Low response rates are generally obtained in household expenditure surveys collecting data via diary keeping. It is reasonable to question the quality of resulting estimates of population characteristics. However, low response rates do not necessarily mean invalid estimates and the opposite may be equally true. This first may be the case particularly for surveys involving diaries, where the respondent records events during a time interval ahead. This implies the study is prospective meaning the choice of responding is made prior to recording events. Thus, what is to be observed cannot be assumed intervening the choice to respond or not. Dropouts during the data collection period may on the other hand be directly affected by realized values on study variables.
The prospective character of diary surveys is detailed upon showing the possibility to derive valid estimates using auxiliary variables. A main issue addressed is what auxiliary variables to use and how to use them in estimation. It is suggested economic and behavioral theory to be used for the modelling of response probabilities. An illustration with an example from the Swedish household survey is presented, and includes a model derived from economic theory suggested for modelling of response probabilities. Finally, the importance of having the response set covering the range of population values of study variables is shown. The implications of results for the design of diary-based surveys are discussed.