Asking sensitive questions
By: Freya St. John, Lecturer in Conservation Science, Bangor University
Many approaches to conserving and managing natural resources depend upon rules which restrict access to resources. However, the existence of rules alone does not guarantee compliance and information on rule breaking behaviour is needed in order to inform the design of interventions. Unfortunately, estimating the prevalence of rule-breaking and understanding what motivates those involved constitutes a major challenge as understandably, rule-breakers are generally unwilling to discuss their motives, or admit to rule-breaking, for fear of punishment or shame. A number of indirect methods have been used to estimate rule breaking in conservation. For example, satellite imagery has been used to assess deforestation rates and fish landing statistics have been compared to quantities found on markets. However, such approaches tell us little about the characteristics or motivations of rule breakers.
Across numerous disciplines, face-to-face questionnaires are used extensively to investigate gather data from people. However, when the topic of investigation is sensitive, conventional questionnaires are particularly susceptible to two forms of bias: Social desirability bias and non-response bias. To address this, a number of specialised methods for asking sensitive questions have been developed within the social sciences (Nuno & St. John 2015). These include the Unmatched Count Technique (UCT) and the Randomized Response Technique (RRT), described below. Despite some recent applications, RRT & UCT have rarely been used to study illegal resource extraction, yet they have the potential to unlock an entirely new perspective on this global phenomenon. Freya St. John, Lecturer in Conservation Science at Bangor University presented these methods during two very well received sessions at the 2017 Wildlife Trade Symposium.
Unmatched count technique (UCT): When using UCT, respondents are randomly allocated into one of two groups, baseline and treatment groups. Participants in the baseline group are presented with a list of non-sensitive items whilst respondents in the treatment group are shown this same list with one additional sensitive item added to it. All respondents are asked to indicate how many items on the list apply to them, however, they never reveal which items are relevant to them. Differences in the means between baseline and treatment groups are used to estimate the prevalence of the sensitive behaviour (Nuno & St. John 2015). For example, to investigate the prevalence of hunting as a livelihood strategy a control card might ask respondents to state how many of the following occupations were done by people in their household: livestock keeping, agriculture, small business, and teaching. The treatment card would include the extra activity of wildlife hunting (See Nuno et al 2013). The simplicity and ease of use in areas of high illiteracy are two main advantages of UCT (Nuno et al. 2013). Unfortunately, it has been shown to be of limited use for very rare behaviours due to wide standard errors around estimates (Tsuchiya et al. 2007).
Randomized response technique (RRT): Depending upon the result of a randomising device such as a dice or coin, the randomised response technique first described by Warner (1965) requires respondents to either answer a sensitive question truthfully, or to provide a ‘forced’ answer. For example, respondents may be instructed to answer the question ‘Did you kill a duiker in the last 12 months’ truthful if they roll a die and it land on 1, 2, 3 or 4. However, if the die lands on 5, they are ‘forced’ to give the prescribed answer ‘yes’ and if it lands on 6, they give the prescribed answer ‘no’. The result of the randomising device, in this example a die, is never revealed to the researcher. In this manner truthfully ‘yes’ responses cannot be distinguished from forced ‘yes’ responses so incriminating information cannot be linked to individuals (St. John et al. 2012). Thus, the randomising device offers protection to the respondent, assuring them that their positive answers cannot be interpreted as an admission of guilt.
Article edited by: Melissa Arias and Nafeesa Esmail