Class politics in the sandbox? An analysis of the socio-economic determinants of preferences towards public spending and parental fees for childcare

Funding information H2020 European Research Council, Grant/ Award Number: 311769; Open access funding enabled and organized by Projekt DEAL Abstract This article analyses the socio-economic determinants of public preferences towards public spending and parental fees for childcare and how they are conditioned by institutional contexts. Previous studies of childcare policy preferences have focused on attitudes regarding the provision of care. However, when it comes to questions of financing, we know astonishingly little about how supportive individuals actually are of expanding pre-school early childhood education and care, and how support varies across different socio-economic groups in society. This is an important research gap because childcare provision and how it is financed have redistributive implications, which vary depending on the institutional design of childcare policy. Using novel and unique survey data on childcare preferences from eight European countries, we argue and show that preferences towards expanding childcare are more contested than it is often assumed. The institutional structure of childcare shapes how income matters for preferences towards how much should be spent and how provision should be financed. Where access to childcare is socially stratified, the poor and the rich develop different preferences towards either increasing public spending or reducing parental fees in order to improve their access to Received: 12 August 2019 Revised: 24 April 2020 Accepted: 30 June 2020


Abstract
This article analyses the socio-economic determinants of public preferences towards public spending and parental fees for childcare and how they are conditioned by institutional contexts. Previous studies of childcare policy preferences have focused on attitudes regarding the provision of care. However, when it comes to questions of financing, we know astonishingly little about how supportive individuals actually are of expanding pre-school early childhood education and care, and how support varies across different socio-economic groups in society. This is an important research gap because childcare provision and how it is financed have redistributive implications, which vary depending on the institutional design of childcare policy.
Using novel and unique survey data on childcare preferences from eight European countries, we argue and show that preferences towards expanding childcare are more contested than it is often assumed. The institutional structure of childcare shapes how income matters for preferences towards how much should be spent and how provision should be financed. Where access to childcare is socially stratified, the poor and the rich develop different preferences towards either increasing public spending or reducing parental fees in order to improve their access to childcare. The findings in this article suggest that expanding childcare in systems characterised by unequal access can be politically contested due to diverging policy priorities of individuals from different social backgrounds.
attitudes, childcare, politics, preferences, public opinion, social investment 1 | INTRODUCTION Scholarly interest in early childhood education and care (childcare, hereafter) as a core element of the social investment approach has increased considerably during recent years (Garritzmann, Häusermann, Palier, & Zollinger, 2017;Hemerijck, 2013Hemerijck, , 2017Morel, Palier, & Palme, 2012). Researchers commonly argue that expanding formal early childhood education and care supports the early development of children and their educational achievements later in life and it facilitates the reconciliation of work and family life for parents with young children. Even though governments of different partisan hues have expanded childcare provision in several European countries in recent years, there is still a wide variation across countries regarding particular policy and reform trajectories (Morgan, 2013).
More specifically, there are significant differences regarding the availability and affordability of childcare, which has implications for class-related differences in access to and usage of childcare (Van Lancker, 2013, 2018Van Lancker & Ghysels, 2016). The persistence of social inequality in access to childcare implies that policy reforms of this sector have different redistributional implications.
These redistributional implications in turn shape the politics of childcare reforms. Several scholars have argued that expanding childcare should be widely popular and that it would allow governments to pursue strategies of 'affordable credit claiming' (Bonoli, 2013) as it is both more popular and less expensive than other kinds of social policies (see also Hemerijck, 2015). More recently, scholars have begun to examine individuallevel preferences towards childcare in greater detail, focusing on preferences towards childcare provision (Banducci, Elder, Greene, & Stevens, 2016;Busemeyer & Neimanns, 2017;Chung & Meuleman, 2017;Goerres & Tepe, 2012;Meuleman & Chung, 2012). This research focus can to a large extent be attributed to data availability as most studies rely on an item included in the European Social Survey on preferences towards government responsibility to 'ensure sufficient childcare services for working parents' (ESS, 2008(ESS, , 2016. A general finding of this research is that individuals are highly supportive of a strong role of the government in childcare provision (e.g., Meuleman & Chung, 2012).
However, we argue that studying preferences towards childcare provision likely underestimates the potential of conflict in preferences regarding how governments should expand childcare and how expansion should be financed.
Developing a better understanding of contentious issues related to the financing of childcare as this article does could significantly help to explain the apparent mismatch between supposedly wide-spread support for childcare on the one hand and persistent cross-country differences in the provision of childcare. To name a few examples of potentially contentious issues, governments might focus on expanding public childcare spending (e.g., to expand enrolment or increase quality), but if parental fees are and remain high, this additional public spending will primarily benefit higher-income parents, because lower-income parents find it difficult to afford the high parental fees (Abrassart & Bonoli, 2015). 1 Alternatively, governments might focus on reducing parental fees. Yet, if childcare is hardly available, this will be of little help for parents who cannot enroll their child because of the shortage of supply.
With low availability of childcare, again higher-income parents are likely to be better positioned to enroll their children compared to lower-income parents because they are better equipped to enroll at alternative providers or because more subtle, socially adverse selection criteria may exist that benefit the better off (Del Boca, Pronzato, & Sorrenti, 2016;Hogrebe, 2016).
In this article, we address the question of how these institutional differences in availability and affordability of childcare affect public preferences and attitudes towards financing childcare reform. 2 We take into account all types of public and private forms of formal early childhood education and care for the whole age range from age zero up to the beginning of compulsory schooling. 3 Inspired by the extensive literature on policy feedback (Busemeyer, Abrassart, & Nezi, 2019), we expect that attitudes and preferences are strongly shaped by institutional contexts, that is, the availability and affordability of childcare. The analysis in this article makes use of the INVEDUC ('Investing in Education in Europe') survey (Busemeyer, Garritzmann, Neimanns, & Nezi, 2018), which is a representative survey of public opinion in Western European countries (United Kingdom, Ireland, Sweden, Denmark, Italy, Spain, Germany and France) conducted in 2014. Most importantly, the INVEDUC survey includes questions on attitudes for and preferences towards public spending and parental fees for childcare, which provide a more detailed measurement of the different dimensions of childcare policies compared to commonly available survey data in the ESS, which focuses on provision of care.
Our analysis supports the argument that availability and affordability of childcare provision at the macro level shape public preferences towards childcare, in particular the extent to which they are structured along income lines.
We empirically identify three clusters of countries with distinct barriers of access to childcare across income groups: First, where access to childcare is socially stratified due to low affordability, higher-income individuals are the strongest supporters of additional public childcare spending, whereas lower-income individuals demand lower levels of parental fees in order to improve access to childcare for them. Second, where access to childcare is socially stratified due to low availability, the pattern is reversed. Lower-income individuals demand more public spending, supposedly to finance additional places, whereas higher-income individuals oppose parental fees, which is, we argue, because they are more likely to send their children to private care providers for which they need to pay higher levels of parental fees. Finally, in countries where access to childcare is relatively open and affordable, attitudinal patterns resemble more the classical conflict about redistributive policies with the rich more opposed to further public spending and more in favour of fees. In general, our findings imply that the politics of expanding childcare are more contested than commonly assumed due to diverging preferences of individuals from different social backgrounds and varying institutional contexts.
In the next paragraphs, we shortly discuss the existing literature on individual-level preferences towards childcare provision. We then map the countries included in our study in terms of their variation in inequality in access to childcare, depending on their levels of availability and affordability of childcare. Based on this country-clustering, we develop our hypotheses regarding how the socio-economic position of an individual should matter for her preferences towards public childcare spending and parental fees. We put these propositions to an empirical test in the subsequent multivariate analysis of childcare preferences.

| DETERMINANTS OF ATTITUDES AND PREFERENCES TOWARDS CHILDCARE POLICIES
The few existing studies on the determinants of individual-level preferences towards childcare provision have identified several factors that matter in shaping preferences (Banducci et al., 2016;Busemeyer & Neimanns, 2017;Chung & Meuleman, 2017;Goerres & Tepe, 2012;Meuleman & Chung, 2012). Similar to findings from research on preferences towards other areas of social policy and redistribution (e.g., Busemeyer, Goerres, & Weschle, 2009;Jaeger, 2009;Van Oorschot, 2006), these determinants can be broadly grouped into being related to material selfinterest, norms and values, and contextual-level factors. 4 With regard to the role of material self-interest, individuals that are more likely to benefit from childcare provision are likely to support a strong role of the government in childcare provision. This includes women with a strong attachment to the labor market, individuals in their childrearing age, those with small children (Meuleman & Chung, 2012), and single parents (Busemeyer & Neimanns, 2017). In addition, becoming a parent has been found to alter preferences towards childcare, although this effect varies between men and women (Banducci et al., 2016).
Furthermore, individual social class position more generally has been found to matter for preferences, but findings to date remain inconclusive (Borck & Wrohlich, 2011;Chung & Meuleman, 2017;Meuleman & Chung, 2012;Mischke, 2014). Individuals in a lower socio-economic position are expected to be more likely to support government intervention in childcare provision because they rely to a higher extent on well-functioning public services whereas individuals in a higher socio-economic position should be rather opposed to government provision because they carry a higher share of the associated tax burden and are better placed to rely on the private provision of care.
In a study of 22 European countries, Meuleman and Chung (2012) find a negative effect of subjective income perceptions on preferences towards government responsibility to ensure childcare. However, using the same data, Mischke (2014) finds that this effect does not hold across countries.
Contextual-level factors are an additional determinant of preferences towards childcare, and they help to explain why the role of individual-level determinants may vary across countries. Studying childcare preferences in East and West Germany, Goerres and Tepe (2012) identify a persistent regime socialisation effect for individuals who were socialised in the former German Democratic Republic where maternal employment and public childcare were widespread and well-accepted. They find that those individuals continue to be more supportive of a strong role of the government in childcare provision. Similarly, Chung and Meuleman (2017) find support among parents for a strong role of the government in childcare provision to be larger in countries where provision is already more generous. For several further contextual-level variables they do not find statistically significant associations with country-average preferences. Thus, generous childcare policies appear to generate selfreinforcing feedback effects which increase overall support and acceptance of these policies (cf. Busemeyer et al., 2019;Neimanns, 2020).
However, to date no study has systematically examined how these feedback effects of childcare policy affect preferences across different socio-economic groups in society, that is, whether a more generous childcare provision attenuates or intensifies conflict in preferences between different socio-economic groups. An additional restriction of previous comparative studies is that they (have to) work with the question from the European Social Survey on the government's responsibility for providing childcare. Support for government responsibility in childcare could go along with support for additional public spending, but not necessarily so, if, for instance, individuals would like to see the government pushing for more private provision and higher fees.
Relatedly, there isso farno internationally comparative study of preferences regarding parental childcare fees. In this article, we study these different aspects related to the financing of childcare in order to provide a better understanding of the redistributive politics of childcare expansion. In the next section, we develop our argument in greater detail.

| CLUSTERS OF CHILDCARE PROVISION AND THEIR EFFECTS ON PREFERENCES TOWARDS PUBLIC SPENDING AND PARENTAL FEES FOR CHILDCARE
Broadly speaking, our article studies the association between income-related self-interest and attitudes towards childcare policies, paying particular attention as to how institutional contexts condition this association across countries. Regarding the institutional context, we focus on how the availability and affordability of childcare affect social stratification in access to care, relying on the relative index of inequality in childcare use developed by Van Lancker (2018). In the following, we examine how the eight countries included in our study distribute according to these criteria. Subsequently, we develop theoretical expectations on how these institutional characteristics shape attitudes and preferences on the micro level.

| Identifying clusters of childcare provision
The relative index of inequality in childcare use developed by Van Lancker (2018) is the most comprehensive available indicator of social stratification in access to childcare. This index measures to what extent fulltime equivalent childcare usage depends on families' equivalised disposable income. A value of zero would imply equal childcare use across income groups (no income-related differences), whereas positive values mean that more affluent parents are overrepresented among the users of childcare (negative values, that is, over-representation of low-income parents, do not occur empirically). The index reveals that in most countries, more affluent families are substantially overrepresented among the users of childcare (Van Lancker, 2013, 2018.
The inequality index is strongly related to and determined by variation in two sub-dimensions: availability and affordability of childcare. 5 Regarding availability of childcare, this factor is often particularly critical for parents from lower-income backgrounds to access childcare. In the case of low availability, governments can target spending towards socially disadvantaged individuals and introduce admission criteria to allocate childcare slots. While these criteria can in principle be designed to prioritise access to low-income parents, in practice, higher-income parents are likely to be in an advantaged position because employment criteria may outweigh social desirability criteria when rationing access (Del Boca et al., 2016), or because those parents' higher social capital makes it easier for them to navigate through the sometimes vague and complex admission procedures (Hogrebe, 2016). 6 Thus, lower availability of childcare provision is likely to be associated with higher social inequality in access to care (Van Lancker, 2013, 2018. With regard to affordability, parental fees have been identified as a central factor shaping social stratification in access to childcare (Abrassart & Bonoli, 2015). Particularly lower-income households will be easily priced out if fees are high. A progressive fee structure, which shifts the burden of private contributions to the higher income groups, can counteract this tendency (ibid.). However, it is not only the progressivity in the design of parental fees that matters for social stratification in access to childcare. Rather, it is the mix of public and private providers that affects the degree of overall progressivity of parental fees. If publicly provided care options are not sufficient to meet the overall demand for childcare places, private providers can fill this gap, which, however, often require parents to pay higher fees (Da Roit & Sabatinelli, 2013;Del Boca, Locatelli, & Vuri, 2004;Naldini & Jurado, 2013). It is often the more affluent parents that make use of such private alternatives. As a consequence, the structure of parental fees becomes (even more) progressive due to the socially stratified use of public and private providers. Hence, and quite obviously, lower affordability (i.e., higher fees) goes along with income-related inequalities in access. Inequality in childcare enrollment is lowest in Sweden and Denmark. Here, children of parents from lower and higher income groups are nearly equally likely to be enrolled in formal childcare institutions. As Van Lancker (2013) notes, the success of achieving relatively equal access to childcare in these countries relies on a supply-side strategy of childcare expansion combined with a well-defined legal entitlement to obtain a care place. Parental fees are income-related and are kept at relatively low levels (see also Table A.1). Thus, both availability and affordability of childcare are high in Sweden and Denmark. The income-related fee structure ensures that childcare spending is highly progressive in fiscal terms. However, the Scandinavian design of childcare provision requires a high budgetary commitment: levels of public childcare spending are highest in these two countries (Table A.1).
The UK and Ireland are located at the other end of the scale with high levels of inequality in childcare enrollment. These two countries constitute a second country-cluster. Even though the level of public childcare spending in the UK ranks fourth in country-comparison, the fact that parents pay up to half of their income on childcare demonstrates that affordability of childcare in this country is low (Penn, 2013; Table A.1). As lower-income families are easily priced out of provision in such an institutional environment, childcare (and public spending devoted to it) becomes highly regressive in fiscal terms.
The remaining four countries with medium levels of inequality in childcare coverage constitute a third countrycluster (Italy, Spain, Germany, and France). With the partial exception of France, these countries are traditionally characterised by a high involvement of families in taking care of small children (in particular for children below the age of three) and by little public resources devoted to the provision of out-of-family childcare options (Morgan, 2013). Only more recently, and with varying efforts, governments tried to expand provision of formal childcare, but spending levels continue to lag behind in international comparison (Table A.1). Thus, the availability of childcare is still low and many families have difficulties to find a care place.
Unlike the high-inequality countries, parental fees are at medium levels on average in these four countries (Table   A.1), suggesting that insufficient availability might be a more constraining factor than affordability of care. However, this aggregate perspective masks the substantial degree of heterogeneity in service provision in these countries.
Related to the scarcity of places in public institutions, private service providers experienced considerable growth in recent years (Da Roit & Sabatinelli, 2013 for Italy; Naldini & Jurado, 2013 for Spain; Fagnani, 2013 for France; Bönker, Libbe, & Wollmann, 2016 for Germany), and they tend to charge higher levels of parental fees. 7 Such developments contribute to institutional stratification in childcare enrollment with lower-income families using disproportionally the less expensive public providers and higher-income families relying more on more expensive private provision. As a consequence, childcare spending becomes relatively progressive, but in a context of dualised provision: Higher-income families contribute with their tax payments to public provision while at the same time paying fees, which are often income-related, and sometimes include top-up fees for privately provided care places. 8

| Hypotheses
The discussion in the previous section highlighted how differences in the availability and affordability of childcare are associated with different patterns of social inequality in childcare usage in the different and distinct country clusters of childcare provision. In the next step, we now formulate expectations on how the particular interplay of For each of these socio-demographic groups, we expect income to matter depending on the specific context of childcare provision and its redistributive implications.
In the second step, we hypothesise about how the institutional context mediates the association between selfinterest and preferences. Starting with the first country cluster that combines high availability and high affordability of childcare (Denmark and Sweden), social inequality in childcare usage is low and access to care has become nearly universal. While parents contribute only a small share of overall funding in the form of parental fees, public expenditure and taxes are at relatively high levels to finance provision. In such a context, where public provision of care is clearly dominant, the lower-and middle-income groups have become dependent on public provision and should support continuing government involvement in the form of high public spending and low fees for childcare (cf. Busemeyer & Iversen, 2014. The higher-income groups, in contrast, still have to pay income-related fees plus taxes to finance the comprehensive level of childcare provision and are clear net-contributors to the progressively designed system of childcare provision. Accordingly, rich households should prefer to strengthen parental fees in order to be able to reduce (or not further increase) public spending and taxes. For both public and private spending, there should be a clear income effect with high income individuals opposing more public and supporting more private spending for childcare provision.
H1: The more universal access to childcare is, the more do higher-income individuals oppose more extensive government involvement in childcare provision, that is, the more do they oppose public childcare spending and support parental fees.
In the two remaining country clusters, access to childcare is more socially stratified, but for different reasons. In a context of low affordability (the high-inequality country cluster of the UK and Ireland), affluent parents are the primary beneficiaries of public money spent on childcare services. While poor parents are likely to be priced out of formal childcare services and have to rely on informal forms of care, rich parents have access to this kind of care but still pay high shares of their income to purchase a care slot. As they disproportionally benefit from these services, they are likely to demand more public expenditure on childcare services in order to reduce costs for them. For the poor to be able to benefit from public childcare expenditure, childcare fees would need to be lower in order to address the affordability problem. Lower fees are thus a precondition to gain access to childcare services and to benefit from public childcare spending. For the rich, high fees are less of a concern. They could support increasing direct subsidies to and public spending on childcare instead of lowering fees as long as access to services remains stratified.
In sum, in this country cluster, lower income households should be the strongest supporters of abolishing fees, whereas higher income households should be particularly supportive of additional public spending on childcare.
H2: The more unequal access to childcare is due to low affordability, the more do low-income individuals oppose parental fees and the more do high-income individuals support public childcare spending.
Finally, in the country cluster characterised by low availability (the medium-inequality country cluster of Germany, France, Italy and Spain), there tends to be a chronic excess demand for childcare places. Where and when governments apply social selection criteria to distribute scarce childcare slots, they may, not necessarily though, reduce social stratification in access to public service provision by prioritising access for socially disadvantaged groups (Del Boca et al., 2016). In such a situation, rich individuals could support the establishment of private alternatives to public provision, even if these go along with higher fees (Da Roit & Sabatinelli, 2013;Del Boca et al., 2004;Naldini & Jurado, 2013). This could lead to the emergence of a dualised pattern in attitudes and preferences, where the poor support and benefit from expanding less expensive care options in the public sector, whereas the rich demand the expansion of private options and have little interest in expanding public provision. In this context, the rich should have a strong interest in lowering fees in order to avoid the double payment problem of having to pay for public provision via general taxes and, potentially, additional fees for private provision. To sum up, in countries where the scarcity of public or publicly subsidised childcare places leads to a more stratified system of public and private service provision, lower income households should prioritise additional public spending to expand access (and increase availability), whereas higher income households should support lower fees.
H3: The more unequal access to childcare is due to low availability, the more do lower-income individuals support public childcare spending and the more do higher-income individuals oppose parental fees.

| Data and methods
To analyse preferences towards public spending and parental fees for childcare, we rely includes design and selectivity weights to account for selection bias. We include these weights in the descriptive statistics. In the main regression analyses below no weights are used, as these can bias the standard errors and are not necessary because the included control variables account for sampling selectivity (Schmidt-Catran, 2016).
Respondents' preferences towards public childcare spending and parental childcare fees are used as dependent variables. Both variables are measured as categorical variables in the INVEDUC survey. We dichotomise both variables to simplify the interpretation of the results by being able to predict marginal effect estimates based on logistic regression models. The results do not change substantially if we use the original coding of the variables and estimate ordered logistic (for public spending) or multinomial logistic (for parental fees) regression models instead (Tables A.7 and A.8 in the online appendix).
For preferences towards spending, respondents were asked the following question: Please tell me whether you would like to see more or less government spending in each of the following areas. Keep in mind that 'more' or 'much more' might require a tax increase: Pre-school and early childhood education. 9 We recode this variable and distinguish preferences for 'more' or 'much more' spending (1) from preferences for the 'same as now', 'less' and 'much less' spending (0). 10 With regard to preferences towards parental fees, respondents were asked: Some people believe that early childhood education should be free for everyone while others believe that parents should pay. What do you think? 1: It should be free for all; 2: It should be free only for those who cannot afford it; 3: All parents should pay the same fees.
As only a small share of the respondents (around 12%) chose the option that all should pay the same fees, we decided to merge categories 2 and 3 to which we assign a value of 0 and contrast this to respondents that prefer no fees at all (coded as 1). In terms of taxes that need to be raised to finance public childcare spending, exempting only the poor from childcare fees should imply only minor changes compared to a situation where all pay the same fees. Respondents' net household income is measured in quintiles as defined by country-specific income distributions.
We present the results using marginal effect plots and report the predicted probabilities of supporting public spending increases and rejecting parental fees by income quintile. The marginal effect plots are based on logistic regression models for each of the two dependent variables including an interaction between income and country dummies. We present separate marginal effect plots for each country cluster in order to deal with the potential problem of small sample size (when focusing on individual countries) while still being able to identify divergent income effects across country clusters. We also assess the associations at the level of individual countries to probe the consistency of income effects within our three country clusters. The full results of the logistic regressions are reported in Table A We control for a list of standard control variables in the logistic regression models (Busemeyer & Neimanns, 2017;Goerres & Tepe, 2012): age, gender, having kids above or below the age of 10, being a single parent, household size, employment status (interacted with gender), public sector employment, whether a respondent lives in an urban or rural area, citizenship, and general attitudes towards female employment. In additional models, we also include region dummies (at the NUTS-2 level) to address the issue of regional and local variation in childcare provision (Goerres & Tepe, 2012). The detailed operationalisation of the variables is presented in Table A.3 in the online appendix.
Despite the relatively small number of countries, the INVEDUC data provide the unique opportunity to test the hypotheses spelled out above. As mentioned above, partly due to data limitations, most existing studies focused on analysing preferences towards childcare provision. We contribute to those studies by examining preferences towards the financing of childcare, that is, preferences towards public spending and parental fees, and how these preferences vary across socio-economic backgrounds, depending on the specific context in which provision takes place. We proceed in three steps: first, we discuss how preferences towards provision, public spending, and fees relate to each other, that is, whether they measure different aspects of childcare preferences or rather one underlying latent dimension. Second, we briefly examine the distribution of country-average preferences to evaluate whether it is possible to identify feedback effects on aggregated public opinion. Third, we present income effects across the three country clusters to discuss how the context of childcare provision shapes childcare preferences across individuals from different socio-economic backgrounds.

| Results
As a first step of the analysis, we briefly assess to what extent it is necessary to study preferences towards the various aspects of childcare policy relating to provision, public spending, and parental fees as separate issue. To this end, we first compare the income effects of regression models based on the INVEDUC data for public spending preferences to those from similar models based on the ESS data which includes a question about the government's responsibility in ensuring sufficient childcare provision. This comparison shows that income effects diverge for several countries between the two surveys (Table A.5 in the online appendix). In particular for the countries with low affordability (UK and Ireland) and for the countries with high availability and high affordability (Sweden and Denmark), differences in preferences between the poor and the rich vary across the two items. While we discuss the substantial effects in more detail below, this comparison demonstrates that preferences towards provision and public spending measure different aspects of preferences towards childcare policy. 12 Second, we probe how preferences towards public childcare spending and parental childcare fees relate to each other in the INVEDUC survey. The bivariate correlation coefficient between both items for the whole sample is low (r = 0.043) and the country-specific correlation coefficients are insignificant in five of the eight countries. This finding suggests that individuals have distinct preferences towards public childcare spending and parental childcare fees that are not strongly related to each other.
Taken together, these findings indicate that policies related to the provision, public spending and fees of childcare should be regarded as separate issues, tapping into different aspects of a multi-dimensional phenomenon. This also demonstrates the limits of previous research that has focused on the aspect of provision as discussed above.
We now proceed with the descriptive and multivariate analysis of childcare preferences towards public spending and parental fees. Online appendix Table A.2 lists summary statistics of our two dependent variables. Figures 2 and   3 plot the predicted probabilities of supporting additional public spending ( Figure 2) and rejecting parental fees (Figure 3) across the country clusters (upper panels) and single countries (lower panels). We report the results of Wald tests to assess the statistical significance of the differences in the marginal effect estimates.
To begin, we evaluate average support levels of increasing public spending and abolishing parental fees for childcare, and whether average preferences vary according to country-specific levels of availability and affordability of childcare. A first notable observation is that individuals seem to be divided on the question of whether childcare provision should become more generous. On average, around 50% of individuals are supportive of increases in public childcare spending whereas the remaining 50% of individuals prefer levels of spending to remain the same or to F I G U R E 2 Predicted probabilities of supporting more public childcare spending by income quintiles, country clusters, and countries. Notes: Predicted probabilities and 95% confidence intervals based on models in Table A.4 in the online appendix, holding constant all other variables at their means decrease (Table A.2). With regard to parental fees, a slight majority of 54% of respondents prefers that there should not be any childcare fees at all (Table A.2). These values are remarkable because previous studies have documented high levels of support for expanding childcare (Bonoli, 2013;Meuleman & Chung, 2012). However, when it comes to preferences towards the financing of childcare in the form of public spending or private spending, that is, parental fees, the public seems to be more divided.
Next, we examine how preferences towards public spending and parental fees at the country-average level relate to inequality in childcare provision as indicated by levels of availability and affordability of childcare. We do not find meaningful associations in this step of the analysis (Figures 2 and 3; Table A.2). Average preferences vary widely across countries, irrespective of availability and affordability of childcare. Support to increasing public spending is strongest in Germany and Sweden where more than 60% of respondents are in favour of additional public spending on childcare. In France, the UK and Denmark, only a minority of respondents supports spending increases (30%, 45% and 48%, respectively). This variation, however, is independent of the level of inequality in childcare coverage in these countries. With regard to parental fees, citizens in Spain, France and the UK are particularly opposed to parental fees. Around 60% of respondents reject fees in these countries. The acceptance of fees is highest in Denmark where less than 20% of respondents prefer that access to childcare should be free. Once again, average preferences vary strongly across countries, but largely unrelated to the level of availability and affordability of childcare. As mentioned above, we do not necessarily expect strong associations between the institutional context and average levels of support for public spending and parental fees (cf. Chung & Meuleman, 2017), but we rather expect contextual factors to influence the micro-level association between income and preferences, that is, the extent of the income cleavage in preferences.
We now go on to examine income effects in preferences towards public spending and parental fees for childcare across the different country clusters as defined above. We focus on discussing preferences by income at the level of country clusters (upper panels in Figures 2 and 3). We briefly refer to the country-level effects (lower panels in Figures 2 and 3) further below to validate that within the country clusters, income effects are relatively consistent.   Figures 2 and 3).
The likelihood of rich individuals in Sweden and Denmark to support additional public childcare spending is 10 percentage points lower than for individuals in the middle-income category (upper graph in Figure 2). With regard to parental fees, the likelihood of rich individuals to support fees is 15 percentage points higher as compared to the lowest income quintile (upper graph in Figure 3). Both differences in the marginal effect estimates are statistically significant at the 95% level. This finding of rich individuals being opposed to increasing public spending and to reducing childcare fees provides clear support to Hypothesis 1.
When examining countries with low affordability of childcare (i.e., high levels of parental fees) (centre plots in Figures 2 and 3), we see that income effects are absent for the issue of public spending. As expected, support for additional public spending in the UK and Ireland is strongest in the highest income quintile, but the differences are not statistically significant. Once we take a look at preferences towards parental fees, we see the reversed pattern.
Those in the lowest income quintile most strongly support the notion that there should be no fees at all. The difference in support between the lowest and the middle-income quintile is statistically significant and amounts to 10 percentage points. These findings suggest that the priority for low-income respondents is to reduce fees, which is a precondition for them to be able to access childcare. In contrast, given the stratifying effect of high fees, the rich are in favour of additional public spending from which they benefit to a disproportionate extent in this context. This evidence is in line with Hypothesis 2.
We find the opposite pattern in preferences for countries characterised by low availability where stratification comes in the form of dualised childcare provision (right-hand plots in Figures 2 and 3). In these countries, which are represented in our sample by Italy, Spain, Germany and France, low-income individuals most strongly support additional public spending on childcare (the difference between the lowest and the highest income quintile is 10 percentage points and is statistically significant), of which they are likely to benefit given the relatively low cost of public services for them. The middle-and higher-income groups, in contrast, have stronger preferences to lower fees instead. Support among respondents in the middle-income group is 6 percentage points above the lowest income quintile in this regard and the difference is statistically significant. This pattern for the low availability countries is in line with expectations. Middle-and higher-income groups pay income-related fees, which can amount to considerable shares of their incomes (Table A.1), and they have a higher likelihood to attend private childcare centres for which they often need to pay additional top-up fees. Taken together, dualised childcare provision leads lowerincome individuals to support additional public spending whereas it leads higher-income individuals to demand lower parental fees, confirming Hypothesis 3.
Note that when evaluating income effects by country, the differences in preferences across income groups are insignificant in some of the countries (cf. Table A.5 in the online appendix), which is to some extent due to the lower number of observations in the individual country cases. The important point to consider here is that within the country clusters, the income effects point into the same directions. The lower panels in Figures 2 and 3 on the countryspecific preferences by income quintiles largely confirm this claim. Thus, our reasoning does not imply that in each and every country we would observe a high degree of conflict in preferences across income groups. We argue instead that across countries and within country-clusters, income-related differences in preferences sum up and lead to clearly identifiable socio-economic patterns of preferences.

| Robustness issues
Our results are robust to alternative model specifications. While we dichotomised our two dependent variables to simplify the presentation of our results, the results hold if we estimate ordered logistic (for preferences towards public spending) or multinomial logistic regression models (for preferences towards parental fees) instead (Tables A.7 and A.8 in the online appendix). One limitation of our analysis is that we are unable to capture adequately subnational variation in availability and affordability of childcare (Goerres & Tepe, 2012). To address this issue, we reestimated our models including region dummies (at the NUTS-2 level). Adding these dummies basically leaves the results unaltered (see Figure A.1 in the online appendix). All statistically significant differences in the marginal effect estimates discussed in the main analysis remain significant when controlling for region. Although this procedure can only imperfectly address the issue of regional and local variation in childcare provision, this finding adds further confidence to the results.

| CONCLUSION
This article analyses the socio-economic determinants of preferences towards childcare. Previous research has largely focused on preferences towards childcare provision more generally, finding high levels of public support (e.g., Meuleman & Chung, 2012). Using new data from a cross-country public opinion survey we add to this research by studying preferences regarding the financing of childcare, that is, preferences about public spending for childcare and support for parental fees. We argue and show empirically that individuals hold distinct preferences on these different dimensions of childcare policy. We find, first, that, compared to previous studies, public support of childcare is much more divided once we take into account issues related to the financing of childcare. Second, our study confirms that material self-interest is strongly related to childcare policy preferences, which may also be because we focus on the distributional aspects and financing of care in this article.
Third, our analysis shows that the institutional context of childcare provision influences attitudinal patterns and thereby political support coalitions for different policy reform options. First, in countries where access to childcare is close to universal (Sweden and Denmark, in our sample), childcare preferences resemble the classical class conflict across the different dimensions of childcare policy. High-income individuals oppose higher levels of public spending and oppose slashing parental fees, whereas low-to middle-income individuals are more in favour of additional public spending and free access to childcare. Second, in countries with stratified access to childcare, preferences of the rich and poor diverge between the different policy alternatives of expanding public spending and lowering fees. Where childcare is hardly affordable for a considerable share of families due to high fees (the UK and Ireland), individuals from the highest income group are the strongest supporters of additional public childcare spending whereas individuals from the lowest income group rather support lower fees to get access in the first place. Where the availability of places presents the most challenging barrier of access to childcare (Italy, Spain, Germany and France), in contrast, low-income individuals most strongly support additional public childcare spending, and individuals with middle incomes most strongly reject parental fees. We argue that these differences in the socio-economic determinants of support for childcare reflect variation in who benefits from the current status quo of childcare provision, and which reform measures help individuals in either reducing barriers in access to childcare, or in maintaining and improving their relatively privileged position in the current system of childcare provision.
The findings in this article have important implications for our understanding of the politics of childcare expansion. Once taking into account aspects related to the financing of childcare, the politics of expanding childcare can turn out to be more contested politically than often portrayed in notions of childcare expansion as an instrument of 'affordable credit claiming' for governments and political parties (Bonoli, 2013;Hemerijck, 2015). Furthermore, depending on the institutional context, it makes a difference which policy instruments governments use to expand childcare. If governments concentrate on expanding public spending (in the low affordability countries) or reducing parental fees (in the low availability countries) while not addressing barriers of access for socially disadvantaged groups, this will benefit in particular (the preferences of) those individuals already enjoying privileged access to childcare. Rather than mitigating inequality, the expansion of public childcare might actually reinforce social stratification in access in these countries, making the benefits of universal childcare provision increasingly difficult to achieve.
in the online appendix for the results). This comparison demonstrates that previous studies have likely underestimated the income-related potential of conflict in preferences applying specifically to the case of childcare.