Motivations for changing fertility plans and behaviours during the COVID-19 pandemic in Italy

ABSTRACT

This study accounts for the heterogeneous consequences of the Covid-19 pandemic on fertility plans and behaviours, by focusing on the motivations for suspended pre-Covid fertility plans and on those for new fertility plans that arose during the pandemic. We rely on unique data collected with a repeated cross-sectional survey conducted in April/May 2021 and October/November 2021 on a sample of young Italians (aged 18–34). We estimate a set of multinomial and logit models to examine some correlates of fertility plans and behaviours. Then, we provide a more qualitative analysis of the reasons behind the resulting patterns of associations. Changes in fertility plans and behaviours from pre-COVID intentions clearly show that the economic recession burdens unequally individuals and their opportunities to cope with obstacles to both work and family involvement. At the same time, those who started to plan childbirth during the pandemic, frequently cite as important motivations the increased opportunities to enjoy the family life, the more balanced work and family involvement, the higher share of domestic tasks in the couple, and the improved relationship quality. Our results suggest the need for exploring also positive channels through which the Covid-19 crisis had provided opportunities for planning new births.

KEYWORDS:

• Fertility behaviours
• fertility plans
• fertility motivations
• COVID
• Italy

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 For calculating the quota EUROSTAT data have been used. After completing the survey phase, the resulting sample (from an IPSOS panel of individuals) was compared with the reference universe. To better balance the slight differences found, a weighting procedure was therefore applied. The weighting procedure used is the Rim/Rake Weighting (also known as iterative proportional fitting) which estimates the individual weights by means of an iterative proportional fitting algorithm. The first iteration calculates the weights to match the totals of the first dimension (first weighting variable), the second iteration matches the totals of the second dimension, and so on. These steps for all dimensions are performed repeatedly unless convergence is achieved within an acceptable margin of error.

2 In this case, the polarity of the dummies has been reversed: in particular, the ‘perception of a possible individual/family income improvement in the future’ takes value 1 in case the individuals answer ‘much positively’ or ‘a bit positively’; regarding the experienced changes in financial situation, here the dummy takes value 1 in the case of an improvement.

3 These intervals have a confidence level of 83.55.

4 A question about the main reason for being inactive or unemployed is asked to those who declared they are not working or studying. Next to a set of standardized answers an open category ‘Other’ provide the possibility to better explain the reason for not working (open answer). Answers can be collapsed under two labels: ‘Difficulties related with finding an occupation’ and ‘Family related reasons’, while health-related reasons are cited only twice among all the unemployed and inactive respondents.

5 To support our decision, we performed other three models in which we included the three covariates for the income vulnerability one by one: even in this case only perceiving the individual income at risk has been found significantly related to the fertility plans. The path for the perceived family income at risk mirrors the one for the individual income.

6 We show similar results by plotting the predicted probability after running a multinomial model with an interaction between the employment condition and the perceived individual income as at risk. Because of fewer observations within each employment categories results are shown in the Appendix (Figure 2A), to support our interpretation.

7 This result is robust also to the inclusion of a control for the fact that the individual was cohabiting with the partner, before the beginning of the pandemic, or whether the cohabitations started during the pandemic.

8 The covariate has been added also with two other alternative specifications: as the dummy for having experienced a decline in income, and as a three categories variable distinguishing between increase, stability, and decline in income. With no one of these operationalizations the relationship with the dependent variable comes out to be significant.

9 A negative significative relationship is present using the opposite definition i.e. the expectation of an income decline.

10 The same result appears also by the predicted probabilities after running a multinomial model with an interaction term between the employment condition and the expected income increase. Because of the small sample size, the differences among the employment categories are not significant, while the trend for the stable and unstable workers is. Results upon request to the authors.