Associations between early family environment and ideal number of children

C hildhood family is vital for the formation of fertility preferences and attitudes towards family life. Yet previous studies mainly focused on structural aspects of the family, whereas the role of perceptions of one’s family in relation to fertility preferences remained largely understudied. This study examined how different aspects of the early family environment (i.e. relationships with parents, happiness in childhood, parental conflicts, family resources, as well as family structure) are related to the ideal number of children reported in adulthood. Using representative cross-sectional survey data from the Finnish Family Barometer 2015, the sample comprised men and women aged 20–45 with and without children. Poisson regression models indicated that a higher number of siblings was associated with a higher ideal number of children, whereas living in a single-parent household and overall negative perceptions of parents were related to a lower ideal number of children independent of various socio-demographic characteristics. Further analyses showed that these family characteristics were associated with the ideal number of children mainly among childless people but not among parents. The findings suggest that the early family environment is related to the formation of the ideal number of children, especially for childless people.

Since the Great Recession, the total fertility rate (TFR) has declined in several high-income countries, with Finland experiencing one of the most pronounced declines (TFR falling from 1.87 in 2010 to 1.32 in 2022, Official Statistics Finland, 2023. Fertility has declined among women of all age groups and different parities, but decreasing first births accounted for about 75% of this decline (Hellstrand et al., 2020). The tempo-adjusted TFR has also declined suggesting that this recent fertility fall is not fully explained by postponement of childbearing (Hellstrand et al., 2020). At the same time, the ideal number of children has declined among people from recent birth cohorts (i.e. 1985-1994) compared to those born in the 1970s and early 1980s, which was driven by substantially higher prevalence of people not wanting to have children from the recent cohorts Correspondence should be addressed to Kateryna Golovina, Helsinki Collegium for Advanced Studies, University of Helsinki, Fabianinkatu 24, 00014 Helsinki, Finland. (E-mail: kateryna.golovina@helsinki.fi).
Kateryna Golovina was supported by the Helsinki Collegium for Advanced Studies and the Finnish Cultural Foundation (grant number 00200993); Markus Jokela was supported by the Academy of Finland (grant number 345186). Data collection was supported by the Alli Paasikivi Foundation. All procedures performed in studies involving human participants were in accordance with the ethical standards of the relevant institutional research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. Informed consent was obtained from all participants included in the study. The data used in this study is available for researchers from the Finnish Social Science Data Archive (https://www .fsd.tuni.fi/en/). R syntax and the report on statistical analyses are available in a repository at the Open Science Framework (OSF): https://osf.io/e4qb2/. (Golovina et al., 2021). Perceived uncertain life situation and unwillingness to change current lifestyle emerged as the strongest factors behind the decision to postpone or not to have (more) children among Finns (Savelieva et al., 2022). This suggests that attitudes towards childbearing have changed among younger people and may have contributed to the recent fertility decline in Finland.
Therefore, understanding how ideal family size is formed and which factors contribute to it is crucial. Given that people in high-income countries typically underachieve rather than overachieve their fertility preferences (Beaujouan & Berghammer, 2019;Morgan & Rackin, 2010), the initially low ideal number of children is likely related to even lower fertility rates. Since childhood family is vital for a person's formation of values and attitudes towards family life (Balbo et al., 2013;Mönkediek & Bras, 2018), we examined how different aspects of the early family environment are associated with the ideal number of children using representative survey data from Finland. In addition to a well-studied family structure, we also focused on less researched perceptions of one's family, such as relationships with parents and happiness in childhood.

Terminology and theoretical framework
Ideal number of children (also known as ideal family size) is a preferred number of children a person would have with perfect fertility control and no constraints on fertility (Philipov & Bernardi, 2012). It can be asked in general (e.g. "What would be an ideal number of children for a family?") and reflects societal norms. In contrast, personal ideal number of children, the concept used in this study, is conceptually similar to a desired number of children, which is usually measured by asking how many children people wish to have (Philipov & Bernardi, 2012). Fertility ideals and desires are more abstract concepts reflecting one's attitudes and wishes about having children, whereas fertility intentions are concrete plans regarding childbearing (Philipov & Bernardi, 2012;Thomson, 2015). Fertility ideals and desires are considered the foundation for fertility intentions, which are realised within conflicting life goals and various constraints over the life course (Bachrach & Morgan, 2013;Miller, 2011). Ideal number of children is more stable over time than intended number of children, but it declines with age and may change after childbirth (Heiland et al., 2008;Kuhnt et al., 2017). In general, individual characteristics, attitudes and social ties are associated with changes in personal ideal number of children (Ray et al., 2018). Although the concept of ideal number of children was criticised for its low face validity and ambiguity (Philipov & Bernardi, 2012), the measure of personal ideal number of children is considered an important predictor of fertility behaviour (Bongaarts, 2001;Miller, 2011;Philipov & Bernardi, 2012;Yeatman et al., 2020).
Within the cognitive-social model of fertility intentions (CSM; Bachrach & Morgan, 2013), personal ideal number of children captures a cognitive image of the childhood family (Rackin & Bachrach, 2016). The CSM postulates that fertility behaviour results from both conscious, deliberate intentions and non-conscious, emotionally laden cognitions (Bachrach & Morgan, 2013). People have schemas-ideas, values, beliefs and patterns of thinking-they employ to guide their intentions and behaviour. In other words, childhood experiences in own family form schemas related to childbearing and family life, which, in turn, form the basis for fertility intentions. According to the CSM, positively valued schemas (e.g. positive image of parenthood) are more likely to be embodied in self-image and motivate behaviour than negatively valued schemas (Bachrach & Morgan, 2013).

Previous research
The idea that childhood family influences fertility preferences-an umbrella term for fertility ideals, desires and intentions-is not new: growing up in a larger family is associated with higher fertility preferences compared to a smaller family (e.g. Axinn et al., 1994;Buhr et al., 2018;Dey & Wasoff, 2010;Heiland et al., 2008;Rackin & Bachrach, 2016). However, most of the previous studies focused on the role of family structure-e.g. living with two parents in childhood or the number of siblings-on fertility preferences (Berrington & Pattaro, 2014;Buhr et al., 2018;Eschelbach, 2015;Heiland et al., 2008;Kotte & Ludwig, 2012;Rackin & Bachrach, 2016). The role of other aspects of the early family environment remains largely understudied. Little is known about how family resources, relationships with parents and overall family life experiences, that is, happiness felt in their childhood family, are associated with the ideal number of children. In light of the CSM, focusing on these emotionally laden experiences in childhood family is important, since more positively valued schemas of childhood life are more likely to be operationalised in intentions and realised in life.
Only one study examined overall family life experiences and fertility ideals: Kotte and Ludwig (2012) included childhood happiness as a covariate to their model and found that higher childhood happiness was associated with higher fertility ideals. Moreover, Rijken and Liefbroer (2009) focused on realised fertility and found that positive family experiences were associated with a higher number of children. Regarding relationships with parents, previous studies examined only available social support from parents in adulthood in relation to fertility ideals (Chipman & Morrison, 2015) and intentions (Tanskanen & Rotkirch, 2014). However, it remains unclear how the relationship with parents in childhood is related to the formation of the ideal number of children.
Finally, associations between socio-economic status (SES) and fertility preferences have been extensively examined (for meta-analysis, see Testa & Stephany, 2018), although childhood SES in fertility preferences was less investigated. Berrington and Pattaro (2014) found a weak association between higher parental SES and a higher desired number of children. It is possible that financial difficulties in childhood family are associated with lower fertility preferences through increased family stress. According to the family stress theories (Malia, 2006), financial difficulties might increase family stress, leading to lower family well-being and subsequently to negative views of family life by offspring.

Present study
This study examined how various aspects of the early family environment, including family structure, relationships with parents, family life experiences and family resources, are associated with an ideal number of children among Finnish adults of childbearing age. We examined these associations separately for participants with and without children, since early family environment may have different implications for a decision to have a(nother) child for childless people and parents (Balbo et al., 2013;Heiland et al., 2008;Kuhnt et al., 2017). Based on the CSM and previous research, we suggested four hypotheses: (1) Lower number of siblings, parental divorce and growing up in a single-parent household are associated with a lower ideal number of children; (2) Perception of at least one parent as not a good parent is associated with a lower ideal number of children; (3) Unhappy childhood and conflicts in family are associated with the lower ideal number of children; and (4) Long-term financial difficulties in the family during childhood are associated with the lower ideal number of children.

METHOD Data
We used representative cross-sectional survey data from the Finnish Family Barometer 2015, conducted by Väestöliitto, the Family Federation of Finland (Miettinen, 2015a), which includes data on the ideal number of children and early family environment. The survey was conducted online by KANTAR TNS Gallup Oyj using their panel of participants aged 20-50 years with no more than three children consisting of approximately 50,000 households representing the Finnish population (excluding the Åland Islands), among whom participants were recruited during the time of the survey. The sample size for the online survey was predetermined and the data collection was finished once the sample size was reached (n = 3180). The sample was homogenous by origin with 95.8% of the participants reporting Finnish as their native language.
For this study, the sample included participants of common childbearing age (20-45 years) to minimise the over-representation of people not wanting to have children among older participants (n = 2380). The participants who did not report their ideal number of children (n = 303), childless participants who or whose partner was pregnant at the time of the survey (n = 25), as well as those who had inconsistent responses on their number of children (n = 4) were excluded from the sample, resulting in n = 2048. After analysing missing values, the final sample was n = 1939.

Measures
Ideal number of children was self-reported using the question: "People have different expectations regarding their own family size. What is your own ideal number of children?". The item was coded as 0, 1, 2, 3, 4, 5 or 6 or more children.
Early family environment was measured with the following aspects: (a) number of siblings, parental divorce and living in a single-parent household during childhood (i.e. family structure), (b) perception of one's parents as "good parents" (i.e. relationships with parents), (c) childhood happiness and parental conflicts during childhood (i.e. family life experiences) and (d) long-term financial difficulties (i.e. family resources). The number of siblings was self-reported with the question "How many siblings did you have in your childhood home?" and the answers were recoded as having 0, 1, 2 or 3+ siblings. Divorce of parents was asked with "Were your parents divorced before you turned 18?" (0 = no and 1 = yes). Living in a single-parent household was asked with "Did you grow up in a single-parent household?" (0 = no and 1 = yes). Childhood happiness was asked with the question of whether participants considered their childhood as happy (0 = unhappy childhood, 1 = happy childhood). Parental conflicts were asked with the question whether there were serious conflicts between parents during childhood (0 = no, 1 = yes). Perception of parents as good was asked with two questions "Would you describe your mother as a 'good parent'?" and "Would you describe your father as a 'good parent'?" (0 = no, 1 = yes). The summary variable was computed by summing two items and ranged from 0 to 2 (0 = neither parent perceived as a good one and 2 = both parents perceived as good). In case of unanswered questions, it was treated as missing. Finally, financial difficulties were measured with a question of whether family had long-term financial difficulties in childhood (0 = no financial difficulties and 1 = long-term financial difficulties).
Background characteristics known to associate with fertility in Finland were included as independent control variables: age, gender, area of residence, income, partnership status and level of education (Miettinen, 2015b). Age was calculated by subtracting respondents' birth year from the year of the survey. Area of residence was coded as a continuous variable: 0 = capital metropolitan area, 1 = other city centre or suburb with over 50,000 residents, 2 = city centre or suburb with less than 50,000 residents, 3 = sparsely populated residential area in a city or other and 4 = a sparsely populated area. Income was coded as a continuous variable representing personal yearly net wages: 0 = less than 10,000€, 1= 10,000-20,000€, 2= 20,001-30,000€, 3 = 30,001-40,000€, 4 = 40,001-50,000€ and 5 = over 50,000€. Partnership status was coded as a binary variable (0 = not in relationship, 1 = in relationship).
Education was used as a continuous variable in the analyses (0 = comprehensive school, 1 = upper secondary, 2 = bachelor-level and 3 = master-level or higher).

Statistical analysis
Poisson regression was used to test all the hypotheses, given that the ideal number of children is a count variable and was not normally distributed based on the graphical assessment (histogram and q-q plot). To control the effect of slight over-dispersion, robust standard errors were used for parameter estimates in all models. Sample weights were included in all models for the total sample. First, the associations between the ideal number of children and all aspects of the early family environment were analysed with separate regression models in the total sample, first without interactions and then with an interaction term between each aspect of the early family environment and parenthood status. Then, we examined the associations between those aspects of the early family environment which had statistically significant interaction effects with parenthood status in separate models for a childless sample and a parent sample. Finally, a model including all aspects of the early family environment which had a statistically significant independent association with the ideal number of children was examined in the total sample. All models included background factors as control variables. Effect sizes were reported by incident rate ratios (IRR) with 95% confidence intervals. All statistical analyses were conducted in R version 4.0.3 (R Core Team, 2020) using package bucky (Tahk, 2019).
Handling of missing data. In the narrowed sample of 2048 participants, 14.0% did not report income, 1.3%area of residence and 0.8%-education. We conducted multiple imputations by chained equations using the package mice (Buuren & Groothuis-Oudshoorn, 2011) to impute missing values for income, area of residence and education level, using the other control variables (age, gender, partnership status, employment status and number of children) and all measures for the early family environment as predictors. If a missing value could not be imputed, the case was omitted from the analyses. After the imputation, the final sample comprised 1939 cases (0.7% of values were missing in parental divorce, 1.8%-living in a single-parent household, 5.4%-parental conflicts, 9.9%-perception of parents, 6.9%-childhood happiness and 6.7%-financial difficulties). All regression analyses were conducted using the pooled estimates of 20 imputed datasets using complete cases. Table 1 shows the unweighted descriptive statistics of the total sample, childless and parents samples after imputing missing values. The mean age of all participants was 34.9, 68.5% were women and 66.0% had a partner. The average number of children was 0.9. The most common ideal number of children was two (44.5%). Parental divorce, living in a single-parent household and parental conflicts were positively correlated with each other, with strengths ranging from moderate to strong (Table 2). Childhood unhappiness was moderately correlated with long-term financial difficulties in childhood (r P = .29, p < .001).

Independent associations and interaction effects
Number of siblings, living in a single-parent household and the perception of parents were associated with an ideal number of children in the total sample (Table 3, univariate models). For participants with three siblings, the ideal number of children was 1.32 times higher than for those without siblings. Living in a single-parent household was associated with 0.89 times lower ideal number of children compared with those not living in one. Perceiving neither one's mother nor father as a good parent was associated with 0.78 times lower ideal number of children compared to those who perceived both of their parents as good ones. There were no independent associations between the ideal number of children and parental divorce, childhood unhappiness, parental conflicts and financial difficulties.

Childless and parents samples
In the childless sample, the ideal number of children was 1.43 times higher for those with two siblings and 1.75 times higher for those with three siblings compared to those without siblings (Table 4, Model 1). Also, the ideal number of children was 0.77 times lower for those living in a single-parent household compared to those not living in one (Table 4, Model 2) and 0.74 times lower for those who did not perceive their childhood as happy compared to those who did (Table 4, Model 4). Perceiving only one parent as good was associated with 0.80 times lower ideal number of children in the childless sample compared to those perceiving both parents as good. Perceiving neither   parent as good did not reach statistical significance in either the childless or the parent samples, probably due to the low number of participants (n = 56 for childless, n = 61 for parents). There were no significant associations in the parent sample, but there was a trend for having 3 or more siblings (Table 4, Model 1) and perceiving neither parent as good (Table 4, Model 3).

Multivariable model
When all the variables with a statistically significant independent association were included in a multivariable model, living in a single-parent household was no longer associated with the ideal number of children (Table 3, multivariable model), partly due to the correlation between living in a single-parent household and the number of siblings. Perceiving neither parent as good was associated with 0.77 times lower ideal number of children, similar to the univariate model, whereas the association between the number of siblings and the ideal number of children became stronger.

DISCUSSION
This study examined the associations between the early family environment and the ideal number of children among Finnish adults of common childbearing age with and without children. The results were in line with most of our hypotheses. People who had two or more siblings had a higher ideal number of children compared to Note: CI = confidence interval; IRR = incident rate ratio. All models were adjusted for age, gender, area of residence, income, partnership status, level of education and number of children. Statistically significant (p < .05) associations are in bold. * p < .05. * * p < .01. * * * p < .001. Note: CI = confidence interval; IRR = incident rate ratio. All models were adjusted for age, gender, area of residence, income, partnership status, level of education and number of children. Statistically significant (p < .05) associations are in bold. * p < .05. * * p < .01. * * * p < .001.
people without siblings and the more siblings one had, the higher the ideal number of children. Those who lived in a single-parent household in childhood had a lower ideal number of children in adulthood compared to those who did not. Confirming our second hypothesis, the perception of both parents as not good was associated with the lower ideal number of children. Our third hypothesis was partly supported: childhood happiness was related to a higher ideal number of children only among childless people, whereas parental conflicts were not related to the ideal number of children. Regarding our fourth hypothesis, there was only a trend for the association between financial difficulties and a lower ideal number of children for childless individuals.
Our results support previous studies showing the association between a higher number of siblings and a higher fertility preference (Axinn et al., 1994;Buhr et al., 2018;Eschelbach, 2015;Heiland et al., 2008;Rackin & Bachrach, 2016). We also found that individuals having one sibling and those without siblings do not differ in their ideal number of children. This finding supports previous studies showing that wishing for a family size like one's own is stronger in larger families than in smaller ones (Buhr et al., 2018). However, due to the prevailing two-child norm in most Western countries (Sobotka & Beaujouan, 2014), people without siblings may still report two children as their ideal. This may be related to a belief that an only child would be lonely or even spoiled, the preference for certain gender composition and the idea that having two children ensures care in elder years (Brinton et al., 2018;Sobotka & Beaujouan, 2014).
Furthermore, our findings of the association between living with one parent during childhood and a lower 8 KARHUNEN, JOKELA, GOLOVINA ideal number of children in the total sample accord with previous studies (Heiland et al., 2008;Rackin & Bachrach, 2016). In contrast, parental divorce was not associated with the ideal number of children despite its strong correlation with living in a single-parent household. It is possible that divorce leads to more diverse family life outcomes compared to living in a single-parent household. Thus, parental divorce itself does not influence the formation of the ideal family size, but the related negative consequences are possibly related to it. To continue, the association between the perception of parents in childhood and the ideal number of children is a novel finding, but it conforms with the previous studies showing that parental emotional support is associated with higher fertility preferences (Chipman & Morrison, 2015;Tanskanen & Rotkirch, 2014).
It should be noted that the associations between the number of siblings, living in a single-parent household, perception of parents and childhood happiness with the ideal number of children were stronger for childless people, whereas for parents there were no associations. For parents, there was a trend towards a higher ideal number of children when having three or more siblings and a lower ideal number of children when perceiving neither parent as good. These findings suggest that the early family environment is associated with ideal family size more strongly for childless people compared to parents. This might reflect the development of the ideal number of children through the lifespan: before entering parenthood, the ideal number of children is formed based on own childhood experiences, whereas after entering parenthood, the family-related schemas are reconstructed based on personal experiences as a parent (Heiland et al., 2008;Kuhnt et al., 2017).
More broadly, our findings suggest that the early family environment is associated with attitudes related to fertility. We examine different aspects of the early family environment and provide novel evidence for the cognitive-social model of fertility intentions, namely that the emotional content of family schemas matters in fertility preference formation. We found that living in a single-parent household, the perception of one's parents as not good, as well as perceiving one's childhood as unhappy were all associated with the lower ideal number of children for childless people. In other words, more negative emotions related to the recollection of one's childhood were linked to a lower ideal number of children.

Strengths and limitations
Our main strength was examining several aspects of the early family environment, especially those receiving little attention previously. In addition to the well-studied family structure, we examined the role of relationships with parents, family life experiences and family resources in association with the ideal number of children. Taking these perceptions of the family into account is crucial to better understand the role of emotions in fertility preference formation. Also, by examining both childless people and parents, we were able to uncover that these groups differ from one another regarding how early family environment is associated with fertility ideals.
Our study has several limitations. First, the study is cross-sectional, so no conclusions about causality can be made. Since the early family environment was measured retrospectively, it might be influenced by recall bias and ex-post rationalisation. That is, both childless and participants with children might have re-evaluated their perception of childhood family experiences after growing up or becoming parents. Second, our sample was restricted to people with a maximum of 3 children, so the associations between the early family environment and the ideal number of children for people with a higher number of children are unclear. Third, the measures used for childhood happiness, parental conflicts and financial difficulties were crude. For example, childhood unhappiness may be caused by sources unrelated to family life (e.g. bullying and illness), a confounding effect which was not possible to control in this study. Finally, although the sample was weighted to correspond to nationwide population characteristics in age, gender, residence, number of children and educational level, it is possible that participants were selected on some other traits, for instance, related to the theme of the survey, which could create biases that weighting cannot account for.

Conclusion
Different aspects of the early family environment are associated with the ideal number of children, including family structure and the overall perceptions of childhood family. This suggests that family of origin provides a model for one's own family and shapes attitudes towards family life, especially before entering parenthood. Our results support the cognitive-social model of fertility intentions by providing evidence for the association between negative views of one's family and the lower ideal number of children. Future studies with qualitative data are needed (e.g. based on in-depth interviews or focus group interviews) to further explore how the experiences in own early family environment and overall perception of it are related to the formation of ideal family size and to disentangle whether other life events might confound this association. Also, studies with longitudinal data could further investigate other factors influencing people's feelings about family life and parenthood. Considering recent fertility decline and decreasing fertility preferences, it is even more important to study what motivates people to have children, not only what enables them to realise their fertility desires.
Manuscript received December 2022 Revised manuscript accepted July 2023