and a relatively small number of elderly people is likely to have more births than deaths even if the average couple does not have as large a family as is needed to maintain the population. On the other hand, an area with few young married couples and many elderly people may have a high net reproduction rate accompanied by more deaths than births. For such areas the crude birth and death rates would present an entirely different picture of the implicit level of population growth than would the net reproduction rate.

Limitations of the conventional net reproduction rate.--Conventional net reproduction rates, i.e., female age-specific net reproduction rates, such as those presented here, do not allow for all of the many factors which may affect fertility rates. Birth rates among women of each age may be unusually low during an economic depression as a result of a low marriage rate and the postponement of childbearing by many couples. They may be higher than usual during a period of prosperity or under the stimulus of certain wartime conditions. In recognition of such situations which cause annual fluctuations in a rate designed to reveal the secular trend in fertility, various refinements of the net reproduction rate have been proposed in the last few years.2 Unfortunately, the refined rates require data which are not available on a county basis. The conventional net reproduction rates reported in this report are the most refined type of measure available for small-area analysis of fertility.

basis. Life tables for geographic divisions in the period from 1930 to 1939 were used in lieu of life tables for individual States, since the latter were not available at the time of computation. Since age-specific rates of young children per 1,000 women were not available on a county basis, such rates were estimated by an indirect standardization procedure identical with the procedure ordinarily employed with birth statistics when only the total number of female births and the distribution of women by age is available. In the present case, county data were available on the number of female children under 5 years old and on the number of women by age. Age-specific rates of young children were available on a State basis from a sample of the 1940 census and were converted by the indirect standardization process into estimates for each county.

Reliability of the estimates.--The rates are necessarily approximate because of the use of indirect standardization, the use of geographic division life tables, and the estimation of completeness of enumeration of children. Similar problems would arise in the computation of rates based on birth statistics.

The method of indirect standardization made possible a complete set of estimated age-specific rates of young children even for counties having very few women of childbearing age. In the interest of obtaining a complete set of rates, the calculations were made without regard to the size of the population in any county. Most of the results for small counties look reasonable in the light of information available regarding the social and economic structure of their population. However, the smaller the population, the greater becomes the likelihood of large relative fluctuations in fertility from year to year, so that a 5-year net reproduction rate may be subject to a large element of chance variation. The accompanying table includes data on the number of white women of childbearing age in each county as an aid to the reader in determining for which counties the net reproduction rates might be most subject to such chance variation.

Color.--Persons of Mexican birth or ancestry who are not definitely Indian or of other nonwhite race are classified as white. This report does not present data for nonwhites (Negroes, Indians, Chinese, Japanese, and other nonwhite races).