4.2. Effort and costs in data accumulation

Manual data entry is costly and time consuming. Although Optical Character Recognition systems for automatic data entry are available, the fastest mode of data collection and accumulation is when beekeepers directly answer the questionnaire using an internet database. Furthermore, after the end of the data collecting period, the respondent can receive feedback and evaluate his/her losses or other aspects of beekeeping experience relative to data accumulated from other participants in the survey. Such systems can encourage participation by other respondents and so achieve a higher response, and are available nowadays for affordable prices. However this approach will fail to achieve a representative sample and may be a source of selection bias if the availability of internet access is associated with questions of underlying interest.

As none of the discussed survey methods is flawless, nowadays survey organisers often use a combination of data collection modes (a mixed-mode survey) to offset the weaknesses of one mode with the strength of another (Brodschneider et al., 2010; Topolska et al., 2010; Soroker et al., 2010, van der Zee et al., 2012). Data validity and accuracy can be improved by interviewer-administered questionnaires of a selected group of beekeepers, ideally randomly selected, by following up a postal or email survey by a telephone interview or offering the opportunity to clarify any points of difficulty. Such a limited follow-up can also sometimes reveal the kinds of bias incurred by the more extensive survey. However, unless follow-up is so limited as to produce little information, it is a very costly option.

Repeating annual surveys among the same group of beekeepers will provide information on time trends, either by simply sampling the same population or possibly by following the same sample of beekeepers through time (i.e. using a panel design), when this is feasible. Use of panel surveys does require some replacement of panel of members who are no longer available to participate in the survey, while trying to keep the sample representative. Comparing the results of self-administered surveys which are widely distributed and interviewer-administered surveys of a selected group of beekeepers will enable evaluation of the extent of colony losses area/countrywide and indicate the reliability or otherwise of a non-randomly selected sample. It may also identify any special cases that require further study, such as extreme losses in specific geographical areas that were overlooked by random sampling and in the event that identification of such areas is the purpose of the study.