Should we design extended or straightforward questions for small stock when records are unavailable?

Abstract

Data from two closely related questions in a survey on rabbits is analyzed in order to determine whether results from these two groups of questions would yield similar results about numbers of rabbits kept by a household. One question seeks a straightforward answer about numbers of rabbits kept while the other group of questions breaks the question into several questions seeking numbers of rabbits disaggregated by sex and age. This is prompted by the fact that record keeping is not a very common undertaking in a small holder agricultural setting in Kenya and that in their absence, farmers may not recall precisely how many rabbits they own unless a headcount is performed. A paired sample t test is implemented to detect any significant under-reporting of rabbit numbers based on numbers from the straightforward question which we hypothesize would yield numbers far less than what is on the farm. The results show that such under-reporting is not serious enough. The conclusion is that between the two question modes implemented in the survey, the straightforward question is suitable as it is time saving when the survey data required does not include numbers disaggregated by sex or age of rabbits.