Volume 39 - Article 3 | Pages 61–94
Prospective versus retrospective measurement of unwanted fertility: Strengths, weaknesses, and inconsistencies assessed for a cohort of US women
|Date received:||15 Apr 2017|
|Date published:||06 Jul 2018|
|Keywords:||fertility, unintended pregnancy, unwanted pregnancy|
Background: Unwanted fertility is the key concept necessary to assess the potential impact of more perfect fertility control. Measuring this continues to be a significant challenge, with several plausible competing measurement strategies. Retrospective strategies ask respondents, either during pregnancy or after birth, to recall if they wanted a(nother) birth at conception; these reports are likely to be biased by an unwillingness to label a pregnancy or birth as unwanted (rationalization bias). Prospective strategies avoid this bias by questioning respondents prior to pregnancy, but reports are obtained months or years before pregnancy and so may not accurately reflect wantedness at conception.
Objective: We describe systematic errors associated with each strategy, show correspondence between strategies, and examine predictors of inconsistency.
Methods: Using the 1979 National Longitudinal Survey of Youth, we compare retrospective and prospective reports for 6,495 births from 3,578 women.
Results: The prospective strategy produces a higher percentage of unwanted births than the retrospective strategy. But the two reports of wantedness are strongly associated – especially for the second birth (vs. other births) and for women with stable (vs. unstable) expectation patterns. Nevertheless, discordant reports are common and are predicted by women’s characteristics.
Conclusions: Retrospective measures are biased by rationalization; prospective measures are biased when women change their expectations prior to conception. For practical and theoretical reasons, we argue that retrospective measurement is more promising for assessing wantedness.
Contribution: We highlight shortcomings in both approaches. Demographers may find ways to measure wantedness more accurately, but many of the measurement problems seem intractable.
Most recent similar articles in Demographic Research