Post Distribution Monitoring Survey in Kalobeyei Settlement, March 2020 [Kenya]

The present PDM was conducted under time and resource constraints related to COVID-19 emergency. Due to the restrictions on public gathering and partial restrictions on movements, the survey findings could not have been triangulated with the FGDs or market assessment, which will be an integral part of all subsequent PDMs.

The PDM household survey data collection took place on 25 - 27 March 2020. ProGres V4 data of Kalobeyei persons of concern was used as a sampling frame, with a sample drawn using stratified random sampling based on random numbers generation. The original sample included over 400 households (adjusted for a non-response rate) aiming at a confidence level of 95% with a confidence interval of 5. However, the enumerators managed to conduct 457 interviews with respondents added through convenience sampling. Due to poor quality of some of the records, however, only 388 data entries were validated. This nevertheless allows us to remain within the same degree of precision in the inference, although affected by a bias linked to a non-probability sampling.

UNHCR Nairobi CBI and Information Management Officers developed and coded the data collection tools and facilitated, together with Data Management Kakuma, the sampling. ProGres V4 data of Kalobeyei persons of concern was used as a sampling frame, with a sample drawn using random numbers generation. Unfortunately, because Assistance Record has not been updated in the V4, some of the respondents that appeared on the lists turned out to be non-recipients (newly relocated refugees, recently split households, etc.). As a result of this sampling frame error and a high non-response rate (including recipients not present at their places of residence), despite the sample size having been initially adjusted for the non-response rate, the teams opted to replace them with available respondents found on the ground (convenience sampling). The decision was taken as replacing the respondents with the ones from the original sample was logistically unfeasible as enumerators were allocated to particular locations and would have risked exhausting the sample without reaching saturation, while further delay was unacceptable in the face of evolving COVID-19 pandemic and the risks that the disease might reach the camp thus halting the exercise.

The original sample included over 400 households (adjusted for a non-response rate) aiming at a confidence level of 95% with a confidence interval of 5. However, the enumerators managed to conduct 457 interviews with respondents added through convenience sampling. Due to poor quality of some of the records, however, only 388 data entries were validated. This nevertheless allows us to remain within the same degree of precision in the inference, although affected by a bias linked to a non-probability sampling.