A fall in reported wildlife consumption can mean two very different things. People may have changed what they do, or become less willing to tell a researcher about it. For conservationists trying to understand pressure on giraffes, mistaking one for the other would make it difficult to judge whether conditions were improving.
Kirstie A. Ruppert and colleagues addressed that problem in a study published in 2021. Working with communities in northern Kenya, they asked about giraffe meat consumption in three ways. Two deliberately obscured each person’s response while allowing the team to estimate how common the behaviour was across a group.
A yes that does not identify a person
In the randomized-response method, a participant rolled a coloured die inside an opaque cup. The interviewer could not see the outcome. Some colours required a yes or no regardless of the person’s experience; others required a truthful answer about eating giraffe meat during the preceding year.
A yes therefore did not tell the interviewer whether the respondent had eaten giraffe. It might simply have been required by the die. Across many interviews, however, the known probabilities allowed the researchers to subtract the expected forced answers and estimate the underlying prevalence.
The second method used illustrated lists of animals. One group received four animals; another received the same four plus giraffe. Respondents reported only how many they or their household had eaten, without naming which ones. The difference between the two groups’ average counts provided an estimate for giraffe consumption.
These approaches depend on people understanding and trusting the rules. Local research assistants helped translate the questionnaire into Maa, check the translation and practise the tasks with participants. Interviews did not proceed to the specialised questions when comprehension appeared inadequate.
Two surveys, three estimates
The first survey ran from November 2016 to July 2017; the second from July to December 2019. Sampling covered community conservancies in Laikipia and Samburu counties. Because reliable household lists were limited and residents moved with their livestock, the team used quotas based on village size and an intended balance of men and women.
The analysis included different numbers of completed answers for each method: for example, 529 and 674 randomized responses in the first and second periods respectively. These were comparisons of samples at two times, rather than a record of the same individuals changing their behaviour.
The randomized-response estimates fell from 54.2% to 18.4%. Their 95% confidence intervals were 47.7–60.6% and 7.3–29.5%, respectively. The gap between those ranges was consistent with a substantial decline, even though the second estimate remained imprecise.
Direct questioning produced much lower estimates: 22.6% and 3.0%. List counting also declined, from 27.7% to 1.2%. Its 2019 interval extended below zero, a mathematical consequence of uncertainty around a difference between two averages. It does not imply negative consumption. It shows that this method could not precisely estimate a low prevalence in that survey.
The methods agreed on the direction of change, while disagreeing about the level. That is stronger evidence for a decline than a direct question alone, but it does not identify one estimate as a perfectly measured true rate.
The questions were similar, not identical
The comparison needs several distinctions. The list question concerned anyone in the household, whereas the dice question concerned the respondent. Direct questions included other giraffe parts as well as meat in the combined analysis. The researchers also altered the die between surveys, increasing the probability that participants were instructed to answer truthfully.
The paper explains these differences and the probability adjustment. They matter because the three results should not be treated as interchangeable measurements of exactly the same behaviour. The interviews also asked about the previous year, so recollection and willingness to follow the instructions remained part of every estimate.
The achievement was to collect converging evidence about a behaviour that is hard to observe openly, while protecting respondents from having to identify themselves as participants in it. The wider study of giraffe use and trade by Matilda Dunn and colleagues identified domestic consumption as a recurring concern, but did not measure its prevalence across households. Ruppert’s team addressed that different, more local question.
What changed around the giraffes?
The surveys coincided with expansion of Twiga Walinzi, a community-based giraffe conservation programme involved in ecological monitoring, outreach and collaboration around wildlife protection. Its presence offers plausible routes to change: people may encounter different information, social expectations or perceived chances that illegal activity will be detected.
The study did not include a comparable population unaffected by the programme, so it could not isolate the programme’s effect. Enforcement, livelihoods, food availability and the number or movements of giraffes could also influence access to giraffe meat. A decline in consumption would have a different conservation meaning if giraffes had simply become harder to find.
Repeating comparable, well-understood questions alongside wildlife monitoring would help separate those possibilities. Assessing respondents’ trust in each method would also clarify why estimates differ. The paper demonstrates a way to detect change in a sensitive activity; connecting that change to its causes and to giraffe survival is the next task.
Sources
Kirstie A. Ruppert, Carly C. Sponarski, Elizabeth O. Davis, Symon Masiaine, Lexson Larpei, Jesus Lekalgitele, Ruth Lekupanai, Johnson Lekushan, Joseph Lemirgishan, Daniel Lenaipa, Jonathan Lenyakopiro, Sebastian Lerapayo, Maxwell Lororua, Jenna Stacy-Dawes and Jenny A. Glikman. (2021). Use of specialized questioning techniques to detect decline in giraffe meat consumption. Journal for Nature Conservation. DOI: 10.1016/j.jnc.2021.126029. Read the full paper.
