1 Introduction

The fertility decline in Western societies has been the subject of research for decades (e.g. Lesthaeghe and van de Kaa 1986; Westoff 1983). Particularly in the early stages of the transition to lower fertility rates, metropolitan areas have been forerunners of fertility change (Lesthaeghe 2010; Walford and Kurek 2016). As a result, fertility is usually lower in urban than in rural regions (e.g. Kulu and Washbrook 2014; Trovato and Grindstaff 1980). Austria is no exception to this trend: After the baby boom of the early 1960s, when the total fertility rate (TFR) was approximately 2.8 children per woman, fertility declined sharply until the mid-1980s (Prskawetz et al. 2008; Sobotka 2015). Since then, the TFR fluctuated between 1.3 and 1.5, reaching its lowest point in 2001. The TFR of its capital, Vienna, was already below 2 in the 1960s, reaching a low of 1.2 as early as 1977. Since the early 2000s, the capital’s TFR has always been close to the national one (Zeman et al. 2019; Sobotka et al. 2012).

The decrease in TFR has raised questions about the changing circumstances of childbearing. One of these questions relates to the role of social values in influencing the desire of young adults to have children. To the surprise of many scholars, the two-child ideal has remained dominant in many Western countries (e.g. Sobotka and Beaujouan 2014), including Austria (Riederer 2005; Buber-Ennser et al. 2023). Persistently low fertility rates did not translate into fertility ideals below two. In addition, studies repeatedly found that women report wanting more children than they will eventually have (e.g. Beaujouan and Berghammer 2019; Harknett and Hartnett 2014). Since personal fertility desires are more dependent on individuals’ living conditions than ideals which reflect highly persistent collective norms and values (Thomson 2015), it is likely that the changing societal context and life circumstances contributing to persistently low fertility rates may have also reduced the desired number of children.

Against this background, the present study asks whether and how social change affected the desired number of children in Austria. We focus on the development of urban-rural differences in the desired number of children. While urban-rural differences in fertility (e.g. Kulu et al. 2007; Vobecká and Piguet 2012), the realization of fertility intentions (e.g. Riederer and Buber-Ennser 2019, 2021), or fertility timing (Buelens 2021; Riederer and Beaujouan 2024) have already been examined, research on urban-rural differences in family planning remains sparse, in particular regarding historical developments and the impact of social change. Addressing this research gap, the present study provides a comprehensive examination of urban-rural differences in the desired number of children. We employ the Austrian Microcensus/Labour Force Survey which repeatedly included a module with information on the desired number of children (i.e. the number of children already born plus further desired children) and provides representative samples of women over a long time span, ranging from 1986 to 2021. Our analysis contributes to the expanding body of literature that integrates spatial heterogeneity in fertility studies (e.g. Campisi et al. 2020; Riederer and Buber-Ennser 2019; Nisén et al. 2021; Clark et al. 2024). This approach is crucial for identifying the diverse preferences in regional contexts, which have also been differently affected by social change (e.g. by immigration). Understanding such heterogeneity is essential for developing effective family support.

After theoretically grounding fertility decline as part of general social change, we discuss potential reasons for differences in the desired number of children between urban and rural regions. Next, we detail how various social developments, ranging from late partnership (and family) formation to increased immigration, may be linked to changes in the desired number of children and differences between urban and rural regions. Following a description of our database and analytic strategy, we investigate urban-rural differences and the developments in the desired number of children among women of main reproductive age (20–40). In our analysis, we differentiate between the capital city of Vienna, other urban regions and rural regions in Austria. We employ multiple regression models to explain urban-rural differences and the decline in the desired number of children over time. Using decomposition analysis, we further examine the role of social change. Our final discussion addresses the observed regional convergence due to a decrease in the desired number of children in rural areas, emphasises the likely adaption of the desired number of children to lower realized fertility across the life course and highlights important implications of our study for future research.

2 Background

2.1 Social change and fertility: The Second Demographic Transition (SDT) framework

The most prominent theoretical framework relating fertility decline to social change is the SDT theory (van de Kaa 1987; Lesthaeghe and van de Kaa 1986). Although structural and technological factors are considered (Lesthaeghe 2010; van de Kaa 1994)Footnote 1, the SDT theory understands value change as the main driver of demographic behaviour. Goldstein et al. (2003, p. 481) even note that changes in fertility preferences are understood as causal driver of fertility decline in this literature. Following the seminal works of authors like Inglehart (1990) or Ariès (1980), the SDT theory assumes essential shifts from altruistic to individualistic values and, correspondingly, from the “child-centred” to the “couple-centred” couple. These shifts are related to gains in independence in general, and female empowerment in particular.

The educational expansion is widely understood as the major societal trend contributing to lower and later fertility. Economic development and skill-biased technological change demand longer periods of formal education and training. In turn, prolonged career preparation leads to later labour market entries, delayed couple formation, postponement of fertility to later ages, and higher probabilities of childlessness in modern service and knowledge societies. At least as important as the increase in educational attainment, however, is female empowerment: “The gender revolution in a broad sense appears to be one of the most important factors driving the trends associated with the second demographic transition” (Sobotka 2008, p. 211). Nowadays, women can perceive themselves to be childfree rather than childless. “As educational, employment and career opportunities opened for women, they could choose between motherhood and other activities” (Mills et al. 2011, p. 849), or opt for both (Goldin 2020), probably leading to a lower desired number of children.

Altogether, living conditions and lifestyles have changed dramatically since the 1960s in Europe. Induced by both improved economic conditions and increased individualization, more and more people of all ages are living in one-person households (Chandler et al. 2004; Snell 2017). Whereas prolonged education encourages extended periods of living alone or in cohabitation, female empowerment and individualism are also likely to contribute to spreading union instability, as women who are no longer economically dependent on their partner can afford to get divorced. Fertility after re-marriages occurs usually at higher ages (Thomson et al. 2012) and, as fertility desires adapt to life circumstances (Thomson 2015), these trends may also contribute to a lower desired number of children.

Finally, many Western European societies have become immigrant societies during the last decades (Castles et al. 2014; Mau and Verwiebe 2010; Zimmermann 2005). From the perspective of the SDT theory, increasing immigration substitutes decreasing fertility to sustain or even increase population size (Lesthaeghe 2010; Zaidi and Morgan 2017). In addition, immigrants also contribute to fertility. Especially, immigrants of non-European origin often have larger family sizes than the native population (Kulu et al. 2017; Toulemon 2004). It is thus likely that immigration also contributes to variation in the desired number of children.

Albeit not all premises and forecasts of the theory have been confirmed by empirical examination (Sobotka 2008; Zaidi and Morgan 2017), the SDT offers a purposeful framework for our research. First, it explicitly refers to ideational factors and fertility preferences. Second, it identifies a number of important trends that are relevant to explain changes in the desired number of children. Third, these trends are also relevant for the explanation of urban-rural differences.

2.2 Urban-rural differences in fertility

Numerous studies have shown that urban regions are characterized by lower fertility (e.g. Campisi et al. 2020; Kulu and Washbrook 2014; Sharlin 1986; Trovato and Grindstaff 1980) as well as higher levels of postponement and late fertility than rural regions (e.g. Buelens 2021; Kulu et al. 2007; Riederer and Buber-Ennser 2019). Indeed, “metropolitan areas are often emphasised as the forerunners of the SDT, where attitudes and norms towards the family are more flexible and alternative couple and fertility patterns develop quickly, whereas traditional family views and behaviours remain longer predominant in rural areas” (Riederer and Beaujouan 2024, p. 2).

Empirical evidence is generally supportive of this claim but somewhat mixed. In Belgium, urban industrial regions have already been forerunners of the first demographic transition and the SDT followed the same pattern of spatial diffusion (Doignon et al. 2020; Lesthaeghe and Neels 2002). In France, however, cultural and political developments have been essential for both demographic transitions. As a consequence, mainly influential cities like Paris or Bordeaux have been at the forefront of the SDT (Lesthaeghe and Neels 2002). Nevertheless, Doignon (2020) adds important evidence by showing that many other French cities experienced profound catch-up processes. Findings for Finland confirm the relevance of the level of urbanity of regions and the mediating function of its cultural and political characteristics (Valkonen et al. 2008). Walford and Kurek (2016) found supportive patterns of a spread of SDT from urban core regions to peripheral areas in England and Wales as well as in Poland between 2002 and 2012. Finally, Clark et al. (2024) observed urban-rural convergence between 1988 and 2018 in the US whereas fertility gaps remained significant in Canada (1990–2017).

Austria is characterised by general trends of the SDT: significant growth of single households, decreasing marriage rates, ascending divorce rates, an increasing proportion of non-marital births, a decreasing TFR and rising age at childbearing; and Austria has also become an immigrant society (Buber-Ennser et al. 2021). Most of these changes started between the mid-1960s and early 1970s and peaked in the early or mid-2000s. Herein, Vienna has been a precursing region in many aspects. As already mentioned in the introduction, fertility decline happened much earlier than in the rest of the country (Zeman et al. 2019). Previous research has further shown that women in rural areas are more likely to fulfil their childbearing plans than women in Vienna (Gisser et al. 1985; Riederer and Buber-Ennser 2021). Other indicators like the divorce rate (City of Vienna 2023a) or the share of one-person households (Riederer et al. 2021) also increased earlier (or faster) in Vienna.

In general, urban communities in Austria are characterized by a lower fertility rate and a higher average age at the birth of the first child than rural communities (Kulu 2006; Wisbauer and Klotz 2019). In addition, family attitudes are still more traditional in rural regions and small towns than in larger cities (Beham-Rabanser et al. 2019). Yet, in recent decades, some convergence between urban and rural regions can be observed in many of the discussed aspects. An exception concerns immigration. The proportion of foreign nationals differs greatly between urban and rural municipalities (Wisbauer and Klotz 2019) and it has risen significantly during the last decades, especially in large urban centres. In January 2023, more than 44% of the Viennese population was of foreign origin (City of Vienna 2023b). The proportion of births in Austria by foreign-born women has increased significantly since the 1990s. Although the average TFR of immigrant women in Austria dropped below 2, some immigrant groups are still characterised by a high TFR (Buber-Ennser et al. 2021).

2.3 Research questions and hypotheses

Against this background, our first research question concerns the existence and development of urban-rural differences in the desired number of children. Previous research repeatedly revealed fertility differences between urban and rural regions in Europe, as well as decreasing fertility and increasing regional convergence. The relevance of cultural factors and attitudes towards childbearing for the SDT suggests that the desired number of children should have also decreased. Cities have been at the forefront of the SDT, but the trend towards convergence means that rural regions might have caught up. We thus hypothesize that the desired number of children is lower in urban than in rural areas (H1a) and that the urban-rural difference decreased over time (H1b).

SDT theory allows us to identify a number of relevant trends that are related to the desired number of children and can be described by changes in the position in the life course (indicated by age and parenthood status), in living arrangements (household type), in education and employment, and in immigration. Our second research question assesses their role in explaining differences between urban and rural areas. We assume that these trends change the composition of the rural and/or urban populations, and thus, affect the development of regional differences in the desired number of children. Specifically, we hypothesize that these factors affect the desired number of children (H2a), are partly responsible for urban-rural differences (H2b) and contribute to the convergence in the desired number of children over time (H2c).

Position in the life course

Since personal fertility desires are repeatedly adapted to the current life situation, birth postponement may contribute to a decrease in the reported desired number of children. The desired number of children of younger women is expected to be mainly driven by highly stable societal ideals, in the context of Austria the dominant two-child norm (Thomson 2015). As the age at first birth increases, more and more women remain childless until their thirties, when they may revise their desired number of children. In general, the desired number of children is adjusted to the size of the own family (Kuhnt et al. 2017). Thus, the position in the life course (i.e. age) and parenthood status are likely to affect the desired number of children (H2a). As cities have been forerunners regarding trends of postponement, increasing age at childbearing and higher shares of childlessness, we assume that life course position and parenthood status will contribute to urban-rural differences in the desired number of children (H2b). However, increasing postponement and childlessness in rural regions may change the composition of the rural population over time and contribute to a convergence in the desired number of children (H2c).

Living arrangements

Postponement is also linked to changing living arrangements due to new lifestyles and later couple formation. As in many Western countries, single-households have become more widespread in Austria in recent decades (Fritsch et al. 2023). Living alone, however, is most pronounced in cities and associated with later couple formation and fertility postponement. If persons living alone report a lower desired number of children (H2a), the larger proportion of one-person households in cities is expected to contribute to urban-rural differences in the desired number of children (H2b). Furthermore, the declining urban-rural difference in the share of one-person households in Austria (Riederer et al. 2021) is expected to contribute to declining urban-rural differences in the desired number of children (H2c). Similar arguments can be brought forward regarding increases in non-marital cohabitation. Although nonmarital childbearing is increasing, married couples are still likely to have stronger family orientations and a higher desired number of children than cohabiting couples. Herein, authors argue that “marriage remains the preferred union type in which to have and raise children” (Guzzo and Hayford 2014, p. 547) and show that cohabiting individuals who understand cohabitation as a prelude to marriage are more likely to want children (Hiekel and Castro-Martín 2014).

Education and employment

Prolonged periods of education and increasing female employment are perceived as major causes of postponement and childlessness (Mills et al. 2011; Vasireddy et al. 2023). Consequently, highly educated and employed women are expected to report a lower desired number of children (H2a). As cities are characterised by a higher proportion of highly educated and employed women, they will likely contribute to a lower desired number of children in urban areas (H2b). If trends of increasing female education and employment spread from cities to rural regions, these trends are expected to contribute to a convergence in the desired number of children (H2c). Although there is still a negative gradient between education and fertility in Europe (Nisén et al. 2021), the association between education and the desired number of children may vary by region, economic development and welfare state regime. Yet, the link between education and fertility behaviour has become complex (Vasireddy et al. 2023) and effects of education on the desired number of children may have weakened in recent years.

Immigration

Although immigration to Austria and, in particular, Vienna is very heterogeneous, we expect that foreign citizens in Austria report a higher desired number of children than Austrian citizens (H2a), as immigrants from third countries may be affected by higher fertility ideals in their home countries. The increasing proportion of immigrants in cities, and particularly in Vienna, may thus counteract the expected general trend of a decrease in the desired number of children. In contrast to other aspects of social change discussed, which are expected to lead to a decline in the desired number of children in rural regions, increased immigration is expected to contribute to regional convergence by stabilising the desired number of children in urban regions.

3 Data and methods

3.1 Data

The data used in this study stems from a special module focused on families that has been added to multiple waves of the Austrian Microcensus/Labor Force Survey (LFS). This module collects information on individuals’ childbearing and fertility desires, and has been conducted approximately every five years since 1986. The Austrian Microcensus data is a representative survey, using a one per cent sample of Austrian households. In early waves, data were collected through face-to-face interviews. Later, the mode switched to computer-aided telephone interviews. In recent years, respondents have mainly been interviewed via computer-assisted web interviews. Participation in the core Microcensus is compulsory, whereas participation in additional modules is voluntary. Nevertheless, the response rates are significantly higher than those of other large-scale Austrian social surveys. Furthermore, other social surveys that collect fertility information typically span only a few years and lack regional information beyond the provincial level. Thus, the census data stands out not only for its long time horizon and representativeness but also for the provision of information on urbanization (to be elaborated on later).

Our analytical sample comprises 28,426 women in their main reproductive ages (i.e. 20 to 40 years), born between 1945 and 2001, who provided answers on questions regarding the desired number of children by themselves (i.e. proxy interviews were excluded). We utilize seven out of the eight survey waves that include the family module to analyse the development of urban-rural differences in the desired number of children between 1986 and 2021. The 2006 wave is excluded due to different wording of survey items and data inconsistencies. All the waves include weights provided by Statistics Austria adjusting for sex, age, size of household, nationality, federal state, and employment status. In addition, we consider weights adjusting for cohort and parity to account for a potential fertility bias in the data.

Our dependent variable is the desired number of children. The measure is calculated by combining the number of biological children ever bornFootnote 2 with the number of further desired biological children (step-, adoptive-, and foster children are not considered). For the latter, respondents are asked about the number of children desired in the long termFootnote 3. They have the options to provide a specific desired number, to refuse to answer, or to indicate uncertainty by responding with “I do not know”. The latter were asked in a second question to provide a rangeFootnote 4. If they did so, we calculate the mean of the two values; otherwise, it is assumed that undecided respondents do not want more children following Sobotka (2009). Our measure has two main caveats. First, we do not know whether previous births (number of children ever born) have been desired. Second, future events may change the desired number of children. Measures that are related to concrete intentions or expectations regarding a specific birth or a specific, short period of time may be more reliable. Nevertheless, such measures do not allow to assess the overall desired number of children (see Thomson 2015). Notably, the field phase of the 2021 survey coincides with the COVID-19 pandemic, particularly a period characterized by strongly eased containment measures but increasing numbers of infections towards the end of the field phase. A prior study has shown that only a minority of respondents altered their fertility intentions due to the pandemic (Buber-Ennser et al. 2023). Moreover, these changes primarily pertained to timing rather than the number of children. Consequently, we posit that these changes are unlikely to substantially bias our measure.

Our main variables of interest are an urban-rural typology as well as the year in which the survey was conducted. To analyse urban-rural difference in the desired number of children, we use available categorisations of women’s place of residence, differentiating between a low, medium or high degree of urbanization (details are elaborated in Table A.2). Unfortunately, there have been methodological shifts in the classification of the urbanization variable across different survey waves. In 1991, 1996 and 2001, the degree of urbanization has been measured at the municipal level, whereas the classification is based on 1 km2 grid cells in 2012, 2016 and 2021. Checks revealed that the allocation of some locations to the medium or high category varied across the survey waves. Therefore, we collapsed these two categories. For 1986, we had to approximate the indicator using available data on municipality size, with rural areas defined as those with fewer than 5000 inhabitants and urban areas as those exceeding this population threshold. Finally, we decided to distinguish Vienna, being the sole metropolis of the country, from other urban regions, resulting in a variable with three distinct categories: rural regions (category: low urbanization), urban regions (categories: medium to high urbanization), and Vienna.

In our analyses, we further consider a number of sociodemographic indicators as explanatory variables. These include age group (20–29, 30–40), parenthood (yes, no), household type (married couple household, cohabiting couple household, one-person household, otherFootnote 5), citizenship (Austrian, non-Austrian), level of highest educational attainment (low, basic, intermediate, highFootnote 6), and activity status at the time of the interview (employed, student, otherFootnote 7).

In the pooled sample, most respondents are from urban areas (44.9%) and Vienna (21.4%), with a smaller segment from rural areas (33.7%) (Table 1). The sample is slightly skewed towards women aged 30–40 and those who already have children. The majority live in couple households, hold Austrian citizenship, have basic (i.e. lower vocational) education, and are employed. The sample sizes of the various waves vary between roughly 2100 and 7200 respondents (Table A.2). During the last 35 years, we observe notable changes: the share of women aged 20–40 living in Vienna increased whereas the proportion living in other urban regions decreased. The percentage of respondents living in rural areas was rather stable and only slightly increased between the mid’ 1980s and 2021. The age structure of interviewees changed in the sense that women in their thirties comprised 46% in 1986 but 61% in 2021. Further, women were more frequently childless or child-free in 2021 than in 1986. In addition, there is a shift towards more single-person households, more non-Austrian citizens, and higher involvement in education and employment. Lastly, educational attainment has seen a considerable increase, marked by a decline in respondents with lower education levels and a substantial rise in higher education levels.

Table 1 Sample characteristics (in %)
Full size table

3.2 Analytic strategy and methods

We begin our analysis with a descriptive examination of the development of urban-rural differences in the desired number of children in Austria and show how regional trends deviate from the national trend. Our further statistical analysis comprises the following steps. In the first step, we employ regression analyses to investigate whether the desired number of children is lower in urban areas (H1a), whether our explanatory factors are related to the desired number of children (H2a) and whether these factors explain the differences between urban and rural areas (H2b). In this step, we focus on three selected survey years covering the beginning, the middle, and the end of our observation period (i.e. 1986, 2001, and 2021). In the second step, we pool data from all available survey years to elucidate changes over time, testing a potential decrease in urban-rural differences (H1b). Assuming convergence, trends are expected to differ by region type. Therefore, we also examine separate regression analyses for each region type (i.e. rural regions, urban regions, Vienna). They allow us to assess whether changes in the composition of regional populations contributed to changes in the desired number of children (H2c). In our third step, we further extend this analysis. Decomposition models based on data from the first and the last survey wave (i.e. 1986 and 2021) are used to assess the extent to which compositional changes in single sociodemographic characteristics contributed to changes in the desired number of children. The results should inform which social trends were most relevant for the assumed convergence between urban and rural regions.

In regression models in the first and the second step of the statistical analysis, we follow a stepwise modelling procedure. We first estimate a regression model including only the urban-rural typology or the survey wave, before adding other explanatory variables in a second regression model. A comparison of the coefficients in the first and the second models allows us to assess the extent to which differences in other explanatory variables can explain variation in the desired number of children between urban and rural regions or between survey waves, respectively. Wald tests are used to assess the statistical significance of the differences in the coefficients. Furthermore, we report heteroskedasticity-robust standard errors in all our regression models to account for the presence of heteroskedasticity as indicated by the Breusch-Pagan test.

While the first step only considers three specific time periods, we provide extensive models that simultaneously assess urban-rural differences and their changes over time in the second step. Specifically, we use an interaction between the urban-rural typology and survey wave to provide statistical tests for changes in urban-rural differences over time. This allows us to confirm previous regression results using the broadest available data basis (N = 28,426 women). Finally, motivated by the increasing age at childbearing, we examine a three-way interaction with age group to gain insights into the role of postponement for urban-rural differences in the desired number of children.

In the third step, we employ the Kitagawa-Blinder-Oaxaca decomposition to analyse the change in the mean desired number of children between 1986 and 2021 (for details, see Jann 2008). Decomposition analyses break down the difference between the two survey years into an explained and an unexplained component. The explained component comprises the portions of this difference that can be attributed to changes in each explanatory variable in the statistical model. Deviating from the procedure of the regression analyses, we include all our explanatory variables in the first decomposition model. In the second decomposition model, we exclude parenthood status. Factors such as partnership status or employment status likely affect parenthood status. Thus, influences of these factors might be overseen if parenthood status is included in the decomposition model.

4 Results

4.1 Descriptives

The urban-rural typology provides a nuanced perspective on trends in the desired number of children that diverge from national averages (Fig. 1). In 1986, there was a stark difference in women’s desired number of children based on their place of residence. Women in rural regions aspired to have on average 2.3 children, while those in urban areas and in Vienna wanted 1.9 and 1.7 children, respectively. Over time, a steady decline in the desired number of children is evident among women in rural regions, reaching the 1986 urban average of 1.9 children by 2021. During the same period, women’s desired number of children also decreased in urban regions, however less markedly, by only 0.2 children. Viennese women consistently reported the lowest numbers throughout all the survey years. These figures remained relatively stable, albeit somewhat fluctuating over time, and slightly decreased from 1.7 children in 1986 to 1.6 children by 2021. In summary, we observe convergence between regions along the urban-rural axes. Nevertheless, there remains a distinct gap between urban and rural areas. Contrary, the differences in the desired number of children between urban regions and Vienna almost completely diminished, as visualised by the overlapping confidence intervals.

Fig. 1
figure 1

Mean desired number of children by urban-rural typology and national trend (1986–2021). (Figures show means and corresponding 95% confidence intervals. Data: Austrian Microcensus/LFS 1986–2021 (weighted))

Full size image

4.2 Estimation results

Linear regression models of urban-rural differences in 1986, 2001 and 2021 confirm descriptive findings (Table 2). First, across all models, the highest desired number of children is found for rural regions and the lowest for Vienna, as indicated by negative AMEs for urban areas and for Vienna, which are largest in size for Viennese women. Second, there is a decline of urban-rural differences over time: The differences between rural region and urban regions as well as between rural regions and Vienna are largest in 1986 (−0.361*** and −0.635***, respectively) and smallest in 2021 (−0.205*** and −0.288***, respectively). Supporting hypothesis H2b, urban-rural differences decrease when further explanatory variables are included, therein suggesting that regional differences in these characteristics explain a relevant part of urban-rural differences in the desired number of children (models 2, 4 and 6 in Table 2). For instance, the difference of −0.635 children between rural regions and Vienna in 1986 would decline to −0.471 children if women in both regions would be identical with regard to the included socio-demographic characteristics (model 1 vs. model 2). Although rural-urban as well as rural-Vienna differences decrease, they remain highly statistically significant in 1986 and 2021 (p < 0.01). Remarkably, coefficients measuring urban-rural differences are no longer statistically different from zero in the model for 2021, indicating that differences in explanatory variables almost fully account for urban-rural differences. In 2021, the differences between rural regions on the one hand and urban regions as well as Vienna on the other hand, decline from −0.205*** to −0.073 and from −0.288*** to −0.091, respectively (model 5 vs. model 6).

Table 2 Estimated coefficients of linear regression models for the desired number of children, survey years 1986, 2001, and 2021, AMEs
Full size table

In line with hypothesis H2a, the associations between our explanatory variables and the desired number of children are largely as expected (Table 2): Women aged 30+, non-parents and women who live alone report lower numbers than women below age 30, parents and married women, respectively. These differences seem to increase slightly over time. Cohabiting women do not differ from married women in 1986 but show a lower desired number of children than married women in later years. Regarding employment status, results indicate that non-employed women report a higher desired number of children than employed women, with the association being no longer statistically significant in 2021. Finally, women with foreign nationality report a higher desired number of children than Austrian citizens.

Education is also linked to the desired number of children. However, findings are more complex. We observe a U-shaped-relationship in 1986 and 2001, indicating a higher desired number of children among both low and high educated women. In 2021, we observe a positive association between education and the desired number of children, as women with an intermediate or high level of education report larger numbers than women with lower or basic education. However, this pattern of results is only revealed in multiple regression models. Bivariate analyses show a negative educational gradient in all survey years except 2021, indicating that other covariates (age, parenthood etc.) are responsible for the lower desired number of children of more educated groups (results are available upon request).

Separate analyses of the development of the desired number of children by type of region (Table 3) confirm that there has been hardly any change in the desired number of children in Vienna (model 5), whereas we can observe a modest decline in urban regions (model 3) and a strong decline in rural regions (model 2). Findings also support hypothesis H2c: The decline in urban regions (model 3) seems to be almost completely explained by the explanatory variables, as hardly any differences between survey waves remain statistically significant in model 4. Changes in explanatory characteristics also account for large parts of the tremendous decline in the desired number of children in rural regions. The decrease of 0.417 children (model 1) between 1986 and 2021 is reduced to 0.158 in model 2.

Table 3 Estimated coefficients of linear regression models for the desired number of children, pooled sample 1986–2021, AMEs
Full size table

Analyses of pooled data corroborate previous findings (for details, see Table A.3 in the appendix). In particular, statistical tests for differences in urban-rural differences between 1986 and the other survey years confirm the decrease of urban-rural differences over time.Footnote 8 Furthermore, the models emphasise that rural regions are distinctively evolving. The introduction of a three-way-interaction between the urban-rural typology, survey year, and age group additionally reveals that changes in the desired number of children in rural regions are mainly driven by developments in the group of women aged 30–40 (see Figure A.1).

Further supporting hypothesis H2c, decomposition analyses based on data for 1986 and 2021 confirm that decreases in the desired number of children in both rural and urban regions are largely explained by compositional differences between survey years. According to these analyses, models including all explanatory variables account for 63 and 70% of the decline in rural and urban regions, respectively (models 1 and 3 in Table A.4). As the desired number of children is almost identical in 1986 and 2021 in Vienna, decomposition analyses for Vienna are less meaningful.

Given the unique role of rural regions in explaining the convergence of the urban-rural divide in the desired number of children, the further discussion of our decomposition analyses will focus on rural regions (Fig. 2). The main factor for the decline in the desired number of children is the decrease in the proportion of parents over time, followed by educational change (mainly due to decreasing shares of lower educated women). These two factors alone account for more than half of the decline in the desired number of children, leading to a decrease by −0.24 children (total decrease: −0.42 children). Excluding parenthood from the analysis, we can observe that changes in activity type (increasing female employment) and in household type (decreasing shares of married couples)Footnote 9 become more important (Fig. 2 and model 2 in Table A.4). This might indicate that both affect the development of the desired number of children through increasing periods of non-parenthood.

Fig. 2
figure 2

Main results of the decomposition analyses on changes in the desired number of children in rural regions (1986 versus 2021). (Decomposition analyses are shown in more details in Table A.4 in the appendix. Data: Austrian Microcensus/LFS 1986 and 2021 (weighted))

Full size image

Findings for urban regions are similar to those for rural regions. However, there is one remarkable difference. Decomposition analyses suggest that in urban regions and in Vienna, the increasing share of people with a foreign nationality counteracted the decline in the desired number of children (models 3–6 in Table A.4), supporting our assumption that increased immigration may have contributed to stabilise the desired number of children in cities.

Decomposition analyses indicated that increasing periods of non-parenthood contribute most to decreases in the desired number of children and regression analyses emphasized the change in the group of women aged 30–40. To investigate this further, we scrutinized the components of the desired number of children, i.e. the number of children already born and the number of further desired children, by age group (see Figure A.2 in the appendix). Austrian women generally experienced a decline in the number of children already born. This decline, however, tends to be largely offset by an increase in the number of further desired children. This compensation does not occur for the 30–40 age group of women residing in rural areas. Among these women, increases in the number of further desired children do not counterbalance the decline in children already born. Considered together, these three findings suggest that postponement of childbearing may lead to an adaption of the desired number of children with increasing age.

5 Discussion and conclusions

This paper focuses on the development of urban-rural differences in the desired number of children. For our analyses, we employ a special module that has been repeatedly added to the Austrian Microcensus since 1986 to study a time span of 35 years. Such a long time-horizon stands out as previous research on urban-rural differences has been largely restricted to cross-sectional data or shorter periods (notable exceptions are Kulu et al. 2007, for the Nordic countries, and Clark et al. 2024, for Canada and the US). Our findings provide novel evidence on heterogeneous trends in the desired number of children across Austria, thereby enhancing our limited understanding of family change in rural regions of high-income countries (cf. Clark et al. 2024, p. 4). Our study further complements research on urban-rural differences in total or cohort fertility (e.g. Vobecká and Piguet 2012; Campisi et al. 2020), fertility timing (Buelens 2021; Riederer and Beaujouan 2024), and the realization of short-term fertility intentions (Riederer and Buber-Ennser 2019, 2021).

Consistent with observed trends in total fertility, we find urban-rural differences in the desired number of children which have declined over time (H1a and H1b confirmed). The process of convergence can be illustrated well by comparing developments in rural regions and Vienna. In the mid-1980s, the desired number of children of women in rural areas (2.3 children) was significantly higher than the replacement level fertility (2.1). Even in 2001, women from rural areas reported a mean desired number of children slightly above the two-child norm. Since then, however, the desired number of children has fallen below the value of two (1.9 in 2021). This is in stark contrast to Vienna, where the average desired number of children has hardly changed between 1986 and 2021 (1.7 vs. 1.6).

Our analyses further confirmed that the position in the life course (age), parenthood, living conditions (single vs. couple household), migrant background, education, and employment status affect the reported desired number of children (H2a). Associations were as expected: A lower desired number of children is associated with higher age, non-parenthood, living alone or in cohabiting couples, Austrian nationality, medium education, and employment. Noteworthy, the association with education reveals a (to some extent) U‑shaped pattern with both low and high educated women tending to report a higher desired number of children (if parenthood is controlled for). This is consistent with the literature showing that women with low social status (traditional norms, lower career aspirations) and high social status (affordability, work-family reconciliation) tend to have larger families (Nitsche et al. 2021; Wood et al. 2014). For example, studies for France (Compans 2021) and the Nordic countries (Andersson et al. 2009) showed that higher educated women also tend to recuperate childbearing plans later in life. This aligns with our further results suggesting that age and parenthood have become more relevant for the desired number of children over time. In line with our expectations (H2b and H2c), the considered explanatory factors were also relevant for urban-rural differences and their development over time (i.e. differences between survey waves).

Findings (regarding age and parenthood) suggest that fertility postponement had a considerably large impact on the change in the desired number of children. The decline in the desired number of children and the convergence between rural and urban regions has been mainly observed among women aged 30–40, and increases in non-parenthood have been the major driver of the decrease in the desired number of children in both rural and urban regions. It is plausible that non-parents report a lower desired number of children than parents, in particular with increasing age when the desired number of children may be adapted to actual life circumstances. Another reason besides adaption is that non-parents likely include a higher proportion of persons not intending to have children at all.

Our study provides evidence for the considerable role of social change. Collectively, the examined factors of social change accounted for the majority of the decline in the desired number of children in both rural and urban regions. Nevertheless, they can hardly explain the observed stability of the desired number of children in Vienna. Although decomposition analyses confirm the assumption that increasing immigration counteracted the otherwise expected further decline of the desired number of children in Vienna, unobserved aspects remain to be discovered. For example, if the expansion of childcare infrastructure (Riederer and Buber-Ennser 2021) and the rising number of births among women in their late reproductive years (Riederer and Beaujouan 2024) have reinforced the belief in the feasibility of later transitions to parenthood and work-family balance, this could have contributed to stabilizing the desired number of children in Vienna as well.

Overall, our findings are in line with SDT theory (Lesthaeghe 2010; van de Kaa 1994): they are (a) demonstrating a decline in the desired number of children, the forerunning function of cities, and regional convergence, and (b) emphasizing the relevance of changes in living arrangements, female education, and female employment for developments of the desired number of children. Yet, similar to previous research (Sobotka 2008; Zaidi and Morgan 2017), a potential inconsistency remains. In our study, this refers to the assumed relevance of ideational change. On the one hand, our results suggest that decreases in marriage contributed to a decreasing desired number of children, indicating the relevance of ideational change (assuming that married couples exhibit stronger family orientations). On the other hand, the desired number of children among women in their twenties has remained largely unchanged. This suggests that the desired number of children did not adapt to ideational change and the low levels of fertility. Instead of conforming to historical trends of decreasing fertility, our results indicate increased phases of non-parenthood and later adaptions of the desired number of children to the individual life situation. In this respect, changes in social and economic structures influencing (a) incentives and opportunity costs of childbearing and (b) conditions facilitating or impeding the realisation of fertility intentions may be decisive (Riederer and Buber-Ennser 2019). Future research should thus investigate processes of postponement and adaption using longitudinal data and consider regional differences in structural aspects. Another direction for future research involves extending the examination of ideational change. If appropriate measures are available, the direct effects of changes in attitudes and values on the desired number of children could be assessed. Furthermore, research could also be extended to analyse the desired number of children in the context of other life goals. It is quite conceivable that ideational change does not only affect fertility desires, but also their significance in comparison to other desires (Testa and Bolano 2019).

Our study has further limitations that have to be acknowledged, most of them referring to non-availability of comparable time-series data. First, we discuss only urban-rural contrasts, albeit previous research has shown that there is rather an urban-rural continuum characterised by a broad spectrum of differences in fertility behaviour. Nevertheless, as indicators of urbanisation varied over time, we had to restrict us to the categorial measure used. Second, all existing measures of fertility preferences have both advantages and disadvantages. Sensitivity analyses using more than one measure would be desirable. However, only the desired number of children has been available to analyse changes. Similarly, many measures used to reflect social change are only proxy variables. More detailed measures would be preferable but have not been available in the survey or at least not in earlier survey waves (i.e. before 2012). One example is migrant background. Available differentiations between immigrant groups became more fine-grained throughout the decades but are not comparable across time and information on country of birth is only available for the three most recent waves.

Despite these limitations, our research adds relevant insights into regional variations and the process of urban-rural convergence in the desired number of children. While public discourses often contend that urban-rural differences have largely disappeared, our study reveals disparities to persist. The findings may reflect diverse preferences, specific needs, resources and circumstances of women in different geographic contexts. Consequently, researchers and policymakers should recognize the relevance of considering this heterogeneity. Urban-rural differences in the desired number of children are also likely to be significant in other high-income countries, yet their exploration warrants more attention.