Mission Impossible? : Exploring the Promise of Multiple Imputation for Predicting Missing GPS-Based Land Area Measures in Household Surveys
Research has provided robust evidence for the use of GPS technology to be the scalable gold standard in land area measurement in household surveys. Nonetheless, facing budget constraints, survey agencies often seek to measure with GPS only plots wi...
| Main Authors: | , , |
|---|---|
| Format: | Working Paper |
| Language: | English en_US |
| Published: |
World Bank, Washington, DC
2017
|
| Subjects: | |
| Online Access: | http://documents.worldbank.org/curated/en/668211499349698549/Mission-impossible-exploring-the-promise-of-multiple-imputation-for-predicting-missing-GPS-based-land-area-measures-in-household-surveys http://hdl.handle.net/10986/27641 |
| Summary: | Research has provided robust evidence
for the use of GPS technology to be the scalable gold
standard in land area measurement in household surveys.
Nonetheless, facing budget constraints, survey agencies
often seek to measure with GPS only plots within a given
radius of dwelling locations. Subsequently, it is common for
significant shares of plots not to be measured, and research
has highlighted the selection biases resulting from using
incomplete data. This study relies on
nationally-representative, multi-topic household survey data
from Malawi and Ethiopia that exhibit near-negligible
missingness in GPS-based plot areas, and validates the
accuracy of a multiple imputation model for predicting
missing GPS-based plot areas in household surveys. The
analysis (i) randomly creates missingness among plots beyond
two operationally relevant distance measures from the
dwelling locations; (ii) conducts multiple imputation under
each distance scenario for each artificially created data
set; and (iii) compares the distributions of the imputed
plot-level outcomes, namely, area and agricultural
productivity, with the known distributions. In Malawi,
multiple imputation can produce imputed yields that are
statistically undistinguishable from the true distributions
with up to 82 percent missingness in plot areas that are
further than 1 kilometer from the dwelling location. The
comparable figure in Ethiopia is 56 percent. These rates
correspond to overall rates of missingness of 23 percent in
Malawi and 13 percent in Ethiopia. The study highlights the
promise of multiple imputation for reliably predicting
missing GPS-based plot areas, and provides recommendations
for optimizing fieldwork activities to capture the minimum
required data. |
|---|