Restriction of the data set to women aged between 15 and 49 years leads to problems of truncation bias when analysis of cohort-period fertility is carried out by marriage and motherhood duration. Because the minimum age for women included in the survey is 15 years, no information is available about the fertility experience of women aged less than 15 years in the five years immediately preceding the survey. In addition, the fertility experience of the 0-4 years duration group in the most recent 5-year period is also under represented for unions or first births taking place before the age of 15 years, and any analyses involving cumulative period fertility are affected since they necessarily involve cumulation of period fertility for the duration group 0-4 (Hobcraft, Goldman and Chidambaram, 1982).
The most important source of truncation however involves the use of the upper cutoff age. At duration 30-34 years since first marriage or first birth, only women who first married or had a first birth before they were 20 years would qualify for inclusion in the data set when the upper cutoff point is set at 49 years. Thus, at high durations distortions occur when comparisons of cumulative cohort fertility ( P) and cumulative period fertility (F) are made. The entire experience of the cohort 30-34 years duration at the time of the survey relates to a group of women who first married or had a first birth before age 20 (Alam and Cleland, 1981:25; Hobcraft, Goldman and Chidambaram, 1982). Due to teenage subfecundity, these women will often have lower fertility in the earliest period, although they will tend to have higher fertility at later durations. On the other hand, estimation of cumulative period fertility involves women with different ages at entry and a variety of fertility experiences, but the range of ages at entry is progressively reduced as the higher duration rates are used. Thus, for the highest duration groups, comparisons between P and F involve non comparable ages at first marriage or first birth.7
To overcome truncation bias as a result of the different ages at entry, analysis could be carried out separately for women according to quartiles of age at first marriage or first birth. Because of small sample sizes, however, especially at the longer durations, it was not possible to carry out separate analyses by quartiles of age at first marriage for all subgroups for which fertility estimates were made. To examine this issue, analysis by quartiles of age at entry was carried out only for the total sample.
Table 3.4 presents cohort-period rates by duration of marriage irrespective of age at first marriage8. The cumulative cohort fertility rate for married women at duration 30-34 years was just under eight, and cumulative period fertility to duration 30-34 years in the most recent period was 7.2. The cohort-period fertility rates in Panel A were only slightly higher than those for all women, except those for duration 0-4 years at end of period where annual cohort-period rates of over 0.30 among married women were several times higher than rates among all women aged 15-19 (less than 0.08). This shows that fertility was very high within the first five years of marriage, with childbearing starting almost immediately after marriage. This perhaps emphasizes the fact that though childbearing did not take place exclusively w ithin marriage, childbearing was the prime reason for marriage (Nukunya, 1975; Fortes, 1978:22; Caldwell and Caldwell, 1985).
progressively w ith duration, and though the average age at entry was below 20 years for all duration groups at survey, there is a difference of about three years in mean age at entry between wom en in the 30-34 years duration cohort and the younger cohorts:
Mean ages at first m arriage and at first birth for m arriage and m otherhood duration cohorts
D u r a t i o n g r o u p at s u r v e y (years)
0 - 4 5 - 9 10-14 15-19 20-24 25-29 30-34
Marriage Cohorts 18.3 18.0 18.0 17.6 17.3 16.5 15.3
Apart from the difference of about one child in completed fertility between all women on the one hand and married women on the other, the major difference between the two groups was the faster build-up of fertility in the early years among married women. By duration 10-14 years, however, the annual cohort-period fertility rate for married women was almost the same as for the corresponding age group among all women aged 25-29 years as shown in Table 3.1.
The coincidence of the cohort-period rates for all women with those for only married women was related not only to the vagueness of the definition of marriage used in this survey9, but also to the fact that by age 25 years or so, almost all women were married. Furthermore, celibacy had never been a bar to reproduction in Ghana. Reported fertility of all women was therefore essentially reported fertility of married women.10
9 The question asked was whether a woman was married or had a regular partner. 'Marriage' was not defined, and the determination of whether a woman's relationship with a man was marriage or not was left entirely to her.
10 The marital and fertility characteristics of women are summarized below: Between 97 and 99.8 per cent of all women aged 25 years and above had ever been married. As early as age 20- 24, almost 85 per cent of all women had ever been married. Thus the composition of the two groups of women in Tables 3.1 and 3.3 was essentially the same.
Proportions of women married and parous A G E All w om en (no.) % e v e r m a rrie d w om en % c u r r e n tly & c o n tin u o u s ly m a rrie d w om en % p a r o u s (all w o m e n ) % p a ro u s (ev e r-m arried w om en) % p a ro u s: (c u rre n tly & c o n tin u o u s ly m arried w om en) 15-19 1371 3 0.4 29.8 21.4 62.7 6 1.4 2 0 -2 4 1 2 2 0 84.6 68.8 75.9 87.7 8 5.6 2 5 -2 9 1011 9 7 .0 73.8 92.3 94.2 9 3.2 3 0 -3 4 802 99.1 67.7 95.6 95.8 9 6.3 3 5 -3 9 703 99.1 66.4 98.4 99.0 9 8 .9 4 0 -4 4 579 99.5 61.1 97.4 97.4 9 7.5 4 5 -4 9 439 99.8 65.8 97.7 97.9 9 8.6
Selecting women who were continuously and once only married introduces some variety into the composition of the sample under study. From 20 years of age, the proportion of women continuously and once married was between 61 and 74 per cent of the total sample of women. The substantially lower proportion was because of the introduction of restricting factors such as continuous marriage and absence of premarital births. On the other hand, when the three groups of women were examined in relation to proportions who had a child, it was seen that by age 20-24, whether a woman had ever been married or not made little difference to her becoming a mother. This further emphasizes the importance of parenthood as opposed to marriage in several African populations (Fortes, 1977:17, 22-28), and also explains why the cohort-period rates for married women were not much higher than those for all women.
Table 3.4 shows that women in all duration groups except duration 0-4 years experienced a decline in marital fertility in the five years preceding the survey, an observation also made with regard to fertility for all women. Although the decline in marital fertility five years before the survey looks questionable, there were not the usual troughs and peaks that would betray the existence of the Potter effect. There was no clear peaking of fertility rates at definite points before the survey. For example, for the duration cohort 20- 24, cumulative period marital fertility (F) for the periods 10-14, 5-9 and 0-4 years before the survey was respectively 5.99, 6.43, and 6.69 (panel C). Completing the cumulative period fertility to duration 30-34 for the periods 5-9, 10-14 and 15-19 years before the survey using the cohort-period rates of adjacent periods, shows that cumulative period marital fertility by duration 30-34 was over eight in each of the three periods.
For example, at duration 25-29, the cumulative period fertility rate was 7.53 in the period 5-9 years before the survey. If it is assumed that the cohort-period fertility rate at duration 30-34 in the period 5-9 years before the survey would be similar to that for marriage duration 30-34 (.107) in the period 0-4 years before the survey, then that rate could be used to complete the cumulative-period marital fertility rate up to duration 30-34 years in the period 5-9 years before the survey. That would give a rate of 8.07. Continuing with the same approach, the corresponding cumulative period marital fertility rate at duration 30-34 in the period 10-14 years before the survey would be 8.32, if the cohort-period fertility at duration 25-29 (.220) and at duration 30-34 (.107) in the period 0-4 years before the survey were used. Whether interest is in the cohort-period marital fertility rates, the cumulative cohort fertility rates or the cumulative period fertility rates, it is apparent that declines only occurred in the five years immediately before the survey for the longer duration cohorts (durations 20-24 and 25-29 years).
Table 3.4 Cohort-period marital fertility rates, and P/F ratios by duration group Duration group
of cohort
Years prior to the survey
at survey 0-4 5-9 10-14 15-19 20-24 25-29 30-34 A
0 - 4
Cohort-period fertility rates: fi
.341 .323 .340 .315 .311 .307 .334 5 - 9 .308 .317 .313 .315 .293 .293 10-14 .289 .303 .290 .302 .283 15-19 .252 .278 .286 .290 20-24 .183 .231 .293 25-29 .136 .220 30-34 .107 B 0-4
Cumulative cohort fertility: P
0.829 0.788 0.775 0.762 0.697 0.691 0.561 5-9 2.328 2.360 2.327 2.272 2.156 2.026 10-14 3.805 3.842 3.722 3.666 3.441 15-19 5.102 5.112 5.096 4.891 20-24 6.027 6.251 6.356 25-29 6.931 7.456 30-34 7.991 C 0-4
Cumulative period fertility: F
0.829 0.788 0.775 0.762 0.697 0.691 0.561 5-9 2.369 2.373 2.340 2.337 2.162 2.156 10-14 3.814 3.888 3.790 3.847 3.577 15-19 5.074 5.278 5.220 5.297 20-24 5.989 6.433 6.685 25-29 6.669 7.533 30-34 7.204 130-34 7.20 8.07 8.32 8.40 D P/F ratios 5-9 0.983 0.995 0.994 0.972 0.997 0.940 10-14 0.998 0.988 0 982 0.953 0.962 15-19 1.006 0.969 0.976 0.923 20-24 1.006 0.972 0.951 25-29 1.040 0.990 30-34 1.109 E 0-4 rppRatios of successive F's 0 947 1.053 0.927 0.987 0.987 1.088 5-9 1.029 0.987 1.006 0.930 1.000 10-14 1.048 0.957 1.041 0.937 15-19 1.103 1.029 1.014 20-24 1.262 1.268 25-29 1.618 F 0-4 RppRatios of successive fs 0.951 0.983 0.983 0.915 0.991 0.812 5-9 1.002 0.986 0.999 0.925 0.997 10-14 1.019 0.975 1.015 0.930 15-19 1.040 0.989 1.015 20-24 1.074 1.039 25-29 1.130
Notes: Lengths of exposure at duration 0-4 for the cohorts 0-4 to 30-34 were respectively :2.43, 2.24, 2.28, 2.42, 2.24, 2.25 and 1.68.
A slight increase in cohort-period marital fertility rates for duration group 0-4 years is observed from the period 25-29 years before the survey. The increase is interesting, and implies an almost certain reduction in the interval between marriage and first birth. Even though marriage for this analysis was indexed from the first postmarital birth, and women with premarital births were excluded from the analysis, the increase in the marital fertility in the first five years of marriage may reflect an increase in births which may have been premaritally conceived and reported as conceived after marriage. This may be related to the increase in the reported level of premarital sexual activity among young people in recent years (Tetteh, 1968), although Fortes (1975) argues that the evidence may be more anecdotal than empirical. A similar observation of an increase in marital fertility rates at marriage duration 0-4 years was made for Korea by Hobcraft et al. (1982), who concluded that the pattern was the result of increasing incidence of premarital conception. A corresponding pattern of an increase in fertility rates in the first five years of marriage observed in the Sri Lanka Fertility Survey, was, however, attributed to the declining incidence of very early marriages, with a consequent diminution of the effects of adolescent subfecundity (Alam and Cleland, 1981:27-28).
Some increase in age at first marriage had probably occurred in Ghana, but average age at first marriage in Ghana had traditionally not been very low, and there is also no evidence of a dramatic increase in the age at first marriage to the extent that has been observed in Sri Lanka. In Sri Lanka most childbearing takes place within marriage (Alam and Cleland,1981), whereas this has not been so in Ghana (Oppong, 1983; Tetteh, 1968). Attitudes about premarital pregnancies range from the situation among the Ga, where a premarital pregnancy or birth carries no moral stigma unless the mother refused to name the father of the child (Azu, 1974:33-34), to that among the Tallensi, where even this causes no problem as the child is
The P / F ratios in Table 3.3 also do not support any evidence of a large-scale
decline in m arital fertility. The ratios w ere close to unity, indicating
constant fertility over time. They d ep arted fairly considerably from unity only w here low cohort-period m arital fertility rates by the 30-34-year d u ratio n cohort had in tu rn led to low cum ulative cohort fertility (?) and co rre sp o n d in g low P / F ratios, especially d u rin g their early period of marriage. For example, the low P / F ratios of 0.940, 0.962 and 0.923 for the 30- 34 year duration cohort occurred at durations where very low fertility rates (Panel A) and equally low cum ulative cohort fertility rates (Panel B) were recorded. On the other hand, the high P / F ratios for duration cohorts 15-19 and above, at 0-4 years before the survey reflect the observed decline in fertility for the longer duration cohorts. The observed decline, is however, not convincing, since it did not extend beyond the five years im m ediately preceding the survey.
The above conclusions are confirm ed by the ratios of successive cohort- period fertility rates (rp i’s) and the ratios of successive cum ulative period fertility (R pi's )- The ratios are alm ost w ithout exception close to unity, except in the five years preceding the survey w here both sets of ratios
in clu d e som e v alu es significantly above unity. From the foregoing
exam ination, although fertility appears to have fallen in the period 0-4 years before the survey, declines in m arital fertility beyond the five years p reced in g the survey w ere eith er very m odest, non-existent or erratic. Because of small sam ple sizes and the lack of significant differences in fertility betw een socioeconom ic g ro u p s (Table 3.2), the analysis and discussion of m arital fertility is restricted to only the total sam ple and not extended to socioeconomic groups.
Estimates of cohort-period fertility by duration of motherhood are presented in Table 3.5. The table confirms the high level of fertility in Ghana. For the 30-34 years duration group, for instance, cumulative cohort fertility at 0-4 years before the survey (panel B) and period fertility at 0-4 years before the survey (panel C) were 7.8 and 6.8, respectively. The ratio of the two estimates and trends in cohort-period motherhood fertility rates (Panel A) suggest that there was a decline in fertility in the five years immediately preceding the survey.
Although the rates did not decline consistently for all cohorts, the pattern of decline is nevertheless clear. At duration 0-4 years at end of period, the motherhood fertility rate for the oldest cohort, 30-34 (0.855), was the highest at that duration. The rates declined consistently to 0.589 for the duration cohort 0-4 years. This shows that the rate of childbearing in the first five years immediately following a woman's first confinement had declined.
Like cohort-period rates for all women and for married women, the cohort-period motherhood fertility rates as well as the cumulative period and cohort motherhood fertility rates had been fairly stable over the years except in the five years immediately preceding the survey when a small decline appeared to have occurred, especially among longer duration cohorts. For example, for each duration group, the cumulative period fertility was lowest for the most recent period. The cumulative fertility for the duration group 20-24 at survey was 6.004, but 6.558 and 6.884 respectively, for the periods 5-9 and 10-14 years before the survey respectively. If the cohort-period motherhood fertility rate to duration 30-34 was completed for the periods 5-9, 10-14 and 15-19 years before the survey using the cohort- period rates of adjacent duration groups, the cumulative cohort-period fertility rate at duration 30-34 would be estimated as 6.81, 8.09, 8.42 and 8.19
Table 3.5:Cohort-period motherhood fertility rates and P/F ratios by duration group Duration group Years before the survey
of cohort ___________________________________ at survey 0-4 5-9 10-14 15-19 20-24 25-29 30-34 A Cohort-period motherhood fertility rates: f j
0 - 4 .589 .614 .627 .639 .687 .695 .855 5 -9 .282 .300 .309 .317 .313 .312 10-14 .266 .279 .294 .293 .299 15-19 .212 .248 .278 .265 20-24 .164 .201 .196 25-29 .101 .146 30-34 .061
B Cumulative cohort fertility: P
0 - 4 1.384 1.443 1.499 1.495 1.505 1.418 1.394 5 - 9 2.853 2.999 3.040 3.090 2.983 2.954 10-14 4.329 4.435 4.560 4.448 4.449 15-19 5.495 5.800 5.838 5.774 20-24 6.620 6.843 6.754 25-29 7.348 7.484 30-34 7.789
C Cumulative period fertility: F
0 - 4 1.384 1.443 1.499 1.495 1.505 1.418 1.394 5 - 9 2.794 2.943 3.044 3.080 3.070 2.978 10-14 4.124 4.313 4.514 4.545 4.565 15-19 5.184 5.553 5.904 5.870 20-24 6.004 6.558 6.884 25-29 6.509 7.288 30-34 6.814 £0-34 6.81 8.09 8.42 8.19 D P/F Ratios 5 - 9 1.021 1.019 0.999 1.003 1.152 1.097 10-14 1.050 1.028 1.010 0.979 1.089 15-19 1.060 1.044 0.989 0.984 20-24 1.103 1.044 0.981 25-29 1.129 1.027 30-34 1.143 E rpp Ratios of successive fs 0 - 4 1.042 1.021 1.019 1.075 1.012 1.232 5 - 9 1.064 1.030 1.026 0.987 0.997 10-14 1.049 1.054 0.997 1.20 15-19 1.170 1.121 0.953 20-24 1.226 0.975 25-29 1.446 F Rpp Ratios of successive Fs 0 - 4 1.043 1.039 0.997 1.007 0.942 0.983 5 -9 1.053 1.034 1.012 0.997 0.970 10-14 1.046 1.047 1.006 1.004 15-19 1.071 1.063 0.994 20-24 1.092 1.050 25-29 1.120
Notes: Length of exposure at duration 0-4 for the cohorts 0-4 to 30-34 were respectivelv: 2.35, 2.35, 2.39, 2.34, 2.19, 2.04, and 1.63.
104 There is thus no clear evidence that fertility had declined, apart from the drastic decline in the five years immediately preceding the survey. Except in the 0-4 years before the survey, the P /F ratios are also close to unity, and indicate fairly stable fertility over the period. The pattern of fertility observed for all women and for married women is thus duplicated for all parous women. In all cases, fertility appeared to have fallen only in the five years immediately preceding the GFS in 1979-80.
Because of small sample sizes and the lack of significant differences in fertility between socioeconomic groups (Table 3.2), the analysis and discussion of marital and motherhood fertility is restricted to only the total sample and not extended to socioeconomic groups. This because the cohort- period method employed allocates women and their births to a large number of cells, some of which end up having either too few women or too few births for any reasonable estimates of fertility to be calculated for socioeconomic groups. Analysis of fertility differentials is, however, carried out in a later section using multiple classification analysis. The dependent variable is total number of children ever born with age, marital and motherhood status restrictions introduced as covariates in order to capture the role of these variables in understanding socioeconomic differentials in fertility in Ghana.
3.3.4 Cohort-period fertility rates by quartiles of age at first marriage