How memory affects reported fishing incomes in household surveys

What is it about?

This paper looks at how well small-scale fishing households remember and report their incomes when they answer household surveys in Cambodia. The authors ran a large randomised experiment where fishing households were asked about their catches, costs, and income using different recall periods: weekly, monthly, seasonal, and yearly. By comparing answers about the same underlying fishing activity across these groups, they measure how much longer recall periods distort reported income and related variables. They find that when people are asked to remember longer periods, they systematically overestimate the quantity and value of fish they caught, and the costs they incurred, while at the same time listing fewer fish species. In contrast, more stable information such as prices per kilogram of fish and the number of days fished is recalled much more accurately even over longer periods. The paper also tests whether data collected by mobile phone reduces these errors and finds that seasonal phone interviews can lessen over-reporting of catch and income compared with face-to-face interviews, although they may introduce some bias in reported prices. Overall, the study shows that survey design choices, especially the length of the recall period and the mode of interview, can strongly affect measured self-employment income in fisheries. These results matter for anyone using household survey data to track livelihoods, poverty, and food security in developing countries, because apparently small design decisions can lead to very different income estimates.

Featured Image

Why is it important?

This paper provides rare experimental evidence on how the length of recall periods in household surveys affects the accuracy of self-employment income data in a low‑income, small‑scale fisheries context. While recall bias has been studied for consumption and agriculture, there has been much less experimental work on self-employment income, particularly in informal fisheries where administrative records are scarce or absent. By implementing a carefully designed randomised control trial with weekly, monthly, seasonal, and annual questionnaires, the authors can isolate the causal effect of recall length, rather than relying on observational comparisons. The findings are striking: longer recall periods can lead to very large overstatements of catch quantities and values—up to around 200–280 percent in some seasonal and yearly specifications—while simultaneously underreporting the diversity of species. At the same time, key variables like prices and days worked are much less affected, revealing that not all variables are equally vulnerable to memory problems. These results highlight that widely used annual or seasonal recall modules can seriously distort income measurement, poverty profiles, and assessments of the economic and nutritional importance of small‑scale fisheries. The paper is also timely for two reasons. First, many countries and international agencies depend on household surveys to monitor progress on development goals, yet lack clear guidance on optimal recall periods for self-employment income. Second, the study speaks to the rapid spread of phone-based surveys: it shows that seasonal phone interviews can reduce some over-reporting relative to in‑person seasonal interviews, suggesting a potentially cost‑effective way to collect more frequent, higher-quality income data, while also warning that phone mode can introduce its own biases in prices. Together, these insights can directly inform the design of future survey programmes and the interpretation of existing datasets in development research and policy.

Perspectives