June 1, 2009
The following table shows mean children ever born (CEB) from the Kenya censuses of 1962, 1969 and 1979 together with the times at which each cohort reached the mean age at childbearing, estimated to be age 30 years.
Table 1: Mean Children Ever Born for Kenya, Censuses of 1962, 1969 and 1979
Sources Kenya Census of Population, 1962 [15-Aug], Volume III, Table X, page 65; 1969 Population Census [24-Aug], Volume IV, Analytical Reports, Table 4.9, page 29; 1979 Population Census [24-Aug], Volume II, Analytical Report, Table 6.1, page 68.
The 1962 time-plot below shows (1) rising completed fertility from 1920 to 1935, (2) a leveling off after 1935, and (3) a consistent “up-down” pattern relative to the trend, with low values from age groups beginning with ages ending in “0”.
Figure 1: Time-plot of Children Ever Born for Kenya, 1962 Census
Re (3), age heaping is commonly more severe for ages ending in “0”. The up-down pattern would be accounted for by (a) lower fertility and/or (b) less complete reporting of CEB for women who misreport age.
Does the apparent rise in fertility before 1935 reflect declining completeness of reporting for older women? The four points in question come from women aged 70-74, 65-69, 60-64 and 55-59 years, respectively.
The plot below showing time plots for both the 1962 and the 1969 data suggests that this is not the case. The 1969 plot lies uniformly above the 1962 plot, even though the 1969 points represent data for older women. At the same time, the 1969 plot adds evidence for rising completed fertility and an attenuated up-down error pattern.
Figure 2: Time-plot of Children Ever Born for Kenya,
Censuses of 1962 and 1969
The plot showing data for all three censuses reinforces these conclusions. The 1979 trend is nearly identical to the 1969 trend displaced 10 years forward in time. It lies slightly above the 1969 plot, suggesting slightly higher completeness of reporting in the later.
Figure 3: Time-plot of Children Ever Born for Kenya,
Censuses of 1962, 1969 and 1979
census. Following cohorts between the two later censuses shows that completeness of reporting increased as women aged during the intercensal period, though this presumably represents more complete reporting overall in the later census rather than age.
Time-plotting these Kenya children ever born data provides several strong conclusions. (1) Completed fertility in Kenya was rising during 1920-1965 at a rate of approximately 0.3 children per woman per decade. (2) Completeness of reporting of children ever born is higher for later censuses, but there is no evidence of appreciable decline with increasing age of woman—even though the oldest women are 70-74 years of age. (3) Differential age heaping on ages ending in “0” and “5” correlates with completeness of CEB reporting and/or level of fertility to create a characteristic pattern of error. Note that there is no competing source of data for the first of these conclusions.
“Time-plotting Life Cycle Events,” by Griffith Feeney, May 30, 2009.
“Time-plotting Children Ever Born Data for the United States”, by Griffith Feeney, May 31, 2009.