Page 60 - IJPS-7-1

Family size preference among women in Nigeria

Government of Nigeria, 2021), and persistence of high fertility preference of five or more children. Non-numeric response
(2.39%) was assigned the median value of 5 (IQR 4) as the IFS variable was not normally distributed. The median value
after assigning it to the non-numeric response was similar to the median without the non-numeric response. Although
Bongaarts (1992) argued that non-numeric responses are likely to be more than the population mean, it is anticipated
that assigning the median value to such responses will not distort the result of this analysis given that the non-numeric
responses were below 3%. Furthermore, in Dodoo and Seal (1994), non-numeric response in fertility preference was more
likely among wives who are younger than their husbands by over 10 years, junior wives in a polygynous union, and those
who have no or low levels of spousal communication on contraceptive use.
2.4. Independent variables

Drawing on past studies, socioeconomic and demographic characteristics of individual women were included as
explanatory variables. Age was categorized in 5-year groups except age 35 – 49 which was recategorized into one group
due to the few cases. A measure of place of residence was rural or urban. Region was measured based on the six regions
in Nigeria: North-central, Northeast, Northwest, Southeast, South-south, and Southwest. The number of respondents’
siblings was a measure of parents’ family size, an indication of intergenerational transmission of family size preference.
Other variables included in the analysis were age at first marriage, the number of co-wives, highest attained education,
year of first cohabitation/marriage, occupation, participation in household decisions, religion, and experience of child
death. Age at first marriage may indicate an inclination to a large or small family size (Mencarini and Tanturri, 2006).
Furthermore, women who entered marriage late may be affected by their declining fecundity if they prefer fewer children
or a desire to “catch up” if they prefer a large number of children (Mencarini and Tanturri, 2006; Upadhyay and Karasek,
2010). The number of co-wives (categorized into monogamy if one cowife, and polygyny for more than one co-wife)
is associated with the desired number of children, due to co-wife competition for the number of children (Feyisetan
and Bankole, 2009). The influence of education at all stages of fertility preference transition is shown in previous
studies (Bongaarts, 2003; 2020), year of first cohabitation/marriage was included, because exposure to similar situations
influences demographic and health behavior (Doctor, 2011). Women are likely to adopt the IFS prevailing among their
peers and in their immediate communities at the time of marriage. The occupation of the respondents was categorised into
five: not working, white-collar (those who worked in offices), sales/services, those engaged in agriculture as employers or
employees, skilled and unskilled manual workers, and others.
Participation in the household decision and attitude to wife beating was included in the analysis as indications of gender
ideology and norms. Participation in household decisions was derived from responses to four questions on the final say in
household decisions on respondent’s health, making large purchases, visits to respondent’s family and relatives, and how to
spend husband’s income. In each question, lone decision by the respondent, a joint decision with the husband or partner, a sole
decision by the husband, and a decision by others were the responses. The four questions were recoded. Participation alone
or joint participation with a partner was regarded as participation in household decision and coded 1, whereas decision by the
husband alone or others was regarded as participation in none and coded 0. The dummy recodes were used to generate an index of
participation in household decisions categorized as participation in none, participation in 1 – 2, and participation in 3 – 4. Missing
values in the participation in the household decision variable (0.24%) were dropped. Attitude to wife beating was measured with
five questions on whether beating a wife is justified when she goes out without telling her husband/partner, neglects the children,
argues with her husband/partner, refuses to have sex with her husband/partner, and burns the food. The response options were yes
(justified), no (not justified), and do not know. The response to the five questions was recoded to generate a composite index of
attitude toward wife beating. A response of no in all the five questions was categorized as not justified, a response of yes and do
not know in 1 – 4 questions was categorized as somewhat justified, and yes in all the five was justified.
Religious affiliation was measured as Catholic, other Christian, and Islam; traditionalist and others were dropped,
because they were few (0.46%). Experience of child death was a dummy variable; experience of child death was coded 1
and no child death 0. Child death influences reproductive preference (Upadhyay and Karasek, 2010). Added as controls
were partner or husband’s family size desire; past studies show that a partner’s desired number of children influences
women’s IFS (DeRose, Dodoo, and Patil, 2002; Izugbara and Ezeh, 2010). Other controls were the partner’s education
and age and the number of living children which are related to fertility preference (Westoff, Bietsch, and Hong, 2013).
2.5. Analytic approach

All the variables were checked for multicollinearity, none of the variables had a variance inflation factor (VIF) of five
and above, indicating that the multocollinearity is not a concern (Pallant, 2020). Description of the study population

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