Subsidising Formal Childcare Versus Grandmothers' Time: Which Policy is More Effective?

2.1 Value Functions

The model is solved backwards. Therefore, I first present the value functions for women in the second period.

3.1 Data

To calibrate the main parameters of the model, data from the European Union Statistics on Income and Living Conditions (EU-SILC) were used. Year 2016 was chosen because, in that particular year, a special module on “Access to Services” with information about childcare costs was made available. The EU-SILC is an annual household survey that provides information on individual and household characteristics, such as employment, education and family status, income, and childcare use (formal and informal).

The analysis is confined to a sample of married individuals because the main focus is the labour supply of married females. In order to analyse individuals who are potentially in the labour force and need childcare, I restricted the sample to women between ages 25 and 45. Women working less than 20 h a month were not considered as employed, given that they need little childcare. Self-employed women were also excluded because of the specificities of their work schedules. To compute the statistics related to childcare arrangements, a smaller sample of mothers with children aged 5 years or younger was considered.

The population was divided into three educational groups: people who have less than a high school degree ( < hs ), people with a high school degree ( hs ) and those with a college degree ( col ).^[13] The matching matrix that assigns spouses to women was generated as described next. I define women of type i and spouse of type j according to their education or schooling levels s such that i , j ∈ s . Based on these three categories, one can define nine household types by the educational attainment of husbands and wives, where Φ ( s , s ) is a particular element of the matching matrix. Table 1 shows the distribution of married households according to the educational level in 2016 as well as female employment rates in each of these nine household types. Some degree of assortative mating can naturally be observed, and female employment increases with education levels.

Table 2 shows that mothers who are helped by grandmothers tend to be less educated. This may be related to their earnings and, consequently, to the affordability of formal childcare. In fact, hourly wages are higher for mothers of young children not helped by grandmothers: 13.2 versus 12.8. Concerning employment, an opposite pattern emerges: the employment rate of mothers of young children who are helped by grandmothers is higher than the employment rate of those who are not helped by them. This is true at all educational levels, but specially for the less educated. The latter may reflect difficulties in finding affordable childcare when grandparents are not available.

In order to further explore this hypothesis I make use of the information included in the EU-SILC special module on “Access to Services” concerning affordability of childcare, unmet needs, and payments. For the sub-sample of mothers with children aged 5 or younger, 92 % of households pay the cost of childcare and 81 % even pay the full cost. Those that do not pay the full cost receive a subsidy from the government. Table 3 shows that more than 40 % of households have some trouble to afford childcare services. Regarding the reasons for not making use of formal childcare, 41 % say that they cannot afford, 3.5 % do not find places available, 1 % do not find places available nearby, 6 % find places available but hours are not suitable, 0.6 % think that the quality is not good enough, and the rest have other reasons (not specified). About 10 % of those make use of their parents instead.

Whether formal childcare is affordable or not, what is true is that most children attend some kind of preschool: 53.7 % in the case of children aged 2 years or younger and 81 % in the case of children aged 5 years or younger (Table 4). It is interesting to note that although 99 % of children aged 3–5 years attend a preschool facility, 13 % of them are also cared for by their grandmothers who, on average, spend 13.4 h a week caring for them, and 2.93 % are cared for by nannies. This additional need of childcare is probably a consequence of long work schedules, which generate a mismatch between the characteristics of the childcare system and the hours of work demanded in most jobs. On the other hand, children aged 2 years or younger spend on average less time with their grandmothers than in formal care: 25.4 h a week versus 29.5. The same can be said for children aged 3–5 years, since public preschool is universally provided.

3.2 Calibrated Parameters

In this section, I explain how values were assigned to the model’s parameters. The policy parameters and the parameters related to childcare costs were taken directly from Spanish data. The value of the discount factor, β , was set to 0.889 so that it is consistent with a real interest rate of 4 % in the reference period.^[14] All remaining parameters, i.e. those of the utility function ( σ , α , σ q , σ k ) , the probabilities of having access to free childcare ( g 1 , g 2 ) , and the parameters of the child quality production function ( ϕ m , ϕ c ) were calibrated to match model moments related to several labour market and fertility statistics for Spain. These statistics were computed using the data described previously.^[15] Although all these parameters were set simultaneously (by means of the method of moments), some parameters have clear counterparts in the data which I proceed to explain now (Table 5).

The employment rate of mothers with children aged 3–5 years was used to match the curvature of consumption in the utility function, σ , given that their decision to participate is, to a large extent, driven by the value of consumption. The percentage of women being mothers is related to the fixed utility cost, σ k , because this is one of the main determinants of fertility. The parameters of the child quality in the utility function: the curvature’s parameter, α , the weight of child quality, σ q , and the weight of mother’s time for child quality, ϕ m , determine labour force participation decisions, especially when children are very young. These parameters were calibrated to match statistics related to the relevance of childcare costs and mother’s time for child quality: the employment rate of mothers with children aged 2 years or younger, childcare costs as a percentage of household disposable income and childcare costs as a percentage of the wage of women without tertiary education. The probabilities of women having access to grandmother-provided childcare in the first and second period, g 1 and g 2 , were set to match the percentage of working mothers using grandmother-provided childcare for children aged 2 years or younger and 3–5 years, respectively.

3.3 Calibration Results

Parameter values were chosen so that the model is consistent with 2016 Spanish data related to the labour market, fertility, and childcare statistics. Table 8 shows that the baseline model is a suitable starting point for carrying out policy experiments because it matches real data relatively well. In addition, it is necessary to assess the model’s performance in matching moments that have not been used for calibration. Table 9 shows that the model accounts reasonably well for most of them. The model captures the fact that working mothers earn a higher wage than the female average. The average hourly wage of childless women is 11.2 euros, while that of mothers is 12.9 euros. Mothers of younger children earn on average higher hourly wages (13.2 euros), probably as a result of selection. Generally, the number of hours of childcare generated by the model is relatively close to the number of hours in the data. Table 3 shows that working mothers of younger children tend to use more hours of grandmother-provided childcare than mothers of older children. The model generates this pattern but underestimates the number of hours that children aged 3–5 years spend with grandmothers and overestimates the number of childcare hours of children aged 2 years or younger. The model also accounts for the fact that children aged 2 years or younger spend more time in childcare facilities (roughly 3 more hours) than with grandmothers. This is probably explained by the fact that women with tertiary education work more hours than women without tertiary education and also make less use of grandmothers’ help (Table 2).^[23]

Concerning distributions, the model is able to capture the qualitative patterns and, to some extent, quantitative numbers. The model replicates mother’s and working women’s distributions relatively accurately. However, it fails to quantitatively account for working mother’s distribution by education: the model overestimates the percentage of working mothers with the highest level of education, and largely underestimates the percentage of working mothers in the lowest category of education.^[24] The model reproduces the behaviour of women with tertiary education particularly well. Both the proportion of college working mothers over college working women and the proportion of college working mothers over college mothers are very close to those observed in the data.

4.1 Childcare Subsidies for Children Aged 2 Years or Younger

In this section, I show the effects of increasing means-tested subsidies for formal childcare for working mothers with children aged 2 years or younger. This experiment was implemented as described next. I set a particular value for the employment rate of mothers of children aged 2 years or younger as a target and computed the change in the value of the childcare subsidy in period 1, θ y , as well as the necessary adjustment in tax rates to make it happen. The target value is set equal to 57.3 % , the European average maternal employment rate for women whose youngest child was aged 2 years or younger in 2016 according to OECD Data.^[25] The corresponding childcare subsidy in period 1 turned out to be three times larger than in the baseline economy and the tax rate increase 2.5 % (Tables A3–A5).

Table 10 shows that the employment rate of mothers with children aged 2 years or younger grows by 6% points, which implies a value of 0.15 for the elasticity with respect to childcare costs. This result is consistent with the values that have been estimated in previous literature, which for Europe range from 0.14 for UK, to 0.46 for Romania.^[26] However, mothers work on average fewer hours. This does not happen just because the mothers who enter the labour market will supply very few hours (composition effects), but rather due to income effects and to the tax adjustment, which make the working activity less rewarding.^[27] This reduction in average hours worked generates a 6 % reduction in the number of hours that children aged 2 years or younger spend in childcare facilities. In addition, more affordable childcare implies a significant reduction in informal childcare use (12.8%).

I now look at disaggregated results according to the educational types of women to observe the model’s predictions along these lines. Table 11 shows that all types of mothers participate more in the labour market. The growth in the proportion of working mothers over mothers is particularly significant for women with low and medium educational levels: 13 and 11%, respectively. This explains the drop in the ratio of mother’s wage to women’s wage, which decreases by close to 3% (see Table 10). This outcome agrees with the results in the related literature. For instance, Del Boca et al. (2009) found that less educated women are more sensitive to changes in the price of formal childcare, and Guner et al. (2020) showed that the change in labour supply is particularly strong for women at the bottom of the skill distribution when child-related transfers are expanded.

It comes as a surprise that for women with tertiary education fertility is not positively affected. According to Del Boca et al. (2009) this type of family policies may positively affect the fertility rate of highly educated women because they reduce the opportunity cost of child-bearing. In fact, this occurs for some college women married to non-college men, but this effect is compensated by the decrease in fertility of some college women married to college men, who do not qualify for the subsidy, but are particularly affected by the tax increase. This is consistent with the results of Haan and Wrohlich (2011), mentioned earlier in the related literature, who concluded that childcare subsidies do not affect fertility on average, though with heterogeneous effects.

4.2 Childcare Subsidies in Periods 1 and 2: the “Double-subsidy”

The childcare subsidy studied in Section 4.1 did not affect the employment rates of mothers of children aged 3–5 years because the target was childcare in period 1. Given the shortage of affordable childcare driven by long work schedules mentioned earlier in the Introduction, an “overtime” subsidy ( p 2 e ) was added to the model to partly cover the cost of after-school hours. This subsidy is targeted to mothers of children aged 3–5 years, which are the ones who may mostly need it, given that regular hours at school in period 2 are already universally provided. The value of this additional subsidy is set so that the aggregate employment rate of mothers of children aged 2 years or younger meets the previous target. Subsidising half of after school hours, 1 2 p 2 e , in addition to the previous childcare subsidy in period 1, (3 θ y ), achieves this target but requires an adjustment in tax rates of about 7 % . The expression of this “overtime” subsidy is given by ∑ x = 1 N n g 2 I { l x , 2 > t 2 r } ( l x , 2 − t 2 r ) p 2 e 2 . This term is only relevant for mothers not helped by grandparents who work longer than regular hours in period 2, and should be added to the right-hand side of the government budget constraint in Section 2.

The effects of this “double-subsidy” are displayed in Table 10. The percentage of women being mothers increases substantially, from 44.6 to 52.2%, especially for women not helped by grandmothers, which grows by 17.5%. As expected, the additional subsidy induces more women to have children, especially women with tertiary education: the proportion of mothers over women rises by 18 % for college women, while for women with low and medium educational levels, it grows by 14 and 10.5%, respectively (see Table 11).

Regarding employment effects, the reduction in the cost of after school hours increases the employment rate of mothers of children aged 3–5 years by 7 % , especially that of women with low and medium educational levels: the proportion of working mothers over mothers rises by 41 and 25% for women with low and medium educational levels, respectively, and by only 2 % for women with tertiary education (Table 11). This is due to the increase in the fertility rate of women with tertiary education, which compensates to some extent the growth in the labour involvement of these types of women. These effects have an impact on the distributions of women according to education levels. Table 11 shows that college mothers gain weight in the distribution of mothers, while women with low and medium educational levels do so in the distribution of working mothers, a finding consistent with the results by Del Boca et al. (2009) mentioned before (Table A7).

Concerning childcare use, cheaper access to formal childcare reduces the percentage of working mothers using free care by 30 and 23% in periods 1 and 2, respectively. As in the previous scenario, the number of hours spent in childcare facilities by children aged 2 years or younger decreases by 6 % because women work on average fewer hours in period 1. Conversely, the number of hours spent in childcare facilities by children aged 3–5 years goes up, because women work on average more in period 2 given that childcare has become much cheaper.^[28]

4.3 Subsidising Grandmothers’ Time

In the following paragraphs, I explain how subsidies on grandmothers’ time are introduced in the model, and how to compute the fiscal cost associated with this policy. To keep the analysis tractable, I abstain from modelling the endogenous decision of whether to provide childcare or not from the point of view of the grandmother.^[29] The idea is that a certain fraction of grandmothers in the economy receive a subsidy to care for their grandchildren. The amount of the subsidy is computed as the product of the daughters’ hourly wages and the daughters’ working hours (for which they need childcare).^[30] The value of the grandmother’s availability parameter is set exogenously, together with the value of the subsidy, to meet the employment targets such that this experiment is comparable to the previous ones. The exercise is done in such a way that taxes adjust to finance the granny subsidy so that the fiscal balance stays the same. Grandmother availability turns out to be four times larger than that in the baseline economy (4 g 1 ), tax rates 9 % higher, and the subsidy received by grannies a third of the daughters’ wages ( ψ = 1 / 3 ).

The bottom panel of Table 10 shows the effects of the rise in the probability of grandmother-provided childcare for mothers of children aged 2 years or younger, from 8.5 % in the baseline to 34 % . This policy generates a 13 % increase in the employment rate of these mothers and a 4 % increase in the employment rate of mothers of children aged 3–5 years. Consequently, the percentage of working mothers with children aged 5 years or younger, who use grandmother-provided childcare, increases from 13 to 45 % . Greater access to informal childcare allows more women to reconcile work and family. These results are consistent with the previous empirical analysis, where it was established that there is a positive association between making use of grandparents and labour force participation (see Table 2 in Section 3.1).

Having greater access to grandmother provided childcare, not only positively affects the female employment rate, but also the fertility rate: the percentage of mothers increases by almost 9 % . As in the previous experiments, the increase in the number of mothers is mostly accounted for by women with tertiary education: the proportion of mothers over women raises by 10% for college women, while it increases by 7 and by 5% for women with low and medium educational levels, respectively (Table 11).

Conversely, the positive effects on employment are concentrated among women with non-tertiary education, in line with the positive association between making use of grandparents and labour force participation for low-education women shown in Table 2 in Section 3.1. The proportion of working mothers over mothers increases by 17 and 12% for women with low and medium educational levels, respectively. Concerning women with tertiary education, the rise in the proportion of working mothers over mothers is more modest for two reasons. First, as in the “double-subsidy” scenario, the rise in the fertility rate compensates to some extent the growth in labour involvement. And second, more than 80% of women with tertiary education were already working in the baseline. For highly educated women the availability of grandmothers is not as essential as for low-education women because they earn higher wages and, consequently, can afford paying for formal childcare. In addition, they face higher opportunity costs of not working.

4.4 Comparing Subsidising Formal Childcare and Grandmothers’ Time

If we compare the results of subsidising formal childcare and grandmothers’ time, the combination of the two types of subsidies for formal childcare, i.e. a “double-subsidy,” turns out to be more effective than the granny subsidy in terms of mothers’ employment and fertility.^[31] For each additional point increase in taxes, the aggregate employment and fertility rates in the double-subsidy scenario increase by 1.35 and 2.25% against 0.9 and 0.96% in the granny subsidy experiment (see these elasticities in the last column in Table 10). However, both experiments require a substantial tax adjustment, 7 and 9 % , respectively. If the aim is just to raise the employment rate of mothers of children aged 2 years or younger, subsidising childcare costs only in the first period is more effective because the fiscal effort is lower: for each additional point increase in taxes, the employment rate of mothers of children aged 2 years or younger goes up by 4.36 against 1.73% in the case of the double-subsidy experiment and 1.42 in the case of the granny subsidy experiment. The reason for this difference in terms of effectiveness is due to the higher fiscal effort needed in the latter scenarios to finance either half of the costs of after-school hours or grandmothers’ time.

Regarding the heterogeneous effects, the subsidy experiments seem to favour the employment rate of women without tertiary education or women married to non-tertiary educated men the most. The reason is quite simple. Subsidies to cover childcare expenses in the first period are means-tested; therefore, high-income families do not qualify for them. By contrast, the greater availability of grandparents does not depend on income. These institutional differences generate variations in the distribution of working mothers by the level of education. In particular, women with tertiary education lose less weight in the distribution of working mothers in the granny subsidy scenario. For instance, the growth in the proportion of college working mothers to college mothers in the granny subsidy experiment is 5.6 % against 2.2 % in the double-subsidy scenario.

The results obtained in this analysis contrast to what other authors have suggested. For instance, Cardia and Ng (2003) claimed that subsidising grandparents’ time is the most effective childcare policy. When they compare the effects of time transfers to that of money transfers, they arrive at the conclusion that money transfers may decrease female labour supply because the income effect can compensate the substitution effect. This result stems from the fact that money transfers are not made contingent on female employment. This is not the case in the present study. It is true that average working hours decrease, but the fact that more mothers participate more than compensates.^[32] Furthermore, Cardia and Ng (2003) do not compute the fiscal cost of the policy, which, here, is found to be quite significant. Rupert and Zanella (2018) also held the opinion that childcare subsidies may not be very successful due to the substitution between informal and formal childcare. This does not occur here. The drop in the percentage of mothers using free care is basically driven by the fact that more women decide to work.

All in all, according to the model, subsidies on grandmothers’ time do not seem to deliver better results than childcare subsidies because they are less effective/more costly. In addition, implementing a granny leave policy in the real world may be very difficult because the reasons why grandmothers do not help with their grandchildren may not be all due to the lack of monetary incentives. Rather, other caring duties (to other children, to their partners or to their own elderly relatives), their physical ability to care for a child (due to the prevalence of chronic illness or other health problems) may also play a role. Another factor that may undermine the importance of monetary incentives is social norms. Southern European grandmothers were likely helped by their own grandmothers and may tend to feel like their duty to help with their care. Last, the satisfaction a grandmother gets from being a “helper” to her family may also be relevant. How much do these factors matter versus opportunity costs is uncertain, and out of the scope of this article, but on the future agenda. Finally, considerations related to inequality and distributional effects of these policies would also seem to favour childcare subsidies versus granny leaves because childcare subsidies improve more the labour market opportunities of women at the bottom of the skill distribution.

5.1 No Subsidy for Working Mothers of Children Aged 2 Years or Younger

Some authors, Bick (2016) among others, claim that the rise in maternal employment is too small to conclude that the lack of subsidised childcare accounts for the low labour force participation of mothers with children aged 2 years or younger in Germany. In this section, the childcare subsidy for working mothers of children aged 2 years or younger is removed to test this hypothesis for the case of Spain.

The upper panel in Table 12 shows that the effects are non-negligible. The employment rate of mothers of children aged 2 years or younger drops by 27%, while that of mothers of children 3–5 years is also significantly reduced (by close to 14 % ). This is in stark contrast to the results obtained in Section 4.1, where the rise in the childcare subsidy only affected the labour force participation rate of mothers with children aged 2 years or younger. This difference is mainly driven by the huge drop in fertility as a result of the removal of the childcare subsidy: the percentage of women being mothers goes down substantially, particularly that of women not helped by grandmothers which drops by 25 % . Note also that in Section 4.1 average fertility was not affected, though with heterogeneous effects. This robustness exercise bring to light the relevance of a minimum level of childcare subsidies to prevent fertility from dropping and the presence of non-linear effects concerning childcare subsidies (Tables A9–A11).

With regard to average hours worked, they raise considerably. This is not driven by the fact that women – not helped by grandmothers and on the verge of participation – stop working when the subsidy is removed. It is mainly driven by women – not helped by grandmothers and married to non-college men – who decide not to have a child and therefore work fulltime; and by mothers (also not helped by grandmothers) who work more hours in the first period as a result of the income effect generated by the elimination of the subsidy (compare Tables A2 and A12 in the Appendix). This last behaviour explains the substantial rise in the number of hours that children aged 2 years or younger spend in childcare facilities, around 7 % , despite the elimination of the subsidy. Furthermore, the fact that childcare costs have become more expensive makes the use of free care by grandmothers more relevant, mainly in the first period: the percentage of working mothers using free care for children aged 2 years or younger rises by 10% points.

These changes in employment and fertility rates are different in magnitude depending on the educational category. The most responsive groups are women married to men without tertiary education and women with tertiary education. For women married to men without tertiary education, childcare subsidies are needed to finance childcare costs and to reconcile work and family. For college women, opportunity costs play a significant role. Table 13 shows that mothers without tertiary education are the main losers in terms of employment because they are more sensitive to the price: the proportion of working mothers over mothers falls by 100 and 78 % for women with low and medium educational levels, respectively; for college women it drops by only 7 % . On the other hand, women with tertiary education are much more affected in terms of fertility because of the higher opportunity costs of having children in terms of forgone wages: the proportion of mothers over women falls by 23 % for college women, while that of non-college women falls by only 16 % .

5.2 No Tax Credit for Working Mothers of Children Aged 2 Years or Younger

A monthly cash benefit of 100 euros for working mothers of children aged up to three was introduced in 2003 (Ley 46/2002 de 18 de Diciembre 2002). Some authors, Azmat and González (2010) and Sánchez-Mangas and Sánchez-Marcos (2008), have found evidence of positive effects of this policy on fertility and on the employment rate of mothers with small children. Table 12 shows to what extent tax credits for working women with children aged 2 years of younger matter for female employment and fertility by removing them. The aggregate effects on employment and fertility are similar to the ones in the previous scenario, but lower in magnitude. The elimination of this transfer not only affects women not helped by grandmothers, as in the previous scenario, but also those helped by grandmothers because this cash benefit is linked to employment, not to childcare expenses. The only requirement for mothers of children aged 2 years or younger is to be employed. This cash transfer is not means-tested (Table A13).

The middle panel in Table 12 shows that the employment rates of mothers in the first and second period fall significantly, 19 and 10 % , respectively. Fertility is also substantially reduced, especially for mothers not helped by grandmothers. On average, the removal of the cash transfer raises the number of hours children aged 2 years or younger spend with grandmothers and in childcare facilities by 6 % since most mothers tend to work longer hours to compensate for the loss of income. A non-negligible number, especially women with college education married to non-college men and not helped by grandmothers, decide not to have a child and work fulltime (compare Tables A2 and A14 in the Appendix). The reasons are twofold: first, these women have very high opportunity costs of not working; and second, the loss of the cash benefit matters a lot given their partner’s low income. As in the previous experiment, the changes in employment and fertility rates are different in magnitude depending on the educational category. Table 13 shows that again, women without tertiary education are the main losers in terms of employment, while women with tertiary education are much more affected in terms of fertility.