Most of us can recall, relatively accurately, the quantities of consumption items we purchase; a kilo of sugar or a litre of milk. However, in many surveys in developing countries, quantities are expressed in non-standard units; such as ‘small heap of tomatoes’ and ‘medium bunch of bananas’. This leads to ambiguous item-unit combinations which are then subject to individual interpretation of what constitutes ‘small,’ ‘medium’, ‘large’.
Consumption data, therefore, has been widely known to contain errors and irregularities. Accurate unit conversion assumptions need to be made to ensure that precise Kcal intake values are obtained. Studies can both under or over-report the consumption within a household and thus draw incorrect conclusions about their poverty status. These conclusions will influence decision making and ultimately lead to distorting the effectiveness of aid resources when allocated.
The use of non-standard units in surveys is important as households can easily relate to them. Insisting that respondents report consumption in metric units will not produce more accurate results as they are not easily relatable. Here, the benefits of CAPI can be taken full advantage of; non-standard units can still be collected and accurate Kcal intake values can be recorded. Over five years ago, we developed innovative features in our CAPI software, surveybe, to counteract common problem areas in the collection of household data. At the time, these features were pioneering and development researchers quickly recognised the significant advantages.
An example of how inaccuracies are brought about by non-standard units is demonstrated in De Weerdt et al. 2011 CAPI vs. PAPI* randomised survey experiment of 1,840 households in Tanzania. A total of 42% item-unit combinations recorded in PAPI were ambiguous. Here, the analyst would have had to interpret the metric unit from the heaps, bunches and pieces that were recorded. In the case of CAPI, only 1.6% of item-unit combinations recorded were ambiguous. With the assistance of images and item-specific units the margin for error is significantly reduced and the computer can calculate an accurate Kcal value as a result (see De Weerdt et al. 2011). Another feature of consumption modules in CAPI is that randomisation is made much easier. For example, if you wanted to only ask household members about 10 foods from a list of 50 possible foods, you can randomise which 10 are shown and display the appropriate images.
EDI’s commitment to development research was demonstrated when we made a conscious decision, over five years ago, to take our model and help other organisations enhance their capabilities to accurately report on poverty statuses by collecting quality household consumption data. The media platform for the global development community, Devex, published the following article ‘A development researcher’s best friend’, praising how CAPI vastly improves the quality and accuracy of consumption data collected in the field. The World Bank has recently published a LSMS Guidebook for The Use of Non-Standard Units for the Collection of Food Quantity. This Guidebook is comprehensive in addressing the challenges of non-standard units and benefits of CAPI that help overcome inaccuracies and irregularities.
For further information on EDI’s experiences in tackling non-standard units we have published two case studies:
‘Improving the accuracy of data through the use of innovative features provided by Surveybe’ First published 2014 and revised 2017
Caeyers, B., Chalmers, N. and De Weerdt, J. 2011. “Improving Consumption Measurement and other Survey Data through CAPI: Evidence from a Randomized Experiment”, Journal of Development Economics 98:19-33. Available here.