From Crisis Tool to Data Revolution: How Phone Surveys Could Reshape Development Policy in Africa

A World Bank study finds high-frequency phone surveys can deliver faster, cheaper data on jobs, food security, health and household welfare, with interviews costing just USD 8–12 compared with USD 230–409 face-to-face. Governments and development partners should adopt mixed-mode statistical systems, combining regular phone monitoring with periodic in-person surveys to control coverage bias, attrition and differences in responses across survey modes.

The World Bank's six-year experience with high-frequency phone surveys in Sub-Saharan Africa shows that a tool introduced as an emergency response to COVID-19 could become an important part of national data systems. Researchers associated with the World Bank's Living Standards Measurement Study (LSMS) and Development Data Group, working with national statistical offices in Burkina Faso, Ethiopia, Malawi, Nigeria, Tanzania and Uganda, examined how phone surveys can provide governments with faster and cheaper information on jobs, food security, health and household welfare. The study is based on the LSMS High-Frequency Phone Surveys and the LSMS Integrated Surveys on Agriculture.

Since 2020, the programme completed 122 survey rounds and 245,224 interviews, following 17,768 households. Interviews generally took place about every two months and lasted around 19 minutes on average. What started as COVID-19 monitoring expanded into employment, agriculture, prices, vaccination, health care, credit, climate risks, migration and conflict.

Faster data at a fraction of the cost

The strongest case for phone surveys is economic. A completed face-to-face household interview across the six countries cost approximately USD 230-409, compared with only USD 8-12 for a phone interview. On a per-interview basis, traditional field surveys can therefore cost around 30 times more. Even after considering that face-to-face interviews contain many more questions, they remain roughly three to four times more expensive per question.

For governments facing limited statistical budgets, this creates an opportunity to monitor economic conditions much more frequently. Instead of waiting years for a major household survey, ministries could use phone interviews to track employment, food insecurity, prices, household businesses and the effects of economic or climate shocks.

Phone surveys, however, cannot replace field surveys. Detailed consumption data, agricultural measurements, anthropometrics and GPS-based land measurement still require longer or in-person interviews. The study therefore supports a mixed system: periodic face-to-face surveys for detailed measurement, combined with regular phone surveys for rapidly changing indicators.

Response rates show why building phone surveys on existing national samples matters. Among households with available contact information, average response across countries and rounds was around 86%, although individual rounds ranged from 47% to 99%. First-round response among LSMS recontact surveys averaged about 74%, compared with roughly 28% among random-digit-dial surveys examined in related research.

Attrition and respondent bias remain major risks

Keeping households in long-running phone surveys is difficult. Cumulative attrition reached approximately 22% in Malawi, 24% in Nigeria, 29% in Tanzania, 36% in Uganda, 39% in Burkina Faso and 53% in Ethiopia. Conflict and instability may have contributed to particularly high losses in some countries.

Still, the researchers found surprisingly little evidence that this attrition created serious bias across most of the household characteristics studied. This suggests that well-designed recontact surveys can remain useful even after losing a significant share of respondents.

Survey design matters. Every additional minute of interviewing was associated with roughly a 1-2% increase in attrition risk. Among unsuccessful interview attempts, 50.7% occurred because phones were switched off, 19.2% because nobody answered and only 6.3% because respondents refused.

Another concern is who answers the phone. Main household respondents tend to be older, male and better educated, potentially distorting employment, health and other individual indicators. In one Nigerian survey round, 84.7% of main respondents reported working, compared with 64.8% of other selected household members.

Proxy answers are useful mainly for easily observable information. Agreement between proxy and self-reported vaccination status reached 94% in Burkina Faso and 93% in Uganda, but agreement on the more subjective issue of vaccine acceptance was only around 53-56%.

Governments and development partners need mixed-mode systems

An even bigger challenge is that people can answer the same question differently depending on whether they are interviewed by telephone or face to face. A Nigerian experiment covering 20 indicators found median differences of around 17-18% between the two modes.

For about 70% of the indicators, telephone responses shifted upward. Respondents were more likely to say "yes", agree with statements, provide extreme values or round numerical answers. Question complexity made the problem worse: inconsistency between telephone and face-to-face responses increased from around 36% for binary questions to 82% for questions offering ten response options.

For governments, the message is that cheaper data should not automatically be treated as equivalent data. National statistical offices should regularly update telephone contacts during household surveys, keep phone questionnaires short and periodically use face-to-face surveys to check the quality of telephone estimates.

International development partners, including multilateral institutions, bilateral donors and UN agencies, can support this transition by financing statistical infrastructure rather than temporary emergency surveys. Investment is needed in mixed-mode survey systems, methodological research, interviewer training, statistical adjustment and stronger national capacity.

New opportunities, but quality must come first

The shift also creates opportunities for telecom operators and survey-technology providers. Private companies could support call-management platforms, airtime incentives, contact verification and digital data systems. Providing phones to excluded households is another possibility. In Nigeria, 995 households across 106 enumeration areas received phones and SIM cards; the intervention cost about 27% of what 12 rounds of phone surveys would cost and showed no evidence that receiving a phone itself changed subsequent responses.

The next priority is to improve methods for correcting coverage and mode effects and to link telephone data more effectively with face-to-face and geospatial information. Researchers also see potential for using phone-based indicators such as assets, housing conditions and food security to estimate changes in poverty between expensive consumption surveys.

The broader lesson is practical: governments do not need to choose between expensive traditional surveys and cheaper phone interviews. A carefully designed combination can provide detailed benchmark data through periodic fieldwork and fast, affordable monitoring through phones. The opportunity now is to turn infrastructure created during COVID-19 into permanent statistical capacity capable of supporting decisions during economic downturns, food crises, climate shocks, conflicts and future public-health emergencies.