Page 61 - IJPS-7-1

Ntoimo

was presented in weighted frequency and percentage. The distribution of the outcome variables by the independent
variables was done using cross-tabulations, and a Chi-square test was conducted to establish an association. Given
that the dependent variable was categorized into three, multinomial logistic regression was used for the multivariable
analysis. The weighting factor provided in the DHS was used for the descriptive and multivariable analyses to adjust for
representativeness. The svy command in Stata was used in both the cross-tabulation and regression analysis to adjust for
the complex survey design of the DHS. All the analyses were two-tailed, the level of statistical significance was P < 0.05,
and statistical analysis was conducted in Stata 13.0 for windows.

3. Results
3.1. Description of the study population
A description of the study population is presented in Table 1. The median ideal number of children in the study population
was 6 (IQR 4). For over 65% of the respondents, five or more children was their ideal number of children. This is a 5.7%
decline from 69.41% in 2008. The number of respondents whose IFS was 0 – 3 and four increased by 28.2% and 4.97%,
respectively, from 2008 (Figure 1). The median age of the respondents was 25 (IQR 9), with the majority aged 20–29.
Most of the women were in a legal monogamous union, married before age 20, and had an average of 2 living children.
The majority (61.5%) had five or more siblings. The distribution of the respondents by their places of residence showed
that 57% resided in rural areas, and over 50% of respondents were Muslims. A larger percentage of respondents (67%)
were working, particularly in sales and services jobs, and slightly above 65% attained any level of education. Many
respondents (35%) participated in no household decision, and about 71% did not see wife beating as justified for any
reason.
3.2. Findings from bivariable analysis

The distribution of the three categories of ideal number of children by the potential explanatory variables is presented
in Table 2. All the explanatory variables were significantly associated with IFS. The ideal number of children by birth
cohort (age) showed that a larger proportion of younger women aged 15–29 desired 5+ children compared to their older
counterparts aged 30–49. The ideal number of children for women who were living together with a partner was lower than
for those in a legal union. There was a variation by region, 0 – 3 ideal number of children was more in the Southwest and
South-south regions than other regions; IFS that corresponds with the country’s recommended 4 is more in the southern
regions than in the Northern regions. IFS for close to half of women in the urban area was <5 unlike those in the rural area.
By religious affiliation, a larger percentage of Catholics and other Christians had IFS 0 – 3 and 4 in contrast with Muslims.
IFS declined with a higher household wealth index. Over 70% of women who were not working and those who had
agricultural occupations had IFS of five and over. The majority of women in white-collar occupations had IFS of <5.
Similar to the pattern in the household wealth index, IFS declined with a higher level of education. The majority of women
who attained no education had IFS of five and over, whereas most of those who attained higher education had IFS <5.
Close to 80% of women who participated in no household decisions had IFS of five and over in contrast with 47% for
those who participated in 3–4 decisions. The majority of the respondents who justified wife beating for any reason also
had large IFS of 5+. Respondents who have 7+ siblings were in the majority among those whose IFS was 5+. About 81%
of women in the polygynous union had IFS 5+ compared to 61% of those in a monogamous union. Most of those who
have experienced child death had IFS 5+. A larger proportion of the respondents who married before age 20 had IFS 5+
compared to 46% for those who married later. IFS seems to be decreasing with the younger marriage cohorts. Those who
married between 2014 and 2018 had fewer IFS than their counterparts who married between 2008 and 2013.
3.3. Results from multivariable analysis

The results of the multinomial logit model predicting the relative risk ratio (RRR) of IFS by the potential independent
variables are presented in Table 3. An IFS of five or more children was the base outcome. Hence, the estimated RRR
reflects the effect of an independent variable on the likelihood of preferring 0–3 or 4 relative to 5 or more children, given
other variables in the model are held constant. Compared to the North-central which was the reference category, the risk
of preferring 0 – 3 instead of 5 was significantly lower in the Northeast and Northwest, but 1.49 and 4.22 times higher
in the South-south and Southwest regions. The relative risk of IFS of 4 was also significantly lower in the Northeast and
Northwest, and the Southeast regions, but higher in the Southwest (RRR 2.35 CI:1.88-2.94). Compared to urban residents,
the relative risk of IFS of 0 – 3 and 4 children instead of 5 or more was significantly lower for rural residents.

International Journal of Population Studies | 2021, Volume 7, Issue 1 55