1. Mobile Phone Surveys: In 2009-10 EDI hosted a Brian Dillon, a PhD candidate from Cornell University. Brian experimented with a novel survey technique, conducting high-frequency interviews with a panel of remote cotton farmers using mobile phones. He wrote up his experiences for the benefit of others pursuing similar endeavours. Downlad note on "Using Mobile Phones to Conduct Research in Developing Countries". 

2. In summer 2009 EDI will randomly introduced a variety of electronic and paper consumption modules to 1,840 households living on the the island of Pemba in Zanzibar to assess the efficacy of new advances in information technology in the collection of survey data.

3. During 2007-2008 EDI implemented a survey of 4,500 households on behalf of the University of Dar es Salaam and the LSMS group at the World Bank that looks at the effect of different types of consumption and labour modules. The objective of this study is to understand the implications of alternative designs to measure household consumption, welfare and labour and to improve on current methods of collecting survey data. Go to project website

4. Shock Modules - download notes on our experciences in administering shock modules (forthcoming in Journal of International Development)

5. Modules to collect data on how people perceive risk (ex-ante) of different income generating activities. This culminated in a research project which looked at why one should, and how one can collect data on risk perception. Download report. (published in the World Bank's Social Protection Discussion Paper Series)

6. As part of the KHDS survey, EDI implemented a study to assess the quality of data collected through household surveys as they relate to time and distance travelled. Over 500 trips were organised in which KHDS respondents were asked to estimate time and distance to a destination (water source or school) and were subsequently requested to walk this same distance with a pedometer to measure the actual time and distance. Initial comparisons of estimated and actual times and distances shows that:

Under an over-estimating
80% of both the time and distance estimates were
over-estimates. In the case of distance 60% of the estimates were at least two
times higher than the actual distance, and 28% higher than five times. The
trend to overestimate was found less severe in the case of time, in which case
47% of the estimates were at least two times higher than the actual distance,
and 11 % were overestimates by more than five times

Correlations
(1) A strong correlation (0.93 to 0.97) was observed between measured distance
and measured time for various trips. This suggests (a) appropriateness of
pedometers as time and distance measuring device for this kind of survey (b) a
well executed survey with appropriately trained enumerators (c) Seriousness on
the part of respondents and supervisors.
(2) A weaker correlation (0.2 to 0.26) between estimated distance and measured
distance suggesting poor perception about distance (kilometers).
(3) A stronger correlation (0.51 to 0.65) between estimated time and measured
time suggesting better perception about time compared to distance.
(4) An even poorer correlation (0.13 to 0.18) between estimated time and
estimated distance.

Tendency to rounding off estimates
While 72 % of the distance estimates
were reported as whole kilometers, 15% were reported as a multiple of 10
kilometers. As for time, more than half (55%) of the estimates were reported as
a multiple of 15 minutes. 85 % of the estimates were multiple of five minutes.
We see a strong tendency to round off while estimating time and distance.

Further work on whether certain socio-economic groups systematically over or underestimate will be done. If measurement error is systematic, then this could lead to erroneous inferences about the relation between access and poverty.

Data collection was funded by the World Bank (TUDTR). Analysis of these data is conducted by Shyam KC. Please contact him for mor details.