32: Reducing the Effects of Item Nonresponse in Transport Surveys

Over the past decade, survey researchers have defined the item nonresponse problem, describing the errors that are, or can be, introduced by item nonresponse. The literature has identified the primary causes of item nonresponse and suggested ways to prevent it. And, there have been substantial advances in standards for reporting and methods to “correct” item nonresponse. This paper reviews that past work and describes some of the current issues and areas of research that are especially relevant to transport surveys.

Item nonresponse has most commonly been used to describe the absence of some subset of data items from a completed survey questionnaire (Zimowski et al., 1997). The data items that are missing could be “ignorable missing data” which are omitted completely at random and thus do not result in biased estimates of the distributions of those items that are missing. In this case, the item nonresponse results only in reduced efficiency of the survey data collection effort. However, it is rarely the case that survey researchers know a priori that the missing data are truly ignorable or, rather, have a distribution different from that of the items that are reported. It is similarly difficult to know how large an impact the missing items might have on either estimates of the parameters of the distribution of that item, or on the estimates of the parameters of functions of that item.