Abstract
A theoretical and analytic model of fertility intentions is proposed which treats “don’t know” responses and other uncertain responses as distinct from more firm intentions. Methodologically, these analyses show that “don’t know” responses need not be treated as missing data, but instead are both valid and meaningful responses. Furthermore, eliminating these uncertain respondents would have the negative effects of distorting across survey comparisons in intentions due to shifts in aggregate uncertainty, reducing the likelihood of accurately detecting shifts in fertility intentions, and lessening the representatives of the sample analyzed. Substantively, in conjunction with Morgan (1981), these results show that the sharp 1965–76 decline in the likelihood of intending more births at parities 2 through 5 occurred as women halted childbearing at minimal acceptable levels and postponed further childbearing. With time (or age), this delayed fertility became fertility about which the respondent was uncertain and, finally, fertility foregone. Since 1970, similar shifts are observed at parities 0 and 1, perhaps foreshadowing an increase in voluntary childlessness and one-child families.
