies on agriculture for food, employment opportunities, and income generation. Thus, in terms of value added, the performance of the agricultural sector has a strong influence on economic growth as well as on the political and social stability in the region. Agricultural techniques in this region is distinctly labor intensive. Therefore, the labor force is greatly influenced by the internal factors highlighted in this discourse. Therein lies the need for the governments of countries in this region to make it a priority to enhance the efficiency of the agricultural sector by paying particular attention to the needs of small-scale farmers, such as providing access to basic factors of production such as land, labor, capital, and entrepreneurship to enhance agricultural output. In so doing, economic strategic planning would generate numerous benefits, including a reduction in poverty rates, the creation of employment opportunities, and improved standards of living, while also empowering the labor force to engage in productive sectors of the economy.

Recently, studies on the effect of agriculture on poverty reduction propose that poverty and agriculture are mutually decided, but the direction of their causality cannot be determined in advance. In his analysis, Machethe (2004) argued that the direction of causality between poverty and agriculture has critical policy implications for poverty eradication and reduction. Meanwhile, the existence of a bidirectional causal nexus between poverty and agricultural value added (AVA) implies a significant dependency on the agriculture sector for poverty reduction. A failure in the agricultural sector will negatively fuel poverty and throw the economy into an array of economic turbulence; therefore, timely efforts to put sound agricultural policies in place would positively induce poverty reduction. However, a unidirectional nexus running from poverty to agriculture or from agriculture to poverty implies less dependency on the agricultural sector. In this regard, poverty reduction policies are said to have little or no impact on AVA. As such, it is probable that no directional causality relationships exist.

In the pool of literature on econometrics, this scenario is referred to as the “neutrality hypothesis,” which signifies that improvement and development of the agricultural sector through pursuance of sound economic growth policies would enhance small-scale farming and create employment opportunities, whereas other agricultural multiplier impacts will have no significant effect on poverty reduction. By implication, the appropriate stakeholders, that is, policy makers that design, organize, and implement small and medium enterprises programs, antipoverty policies, and microfinance institutions of such economies, must narrow their focus to policy measures besides agriculture to reduce and eventually eradicate poverty in the region.

Therefore, Machethe’s (2004) assertion that agricultural development stimulates, slows down, or neutralizes poverty reduction forms the crux of the motivation for this study, which is to investigate the direction of the causality relationship between poverty reduction and agriculture development in Southern Africa. The existing studies show that the relationship between poverty and agriculture differs because the various countries in this region have peculiar characteristics. In addition, the difference is caused by the variation in proxy variables employed and econometric techniques used in other scholars’ empirical estimations. This has resulted in contradicting results that are not robust, sound, or reliable for policy recommendations that cut across the countries in Southern Africa and the world at large. Therefore, depending on the direction of the causality nexus between poverty and AVA, policy implications and recommendation can be inferred with the goals of increasing agricultural productivity, enhancing household earnings, generating employment, and increasing the participation of the poor in the agricultural sector to boost economic performance.

Based on the qualitative study conducted by Machethe (2004), this study seeks to evaluate through a holistic approach and empirically investigate the direction of the causality relationship between poverty and agriculture for countries in Southern Africa. The contribution of this study to existing literature is 2-fold. First, it evaluates the dynamic causal nexus between poverty and agriculture in Southern Africa, taking into consideration the problem of cross-sectional dependency in panel data models. The second-generation panel data method was employed to evaluate the cross-sectional dependency in the panel variables, as advanced by Pesaran (2004), and the panel unit root tests (PURTs), as proposed by Maddala and Wu (1999) and Pesaran (2007), to affirm the nonstationarity of the variables of interest. This was followed by the bootstrap panel cointegration test (PCT) introduced by Westerlund and Edgerton (2007) to confirm the likelihood of a long-run equilibrium relationship of the model. To examine the dynamic causality relationship, the study employed Granger noncausality in a heterogeneous panel-based test suggested by Dumitrescu and Hurlin (2012), whose panel-based Granger causality test is new, more reliable, and appropriate for examining dynamic causality relationships for panel data when compared with an asymptotic approach. The second contribution of this study is that, to the best knowledge of the authors, no previous research has been conducted on the topic using the Granger causality perspective of the variables being examined. This article appears to be the first to evaluate the poverty–agriculture dynamic causality relationship for countries in Southern Africa using a second-generation panel-based methodology, which accounts for cross-sectional dependency. Thus, the novelty of this article rests in the application of a more robust econometric methodology and the empirical results of the existence of a short-run bidirectional causality relationship between poverty reduction and agriculture development for the sampled regions. The “Poverty in Southern Africa: Causes and Effects” section of this study provides the conceptual framework, the “Data and Methodology” section gives a detailed discussion of the data and methodology employed for the study, the “Results and Empirical Discussions” section discusses the empirical results and findings, and the concluding remarks are given in the “Conclusion” section.

Poverty in Southern Africa: Causes and Effects

According to World Bank (2001) report, about 47% of the world population is relatively poor, earning and living on less than US$2 per day, whereas about 20% are extremely poor, earning and living on less than US$1 per day. Poverty remains a vast unresolved economic, social, and political problem in the world. Therefore, complete eradication of this economic bottleneck requires a gradual planning process, and it cannot be resolved overnight.

Poverty rates are generally high in Africa, and the reasons for and extent of poverty vary across countries. According to The World Factbook, about 41% of the population live below the poverty line in Angola, 35.9% in South Africa, 72.3% in Zimbabwe, 57.1% in Lesotho, 60.5% in Zambia, 52% in Mozambique, 52.4% in Malawi, 30.3% in Botswana, and 28.7% in Namibia. This trend is expected to continue if nothing is done to reduce poverty to the bare minimum.

There are several factors that contribute to the high poverty levels in this region, such as poor attitudes toward education, high unemployment rates, inadequate access to productive land, inadequate capital or financial services, neglected female participation in the labor force, poor health facilities, urbanization, gender differentials, and insecurity resulting from civil unrest and armed conflicts across the continent. According to Venkatasubramanian (2001), there is a high rate of school dropouts caused by families’ inabilities to meet the financial demands of education. The Universal Basic Education (UBE) system is in place in Botswana, Malawi, Tanzania, and Uganda, but these countries experienced an approximately 12% drop in education enrollment between 1990 and 2002. Another important report reveals a high unemployment rate of about 20% in Zimbabwe and Botswana, about 30% in South Africa, and 40.5% in Lesotho, whereas in Namibia, the unemployment rate rose from 19% in 1991 to 34.8% at the end of 1997.

The HIV/AIDS epidemic also poses a global threat that stifles socioeconomic development throughout Africa. According to the World Bank, there is a complex relationship between gender inequality, poverty, and HIV/AIDS pandemic. Sub-Saharan Africa has about 30 million people living with HIV in 2002. Some countries in the studied region recorded a rise in HIV rates among the adult population. This is evidenced by statistics; Lesotho records 31% of its population infected with HIV/AIDS, Swaziland showing 34%, Zimbabwe showing 38%, and in Botswana, the HIV/AIDS rate is reported to be 39%. According to United Nations AIDS Program, South Africa, Zambia, and Ethiopia recorded 51%, 69%, and 75% HIV infection rates, respectively. In addition, data show that 30% of HIV victims in Africa live in the southern region. The prevalence of HIV/AIDS is linked to poverty, which is believed to be the root cause of the skyrocketing rates of the scourge. Because the region is heavily labor force oriented, sound health remains the required status for well-meaning jobs; therefore, poor health keeps the young and agile from engaging in productive activities. This makes people vulnerable to exploitation of labor and environmental resources, the consequence of which is environmental tension.

In addition to HIV/AIDS, political unrest contributes significantly to poverty rates in Africa. Since 1970, the continent has experienced 30 different wars. The longest wars were in Mozambique and Angola, both of which were characterized by political strife and armed conflict and endangered nearly one-fifth of Africa’s citizenry. These conflicts have been crucial underlying determinants that have hugely contributed to hunger and poverty within the region. As the continent with the largest mining sector in the world (in South Africa, Swaziland, Angola, Mozambique, Malawi, Namibia, Zambia, and Zimbabwe), such violence fatally affects the mining activities in this region. Casualties reportedly occurring in locations in and around the mines climb to 26,000 annually.

The increasing rural–urban migration in pursuit of white collar jobs or greener pastures also has an adverse effect on the poverty rates. The United Nations Development Programme (UNDP; 2002) reported that in 1975, Malawi, Botswana, Zimbabwe, and South Africa, the urbanization figure measured as a percentage of total population was as low as 7.7%, 12.8%, 19.6%, and 48%, respectively. By 2000, the statistics rose to 14.7%, 49%, 35.3%, and 56.9%. This mass movement has exposed urban dwellers to impoverished living conditions that are best described as random settlements. These random settlements have insufficient facilities and subject residents to environmental pollution and/or industrial hazards. According to the South Africa Statistics, it was reported that, due to the stress associated with living in urban centers, issues such as divorce become eminent and are expected to continue to rise. This breakup of families presents a specific challenge for women that could facilitates impoverishment, relative to men, women do benefits from several resources via marriage.

Theoretical Backing

The physiocracy school of thoughts dated to the 18 century underlines the ideology that agricultural development is key to nation’s economic fortune. They claim that the path to long and sustainable economic growth rests on land agricultural development (Sertoglu, Ugural, & Bekun, 2017). Nobel Laureate Gunnar Myrdal in economics further strengthens this position. He asserted that the nations agricultural section hold the potential and is key driver for long-run economic growth. However, the path to how this believe translate into fruition is been a matter of debate by most developing, developed and emerging government administrators, economist, and development specialist. However, different government administrators such as increased capital and labor stock accumulation (Neoclassical growth model) and increased agricultural farm inputs have explored diverse routes. This is pertinent given the quest by most economies and regions strive to achieve the most minimal poverty rate by using several means of growth and development. For this reason, agricultural advancement is considered a panacea to poverty reduction, particularly, in developing countries.

Data and Methodology

To explore the interaction between poverty and AVA, the study sampled nine Southern African countries, South Africa, Botswana, Malawi, Lesotho, Namibia, Mozambique, Swaziland, Zambia, and Zimbabwe, for the period of 1990 to 2015 based on data availability. Table 1 shows the descriptive statistics for the sampled countries, while the variables under investigation are discussed below:

Table 1.

Descriptive Statistics.

Countries	Variables	M	Mdn	SD	Minimum	Maximum
Botswana	AVA	3.302	2.952	0.963	2.032	5.204
	Poverty	0.615	0.591	0.051	0.558	0.689
Malawi	AVA	35.441	34.549	5.597	25.076	48.904
	Poverty	0.394	0.387	0.044	0.325	0.476
Lesotho	AVA	2159.431	2176.369	157.236	1691.103	2,398.687
	Poverty	0.467	0.468	0.022	0.437	0.497
Namibia	AVA	9.522	9.680	1.417	6.686	11.823
	Poverty	0.587	0.583	0.026	0.554	0.640
Mozambique	AVA	29.178	28.104	4.388	21.386	37.118
	Poverty	0.318	0.325	0.073	0.208	0.418
Swaziland	AVA	9.380	8.830	2.495	6.122	14.223
	Poverty	0.525	0.529	0.019	0.492	0.551
South Africa	AVA	3.354	3.317	0.741	2.322	4.630
	Poverty	0.634	0.636	0.017	0.609	0.666
Zambia	AVA	15.505	15.959	5.787	5.250	34.101
	Poverty	0.468	0.449	0.066	0.396	0.579
Zimbabwe	AVA	16.580	16.534	3.456	7.413	21.788
	Poverty	0.453	0.451	0.034	0.406	0.516

Source. Author’s computation.

Note. AVA = agricultural value added.

AVA: This comprised values added for all outputs from agricultural subsectors such as crop production, husbandry and fishery, forestry, and livestock. The data were retrieved from the World Bank database (online).

Poverty: This study employed the Human Development Index (HDI) as a proxy for poverty. The rationale behind the choice of the composite index variable is seen in the rich composition of the index as it reflects the living standards of individuals and economies (Anand & Sen, 1997; Kelley, 1991; Seth & Villar, 2014). The HDI data were obtained from the Human Development Report (online).

The current empirical study follows four paths. The first path is the test of cross-sectional dependency proposed by Pesaran (2004). This test is necessary to avoid a spurious regression trap and subsequently invalid policy implication(s). The econometric model specification is bivariate. Poverty is modeled as a function of AVA. Second, the PURT is used to check for stability and asymptotic features of the data as put forward by Pesaran (2007). Third, for a long-run equilibrium relationship, this study employs the bootstrap panel cointegration relationship test suggested by Westerlund and Edgerton (2007). Finally, Dumitrescu and Hurlin’s (2012) Granger causality test is employed to ascertain the direction of causality among the variables of interest. The presence of cross-sectional dependency among the paneled countries led to the adoption of second-generation panel econometrics techniques.

Cross-Sectional Dependence Test

Macro panel data are usually plagued by the presence of cross-section dependency (CSD). If a series of panels has CSD, it implies that there are common, unobserved factors that affect the rise of the countries’ variables over their individual time paths (Breusch & Pagan, 1980; Pesaran, 2004). This necessitates testing for CSD to avoid a spurious regression fit. A popular approach used are the Lagrange Multiplier (LM) test developed by Breusch and Pagan (1980), which is applicable when the panel time dimension (T) is greater than the cross-sectional dimension (N) (T > N). Another famous CSD test is the Pesaran (2004) test, applied in cases when T > N or N > T. The Pesaran (2004) test becomes biased when the panel average is zero, and the individual average is statistically different from zero. The LM statistics are given as follows:

C D L M = T Σ \sum_{i = 1}^{N - 1} \sum_{j = i + 1}^{N} ρ^{2} i j : χ^{2} \frac{N (N - 1)}{2} .

(1)

Here, the LM statistics are given as $\sum_{i = 1}^{N} u^{2}$ (sum of squared residuals from the correlation coefficient). The statistics are asymptotically distributed by: chi-square and the parameter is given as $N (N - 1) / 2$ , where N is the number of observations and $ε$ depicts residuals. Furthermore, is estimated as follows:

\hat{ρ} i j = \hat{ρ} j i = \frac{\sum_{t}^{T} = 1 {\hat{Σ}}_{i t} {\hat{Σ}}_{j t}}{{(\sum_{t}^{T} = 1 {\hat{Σ}}_{i t}^{2})}^{\frac{1}{2}} {(\sum_{t}^{T} = 1 {\hat{Σ}}_{j t}^{2})}^{\frac{1}{2}}} .

(2)

Table 2 strongly supports the presence of cross-sectional dependency, which denotes the common unobserved effect for the variables under review. The presence of CSD reveals interdependency within the cross-sectional unit. Eberhardt and Teal (2011) argued that CSD depicts unobserved common shock. Therefore, two distinct kinds of dependency, long-range and spatial dependency, are seen across cross sections, as widely argued in the literature (Anselin, 2001; Moscone & Tosetti, 2010). Spatial dependency takes into account the distance between cross-sectional units, whereas long-range dependency exhibits the same behavior in term of shocks. The assumption of no serial correlation is still valid even in the presence of interdependence.

Table 2.

Pesaran’s (2004) Cross-Sectional Dependency Test.

Variable	CSD test	p value	Corr.	Abs(corr.)
Poverty	15.51*	.000	.507	.546
AVA	12.84*	.000	.420	.436

Note. Null states cross-sectional independence CSD ~ N (0, 1). Corr. = correlation; AVA = agricultural value added; CSD = Cross-sectional Dependence.

Asterisks means statistical rejection of the null hypothesis at 1% significance level.

PURT

The PURT takes into account both the time series dimension and its cross-sectional dimension. The panel unit root is also statistically more powerful than the time series unit root, where only the time dimension is considered (Baltagi, 2008). This is supported by the variability of increased data of both cross-sectional and time series dimensional analysis.

However, panel data unit roots are usually plagued with CSD problems. Thus, panel unit roots are grouped into two categories, the first-generation panel and the second-generation panel tests. The first-generation tests are further divided into two subgroups: (a) homogeneous and (b) heterogeneous unit roots (Im, Pesaran, & Shin, 2003; Levin, Lin, & Chu, 2002; Maddala & Wu, 1999). The first-generation panel model does not take into account CSD as it assumes cross-sectional independency. However, the second-generation panel estimators account for CSD. Thus, the second-generation panel unit root estimator gives more robust, efficient, and consistent panel estimation results. The CIPS test proposed by Pesaran (2007) is asymptotically robust in heterogeneous settings with the null hypothesis of nonstationarity.

PCT

The current study also examines the existence of a long-run equilibrium relationship (cointegration) among the integrated series that has dual dimensions of time (T) and cross section (N). Cointegration analysis has gained huge traffic in empirical literature because most economic/financial theories/hypotheses are long-run induced. Nevertheless, most studies that investigate cointegration models fail to reject the null hypothesis of no cointegration against the obvious theoretical postulate of cointegration. The rationale underpinning the above is that the time series and panel data procedure requires that the estimated parameters be the same, both at their levels and difference forms. This flaw was corrected by Westerlund and Edgerton (2007), who proposed four cointegration estimators to fix the above observed deficiencies to give more reliable and robust estimates. The merits of the fairly new developed Westerlund cointegration estimators over others are based on structural dynamics; the test is not based on residuals like previously known cointegration tests, meaning there is no form of factor restriction. The rationale behind this method is to analyze the null of no cointegration and the assumption that the Error Correction Term (ECT) in a conditional panel is equal to zero. The first two tests (Ga and Gt) were developed to test the alternative hypothesis when the whole panel is cointegrated, whereas the next two (Pa and Pt) test the alternative hypothesis that at least one unit out of the panel is cointegrated. The cointegration model specification is given as

\begin{matrix} Δ y_{i t} = C_{i} + β_{o i} (y_{i, t - 1} - b_{i x i t - 1}) + \sum_{j = 1}^{p_{i 1}} β_{11 j} Δ y_{i, t - j} \\ + \sum_{j = - p_{2 i}}^{p_{3 i}} β_{2 i j} Δ x_{i t - j} + ε_{i t} . \end{matrix}

(3)

Here, the speed of adjustment to the equilibrium path is given as $β_{o i}$ . It is worth mentioning that the penultimate term comprises lags of ∆x, as well as leads, otherwise, we assume exogeneity.

Panel Granger Causality Test (PGCT)

This study uses the heterogeneous noncausality test developed by Dumitrescu and Hurlin (2012) to investigate causality among countries’ panels for the variables under observation. The Dumitrescu and Hurlin (DH) test is used in situations where the cross-sectional dimension thrives, while the time dimension is fixed. However, the test is also applicable when T > N. The structure of the test is built on the vector autoregressive (VAR) model framework with the assumption of cross-sectional dependency. The distinction of the DH test according to the Monte Carlo simulation is that even in the presence of CSD, the test still generates strong and robust estimates. The test is also applicable to both heterogeneous and balanced panels. Furthermore, the DH test displays two distinct distribution features, asymptotic and semiasymptotic. When T > N, the asymptotic distribution is most efficient and robust, while when N > T, the semiasymptotic distribution becomes handy. Therefore, with occurrences of CSD, both simulated and estimated critical values obtained via duplication are utilized.

The linear version of the panel model is as follows:

y_{i t} = \sum_{m = 1}^{m} φ_{i}^{(m)} y_{i, t - m} + \sum_{m = 1}^{m} β_{i}^{(m)} x_{i, t - m} + μ_{i, t} .

(4)

Here, m represents lag length, $φ_{i}^{(m)}$ is the autoregressive parameter, and $β_{i}^{(m)}$ denotes the regression coefficient, which varies within groups. Significantly, the DH test does not use a random process, but rather it is a fixed type of test that stimulates a fixed coefficient model, while the other individual cross-sectional units are independent. The DH causality test is asymptotically normally distributed and accommodates heterogeneity. Thus, the DH test assumes a homogeneous nonstationary hypothesis for causality estimations with heterogeneous models. The null and alternative hypotheses for the homogeneous nonstationary hypothesis are given as follows:

\begin{array}{l} H_{0} : δ_{i} = 0, \forall_{i} = 1, \dots, K \\ H_{1} : δ_{i} = 0, \forall_{i} = 1, \dots, K_{1} \\ δ_{i} \neq 0, \forall_{i} = K_{1} + 1, K_{1} + 2, \dots, K . \end{array}

Here, $K_{1}$ represents the unknown parameters that fulfill the condition $0 \leq K_{1} / K ρ 1$ . However, the ratio between $K_{1} / K$ is expected to be inevitably inferior to one, given that $K_{1} = K$ . This implies no causality relationship for any of the individual cross-sectional panel unit roots. Thus, we fail to reject the null hypothesis of the homogeneous nonstationary hypothesis. However, the converse of $K_{1} = 0$ signifies a causality relationship in all of the individual panel units. This study’s econometrics dynamic causality model specification is given as follows:

\begin{matrix} Δ P o v e r t y_{i t} = ϕ_{1 j} + λ_{1 t} ε_{i t - 1} + \sum_{k} ϕ_{11, i k} Δ P o v e r t y_{i t - k} \\ + \sum ϕ_{12, i k} Δ A V D_{i t - k} + ε_{1, i t} . \end{matrix}

(5)

\begin{matrix} Δ A V D_{i t} = ϕ_{2 j} + λ_{2 t} ε_{i t - 1} + \sum_{k} ϕ_{22, i k} Δ A V D_{i t - k} \\ + \sum ϕ_{21, i k} Δ P o v e r t y_{i t - k} + ε_{2, i t} . \end{matrix}

(6)

Here, K depicts the optimum lag length as automatically chosen by Akaike information criterion (AIC).

Results and Empirical Discussions

In this section, the study reports the results obtained from the panel estimations. Table 3 presents the panel unit root results as proposed by Pesaran (2007). For the unit root test, the null hypothesis of unit root could not be rejected at level for the variables, thus, the variables are integrated of first order, that is, I(1). The test statistics are lower than the critical values at all significance levels. Therefore, the null hypothesis that each variable is stationary at zero order was rejected, meaning the variables are nonstationary at all levels. It is crucial to pay attention to the stationarity properties of variables to avoid the likelihood of making a spurious inference.

Table 3.

Pesaran (2004) Unit Root Tests.

Variable	CIPS statistics	Critical value (%)
Variable	CIPS statistics	1	5	10
AVA	–2.009	–2.51	–2.25	–2.12
Poverty	0.004	–2.44	2.22	–2.1

Note. Variables are nonstationary at all levels. AVA = agricultural value added; CIPS = Cross-sectional Dependence Im Pesaran Shin.

Furthermore, we estimated using Pedroni’s (1999) first-generation PCT, which is usually applied to examine cointegration relationships among variables. This test is specified under the null hypothesis of no cointegration. It accounts for heterogeneity and independence within the cross sections. However, the existence of cross-sectional dependency in our macro panel data indicates that the Pedroni test is not suitable for evaluating cointegration relationships. Eberhardt and Presbitero (2013) argued that not considering the existence of cross-sectional dependency arouses grievous and vague estimate identification problems. Kao’s suggested test was also estimated to confirm the Pedroni cointegration test.¹ It implied no cointegration relationship, confirming the Pedroni test under the assumption of coefficient homogeneity. Neither the Pedroni nor the Kao cointegration tests could reject the null hypothesis of no cointegration relationships among the panel data.

To confirm our empirical results based on the presence of cross-sectional dependency, we proceeded with the second-generation panel bootstrapping cointegration test, as advanced by Westerlund and Edgerton (2007). This test deals with dynamic structures and not residuals. Having confirmed the existence of cross-sectional dependency among the panel countries, the Westerlund and Edgerton (2007) cointegration test became a suitable approach to evaluate the cointegration nexus among the panel countries. Table 4 reports the Westerlund and Edgerton (2007) cointegration results with a bootstrap method that produced sound coefficients, standard errors, a confidence interval, and robust critical values. This study ran 5,000 estimation repetitions for accuracy purposes and to uphold resampling to achieve robust results. From the results reported in Table 4, we could not find evidence in support of a cointegration relationship; thus, we conclude that the variables have no long-run equilibrium cointegration relationships.

Table 4.

Westerlund and Edgerton (2007) Bootstrapping Cointegration Test.

Statistic	Value	Z value	p value	Robust p value
Gt	–1.707	2.433	.993	.750
Ga	–7.461	2.007	.978	.640
Pt	–4.616	1.968	.976	.650
Pa	–5.975	1.447	.926	.580

Note. Null of no cointegration with significance level of 1, 5, and 10 respectively. The simulation was carried out via bootstrapping regression with 100 repetitions. Gt and Ga test the cointegration for each country individually, and Pa and Pt test the cointegration of the panel as a whole.

Table 4 shows no cointegration relationships under the assumptions of whole and within the individual cross sections. To achieve our study objectives, the dynamic causal relationship among the variables of interest was examined.

The DH causality test was estimated to examine the directions of dynamic causality relationships among the variables. This test produces more robust, stable, and reliable results, while it also accounts for cross-sectional dependency among the panel variables. Table 5 reports the results obtained from the Granger causality test. Empirical results show significant Granger causality running from AVA to poverty and from poverty to AVA at a significant 1% level. This signifies a bidirectional causality running from poverty to AVA for Southern African countries. The bidirectional causality is indicative; if agricultural development is enhanced, its multiplier effect would influence poverty reduction. Simultaneously, policies that promote poverty reduction would also influence agricultural development within the region.

Table 5.

Dumitrescu and Hurlin (2012) Panel Granger Causality Test.

Null hypothesis	W statistics	Zbar statistics	p value	Causality
AVA ≠> HDI	4.991***	3.286	.001	Yes
HDI ≠> AVA	5.752***	4.194	.003	Yes

Note. The symbol ≠> signifies null hypothesis that, the variables do not Granger cause one another. However, bidirectional causality is found at 0.01 percent significant level. AVA = agricultural value added; HDI = Human Development Index.

Asterisks denote statistical rejection of the null hypothesis at 1% significance level.

However, this causality relationship only seems to be effective in the short run, as we could not, through the empirical results, provide evidence in support of a long-run equilibrium cointegration relationship between agricultural development and poverty reduction. This confirms the findings of Christiaensen, Demery, and Kuhl (2006, 2011), where the relationship between agriculture and poverty reduction was clarified as a Granger causality relationship.

The empirical results for Southern African countries propose a dynamic causality relationship running from AVA to poverty and from poverty to AVA. This indicates that Southern African countries largely depend on the agricultural sector in their quest for poverty reduction. However, based on the empirical findings, improving agriculture as a measure to reduce and fight poverty levels can stimulate the economy only in the short run. Reduction of chronic poverty in this region requires sound measures that would have a long-term impact on poverty reduction as pursuance of agricultural development is not a panacea to poverty reduction in the long term. This is a wakeup call for governments, policy makers, and private individuals who design, structure, and implement antipoverty programs that agricultural development is necessary, but alone, it is not a sufficient approach to fighting poverty in the region. If any meaningful antipoverty program is to be achieved, policy makers must advocate for short-term and long-term policies and programs for the sampled region. Long-term programs outside of agricultural development that would sustain, spur economic growth, and, by extension, alleviate poverty to improve the welfare of the citizenry should be implemented.

Conclusion

This article utilizes second-generation panel estimation techniques to consider the dynamic causality relationship between poverty and AVA. The study used as a proxy human development index for poverty and AVA for agricultural development for a panel of nine Southern African countries for the period from 1990 to 2015. Our current study could not capture all the countries in the region based on data availability. For sound and robust empirical results and analysis, the panel cross-sectional dependency test proposed by Pesaran (2004) was utilized to investigate the presence of common, unobserved shock, which is often found among panel data. Afterwards, the Pesaran (2007) for PURT that accounts for cross-sectional dependency was examined using the stationarity and stability of the variables. Based on the results, a cointegration relationship using the Westerlund and Edgerton (2007) test was also considered. From the empirical results, we found no evidence in support of a long-run equilibrium cointegration relationship among the variables. Subsequently, the dynamic causality test proposed by Dumitrescu and Hurlin (2012) was evaluated, where the empirical results revealed bidirectional Granger causality running from agriculture to poverty.

Our empirical result is indicative for countries in Southern Africa. Previous empirical studies found a linkage between agricultural development and poverty reduction but failed to explicitly spell out the direction of causality and the time span of this dynamic relationship, that is, whether this relationship only exists in the short run, long run, or both (Anríquez & Stamoulis, 2007; Christiaensen, Demery, & Kuhl, 2011; Machethe, 2004; Pauw, 2007; Peskett et al., 2007). The current study provides two distinct contributions. First, the empirical results from the cointegration relationship observed indicate the absence of a long-run equilibrium relationship between agriculture and poverty for the sampled region. We infer that the relationship between the variables exists, but only in the short run. This is an indication of no co-movement in the long run between agriculture and poverty for Southern African countries. Second, the bidirectional Granger causality running between the variables conforms to the findings reported in the existing literature that agricultural development enhances poverty reduction and, as such, improves living standards (Christiaensen et al., 2011, 2006; Machethe, 2004).

Conclusively, our current study examines the dynamic causality relationship between poverty and agriculture in Southern African. We are of the opinion that, the governments and policy makers in this region should look out of the box of agriculture and open their arms to other problems solving economic policies if any meaningful antipoverty program is to be achieved. Policy makers must advocate for economic policies that would not only be effective in the short run but would have sustainable long-term impact in fighting poverty in the region as a whole. The result of this study have shown that agriculture is a necessary but not a sufficient solution for poverty alleviation in the sampled region. Having this significant result for Southern Africa, we suggest that future research in this field should examine the relationship between agriculture and poverty alleviation, most especially, Africa to substantiate the current position of this study for Africa using the second-generational panel.

Footnotes

Appendix

Pedroni cointegration test results
	Statistics	p	W statistics	p
Alternative hypothesis: Common AR coefficients (within-dimension)
Panel v-statistic	–1.251	.894	–0.983	.8374
Panel rho-statistic	–0.280	.389	–0.214	.4150
Panel PP-statistic	–0.541	.294	–0.446	.3276
Panel ADF-statistic	0.788	.784	0.702	.7587
Alternative hypothesis: Individual AR coefficients (between-dimension)
Group rho-statistic	1.0170	.8454
Group PP-statistic	0.8144	.7923
Group ADF-statistic	0.1302	.5518
Kao cointegration test
	t stat	p
ADF	–1.1312	.1290
Residual variance	7.13E–05
HAC variance	0.000209

Note. No cointegration vectors at all significance levels.

HAC = heteroskedasticity and autocorrelation consistent; ADF = Augmented Dickey Fuller; PP = Philip Perron; AR = Autoregressive.

Acknowledgements

The both authors of this article are immensely grateful to Prof. Dr. Mehmet Balcilar for his tutelage, mentorship, and continuous commitment to scholarly advancement in their careers.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

ORCID iD

Seyi Saint Akadiri

Author Biographies

Festus Victor Bekun is currently a PhD candidate at the Department of Economics at the Eastern Mediterranean University Northern Cyprus. He also obtained his MSc in economics from the same institution. He is also a research assistant at the same department. He teaches basic economic principles of micro and macroeconomics and inferential statistics. His research interest is in the areas of agricultural economics, energy economics, and tourism economics.

Seyi Saint Akadiri holds a PhD in economics from Eastern Mediterranean University. He is interested in research areas such as; growth, energy economics, environmental economics, tourism economics, food policy and forecasting.

References

Anand

Sen

(1997). Concepts of human development and poverty: A multidimensional perspective. In Poverty and Human Development Papers (pp. 1-20). New York, NY: United Nations Development Programme.

Anríquez

Stamoulis

(2007). Rural development and poverty reduction: Is agriculture still the key. Electronic Journal of Agricultural and Development Economics, 4(1), 5-46.

Anselin

(2001). Spatial econometrics. In Baltagi

(Eds.), A companion to theoretical econometrics (pp. 310-330). Oxford: Blackwell.

Baltagi

(2008). Econometric analysis of panel data. Chichester, UK: John Wiley.

Breusch

T. S.

Pagan

A. R.

(1980). The Lagrange multiplier test and its applications to model specification in econometrics. The Review of Economic Studies, 47, 239-253.

Christiaensen

Demery

Kuhl

(2011). The (evolving) role of agriculture in poverty reduction. An empirical perspective. Journal of Development Economics, 96, 239-254.

Christiaensen

L. J.

Demery

Kuhl

(2006). The role of agriculture in poverty reduction: An empirical perspective (Vol. 4013). Africa Region: The World Bank.

Dandekar

V. M.

(1986). Agriculture, employment and poverty. Economic and Political Weekly, 21(38/39), A90-A100.

De Janvry

Sadoulet

. (2010). Agricultural growth and poverty reduction: Additional evidence. The World Bank Research Observer, 25(1), 1-20.

10.

Dumitrescu

E. I.

Hurlin

(2012). Testing for Granger non-causality in heterogeneous panels. Economic Modelling, 29, 1450-1460.

11.

Eberhardt

Presbitero

(2013). This time they are different: Heterogeneity and nonlinearity in the relationship between debt and growth (No. 13-248). Washington, DC: International Monetary Fund.

12.

Eberhardt

Teal

(2011). Econometrics for grumblers: A new look at the literature on cross-country growth empirics. Journal of Economic Surveys, 25, 109-155.

13.

Ellis

Freeman

H. A.

(2004). Rural livelihoods and poverty reduction strategies in four African countries. Journal of Development Studies, 40(4), 1-30.

14.

Fan

Johnson

Saurkar

Makombe

(2008). Investing in African agriculture to halve poverty by 2015. Washington, DC: International Food Policy Research Institute.

15.

Haggblade

Hazell

P. B.

Reardon

(Eds.). (2007). Transforming the rural nonfarm economy: Opportunities and threats in the developing world. Baltimore, MD: Intl Food Policy Research Institute.

16.

K. S.

Pesaran

M. H.

Shin

(2003). Testing for unit roots in heterogeneous panels. Journal of Econometrics, 115, 53-74.

17.

International Council on Social Welfare. (2017). East & Southern Africa Newsletter: December 2016–March 2017. Retrieved from https://www.icsw.org/

18.

Johnston

B. F.

Mellor

J. W.

(1961). The role of agriculture in economic development. The American Economic Review, 51, 566-593.

19.

Kelley

A. C.

(1991). The human development index: “Handle with Care.” Population and Development Review, 17, 315-324.

20.

Kimenyi

M. S.

(2002). Agriculture, economic growth and poverty reduction. Available from http://www.eldis.org/

21.

Levin

Lin

C. F.

Chu

C. S. J.

(2002). Unit root tests in panel data: Asymptotic and finite-sample properties. Journal of Econometrics, 108, 1-24.

22.

Machethe

C. L.

(2004, October 28-29). Agriculture and poverty in South Africa: Can agriculture reduce poverty. Paper presented at the Overcoming Underdevelopment Conference held in Pretoria, South Africa.

23.

Maddala

G. S.

(1999). A comparative study of unit root tests with panel data and a new simple test. Oxford Bulletin of Economics and Statistics, 61(S1), 631-652.

24.

Mbiba

(1995). Urban agriculture in Zimbabwe: Implications for urban management and poverty. Harare, Zimbabwe: Department of Rural and Urban Planning, University of Zimbabwe.

25.

Moscone

Tosetti

(2010). Testing for error cross section independence with an application to US health expenditure. Regional Science and Urban Economics, 40, 283-291.

26.

Munthali

S. M.

(2007, February). Transfrontier conservation areas: Integrating biodiversity and poverty alleviation in Southern Africa. Natural Resources Forum, 31(1), 51-60.

27.

Orr

(2000). “Green gold?” Burley tobacco, smallholder agriculture, and poverty alleviation in Malawi. World Development, 28, 347-363.

28.

Pauw

K. W.

(2007). Agriculture and poverty: Farming for food or farming for money? Agrekon, 46, 195-218.

29.

Pedroni

(1999). Critical values for cointegration tests in heterogeneous panels with multiple regressors. Oxford Bulletin of Economics and Statistics, 61(S1), 653-670.

30.

Pesaran

M. H.

(2004). General diagnostic tests for cross section dependence in panels (CESifo Working Paper Series No. 1229; IZA Discussion Paper No. 1240). University of Cambridge.

31.

Pesaran

M. H.

(2007). A simple panel unit root test in the presence of cross-section dependence. Journal of Applied Econometrics, 22, 265-312.

32.

Peskett

Slater

Stevens

Dufey

(2007). Biofuels, agriculture and poverty reduction. Natural Resource Perspectives, 107, 1-6.

33.

Schneider

Gugerty

M. K.

(2011). Agricultural productivity and poverty reduction: Linkages and pathways. Libraries Test Journal, 1, 56-74.

34.

Schultz

T. W.

(1964). Transforming traditional agriculture. New Haven, CT: Yale University Press.

35.

Sertoglu

Ugural

Bekun

F. V.

(2017). The contribution of agricultural sector on economic growth of Nigeria. International Journal of Economics and Financial Issues, 7, 547-552.

36.

Seth

Villar

(2014). Human development, inequality and poverty: Empirical findings (Working Paper Series No. 14.11). Retrieved from http://www.upo.es/serv/bib/wps/econ1411.pdf

37.

Siegel

P. B.

Alwang

(2005). Poverty reducing potential of smallholder agriculture in Zambia (Africa Region Working Paper Series No. 85). Washington, DC: The World Bank.

38.

United Nations Development Programme. (2002). Kenya human development report 2001: Addressing social and economic disparities for human development. Nairobi, Kenya: Author.

39.

Venkatasubramanian

(2001, December 04). Education and poverty. The Hindu. Retrieved from https://www.thehindu.com/thehindu/op/2001/12/04/stories/2001120400060100.htm

40.

Westerlund

Edgerton

D. L.

(2007). A panel bootstrap cointegration test. Economics Letters, 97, 185-190.

41.

The World Bank. (2001). Attacking poverty: World development report. Washington, DC: Author.

42.

Thompson

Rusastra

I. W.

Baldwin

(2008). Towards a joint regional agenda for the alleviation of poverty through agriculture and secondary crop development: Proceedings of the regional meeting, Bangkok, 21-22 November 2007. Monographs.