In recent years, there has been a marked decline in the amount of hydrological data being collected in many parts of the developing world. This paper reviews some of the arguments in favour of the continued collection of hydrological data. The technical case is based on the need for long record lengths in order to obtain a reasonable estimate for the mean and variability of river flows. Estimates are presented for the dependence of the standard error on record length for the case of the main sample statistics, for reservoir design and for flood estimation. Typical values are presented both in general terms and for some specific historical time series. The current status of hydrological data collection in Africa is then contrasted with the situation in Europe before concluding with the results of several case studies of schemes in Africa and Asia whose operational performance has been affected due to insufficient hydrological data.