Page 118 - IJPS-1-1

Using panel data to examine pregnancy attitudes over time

“very important.” Women reporting values 4-6 at both waves were classified as having
“consistently strong” pregnancy avoidance, women reporting values 1-3 as “never strong,”
women reporting values 4-6 at first wave but 1-3 at second wave were classified as “be-
came weaker,” and the opposite as “became stronger.” Preliminary analyses explored sev-
eral coding schemes and resulted in largely the same findings.
Our explanatory variables included, firstly, changes in union status, which was classi-
fied as: no change; stronger union (for those who got married, started cohabiting or dating);
and union dissolution (including divorce, dissolution of cohabiting union and transitioning
from dating to single).
Employment status had three categories: not employed, employed part-time (<35
hours/week) and employed full-time (≥35 hours/week). Change in employment status was
described using categories: “more work” (transitioning from no job to part- or full-time; or
from part-time to full-time) and “less work” (transitioning from a part- or full-time job into
unemployment; or from full-time to part-time).The survey did not assess whether women
who were not employed had been laid off, were on leave, or were not working by choice.
Our analyses also include the baseline characteristics of age, race/ethnicity, education,
parity, and the age of the youngest child in the household. Sample distributions of all va-
riables are provided in Table 1. Values are reported at each wave for time variant and at
baseline for time invariant ones.
We analyzed the data using descriptive statistics and multinomial regression. Firstly,
independent variables were tabulated against the two outcome variables. Multinomial re-
gression analyses with changes in trying (never trying as the base outcome); and pregnan-
cy avoidance (consistently strong as the base outcome) as the outcome variables were
conducted. The covariates that were not significant at 10% level were excluded from the
final models. We also tested whether education interacted with any of the other characte-
ristics, but the interactions were not statistically significant. The results were illustrated by
calculating fitted probabilities using average marginal effects at representative values

(Williams, 2012) for the outcome categories of interest; that is for having experienced a
change in pregnancy avoidance or in trying to become pregnant. The probabilities were
calculated for each explanatory variable by treating all respondents as though they had the
characteristic of interest, say they experienced a union dissolution, leaving the values of all
other variables as observed. The same calculation was subsequently conducted for each of
the categories of the explanatory variable; that is also to “no change in relationship status”
and “relationship became stronger”, for example. The average of these marginal effects
became the probability of having experienced a change in pregnancy avoidance or in try-
ing to become pregnant (Williams, 2012). We present the results as the predicted probabil-
ities with 95 per cent confidence intervals.

3. Results

Four percent of women decided to start trying to get pregnant and two percent stopped
trying without getting pregnant between baseline and follow-up studies (Table 2). Being in
a romantic relationship that moved to “the next stage” was associated with starting to try
to get pregnant more often (5 percent of women) than union dissolution (3 percent). Con-
sistently working part-time was associated with starting (5 percent) and stopping trying (3
percent). Five to six percent of women who were aged 25 to 29 years, had high school
education, had one child, had infants or toddlers, or were Black, started trying to get preg-
nant between the waves compared to two to four percent of women in the other categories
of these covariates.
Nine percent of women transitioned from strong to not strong avoidance and seven per-
cent from not strong to strong (Table 3). Women who got married or started dating or

International Journal of Population Studies | 2015, Volume 1, Issue 1 112