Overcoming the challenges of rural surveys in developing countries | University of Oxford
Overcoming the challenges of Rural Surveys in Developing Countries
Overcoming the challenges of Rural Surveys in Developing Countries

Image credit: Shutterstock

Overcoming the challenges of rural surveys in developing countries

Field researchers, Dr Giacomo Zanello, Dr Marco Haenssgen, Ms Nutcha Charoenboon and Mr Jeffrey Lienert explain the importance of continuing to improve survey research techniques when working in rural areas of developing countries.

News about big data and artificial intelligence can leave the impression that a data revolution has made conventional research methods obsolete. Yet, many questions remain unanswerable without working directly with (and understanding) the people whose lives we are interested in. In development studies research, survey research methods therefore remain a staple of data generation, and survey data generation itself remains an active field of debate. In today’s blog, four researchers showcase recent methodological advances in rural health survey research and the advantages they bring to conventional research approaches.

Reaching People at the Margins, 25% off! (Dr Marco J Haenssgen, Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine)

Generating representative data from rural areas of developing countries is a real challenge because often we lack detailed and dependable information on the local population, which makes drawing a sample very difficult. However, recent technological revolutions that we are rather familiar with – the Internet, mobile phone technology, satellite navigation – can also facilitate our work in survey research. Satellite maps in particular help us to:

(1) Select villages more rigorously: We can use satellite maps to generate or verify geo-coded village registers (e.g. censuses or the US National Geospatial-Intelligence Agency) to draw geographically stratified samples. Geo-stratification ensures that we do not accidentally select only “easy” villages that represent less constrained lifestyles in the rural population.
(2) Identify and select houses within the villages more inclusively: Conventional methods to draw a sample of households require either a very laborious enumeration process by going from house to house to establish a sampling frame, and/or are likely to exclude households and settlements at the fringes of a village (e.g. a “random walk”). By using satellite images to enumerate all houses in a village, not only do we save a lot of time and money, but we can also ensure that all parts of a village are represented fairly.
(3) Reach survey sites more efficiently: The logistical benefits cut as much as 25% off the conventional survey costs and time, which can save up to £5,000 for a PhD-level survey (400 respondents in 16 villages) and £40,000 for a medium-sized two-country survey (6,000 respondents in 139 villages).

We need to appreciate that satellite-aided sampling approaches are only an addition to our survey toolkit. They do not work well in urban areas, with mobile populations, or in regions that we are not familiar with. But where they work, they are a real alternative to conventional survey approaches and can make projects feasible that would otherwise be prohibitively expensive, without compromising quality.

Taking Energy Measurement From the Lab to the Field (Dr Giacomo Zanello, School of Agriculture, Policy and Development, University of Reading)

How much energy do you burn during the day (at your job, doing household chores, or at the gym) and is this “energy expenditure” in balance with the calories you take in with your food and drinks? Historically, to answer this question, participants had to spend time in a sealed chamber in a lab which measures the change in oxygen levels while performing activities. While this provides an accurate estimate of energy use, this method is quite impractical to understand real-life settings, particularly for remote areas in a developing country context. It is in these contexts where calorie deficits are most pressing, and yet we do not know much about farmers’ energy use, differences across gender and age groups, or variations of energy use across the seasons and during health or climate-borne adversities.

Recent technological advances allow the measurement of energy expenditures of free living populations to a scale and within a budget inconceivable few years ago. Using Fitbit-like accelerometers we can capture people’s movements and use this information to estimate calorie expenditure. By wearing these devices we follow people’s activities throughout the day, weeks, and seasons and use this information to estimate their energy use. This new glimpse into how people spend their energy can improve health research in multiple domains, for example:

• Having a more accurate assessments of the incidence, depth and severity of undernutrition and poverty,
• Estimating energy requirements for specific livelihood activities, or
• Studying the effect of health conditions and illnesses on livelihood activities.

These are just some possibilities, and the data collected through this innovative methodology extends beyond health-focused research. It also enables us to learn more about how labour is distributed within rural households in developing countries, or measure production in the household and the “informal economy” to produce better estimates of the size of rural economies.

Taking energy measurement from the lab to real-life settings is not without complications. We have to make careful decisions about the devices we use (e.g. easy to wear, not requiring user interaction, not attracting too much attention), build a trusting relationship with our research participants, and acknowledge that even accelerometer-generated data only offers a partial view into energy expenditure and daily activities. Yet even this partial view can afford a completely new understanding of people’s rural livelihood.

A Qualitative Research Update for Social Network Surveys (Ms Nutcha Charoenboon, Mahidol-Oxford Tropical Medicine Research Unit)

Health and treatment hardly take place in isolation – people around us influence our behaviour, give us advice, or lend us a ride to the hospital. Public health information campaigns, too, are subject to people’s relationships because they might be communicated further or even be instrumentalised for political purposes. Perhaps it is no surprise then that there are calls for more social network research on health in developing countries, but such research faces difficult questions, like how do we ask elicit the names of people in these networks, and how we can match these names in place where one person might be addressed in several different ways (e.g. “Old Father,” “Leader,” Yod Phet, and Ja Bor).

How can we overcome such difficulties? One possibility is cognitive interviewing, consisting of a set of interview techniques to test and interpret survey questions. Among others, interviewees are given survey questions and asked to “think out loud” on how they understand and answer the question, to paraphrase the question in their own words, or to explain village life and the local context. Such information gives researchers a better grasp of local social networks, living arrangements, and people’s understanding of social network questions. In our study in rural Thailand and Lao PDR, it enabled us to drop irrelevant questions, add questions to map health social networks more comprehensively, and to identify mechanisms to locate named contacts within the village more effectively.

But beware of surprises when you carry these methods over to developing country contexts because they tend to assume Western communication norms. Our research participants felt uncomfortable when asked to articulate their thought processes or to answer “why” questions. To cope with such complications, the methods themselves need to be adapted to context, for example by being more closed-ended and by adopting more conventional semi-structured interview techniques.

Shining New Light on Health Behaviours (Mr Jeffrey Lienert, Saïd Business School and National Institutes of Health)

When people get sick, they do not just make a one-off treatment decision like “I’ll go to a clinic / a private doctor / a pharmacist” and stick to it for the remainder of their illness until they are cured. Rather, they go through several phases. For example, a person might first wait and see if it the illness would not go away by itself, then later decide to buy some painkillers to cope with it, visit a private doctor when things do not get better, then lose hope in modern medicines and visit a traditional healer. We gain a lot of information about people’s behaviour if we collect such data on treatment “sequences.”

Not only is it rare for studies to record treatment sequences at all, but there are also no agreed tools for their analysis. First ground has been broken with sequence-sensitive analyses to produce more accurate typologies of behaviour, but we can go further and apply network analysis techniques to make maximal use of sequential data. More detailed analyses can differentiate between the individual steps, explore whether sequences of behaviour resemble each other across people, and which kind of social network is most decisive for such a resemblance. The downside of these arguably more complex analyses is the technical skill required to perform them, but once these methods become more established, they will be able to us to give more detailed (and realistic!) behavioural profiles of different settings and social groups with revolutionarily new insights for health policy.

Methodological innovation enables easier, more precise, and new ways of understanding human behaviour. That does not necessarily mean “big data” and algorithms. Innovation also arises from new combinations of conventional methods with other established techniques and new technologies. Combining rural health surveys with satellite imagery and accelerometers, social network surveys with cognitive interviewing, and healthcare access data with social network analysis does not just keep the methodological debates in survey research alive. It also enables new research, new questions, and a new view on human behaviour.

This blog entry derives from the authors’ contributions to the ESRC NRCM Research Methods Festival 2018 Conference in Bath, drawing on research from the projects Antibiotics and Activity Spaces (ESRC grant ref. ES/P00511X/1), Mobile Phones and Rural Healthcare Access in India and China (John Fell OUP Research Fund ref. 122/670 and ESRC studentship ref. SSD/2/2/16), and IMMANA Grants funded with UK aid from the UK government (ref. #2.03).