New paper: "Milking the Data"

Quick: how much milk did you drink last year?

If you can answer that accurately, you’re either taking the “quantified self” thing a bit far, or you may have been reading some of our research.

A new paper co-authored by our colleges on the Living Standards Measurement Study (LSMS) team compares different methods for estimating how much milk is being taken from livestock for human consumption.

Alberto wrote about this research last year and the work has been published in Food Policy under an open access license. I think the findings are super-interesting - the authors are trying to understand how to accurately find out from individuals “how much milk did you collect from your animals this year?”

Simply asking that question isn’t likely to get you an accurate answer, but if you had to rely on questions in a survey, which questions would you pick? The study compares the answers provided by different survey “recall methods” in Niger against benchmark data gathered by actually measuring the volume of milk taken (weighing it in a jug... ) one day every 2-weeks over the course of a year.

Measuring volumes of milk is going to be the most accurate method, but it’s too expensive to be done at scale. The study finds that it is possible to get accurate estimates of “milk off take” (milk taken for human consumption) using recall methods, and it compares the different options for questions that survey designers can use.

Improving data quality is critical to understanding the livelihoods of the rural poor—but finding the best data collection method, at the best cost, can be difficult. Agricultural statistics are especially weak in Sub-Saharan Africa, and this research is a welcome addition to the body of knowledge in the area.