Analyses were dependent upon self‐reported measures of alcohol consumption and so potentially subject to a degree of inaccuracy owing to reporting and recall biases 38 and measurement error 39. Furthermore, current drinking was derived from questions that concern consumption during the week prior to interview. It is possible that these provide a poor surrogate for true average weekly consumption, with quantity–frequency questionnaires tending to produce lower drinking estimates than graduated frequency questionnaires 38, which may be more effective at capturing episodic heavy drinking 40. Drinking diaries also show promise as a more accurate means of estimating consumption 41. Regardless, observed issues of longitudinal stability will apply to data obtained using any means of self‐reported questionnaire, with implications for all studies of alcohol, whether cross‐sectional or longitudinal in design. Finally, reported analyses show just one of many different dimensions of drinking behaviour, such as differences by drink type and episodic heavy consumption.
