This paper is concerned with variations in the performance of the maternity care system between local authorities in England and Wales. The main aim is to determine the relation between the output of the system and the corresponding inputs of human and physical resources, with a view to identifying those aspects of the maternity services which are effective in improving performance. There are substantial variations in population characteristics and general environment between local authorities and due account must be taken of these factors in the analysis. Five measures of performance were chosen as follows: (i) perinatal mortality amongst the low weight (less than 2,501 g.) births, (ii) perinatal mortality amongst the higher weight (greater than 2,500 g.) births, (iii) the proportion of low weight births, (iv) overall perinatal mortality and (v) perinatal mortality corrected for birthweight distribution. The only measures of input available on a local authority basis are the proportions of institutional confinements for (i) the low weight and (ii) the higher weight births, and (iii) the proportion of low birthweight infants delivered at home who were eventually transferred to hospital during the first month of life. The characteristics and general environment of each local authority were represented by a set of 88 ‘descriptive variables’, representing features of the local authority not directly related to the maternity services. The performance measures are closely associated with the descriptive variables. About three-quarters of the total variation between local authorities in overall perinatal mortality can be accounted for by differences in the descriptive variables. This is in part a reflection of the close association between perinatal mortality and birthweight distribution, but the descriptive variables are also capable of accounting for a substantial part of the total variation when perinatal mortality is corrected for birthweight distribution. About four-fifths of the total variation in the proportion of low weight births between local authorities can be attributed to differences in the descriptive variables. This confirms that the maternity services as such do not bear a close relationship to birthweight distribution, which is determined largely by the characteristics of population and general environment. On the other hand, only about one-third of the total variation of perinatal mortality in both the low weight and higher weight births can be accounted for by differences in the descriptive variables. This leaves the remaining two-thirds of the variation in these performance measures to be explained by other factors, including differences in the delivery of maternity care. The relation between the measures of performance and input is then considered. When the descriptive variables are ignored, the input measures account for less than one twenty-fifth of the variation in perinatal mortality in the low weight births and for only a similar part of the variation in perinatal mortality in the higher weight births. When account is taken of the descriptive variables, the relationship between the measures of performance and input is weaker still. This demonstrates that the items concerning the maternity services currently available in the routine statistical returns, including the proportions of institutional confinements, bear little relation to performance. It is clear on general grounds that these items may provide only a very superficial indication of the performance of the maternity services and more sensitive measures of input are required. The discrepancies between the actual performance of individual local authorities and the performance predicted on the basis of the descriptive variables should provide valuable pointers to the choice of more relevant items.