On the impact of provincial development policies in South Africa

ABSTRACT Although South Africa is one of the biggest economies in Africa, poverty and income inequality persist and a vast number of households lack access to water and sanitation services. Provincial governments have implemented the Provincial Development and Growth Strategy to improve standards. We evaluate its effects on selected development indicators. Using a generalisation of the Difference-in-Differences method, we study the effects on four development indicators: food security, economic well-being, and water and sanitation security. We use secondary data from the General Household Survey, collected between 2002 and 2017 and conclude that (i) the policies improve development indicators; (ii) the effects are heterogeneous across racial and geographic distributions of households. Long-term economic stimulators, such as employment opportunities and education for vulnerable communities, are needed to improve household welfare across the provinces. Programmes emanating from these development policies should be ongoing and continuously adapted to fulfil the specific needs of the local groups.


Introduction
Most developing countries are battling with poverty eradication and the attainment of economic growth, key points of the Sustainable Development Goals (SDG) adopted by all United Nations member states in 2015 (de Janvry & Sadoulet, 2017;Handa et al., 2018).The growing need for poverty alleviation strategies and provision of basic services has pushed governments to create, modify and implement policies to attain targeted goals.One of the largest economies in Africa, the Republic of South Africa (RSA) is a developing country with a three-tier government composed of the national, provincial, and local levels of government, which all have legislative and executive authority.The nine provinces in the country play a key role in carrying out national imperatives, given the different constraints, resource endowments, and population size.
While South Africa is relatively thriving in various development aspects such as a progressive constitution, expansion of industry and finance sectors, and food security at in policy implementation could lead to certain groups in society benefitting more, further highlighting the importance of considering inequalities in sustainable development (Cole et al., 2017a).These investigations of sub-national inequalities are important for policy planning and sustainable development.Although the PGDS is a wide policy framework, the programmes implemented in the provinces emanate from the PGDS.
All provinces have adopted the framework, with common targets of alleviating poverty, improving economic well-being, and basic needs provision.The provincial governments implement the PGDS considering their resource constraints and opportunities, as well as the population needs.As such, the PGDS impact is a reflection of whether the targets have been met and reflect the overall impact of localised programmes within the provinces.To this end, this study evaluates the impact of the PGDS on social welfare indicators, food security, economic well-being, water, and sanitation access in South Africa.To the knowledge of the authors, this is the first study to empirically evaluate the impact of PGDS on welfare indicators, thus the paper bears three aims: (1) To evaluate the overall impact of PGDS on welfare indicators; (2) To evaluate the potential heterogeneity of the impact of PGDS on welfare indicators across population groups and (3) To evaluate the potential heterogeneity of the impact of PGDS on welfare indicators across geographic areas.This is particularly important in the context of South Africa, where citizens belong to different racial groups and are settled in different geographical types across the provinces.In the next section, we review development policy evaluations studies, which gives rise to the adopted empirical methods presented in Section 3. Section 4 discusses the overall and heterogeneous impacts found across social groups, and Section 5 highlights the conclusions and policy recommendations informed by the results.

Review of development policy evaluation studies
Policy impact evaluation has grown popular across the globe in efforts of improving policy-making and implementation, as well as modify policy targets.It is generally described as the examination of the content and implementation of a policy.Additionally, policy evaluation scrutinises the changes in the well-being of individuals that are brought about by the implementation of a project, programme, or policy (Gertler et al., 2016).It should however be noted that, apart from measuring the cause-effect of policies to evaluate failure or success, evaluation is also performed in terms of costeffectiveness, viability, and acceptability of the stakeholders involved (Ravallion, 2008;Maloney & Gaurav, 2018).Although in many cases some policies exhibit positive results, some targets are not met and, in some cases, remain possible disincentives for the development of the beneficiaries (Azuara & Marinescu, 2013;Garganta & Gasparini, 2015;Gutiérrez & Teshima, 2016).As such, policies are subjected to refinements or to be entirely halted, thus highlighting the importance of policy evaluation (Resnick et al., 2018).
In general, policymakers devise and implement policy with the expectation of attaining equitable outcomes in society, however, this is not always the case due to social, institutional, and political constraints (Qureshi et al., 2015).In their paper about social exclusion, Hoff & Walsh (2018) emphasise that a phenomenon across the world exists, where certain groups do not fully benefit from public policies for reasons beyond structural and institutional barriers.The groups could be those that are previously disadvantaged or those that exclude themselves.Literature has shown that policy evaluation does exhibit the impact but also reveals the heterogeneity across social groups, income groups, countries, and geographic areas as shown in the literature (Djebbari & Smith, 2008;Dercon, 2009;Ravalion, 2008;de Janvry et al., 2010;Hoddinott et al., 2018;Daidone et al., 2019).Daidone et al (2019) provide evidence from several impact evaluations of government-run cash transfer programmes in Sub-Saharan Africa.They used a Differencein-Difference (DiD) estimator to derive the Average Treatment Effect (ATE) and found that, although cash transfers have positive impacts on livelihoods, they are not adequate to sustainably move households out of poverty.The study also found different impact estimators across countries, which were attributed to factors such as the demographic profile of the beneficiary, programme design, and implementation.Using the Random Coefficient Model and the Fréchet-Hoeffding bounds method, Djebbari & Smith (2008) investigated the impact heterogeneity of the Mexican conditional cash transfer programme PROGRESA.The study emphasises that additionally to estimating treatment effects between impacts, the distribution of the impacts is important in evaluating how the programme functions.This can be helpful in identifying the inequalities with the population groups, an issue of importance in development programmes.The key findings of the study are that variation in impacts exists according to household poverty levels.Similarly, Dercon (2009) evaluated the impact of cash-transfer programmes on the livelihoods of beneficiaries with a focus on agricultural activities.And although overall the policies have a positive influence, the study highlights that poverty remains a predominantly rural phenomenon and that the cash transfer programmes are not enough to move rural households out of poverty.In contrast, however, de Janvry et al. (2010) noted that after policy efforts for poverty reduction in developing countries, where urban population shifts exist, rural areas contributed more than half the observed aggregate decline in poverty.Daidone et al (2019) noted that households in rural areas face more constraints in generating sustainable livelihoods.Bearing the implication that policy should be altered to meet the needs of different social groups.
The phenomenon of heterogeneity in policy impacts appears to be prevalent across the world.When the heterogeneous effect of policies amongst groups is excluded from the evaluation, their impact is likely to be over or underestimated (Behrman & Hoddinott, 2005).It is for this reason that literature suggests that the attainment of food security can be strengthened if social justice is introduced to the design and delivery of policy reforms (Turok, 2010).For a developing country, such as South Africa, often shortterm strategies such as PPRS are implemented in the form of food safety nets and nutritional programmes, whilst longer-term policies such as the PGDS are holistic approaches to improve local economic development, social welfare, and food security.Literature provides evidence of the impact of different types of policies (long-term; short-term, complex, and simple) spanning a wide range of sectors across the globe.As aforementioned, the nine provinces in South Africa have each implemented the PGDS and PRS at different periods and are composed of different proportions of population/racial groups and geotypical distributions.Given these disparities, a holistic policy impact assessment and an evaluation of the heterogeneous impacts of the policies are presented in this study.The methodological approach followed is presented in the next section.

Empirical methods and data
Most policy impact evaluation studies use Randomized Control trials, Propensity Score Matching, among others approaches (Khandker and Koolwal, 2010;Gertler et al., 2012). 3,4The DiD is a widely used approach in policy impact analysis (Hoddinot and Skoufias 2004;Janvry et al., 2010;Daidone, 2019) and bears a key assumption there is no systematic unobserved time-varying difference between the treatment and control groups that would cause the outcomes for the comparison group and treated group to have different trends over time.This study uses a generalisation of the Difference-in-Difference (DiD) technique, the roll-out of a programme with panel data to disentangle the impact of Provincial Development and Growth Strategies (PGDS) implemented in a staggered fashion in the nine provinces of RSA.

Differences-in-differences
The DID estimator allows for the estimation of the Average Treatment Effect (ATE) of the policy.The estimator assesses the differences among treated households (i.e.located in provinces where the policy was implemented) and untreated households (i.e.located in provinces where the policy was not, or not yet, implemented): where DY i is the change (first difference) in the development indicator, X i is a vector of control factors, 1 i is the error term, a measures the changes in intercept, and u represents the changes associated with the control factors.
The model was used to estimate the effect of the policy in the provinces between the years 2003 and 2004, the period before and after the policy was implemented, and between provinces that implemented the policy and otherwise.Furthermore, a regression model is estimated to evaluate the effect of provinces stopping/or exiting the policy.Most provinces, including EC, FS, GP, NC, and NW paused the implementation during the year 2014.This enables the estimation of the effect in contrast with provinces that continued with the policy.A variable indicating early adopters of the programme (equal to 1 if the programme is implemented in 2004 and 0 otherwise) was included to establish differences in policy effects between 'early' and 'late' adopters.The analysis of subsamples of early and later adopters shows if there are differences in the impact of the PGDS on welfare indicators across the two groups.

Roll-out technique of average treatment effect
Given that all provinces implemented policy during different periods, the roll-out technique has been used to analyse the impact of large-scale policies with provinces progressively implementing the development policies in a staggered roll-out (e.g.Galiani et al., 2005;Jensen, 2007;de Janvry et al., 2010).This is done by including a full set of dummies for each group and all periods, as well as their interactions.The policy dummy then gives the measurement of the effect of the policy.
The estimated equation is given as: where i, p, t index households, provinces, and years.The binary dependent variable (Y ipt ) is a function of the treatment variable P pt which is the implementation of PGDS during a period t in a given province p, and of a set of k control factors, X k ipt , namely age, sex, and population group X k ipt ; m pt are time-varying province-fixed effects; d and u k are parameters to be estimated, where d is the estimate of average effect of policy on development; 1 ipt is an error term.P pt equals 1 after the p-th province has entered the programme (0 before and once its implementation is stopped).Due to the possibility that differences across the provinces, and time-varying factors might be correlated with the dependent variables, time and provincial fixed were included in all estimated models.The fixed effects also account for dynamic variables such as income which could be increasing over time.The inclusion of which, controls for unobserved timevariant variables (Galiani et al., 2005).The model in equation ( 2) is estimated using different dependent variables, Y ipt , as development indicators: food security, economic wellbeing, sanitation security, and water security.
Although food security is a multi-faceted concept, the variable food insufficiency was used as a proxy for food security (i.e. the outcome of the policy).It is captured by the household's insufficiency of food, with 1 if the household never has insufficient food and 0 otherwise.The same model (equation 2) was estimated, replacing the dependent variable with a household expenditure variable, used as a proxy for economic wellbeing.According to Statistics SA (2019), a poverty line of R561 ($36.1) per person per month was reported, and by using the average of 3.3 people per household in South Africa, the household expenditure variable was set as a binary variable equal to 1 if a household spends above R1800 ($117.4) per month, and 0 below Additionally, water, and sanitation access were used as proxies for water and sanitation security, respectively.Water security is represented by the distance between the household and the main water source, and is a binary variable, where 1 is if the water source is less than 200 metres from the household, and 0 otherwise.Sanitation security is also a binary variable representing how far the main toilet is from the household, where 1 is less than 50 metres and 0, is otherwise.The farther the distance from water or sanitation source, the harder the access: when it takes more time to collect water or to use the toilets, households are likely to be water and/or sanitation insecure (Anderson & Hagos, 2008;Warner et al., 2008;Foute & Sikod, 2012;Pickering & Davis, 2012).The distance to sources has been conventionally used in welfare studies to measure access and studies highlight that distance to water source is one of the main indicators of water poverty and scarcity (Ngasala et al., 2018;Nonkeou et al., 2022).Statistics South Africa presents water and sanitation access in the GHS as category variables with 5 classes ranging from less than 50 metres to more than 500 metres.To fairly distribute cases and improve variation for analysis, we transformed the variable into a binary one.The binary variables have been used as proxies for water and sanitation security.The implementation of the Provincial Poverty Reduction Strategy (PPRS) was included as a control factor in the regressions as it was at times implemented alongside the PGDS across the provinces.The PPRS is designed to promote the empowerment and development of poor communities through poverty alleviation projects (Mensah & Benedict, 2010).
The indicators chosen for the study makeup two of the United Nations' Sustainable Development Goals (SDGs) (Goals 1 and 6).Water scarcity and inadequate sanitation have also been linked to adverse effects on food security (Le Blanc, 2015) and together with poverty reduction, are indicators that the development policies in South Africa aim to improve.As such, the hypothesis tested is if PGDS influences social welfare indicators food security and economic well-being, water, and sanitation security.Considering that PGDS is a long-term policy, including two years as in the DiD estimations might understate or overstate the effect.Including all years may also pose an estimation problem, given that some provinces halt the implementation, to resume at a later stage.For sensitivity analysis, regressions were estimated to evaluate the impact, in the first year of implementation, and each five years after, i. e. 2004; 2009, and 2014.This enables dissemination of policy impacts over time and would indicate if the impact increases or decreases.

Capturing heterogeneity in policy development impacts
Drawing from the literature, Ruben & John (2004) suggest that 'one-size-fits-all' policies are not adequate to address developmental issues and that developmental strategy should be accompanied by measures that will yield sustainable growth and development for beneficiaries.Furthermore, after 25 years of democracy, South Africa is still plagued with income inequality, particularly among racial groups.The South African population is made up of four main racial groups, namely African, Indian/Asian, White, and Coloured.The name 'Coloured' refers to a legal name designated for a group of heterogeneous people who are of mixed ancestry in Southern Africa.Salisbury (2016) highlighted that there exists a large portion of earnings differentials of about 34 and 42 percent, driven by the 'labor market's lower valuation of African and Coloured worker's productive characteristics' thus, making the two population groups more likely to be food insecure.
Additionally, the spatial distribution of households in South Africa has been considered in various development studies (Maseko et al., 2015;Naicker et al., 2015), where the emphasis has been put on in-depth analysis of development at localised levels.As such, this analysis considers the potential heterogeneous effects of the policies between different racial groups and geographical areas, using policy interaction effects.This included racial group and policy interaction variables in the estimation to compare the impacts across the groups, the same was done for geographic areas.For ease of interpretation, an Ordinary Least Squares Regression was used to estimate the effects of policy.In estimating the effect of PGDS, PPRS was used as a control factor in the analysis, as its omission would yield biased estimates, possibly overstating or understating the effect.

Data description
The study uses data from the annual General Household Survey (GHS) in South Africa throughout 2002-17.The GHS is a cross-sectional survey, conducted annually, since 2002.The survey questions are designed to collect information on service delivery and living conditions and cover a range of broad (StatsSA, 2019).More than 20,000 households were included in the surveys each year, across the nine provinces, namely Western Cape (WC), Northern Cape (NC), Eastern Cape (EC), Free State (FS), KwaZulu-Natal (KZN), North-West (NW), Gauteng (GP) Mpumalanga (MP) and Limpopo (L).The comprehensive dataset portrays a pseudo-panel, as the individual datasets do not necessarily contain information about the same individuals across the years.Given that the PGDS is implemented at the Provincial level, the GHS is suitable because it includes households dispersed across and within the provinces, evading selection bias.The data was supplemented with information about the implementation of Provincial Development and Growth Strategies (PGDS) and Provincial Poverty Reduction Strategies (PPRS), from the provincial government's public documents.The following section presents the descriptive statistic as well as the empirical findings of the study.

Descriptive statistics
Table 1 presents the descriptive statistics of development indicators included in the empirical models.The results show that 76.60% of the household never have insufficient food while about 46.76% of the households reported that they are above the poverty line.About 11.30% of households have their toilets positioned over 50 m from the house, while 49.93% of the households have their water source within 200 m from the house.In addition, the distribution of development indicators across geographic areas and racial groups is shown.
The results show the proportion of households that are water and sanitation secure, as well as those who live above the poverty line (economic well-being), across the different

Difference-in-Difference approach
Table 2 presents the regression results of the DiD.The results indicate an improvement in household food security and economic well-being after the policies have been implemented, however, no improvement in sanitation access and a negative impact on water access.This contradicts the expectation that the policy would enhance water security.Given the short time between pre and post-policy, it is highly unlikely that long-term structural improvements, such as toilets and water access, would be significantly improved.As such, further analysis that differentiates across geotypes is included in the upcoming subsections to differentiate the impacts.In contrast, food expenditure and sufficiency, that are relatively more attainable goals in the short-run, were improved after the implementation of the PGDS.
For sensitivity analysis, a regression model is estimated to evaluate the effect of provinces stopping/ or exiting the policy.Most provinces, EC, FS, GP, NC, and NW terminated the implementation during the year 2014.This enables the estimation of the effect in contrast with provinces that continued with the policy.The results, presented in Table 3, indicate that food security declines for households in which provinces have exited, however economic well-being and water security continue to improve for households in provinces that stopped the policy.This indicates that food security is on average more contingent on the policy, such that there could be a decline if the policy ceases to continue.Furthermore, water infrastructure is long-lasting, and once it has been improved, whether a province pauses policy implementation or not, it is unlikely to change.In addition, if the exit variable positively affects water security, this could also be an indication that water security is on average on not dependent on the implementation of the PGDS.The coefficient estimate indicates an improvement in food security for early adopters, which could be an indicator that the early adopters possibly gain resilience (in terms of food security) throughout the years, and that long-term implementation of policy is beneficial.
The negative effect on water security for early adopters shows that provinces that implemented early, largely those with a higher number of households based in rural and informal settlements, still have a long way to go in providing water security.For instance, in the Eastern Cape province, one of the early adopters, Hay et al. (2012) noted that water supply infrastructure cannot meet water requirements and towns experience shortfalls.Furthermore, StatsSA (2016) revealed that only 74.2% of households in the Eastern Cape, which is less than the national average of 84% have access to water, so although they adopted policy early, the provinces still lag in terms of water access.Another justification is that in 2008, the Department of Water Affairs intensified water reforms throughout the country to develop water reconciliation strategies (StatsSA, 2016), these would be relatively more easily integrated into projects which provinces that adopted later are undertaking.There were no significant policy impacts on sanitation security in the estimated model.

Roll-out of programme technique regression
A roll-out technique regression method was then estimated to evaluate the impact of the policy between 2002 and 2017.This gives the average effect of the policy given that provinces implemented it during different periods.The Provincial Poverty Reduction Strategy (PPRS) was also included in the model as a control factor as it was a strategy that was concurrently adopted across the provinces throughout the years.Table 4 shows that food security, economic well-being, and sanitation security improved after the implementation of PGDS.However, a negative effect is found on water security.This shows that PGDS have generally been effective in tackling development indicators.

Time dimension of development policy effects
For sensitivity analysis, regressions were estimated to evaluate the impact, in the first year of implementation, and each five years after, i. e. 2004; 2009, and 2014.This enables dissemination of policy impacts over time and would indicate if the impact increases or decreases over time.The results in Table 5 indicate that the impact of PGDS on food security increases over time, moving from 0.08 in 2004, to 0.2 after 10 years of implementation, a 150% increase in the average impact.This also applies to economic well-being, although there is a slight decrease after 2009.In 2004, PGDS positively impacted water security, however, had the opposite effect 10 years later.The impact on sanitation security was constant between 2004 and 2009.This shows that PGDS has a positive effect on food security and well-being and that the continuation of implementation could help strengthen the resilience of South African households in terms of these indicators.

Heterogeneous effects of development policies
Table 6 presents regression results of the average effect of PGDS while considering the interaction effects of Policy and race.Salisbury (2016) highlighted that there exists a large portion of earnings differentials amongst racial groups.This is consistent with the findings of this study where results show a positive average effect of the PGDS on food security, with significantly higher effects for white households and slightly lower for coloured households5 .The results indicate that African households benefit less in terms of all the development indicators.Consistent with the findings of the StatsSA report on livelihoods (2019), showing an increase in inequality amongst African and white households.Whilst both Policy variables individually improve food security, the results generally show they, on average, have a higher impact on white households, highlighting the heterogeneity.In this model, PGDS generally does not improve economic well-being, but the opposite is true for white households.The policy generally does not have a positive effect on water and sanitation security.Table 7 presents the average effect of policy considering the interaction of policy and the geographic location of the household.The spatial distribution of households in South Africa has been considered in various development studies (Maseko et al., 2015;Naicker et al., 2015), where the emphasis has been put on in-depth analysis of development at localised levels.The results show that PGDS positively impacts food security, with different effects across geographic types.Households in urban and tribal areas are on average more food secure as compared to those in urban informal settlements and those in rural areas.Food security and economic well-being are shown to improve after the implementation of the policy, but not for households in rural and informal settlements.The results are consistent with previously mentioned studies, which highlight the plight of urban informal settlements where poverty levels are higher.This can be grounded based on these households being at the bottom of the social ladder and also bearing characteristics that impede upward mobility (Turok, 2010).
In contrast, the PGDS has positive impacts on water and sanitation security for households in rural and informal settlements.These are areas where interventions are needed most, and the results indicate that in these terms.Rhodes & McKenzie (2018), noted that rural households are worse off in terms of water access, due to infrastructural, and technical constraints, as well as poor service delivery in these areas, however, the implementation of PGDS over time seems effective in improving access to both water and sanitation.The significant positive estimate can also be attributed to the fact that rural and informal settlement households start on a lower base than their geographic counterparts, as such it is expected for the impact to be higher.Sanitation access has improved in urban and urban informal areas after the implementation of PGDS.

Conclusions and recommendations
The Provincial Growth and Development Strategies (PGDS) is a policy framework adopted by provinces in South Africa, to improve the social well-being of citizens.This study focuses on the impact of the broader, long-term PGDS on development indicators and the potential heterogeneity of the impact across social groups, using DID and the Roll-out-of programme approach.We found that the policy generally has a positive average effect on food security and economic well-being.The impact is higher after years of implementation have increased, i.e. 2004 vs. 2014: the prolonged implementation has benefits for the households of South Africa.Hoff & Walsh (2018) highlighted that certain groups might not benefit from public policies for reasons beyond structural and institutional barriers and added that groups could be those that are previously disadvantaged or those that exclude themselves.This is consistent with the findings of this study: heterogeneous effects exist across racial groups, with African and Colored households, previously disadvantaged groups in South Africa, benefitting the least.The findings are also consistent with Salisbury (2016), who highlighted earnings differentials amongst racial groups, with non-white groups earning relatively less than their white counterparts.Rhodes & McKenzie (2018) also found that African households are worse off in terms of water and sanitation access.In addition, we show that the differential impacts are evident across geographic types, insofar households in informal settlements and rural areas benefit less..We also found positive impacts of the PGDS on water security for selected geographical types.However, the average (overall) impact tend to be e negative impact, calling for further interventions on water security.In general, this shows that the policy should continuously be adapted to benefit all racial and geographic groups across the country, according to needs and basis.This can be done in putting emphasis on longterm economic stimulators such as education and health, particularly for the PGDS.The transfer of human capital in rural areas for the sustainable delivery of services, is needed.Efforts should be put into empowering households in informal settlements and rural areas, which are areas of historical neglect, primarily occupied by Africans and Coloured households, two racial groups that bear the lowest average effect of the policy.Furthermore, solutions such as the creation of employment opportunities in rural and informal settlements need to be at the forefront of policy to improve food security in these areas.The continuation of combining short and long-term strategies with programmes with inter-linkages and synergies on targets is recommended to target broader social welfare issues (Outlook, 2011;Turok, 2010;Dredge, 2015;Mamoon & Ahsan, 2017).
A shortcoming of this study is that the data used is a pseudo-panel, as such a generalised average effect is estimated.Due to the policy being implemented by the provincial government, and the broad nature of the policies, it is not possible to have a clear control and treatment group within the provinces, even though not all households in the treated provinces could have benefited from the programmes that emanate from these policies.Nonetheless, the study provides a benchmark for further studies based on more refined data.Future research should focus on specific programmes.and on different aspects of food, water and energy security.Furthermore, individual provincial estimates could shed light on which provinces need to improve the provision of basic services and the enhancement of food security and well-being.This can help in the formulation of relevant context-specific interventions.The study calls for future research on the impacts of policy interventions on the WEFE nexus, to plan effective (and synergic) interventions.

Table 1 .
Descriptive statistics of development indicators across the population and geographic groups.The results indicate that 83.52% and 75.38% of white and Indian households, respectively, live above the poverty line, with only 43.8 and 54.2 African and Coloured households, respectively, living above the line.Coloured households have the less percentage of households who have access to water and sanitation (34.6% and 5.02%).The bulk of the households that have less water and toilet access is based in rural areas, while tribal households have the least percentage of households that spend above the poverty line (27.94%).

Table 2 .
Regression results of the relationship between food insecurity and Provincial Growth and Development Strategies (PGDS) between 2003/2004.Significant at the 1% confidence level, Standard error in parenthesis.Provincial and time-fixed effects are included.Control Factors: Sex, age, population group, PPRS.

Table 3 .
Regression results of the relationship of ceasing the Provincial Growth and Development Strategies (PGDS) between 2014/2015.

Table 4 .
Regression results of the relationship between food insecurity and Provincial Growth and Development Strategies (PGDS).

Table 5 .
Regression results of the periodic impacts of PGDS on development indicators.Significant at the 1%, 5%, and 1% confidence level, respectively.Standard error in parenthesis Provincial and time-fixed effects are included.Control Factors: PPRS, sex, age, population group.

Table 6 .
Regression results of the relationship between food insecurity and PGDS and the PPRS across racial groups.

Table 7 .
Regression results of the relationship between food insecurity and (PGDS) (P1) across geographic groups.Significant at the 1%, 5%, and 1% confidence level, respectively.Standard error in parenthesis.Provincial and time-fixed effects are included.Control Factors: Sex, age.