A Note on the Demographic Base of Household Formations in the 60's

Abstract

The purpose of this note is to call attention to an interpolation procedure used in Bureau of Census illustrative projections of the number of households from 1960 to 1980.1 The present writer turned to these series in order to estimate annual household formation rates in the 1960's. The data there indicate that the annual number of households formed might be very low in years ending July 1960, and July 1961, and might decline in 1965 and 1966. These figures for Series B, a medium high projection, are in the second column of Table 1. Note that the figures for 1960 and 1961 (less than 600,000) are substantially lower than the average rate in the three years ending March 1959 of about 840,000.2 A different interpolation procedure indicates that annual household formation rates should not be particularly low in 1961 and that the number should rise rather smoothly from 1960 to 1970. Although widespread use has not been made of the annual figures that can be derived from the Bureau of Census publication, mention is made of the low formation rate just prior to 1962 in both an article in Fortune and the new book by Bogue.3 The Bureau procedure here was to project the total number of households that would be in existence in July of the years 1960, 1965, and 1970. Figures for intermediate years obtained through an interpolation process, in which preliminary figures were based on the assumption that the year-to-year growth in number of households would follow the same pattern as the growth in the number of men who are in the principal ages of first marriage (age 20 to 24); the final figures shown in Table 2 resulted from the application of a three-year inoving average to these preliminary figures.4 (Emphasis supplied.) It is the argument of this note that the number of households formed (the growth or first differences of the number of households in existence) is more closely correlated with the number of men age 20 to 24 than with the growth or first differences in the number of men in this age group. The argument is that, other things being equal (particularly the relative portion of men in each single year of age within the group), there will be more marriages out of a large group of men age 20 to 24 than out of a smaller group.5 As a matter of fact, the number of men in this age group increases each year from 1960 to 1970 and, other things remaining equal, we should expect a larger number of marriages in each year.6 (The first diferences in number of men in this age group increase and decrease so widely over this period, that it is not surprising that a three-year moving average had to be applied in the Census interpolations.) Interpolations from 1960 to 1970 based on the number of men in this age group are shown in the third column of Table 1. However, (1) the proportion of men at each year of age in the group 20 to 24 changes over this period and (2) the marriage rate at each single year of age varies greatly over even this small range. Therefore, a better basis for interpolation might be an annual series of estimated first marriages based on s'ingle yeai of age marriage rates, and population by single year of age. In connection with other work, the writer had made annual projections of first marriages and remarriages in the 1960's based on the number of females in each year by single year of age and on marriage rates by single year of age developed by Jacobson for 1948.7 Approximately 66 percent of the total increase in new households from 1950 to 1965 is the increase in husband-wife house-