Random digit dialling had never been used in Cambodia. Mobile network penetration had been too low for too long, and by the time it was viable, nobody had tried it. When the Konrad-Adenauer-Stiftung commissioned a nationally representative political opinion survey, I saw an opportunity to test the methodology for the first time in the country.
Cambodia has three mobile networks. Without reliable market share data, we assumed equal representation — random digits generated using each network's defined prefixes, equally weighted in the sampling frame. There was a slight oversampling of urban areas, which we accounted for through post-stratification weighting. The mobile pickup rate was around 14%, of which about 50% led to a completed interview. That gives you roughly one complete interview for every 15 calls — a 7% effective completion rate. On paper, that sounds inefficient. In practice, it is significantly faster than the alternative. Entering 15 phone numbers to get one complete interview is much quicker than sending an enumerator walking to the next randomly assigned household in a face-to-face survey.
RDD also gives you much wider geographic coverage. Instead of sampling specific villages or enumeration areas — and then dealing with the logistics of reaching them — you end up with respondents from districts all over the country. The interviewers were happy with it. Supervisors could easily oversee quality by listening in on calls. Fieldwork was completed remarkably quickly.
There are drawbacks. Interviews had to be kept short, so this works for certain study types but not for long, complex questionnaires. The incentive — a mobile top-up sent remotely immediately after the interview — introduces some bias toward respondents who are more available, have flexible time, or for whom the small incentive is more meaningful. These are real limitations, and I would not use RDD for everything. But for what it does well — fast, nationally representative snapshots on focused topics — it worked extremely well in Cambodia.
The full report is available from KAS. Read Part 2: What the findings revealed →