Should all attrition households in rural panel datasets be tracked? Lessons from a panel survey in Nepal

Research output: Contribution to journalJournal articlepeer-review

Standard

Should all attrition households in rural panel datasets be tracked? Lessons from a panel survey in Nepal. / Walelign, Solomon Zena.

In: Journal of Rural Studies, Vol. 47, No. Part A, 2016, p. 242–253.

Research output: Contribution to journalJournal articlepeer-review

Harvard

Walelign, SZ 2016, 'Should all attrition households in rural panel datasets be tracked? Lessons from a panel survey in Nepal', Journal of Rural Studies, vol. 47, no. Part A, pp. 242–253. https://doi.org/10.1016/j.jrurstud.2016.08.006

APA

Walelign, S. Z. (2016). Should all attrition households in rural panel datasets be tracked? Lessons from a panel survey in Nepal. Journal of Rural Studies, 47(Part A), 242–253. https://doi.org/10.1016/j.jrurstud.2016.08.006

Vancouver

Walelign SZ. Should all attrition households in rural panel datasets be tracked? Lessons from a panel survey in Nepal. Journal of Rural Studies. 2016;47(Part A):242–253. https://doi.org/10.1016/j.jrurstud.2016.08.006

Author

Walelign, Solomon Zena. / Should all attrition households in rural panel datasets be tracked? Lessons from a panel survey in Nepal. In: Journal of Rural Studies. 2016 ; Vol. 47, No. Part A. pp. 242–253.

Bibtex

@article{5cf5d84fa0f1484fb24f5ebbfad00bf2,
title = "Should all attrition households in rural panel datasets be tracked? Lessons from a panel survey in Nepal",
abstract = "Panel surveys are always subject to attrition: the original number of respondents is reduced over time and this process potentially affects the internal and external validity of a study. This is a common challenge in rural panel surveys in developing countries, where {\textquoteleft}attritors{\textquoteright} are typically not tracked and instead excluded from panel surveys. The few existing tracking studies focus on following {\textquoteleft}movers{\textquoteright} who migrated out of the village of origin, even if a significant number of {\textquoteleft}attritors{\textquoteright} can also be {\textquoteleft}non-movers{\textquoteright} who remained in the same dwelling. Non-movers are important for rural livelihood studies, but we do not know much about the implications of the exclusion of this group for rural livelihood studies. Using a panel data set from rural Nepal, reduced from 507 to 428 households over six years, we tracked and interviewed all attrited households in order to examine the characteristics of non-movers (in comparison to movers and non-attritors) over time and thus assess the effect of their exclusion on static and dynamic livelihood modelling. The majority of attritors were found to be non-movers. Non-movers, movers, and non-attritors are relatively similar in terms of asset endowment, household (and household head) characteristics, and livelihood activities in the initial year of investigation, but the three groups behave differently in the last year of investigation. Different household socio-economic factors determine households' probability of being a mover or non-mover. These disparities have resulted in significant differences in parameter estimates of static and dynamic rural livelihood models. The exclusion of movers and non-movers from the overall sample affects actual household livelihood transitions (in terms of farm income share) differently: the farm income share of households is overestimated when movers are excluded and underestimated when non-movers are excluded. However, the livelihood models were not so severely biased as to affect the conclusions reached for the population based on the non-attritors sample. This may be due to the small size of the movers and non-movers samples and the heterogeneity within the attritors sample in the data. The additional cost of tracking non-movers was very low and this sample is important for rural livelihood studies. Hence, the non-movers sample should always be tracked. The cost of tracking movers was also low, though much larger than the one for non-movers, but this sample is less important for studies that aim to understand rural livelihoods. Hence, the decision whether to track the movers sample depends on the purpose of the study.",
author = "Walelign, {Solomon Zena}",
year = "2016",
doi = "10.1016/j.jrurstud.2016.08.006",
language = "English",
volume = "47",
pages = "242–253",
journal = "Journal of Rural Studies",
issn = "0743-0167",
publisher = "Pergamon Press",
number = "Part A",

}

RIS

TY - JOUR

T1 - Should all attrition households in rural panel datasets be tracked? Lessons from a panel survey in Nepal

AU - Walelign, Solomon Zena

PY - 2016

Y1 - 2016

N2 - Panel surveys are always subject to attrition: the original number of respondents is reduced over time and this process potentially affects the internal and external validity of a study. This is a common challenge in rural panel surveys in developing countries, where ‘attritors’ are typically not tracked and instead excluded from panel surveys. The few existing tracking studies focus on following ‘movers’ who migrated out of the village of origin, even if a significant number of ‘attritors’ can also be ‘non-movers’ who remained in the same dwelling. Non-movers are important for rural livelihood studies, but we do not know much about the implications of the exclusion of this group for rural livelihood studies. Using a panel data set from rural Nepal, reduced from 507 to 428 households over six years, we tracked and interviewed all attrited households in order to examine the characteristics of non-movers (in comparison to movers and non-attritors) over time and thus assess the effect of their exclusion on static and dynamic livelihood modelling. The majority of attritors were found to be non-movers. Non-movers, movers, and non-attritors are relatively similar in terms of asset endowment, household (and household head) characteristics, and livelihood activities in the initial year of investigation, but the three groups behave differently in the last year of investigation. Different household socio-economic factors determine households' probability of being a mover or non-mover. These disparities have resulted in significant differences in parameter estimates of static and dynamic rural livelihood models. The exclusion of movers and non-movers from the overall sample affects actual household livelihood transitions (in terms of farm income share) differently: the farm income share of households is overestimated when movers are excluded and underestimated when non-movers are excluded. However, the livelihood models were not so severely biased as to affect the conclusions reached for the population based on the non-attritors sample. This may be due to the small size of the movers and non-movers samples and the heterogeneity within the attritors sample in the data. The additional cost of tracking non-movers was very low and this sample is important for rural livelihood studies. Hence, the non-movers sample should always be tracked. The cost of tracking movers was also low, though much larger than the one for non-movers, but this sample is less important for studies that aim to understand rural livelihoods. Hence, the decision whether to track the movers sample depends on the purpose of the study.

AB - Panel surveys are always subject to attrition: the original number of respondents is reduced over time and this process potentially affects the internal and external validity of a study. This is a common challenge in rural panel surveys in developing countries, where ‘attritors’ are typically not tracked and instead excluded from panel surveys. The few existing tracking studies focus on following ‘movers’ who migrated out of the village of origin, even if a significant number of ‘attritors’ can also be ‘non-movers’ who remained in the same dwelling. Non-movers are important for rural livelihood studies, but we do not know much about the implications of the exclusion of this group for rural livelihood studies. Using a panel data set from rural Nepal, reduced from 507 to 428 households over six years, we tracked and interviewed all attrited households in order to examine the characteristics of non-movers (in comparison to movers and non-attritors) over time and thus assess the effect of their exclusion on static and dynamic livelihood modelling. The majority of attritors were found to be non-movers. Non-movers, movers, and non-attritors are relatively similar in terms of asset endowment, household (and household head) characteristics, and livelihood activities in the initial year of investigation, but the three groups behave differently in the last year of investigation. Different household socio-economic factors determine households' probability of being a mover or non-mover. These disparities have resulted in significant differences in parameter estimates of static and dynamic rural livelihood models. The exclusion of movers and non-movers from the overall sample affects actual household livelihood transitions (in terms of farm income share) differently: the farm income share of households is overestimated when movers are excluded and underestimated when non-movers are excluded. However, the livelihood models were not so severely biased as to affect the conclusions reached for the population based on the non-attritors sample. This may be due to the small size of the movers and non-movers samples and the heterogeneity within the attritors sample in the data. The additional cost of tracking non-movers was very low and this sample is important for rural livelihood studies. Hence, the non-movers sample should always be tracked. The cost of tracking movers was also low, though much larger than the one for non-movers, but this sample is less important for studies that aim to understand rural livelihoods. Hence, the decision whether to track the movers sample depends on the purpose of the study.

U2 - 10.1016/j.jrurstud.2016.08.006

DO - 10.1016/j.jrurstud.2016.08.006

M3 - Journal article

VL - 47

SP - 242

EP - 253

JO - Journal of Rural Studies

JF - Journal of Rural Studies

SN - 0743-0167

IS - Part A

ER -

ID: 165001565