In the conduct of sample surveys in transport, there will inevitably be some level of non-response. The issues of non-response are fundamentally connected to the questions of reducing survey bias and increasing the accuracy of sample estimates. This is because non-respondents in sample surveys have often been shown to have significantly different characteristics from those of the respondents. These differences are in terms of the sociodemographic characteristics and, more importantly, in terms of travel behaviour characteristics. For example, non-respondents to household interview surveys tend to travel more than respondents to such surveys, because one of the main reasons for their non-response is that they are out of the house (travelling) when the interviewer calls to perform the interview. If due allowance is not made for this known difference, then estimates of total travel and travel distance will be under-estimated from such surveys. This will then result in under-estimates of emissions and fuel consumption in the survey area. Other types of non-response bias are associated with other types of survey method. This paper considers a range of issues associated with non-response to mailback travel surveys. It begins by reviewing some of the previous research on non-response in mailback travel surveys, in particular methods of correcting for this non-response. It then proposes a number of different behavioural mechanisms that might be associated with non-response patterns in those surveys, and suggests that previous methods of correcting for non-response in mailback travel surveys may be in error. The paper concludes by suggesting further steps that must be taken to fully understand the behavioural mechanisms underlying non-response in mailback travel surveys. (a).
Abstract