The availability and quality of demographic data in developing countries are far from adequate. The introduction and improvements of techniques for estimating mortality from nontraditional sources of data and for correcting the shortcomings in traditional data are indispensable. Data on deaths in a period but with an unknown completeness of coverage is usually available through vital registration or from single or multi round household surveys. The growth balance method makes use of such data and provides an estimate of the extent of the under-registration of deaths. An extensive study of this method, regarding the effect of deviations from the underlying assumptions and possible modifications to overcome its shortcomings, is presented. This study reveals that the method is generally robust to patterns of mortality change similar to those in developing countries and also to recent changes in fertility. Possible modifications to allow for certain types of changes in mortality and fertility are also presented. A modification of the method to allow for the effect of migration is introduced and applied to actual data of Kuwait. The effect of differential under-registration of deaths on the method is discussed and a procedure to estimate this differential under-registration is proposed. This procedure is applied to hypothetical data as well as to data on Iraq. A model of age error and the general likely effect of this error on the growth balance estimate are discussed. Several practical considerations are also dealt with, such as the effect of graduating the age and death distribution before applying the method, the appropriate method of fit and an alternative formula that may be used. Finally, as an illustration of the interaction of several deviations from the underlying assumptions and the suitability of the technique and the adjustments procedures suggested, a general application using hypothetical data and actual data for Guinea is presented.