5.2. Potential bias of the survey sample

The responsiveness of the target population can be biased. For example, beekeepers suffering higher colony losses might be more likely to respond to surveys than those suffering fewer losses, although other biases are also possible. For example, Brodschneider et al. (2010) found a potential bias in reported colony losses from the same region collected by different media. Respondents who returned a postal questionnaire from a beekeeping journal reported higher proportions of losses than those responding online or at a convention in this particular region. This suggests that responses of some groups (such as from visitors of a convention) may not constitute a statistically representative sample, being from a different population than the target population. In such situations a mixed sampling approach must be considered. This enables comparison of the outcome of the different sampling methods, which should be reported in the final report. This could be due to different experiences of the target groups and hence a different level of motivation of the beekeeper to respond. A randomized sample may suffer from the same non-response problems, but maybe to a lesser extent because the approach is more focussed on the individual beekeeper. It may be possible to overcome this problem by increasing the response rate via encouragement of broad participation in the survey and the use of reminders. Using mixed media surveys or surveys using random surveying of the population of beekeepers to achieve a more representative sample may help, but the underlying problem remains that of the association of the response rate with the underlying questions of interest to the survey organisers. Response rates can be rather low. Dahle (2010) quotes a 15% response rate from surveys sent out in a beekeeping journal. Van der Zee (2010) quotes a 7.5% response rate from beekeepers who were invited in a national beekeeper journal to participate in an internet survey as well and they found differences between the results of these surveys and those from random surveys. In Denmark, response rates of up to 33% have been achieved (Vejsnæs et al., 2010). In the Netherlands, an average of 22% of the beekeepers surveyed from 2006-2012 (van der Zee et al., 2012) responded to a mixed mode approach of a questionnaire included in the 2 national beekeeper journals. The letters could be sent back without charges or through an email with a personalized link to the questionnaire on the internet.