Description: This data set contains the responses from 476 participants regarding household demographic information, poverty perceptions, household income and expenditure patterns, food and nutrition security, asset status of the household, health status, children and education, unemployment and poverty, employment and income generating history of household members, credit risk, general overview of the EPWP project, EPWP employment, skill and training profile, and personal/household benefit from EPWP.

The study targeted 400 EPWP beneficiaries and the actual realised sample was 376 EPWP current and past beneficiaries from 15 EPWP projects in 4 KZN districts. The study also targeted 100 non-beneficiaries for comparisons.

The data set contains 476 cases and 1087 variables. Abstract: Public employment programmes (PEPs) or (public works programmes (PWPs) remain a popular policy instrument globally and particularly in developing countries as a short term measure for tackling unemployment and alleviating poverty. They are premised on the notion that employment will directly impact household livelihoods through access to wages, while it will more broadly act as a stimulus for the economy.

Public employment programmes transfer impacts in three key ways, namely:

transfer of wages,

asset accumulation at the individual and household levels and

assets that are created from projects for the community at large.

A socio-economic impact assessment of EPWP Phase 2 in KwaZulu-Natal and a measure to estimate the total rand value of each EPWP job in the province was commissioned by the Provincial Department of Transport (DoT), the lead agency responsible for the co-ordination and monitoring of the Expanded Public Works Programme (EPWP) in the province.

The evaluation adopted a mixed methods design that drew on the strengths of quantitative data collection and analysis involving a purpose built comprehensive household survey. Qualitative data collection methods and analysis was widely used on focus groups and key informant interviews. The survey was administered to a total of 376 EPWP participants both current and past, from fifteen projects in four districts across KZN namely Zululand, eThekwini, uGu and uMzinyathi. In addition 100 non beneficiary (comparison group) stakeholders were interviewed. The beneficiaries were identified from projects across all four EPWP sectors.

Face-to-face interview

Observation

The survey was administered to a total of 376 EPWP participants both current and past, from 15 projects in four districts across KZN and 100 comparison participants (non-EPWP) from across four EPWP sectors.

A multi-stage purposive non-probability sampling approach was adopted for this study. The primary sampling unit was an EPWP project as recorded on the EPWP Web based information management system, the output of which is referred to as 'data dumps'. The sampling approach took account of the differentiation in respect of EPWP projects, across the four sectors, geographical spread in KZN, inclusion of both urban and rural projects. At an EPWP project level a minimum of 25 interviews were conducted with beneficiaries to allow for heterogeneity in the sample selection to include youth, women and people living with disabilities. The sample size selected was informed by the need to have sufficient number of observations per site in order to be able to make inferences and to assess any trends and patterns emerging and also by cost and time constraints. Beneficiary samples were drawn from project beneficiary lists provided by site managers

Three major constraints were faced in respect of the survey, firstly the lack of baseline socio-economic data on EPWP participants. To address this challenge questions were included in the survey to help ascertain the changes experienced by beneficiaries once they had entered the programme compared to their situation prior to entry. The reliability of this data is noted as it is acknowledged that recall is not always optimum.

The second challenge relates to a methodological issue of attribution of causality and related to the fact that a comparison group was not constructed at the time at which the intervention was implemented. Experimental impact evaluation design studies compare the observed impacts on the group that received the intervention to a group which did not receive the intervention to ensure that the observed outcomes are a direct result of the programme itself and would not have occurred without the programme (Shahidur et al., 2010). This group is referred to as the comparison group which is randomly selected. No baseline data was ever collected on EPWP Phase II matching it difficult use the more rigorous propensity score techniques. The non-collection of baseline data remains a big challenge of programme evaluation in South Africa and the evaluator is almost always forced to use so-called ex-post matching techniques. According to Gertler et al (2011) matching can be achieved when the programme assignment rule is known by applying matching on the rule. To create a comparison group post implementation, matching methodology was applied to four of the sampled sites. The main criteria used to identify comparison groups was to identify areas with a similar socio-demographic profile as EPWP beneficiaries such as residents of a neighbouring ward or a section of the ward where no EPWP intervention was implemented. A sample size of 100 (being 25%) of the beneficiary sample was agreed on and accordingly sampled. Although it is acknowledged that this approach is less than ideal under the given circumstances this was seen as a fair compromise.

The third challenge was in identifying and establishing contact with exited EPWP beneficiaries. Here the challenge was poor record keeping by implementing agents as well as outdated contact details made contact with this group. To address this challenge the study relied on implementing agents who were able to identify exited beneficiaries. Hence the selection of past beneficiaries was purposive.