The e ect of child care prices on the labour supply of parents

Transcription

1 The e ect of child care prices on the labour supply of parents Tuomas Kosonen University of Helsinki and HECER Work in progress May 2009 Abstract We estimate the e ect of child home care allowance to parents labour supply. A municipal supplement to home care allowance in Finnish child care system provides exogenous variation to labour supply. The supplement allows us to compare labour outcomes of similar families that receive di erent home care allowances because they live in di erent municipalities. The results for mothers indicate a large statistically signi cant negative e ect to labour force participation when child home care allowance is increased. Acknowledgement 1 I would like to thank Roope Uusitalo, professor Richard Blundell and porfessor Jukka Pirttilä, as well as many seminar audiences for many helpful comments. I am also grateful of the nancial support from Yrjö Jahnsson Foundation, Finnish Cultural Foundation and Palkansaajasäätiö to costs arisinig from a visit to University College London. 1 Introduction Many mature economies face a problem of falling fertility rates (OECD 2007). The problem is that on the other hand these countries would like to increase national fertility rates and on the other increase employment rates of mothers. An interesting feature is that among the EU the Nordic countries have leading gures in both dimensions; employment and fertility rates. The Nordic countries also have etensive public day care systems, which has possibly contributed to current situation. 1

2 It is of policy interest how strong incentives various public day care systems create. To provide reliable answers, one should nd exogenous variation to labour supply from these institutions. In this paper we will estimate an e ect of a policy on mothers labour supply that hopefully provides us exogenous variation. Theoretically, when studying labour supply, the incentives of various policies can be divided into two margins, extensive and intensive margin. It has remained empirical question, though, how large the labour supply elasticities over these margins are. One line of research that has tried to estimate sizes of these elasticities have used variations in in-work bene ts. Such reforms are for example Earned Income Tax Credit (EITC) in USA and Working Families Tax Credit (WFTC) in UK. It usually has been found that the elasticity over extensive margin is greater than the elasticity over intensive margin (Eissa and Liebman 1996, Blundell and MaCurdy 1999, Blundell et al. 2005). There has been empirical research speci cally about how much child care prices a ect the labour supply of parents. Among older papers there has been confounding results, that may partly be a result of various econometric methods (see Blau and Robins 1988 and Blau 2003 for surveys). Among more resent research Lefebvre and Merrigan (2008) and Baker et al. (2005) both study a Canadian reform of They use as identi cation strategy new policy implemented in Quebec according to which monthly price for day care was 5 Canadian $ per child. They compare work outcomes of mothers eligible to subsidised day care price with mothers living in other provinces. Both articles nd positive impact on labour force participation. The problem with two above mentioned papers is that there was only one treated group, Quebec in Milligan and Stabile (2007) have variation in the National Child Bene t in many provinces of Canada. The reforms clawed back bene ts in some provinces while in some other provinces they did not. This provides a case to study how child bene ts that operate like inwork subsidies provide incentives for labour market participation. Milligan and Stabile simulate incomes and bene ts based on observable characteristics and use triple-di erence method to study the e ect of the reforms. They nd that additional 1000 Canadian $ in annual bene ts increased labour market participation by around 4 percentage points. Lundin et al. (2008) use a child care reform in Sweden to identify the e ect of child care prices to labour supply of parents. They perform di erence-indi erences matching estimator. The results indicate that there are no impact from the child care price on labour supply or income of parents. Identifying variation in this study comes from a Swedish municipal level day care reform. Schone (2004) studied cash-for-care subsidy reform. A new system was introduced to Norway that provided subsidies to mothers who take care of 2

3 their children outside of the public day care network. This reform is interesting because it has similar idea than the Finnish reform we are going to study in this paper. Schone uses triple-di erence method to estimate the causal e ect of the reform, since it was introduced everywhere in Norway at the same time. His results indicate a modest negative e ect for mothers labour force participation due to introduction of the bene t. We observe that in Finland the labour supply of women in general is on quite high level compared internationally. However, the participation rate of mothers of young children is very low (OECD 2005). Finnish child care system might have helped to create this situation. It provides publicly provided child care for everybody who wants it and at the same time supports home care. There are spacial variation in municipal home care allowance and many instances when municipalities have changed their subsidy levels. We thus have many reforms in a policy that provide signi cant di erences in incentives to participate to labour force. We provide new features to earlier literature at least in two dimensions. First, the policy reform we are studying provides signi cant variation both over time and across regions in child care related subsidies. Second, we have variation in prices that should increase elasticity over the extensive margin of labour supply. Earlier papers have studied reforms where taxes on extensive margin were reduced. We will be able to present the estimates for participation elasticities using the municipal supplement to home care allowance as an instrument for change in income associated with participation. The rest of the paper proceeds as follows. We present in the following section 2 the source of variation we have for the child care prices and argue why this variation should be exogenous. We will also provide a short description of Finnish child care system. We discuss the identi cation issues and econometric speci cation in section 3. In section 4 there will be description of our data set and descriptive statistics. We proceed to showing estimation results in section 5. In the same section there will be some robustness checks. Our estimate for participation elasticity as well as relevance of the reforms are presented in section 6. Finally there are the conclusion of this study in the last section. 2 Forms of child care Child care can be formulated in many ways. To understand the specialities in Finnish institutions, we rst compare those to intitutions in other countries. Then we move on to describe Finnish day care in more detail and especially how these institutions provide exogenous variation to parents labour supply. 3

4 Employment and Fertility in Selected Countries 90 2,5 Maternal employment rate by age of youngest child Finland Denmark Sweden Canada United States OECD average 2 1,5 1 0,5 0 Fertility rate < Fertility rate 2005 Figure 1: Source: OECD How Finland is doing compared to other countries? In Nordic countries the public day care is extensive. Children are entitled to have a place from public day care center. The price for day care is heavily subsidised by the government. The idea of this system is partly to encourage women to participate labour markets. In Anglo-saxon and Central European countries the day care relies more on private markets. The price a household ends up paying for day care can be much higher there than in Nordic countries. Two key features that countries try to a ect with their child care institutions are mothers employment rate and fertility rate. Aggregated statistics for these can be seen from Figure 1. It is evident that Nordic countries have better employment rates and also higher than OECD average fertility rate, also the USA have higher fertility rate. The rst column stands out for Finland in Figure 1, the employment rate of mothers with youngest children under 2 years old. Interestingly in Finland there is much more generous home care allowance for these families than in Other Nordic countries. The public day care system is as generous as in other Nordic countries. When families are no longer applicable to this allowance, the mothers employment rate shoots back to high gures. 4

5 2.2 Finnish day care In this section we describe how child care is provided in Finland. The idea of Finnish system has been to provide nancial assistance to a parent regardless of the choice a parent makes. After a maternity leave (when the newborn child is 9 months old) a parent can choose essentially from three child care alternatives that are nancially subsidised by the government: home care, public day care or private day care. Every child under the age of 7 (when they start primary school) are entitled to a public day care place if they ask for it 1. The day care fees are controlled by the government and a typical family with two children in public day care paid 380 e per month in 2005 from it. Thus, public day care is the predominant choice in Finland for a typical family. Also private day care is subsidised 2. The private day care allowance works by giving allowance directly to care provider. Municipalities are also able to pay municipal supplement to private day care allowance, if they choose to. In this study we are not interested in whether a family chooses private or public day care. These two choices are similar regarding the employment decision. The form of day care matters for its price, though. Since public day care is usually cheaper than private, we can think of estimating the lower limit of change in net income associated with participation when restricting the analysis to the choice between home care and public day care. When a child under three years of age is taken care of by a parent, he or she is entitled to a child home care allowance. It can be paid until youngest child reaches the age of 3 and who are not in public or private day care. The amounts depend on family characteristics and are between euros per month. It is possible for either of two parents to receive the child home care allowance, although it is dominantly mother who takes up the allowance. The interesting variation for this study comes from the municipal supplement to home care allowance. Some municipalities have decided to pay supplement on top of national home care allowance while other municipalities have not adopted the supplement policy. The municipal supplement have belonged to Finnish child care system from 1980 s. In our observation period, from 1995 to 2005, there were 5 municipalities in the beginning and 65 in the end that had adopted the supplement policy. Municipalities also 1 A law states this. Before 1995 the law stated that every child under the age of 4 are entitled to public day care. 2 This system has been nationwide in place since Between 1995 and 1997 there was an experiment in 33 municipalities that provided similar allowance. Viitanen (2007) describes this experiment in detail. She found a positive e ect to the use of private day care, and little e ect to labour force participation. 5

6 change their policy rules over time. Although there are strict rules how each municipality should provide day care, the are not that much rules about municipal supplement. Thus, there is a lot of variation in the details how each municipality pay the supplement. Typically the municipal supplement is paid per child. It is possible to receive extra supplement if the youngest child has older siblings. The mean monthly supplement level in our data is 200 e and the mean sibling extra is 50 e per family. The municipal supplement, apart from a small number of municipalities, do not depend on family income. There is also a pre-work condition in some of smaller municipalities, according to which to be entitled to the municipal supplement, a parent needed to have a work place prior to parental leave. Because it is the municipality council who decides to pay the supplement, we have exogenous policy to the labour supply of an individual. There are potential problems, though. One is policy endogeneity discussed by Card and Levine (2000) and Lalive and Zweimuller (2004). If a municipality chooses the supplement because they are experiencing bad labour market shocks, the supplement policy itself would be correlated to employment outcomes. Other kind of potential problem arises if people choose the municipalities they live in based on supplement policies. There are not much literacy references about Finnish municipal supplement policy to provide reasons why municipalities decide to adopt it. Potential reasons include political considerations, for example right-wing and center parties in Finland have supported the idea that any choice a parent makes regrading child care must be supported by the public sector, not just day care. Judging from the increasing number of municipalities that have adopted the policy, it seems that it has become popular to have the supplement policy in a municipality. To provide a better answer to problems raised above, we provide descriptive statistics. Two groups are calculated from municipal level data from Statistics Finland. The other group, "Supplement", had the municipal supplement in some or all of the years The other group, "No Supplement", did not have the policy in any of the years. The gure 2.2 shows employment rate trends for women aged between 25 and 39 years for the two groups. The shaded areas and capped bars show con dence intervals around the means. It is apparent from the gure that although there is a small di erence in the means, the trends follow each other pretty well. Thus, there does not seem to be a problem about policy endogeneity. Many municipalities adopted the supplement policy around year 2000, but womens employment rate does not show any signi cant deviation from the trend before that year. 6

7 ,6,65,7,75,8 Employment rates, women 25 to 39 yo Year Sd*(+/-0.5) Supplement Mean Supplement Sd*(+/-0.5) No Supplement Mean No Supplement 7

8 Figure 8 in Appendix shows two migrant statistics relative to inhabitants in a municipality. The left panel shows total inland migration to a municipality and the right panel is corresponding gure for babies aged 0 or 1 year. The di erence between the means is larger than with employment rates, but the trends follow each other pretty well. Even if some of the families moved in hope of getting the supplement, their share of total migration was so small that it does not show up in average statistics. Similarly gure 3 compares fertility rates between treatment and control groups. Again no clear di erence between two trends is found. The fertility is something policymakers would have like to be a ected, but in this case it does not seem to be the case. Based on the two gures shown, there does not seem to be serious policy endogeneity problem here. We perform more rigorous robustness tests with microdata after the main estimation results. 3 Identi cation and econometric strategy We consider here on what conditions our model is identi ed and the potential threats to identi cation when those conditions are not ful lled. We use di erences in di erences (DD) strategy to estimate the e ect municipal supplement have on parents labour supply. We use a linear probability model in our estimations. Our dependent variable is labour supply Y, that is a dummy variable having value 1 when a parent participates and zero otherwise. The key explanatory variable is P (municipal supplement). We estimate the effect of price P iym on labour supply Y iym in year y, municipality m and of individual i. The estimated equation is thus: Y iym = + 1 P iym + 2 X iym + 4 Mun m + 5 Y ear y + iym (1) The aim is to identify 1, the e ect price variable has on labour supply of parents. The other variables in equation (1) are municipal Mun m and year Y ear y level dummies and controls X iym. The model in equation 1 is estimated for a population of parents. If 1 is identi ed properly it tells how much larger probability a parent has to supply labour when P is increased one unit conditional on observable characteristics. Thus, any change in P is allocated to simultaneous change in Y. The only di erence between the groups should be that they have di erent prices P, because they live in di erent municipalities. Thus, the identi cation relies on municipal and year level factors. That is why we specify the controls to include municipal level variables such as average unemployment rate and share of children in day care relative to the number of children in that municipality. We also 8

9 include individual level controls to cope with the individual level variation, this should reduce the variation of error term. 3.1 Identi cation issues To be more precise about the identi cation of the e ect municipal supplement has on labour outcome Y iym, let us say that Yiym 0 is outcome of the control group. Then the model identi es 1 conditional on controls if the following condition holds: E Y 0 iymjm; y; P iym ; X iym = E Y 0 iym jm; y; X iym = m + y + 2 X iym (2) The above equation states the di erence in di erences assumption that selection into treatment should be exogenous to outcome. The assumption that guarantee that this is the case here is that P iym is exogenous to labour supply. We should be sure that municipalities do not select the supplement based systematically on employment situation. Also the aggregate employment time trends should be parallel between treatment and control groups. Finally the composition of groups should be similar; people should not move from municipality to other based on prices. After the estimation results we present some robustness checks as a defence against some of these potential problems. Against potential di erences in macro trends that are not taken care of by control variables, we estimate following equation: Y icym = + 1 P icym + 2 X icym + 4 Mun m + 5 Y ear y + icym This is triple-di erence equation, where c refers to whether the family is entitled to have municipal supplement by the age of the youngest child or not. The third di erence is taken between families whose youngest child is either between 9 months and three years old or between three and ve years old. The families in the older age group should not have any variation in family bene ts across control and treatment groups, mothers in these families are not entitled to home care allowance nor the supplement to that. Now in the term X icym there are also second level interactions between year, municipality and age group dummy of the youngest child. Estimating 1 identi es the e ect municipal supplement has between di erent municipalities and between families of di erent age structure in those municipalities. The third di erence is then substracted from the other di erences clearing out 9

10 any spurious time trend e ects. Thus, this equation deals with the problem of di erent macro trends across municipalities. There are empirical reasons to believe that municipalities pay the supplement randomly, and not related to work outcomes of parents. We refer to Finnish child care institutions that were described in section 2. It seems municipalities have an incentive to o er supplement in order to keep children out of expensive public day care. This might look especially lucrative from municipality point of view, if there is a baby boom in the municipality. However, there is a state grant system, that provides subsidies for the municipalities according to number of children between 0-6 years old lived in the municipality. These grants are not ear-marked, so municipalities receive the grants independent of how many children participated to public day care, leaving some incentives for the municipality to keep children out of public day care. However, especially after 1997, if municipality experiences unfavourable economic conditions, it should actually be better idea to provide supplement for private day care allowance than to home care allowance. If the policy follows these suggested lines, the parents are more likely to work and contribute to tax revenues of the municipality. So, the existence of supplement to private home care allowance is better instrument to reduce the demand of public day care, if that is the only motivation for the supplement. Our second line of evidence is to describe empirically the macroeconomic conditions. The descriptive statistics are shown in gures 2.2 and 8. Since our econometric speci cation allows for municipal level di erences and time trends, we are not worried about di erences in means as such. More worrying would be di erences in time trends between the groups, and such di erences are not visible in the gures. 4 Data and Descriptive statistics We use individual level microdata from years 1994 to Our data comes from multiple sources. The base data, Income Distribution Statistics, comes from Statistics Finland and is an individual level data, containing over observations from about households per year. From these we have for all years about 6000 households that have children between 9 months and 3 years old and about households that have children under age of 6. In the data we have a rich set of variables describing family characteristics, demographics, incomes and bene ts coming from registers and surveys. The rest of the information is on municipal level and it has been linked to the individual data. It comes from a survey to municipalities conducted by Uni- 10

11 versity of Turku, own survey to municipalities, from the Social Insurance Institution of Finland and from Statistics Finland. The data is a repeated cross section on individual level, although there is a rotating panel system. 3 We can also make a municipal level panel from our data. 4.1 Child care subsidies and prices The home care allowance, municipal supplement to it and day care prices are determined by family characteristics and income. We implement these prices to families based on eligibility rules described in section 2. For each family two prices are calculated, the other corresponds taking care of the children at home and the other corresponds to children being in public day care. If there are multiple children only the case that all the children are treated in a similar way is calculated. Below we show tables and gures that describe the relative day care prices as well as other relevant characteristics from the data. To get some handle on how large the implemented level of bene ts are compared to mean net income associated with full time work, we present the yearly averages from our data of these variables in table 1. The Family bene ts column shows the home care allowances, the Day care fee column the public day care fees and the Supplement column the municipal supplement conditional on being eligible to it (zeros not included). Families that are eligible to the supplement would receive on average the Family bene ts plus the Supplement if the children were treated at home. Net incomes are calculated net of taxes from parents of 0-6 years old children that are in full time work, separately for mothers and fathers. Figures in the table 1 are in nominal terms. From the table 1 it is evident that the total cost of going to work, forgoing Family bene t plus Supplement and paying Day care fee, is a substantial part of net income. If a mother is eligible to the supplement, the total bene ts and day care fees added are equivalent to quarter of income. We note that there are other bene ts in Finland that families usually receive, but are not related to children or the decision to participate to labour force as such. We do not consider those potential bene ts, because there is no municipal level variation in them and they do not change the relative price of going to work. 4 In Figure 2 there is scatterplot of the municipal supplement and family 3 In rotating panel each household is surveyed in two consecutive years and each year half of the sample consists of new households. The data is a panel on municipal level Thus, there are two consecutive observations for each individual. 4 Except household allowance has regional level variation. We can control for that in the estimations. 11

12 Trends in mean bene ts and incomes over time Year Family Day care Supple- Net income Net income Number bene ts fee ment if >0 (Females) (Males) of obs Table 1: Mean level of bene ts and incomes. Incomes are calculated from full-time workers that have young children. income for those who are eligible to it. First thing to notice is that there is no apparent income dependency for the supplement. On the other hand having more children leads the supplement to be higher than with just one child. The amounts received ranges between 420 e and close to zero. There are fewer observations in earlier years, because many municipalities did not pay the supplement then. 4.2 Employment rate One substantial part of the story is the employment rate. We construct our dependent variable from variables that show how many months people have been supplying labour either full- or part-time each year. We encode this as being zero or one, having value one when the individual has supplied more than ten months labour in a year. 5 The employment trends over time for mothers by the age of youngest child are shown in Figure 4.2. The gure shows clearly how the employment rate of mothers increases when their youngest child gets older. There is also a downward trend in the employment of mothers of 2 years old child, that could be caused by expansion of municipal 5 The number of months supplying labour - variable was constructed so that most of the mothers had either 0 or 12 months in a year. Thus, changing the de nition of participation between 1 and 11 months doesn t a ect results much. 12

13 Municipal Supplement Income One Child Two Children Three Children Graphs by Year Figure 2: A scatterplot of municipal supplement and day care fees by the number of children in the family over selcted years. 13

14 0,2,4,6 Mothers employment rate Employment trend by the age of child Year 1yo child 3yo child 2yo child 4yo child supplement at the same time. These employment rates are somewhat lower, but consistent, with those found in OECD (2007). They nd in OECD (2007) that the employment rate for women increases rapidly with the age of the youngest child. The descriptive statistics from some of the demographic variables are shown in the table 2. It is divided into two panels corresponding to control and treatment groups. The control group consists of municipalities where there is no supplement policy in any of the years. In the treatment group municipalities have the supplement policy in some or all of the years. The gures in the table are calculated form a population of mothers whose youngest child is between 9 months and three years old. Av degree is average of ve-step measure of education degree, Participation shows average of participation dummy and income is average of individual earned income, not family income (in euros). It is quite low, because participation rate is also low. Overall the treatment and control groups seem to have quite similar caharacteristics. 6 6 We also tried to limit the sample to municipalities where these cahracteristics have even closer match. It did not change estimation results. 14

15 Descriptive statistics. Mothers of 0-3 yo in control and treatment groups Year Control group, N of obs has never Age supplement Av degree Participation Income Children Treatment group, N of obs supplement Age in some years Av degree Participation Income Children Table 2: Descriptive statistics for mothers whose youngest child is between 9 months and three years old. Upper panel for those who live in municipalities that do not have supplement policy in any of the years, lower for those who live in municipalities that have the policy in all or some of the years. Figures show mean level, except the number of observations (upmost statistic). 15

16 COEFF (i) (ii) (iii) (iv) Supplement *** *** *** *** (0.0097) (0.0096) (0.023) (0.0211) Constant ** 10.2** -20.1*** (0.11) (0.34) (4.26) (0.581) Observations R-squared Indiv controls Yes Yes Yes Municipal controls Yes Yes 2nd level interactions Yes Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Table 3: Estimation results, mother s labour supply as a dependent variable. 5 Estimation results We present the main estimation results below in table 3. In each column the dependent variable is mothers labour supply dummy. The measured monetary variables (like the municipal supplement and unearned income) are in 100 euros per month. We performed all the estimates to fathers as well, but did not get any statistically signi cant results for municipal supplement. The results in table 3 are organised as follows: in column (i) there is a di erence-in-di erences (DD) estimate without any controls besides year and municipality dummies. In column (ii) individual level controls are added, including indicator whether spouse works and interactions with indicators for each child and age of mother. In column (iii) there are also municipal level control variables making this estimate the most exible DD estimator shown. The control variables include individual demographic characteristics like age, age squared, four-step measure of school degree and an indicator for each child. The municipal level controls include average unemployment rate and aggregate income in the municipality, share of children in day care and the municipal income tax rate. Finally in column (iv) there is the tripledi erence estimator. The number of observations now increases, since there are also mothers with older children in the estimates. The point estimate of the supplement is a little bit larger than in column three. Standard errors take into account that unobservables are correlated in municipal level. Our preferred estimate in column (iv) indicates that increasing the municipal supplement by 100 e a month causes 10% less women to participate. Since there is probably variation in how mothers respond to municipal sup- 16

17 plement, we interpret this as average treatment e ect on the treated. Those who receive municipal supplement are treated in this case. We included the triple di erence as a check against potential di erences in macro trends. It is reassuring that the result was close to DD estimate. The result in column (iv) seems to be robust with quite exible set of control variables. Since the point estimate do not change much, when conditioning on municipal level variables, the result does not depend directly on macroeconomic conditions of the municipality. We performed many robustness checks that did not change the result in a signi cant way. These were: exclude municipalities with income dependency in the supplement rules, exclude those with previous work condition and exclude any one of the largest municipalities. In one of the checks we only took years after 1997 into account with no signi cant e ect on the results. There are various threats to identifying true average treatment on treated e ect with our chosen strategy, as discussed in identi cation section. As a defence against policy endogeneit, we perform out-of-sample robustness checks in table 4. In column (i) we introduce a pseudo-rule that made mothers whose youngest child was between 3 to 5 years old eligible to the municipal supplement if they lived in a supplement municipality. The estimates are otherwise similar to ones in table 3. The families that have older children seems to be a natural candidate for out of sample check, since their characteristics should otherwise be close to families that have just little bit younger children. We do not get signi cant estimates from out of sample rule in column (i). This was same group that we used for third di erence in triple-di erence estimate. The zero result here shows that there were no delayed e ect for mothers employment from the supplement policy. The result is in line with OECD (2007) statistics that indicate that mothers employment with children of this age is much higher than mothers with younger children. In column (ii) of table 4 we estimate an out-of-sample robustness check for a di erent group. Here we utilise the rotating panel feature of the data and take into estimation sample families that will have a child next year aged 9 months or younger, but do not have this year any children that are between 9 months and 3 years old. Thus, they are not entitled to municipal supplement yet, but live in municipalities that have the policy. This estimate should tell something about potential anticipation e ect. However, the point estimate is not statistically signi cant. This indicates that there is no serious anticipation e ect (although we do notice that the sample size is only 533 in this estimate). In column (iii) we again use the rotating panel feature, this time to take into account the e ect from people who moved. We include this as an indicator to the model and interact it with the Supplement variable. If people 17

18 COEFF. (i) (ii) (iii) (iv) Supplement *** *** (0.0160) (0.0654) (0.023) (0.021) Spouse *** *** works (0.0116) (0.0495) (0.023) (0.015) Moved *Supplement (0.049) Constant *** 10.11*** (5.417) (24.29) (6.587) (3.663) Obs R-squared Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Table 4: Some robustness checks to the municipal supplement estimates. All estimates have mother s labour supply dummy as a dependant variable moved in hopes of getting higher municipal supplement, and do not supply labour because of that, the interaction coe cient should be negative. Our estimated coe cient is positive and again not signi cant indicating that people do not move in search for higher bene ts. In column (iv) there is base-line estimation, but with pseudo-rule, since we needed to simplify rules for the implementation of pseudo-supplement. Coe cient for the municipal supplement is similar to main estimates, the simpli cation of rules do not seem to a ect estimates. In table 5 we present results for how much 100 euros per month of municipal supplement a ects incomes of mothers. The idea of estimating the e ect of municipal supplement has on incomes is that explanatory variable is now continuous and measures change in income throughout the year. The income used here includes all the personal income the mother gets from work, excluding bene ts related to child caring. The estimates are similar to 3, except for the dependent variable. Our preferred estimate, the DD in column (ii) indicates that for every 100 euros a mother receives municipal supplement she earns 77 euros less income. Given the large participation elasticity, one could have expected larger income response. However, the total income that people receive when not participating in Finland is not that much lower to total income when participating. This makes the e ect on income lower. 18

19 COEFF. (i) (ii) Supplement *** -76.6*** (24.52) (12.03) Age 24.36*** (2.824) School 401.0*** degree (16.97) Constant 494.6*** (50.78) (8130) Obs R-squared Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Table 5: Estimation results, mother s income as a dependent variable. 6 Economic relevance of results After presenting the estimation results in previous section, we turn our interest to economic relevance of those estimates. We rst aim at estimating the participation elasticity for mothers and after that present some rough calculations about the costs of these policies to local economies. We do not calculate elasticity direclty from point estimates presented earlier, since then we would consider an elasticity of a subsidy. We instead use information in our dataset on incomes for those who work and for those who do not. This di erence would itself su er from endogeneity problem, so we use municipal supplement as an instrument for the change in incomes associated with participation. To validate the use of this instrument, we have done above a lot of robustness checks to convince that municipal supplement is exogenous to labour supply of an individual. On the other hand it a ects directly the income one gets when the change in participation status is made. To be able to assess and simulate incomes for counterfactual case, we need to make some further assumptions; we only calculate the elasticity for couples whose father is working with average income and whose youngest child is between 9 months and 3 years of age. This is the predominant type of household in the data. We do not have enough observations from single mothers to calculate any robust estimates. For a mother in this type of family the average bene t when not participating and excluding municipal supplement were 520 euros per month. The coe cient should have positive sign because the subsidies enter the income equation with a negative sign. 19

20 COEFF (i) (ii) Change in income *** *** ( ) ( ) Spouse works 0.095*** (0.012) School degree (0.017) Constant -0.59*** -0.8** (0.12) (0.37) Observations R-squared Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Table 6: Labour supply dummy as a dependent variable against change in incomes instrumented with municipal supplement. Estimated for population of mothers with youngest child between 9 months and 3 years old for years The incomes taken into consideration are net of tax full time working income (that was predicted to people) minus day care fees minus home care allowance (and the supplement when entitled to it). This was then instrumented with municipal supplement. The second stage results from 2SLS are shown in table 6. The explanatory variable is again a dummy for participation status. Now the unit is 1 euro per month. In column (i) there is estimate without any extra controls, in column (ii) the estimation included full set of controls. According to these estimates the amount that mothers gain when they participate to labour leads to increase in participation of.08 % from 1 euro per month. The participation elasticity implied by the coe cient in the table 6 can be calculated as dparticipation dincome income = participation income participation = 0: :25 = 2:6 Where 800 euros is the net income of mothers on average in the estimation sample. This elasticity is huge compared to estimates in the literature, but the result seems robust to di erent estimation speci cations. One factor that makes this elasticity so large is the small base line participation rate (0.25) of 20

21 mothers. If the participation rate were closer to levels of women in this age group in general (0.7), we would get elasticities closer to one. The elasticities in previous literature from day care fees are not direclty comparable since they do not use the total change in incomes as explanatory variable. To report some estimates, Baker et al. (2005) estimated participation elasticity of from decreasing child care cost, and Milligan and Stabile (2007) reported elasticity of 0.96 for having earnings as a major source from the Canadian bene t reform. To asses the e ect of municipal supplement to municipal economy, we include here a very rough estimate of costs and bene ts the policy potentially creates to municipalities. We consider a mother living in a couple with two children and in municipality with a typical supplement policy. We compare the costs and bene ts of supplement policy assuming a mother decides not to participate because of the supplement. On bene t side from the municipality point of view the municipality do not need to provide day care for the two children. The cost of day care to a municipality per child is roughly 600 euros minus the amount the mother would pay themselves, 380 euros. In all this makes ( ) = 820 euros saved from day care costs. On the cost side municipalities need to pay the home care allowance plus the supplement and lose income taxes. The average municipal supplement level in 1 is little bit less than 200 euros. The home care allowance is on average 300 euros per family.the monthly gross income of mother who is participating to labour force is on average 2200 euros. The e ective income tax rate in municipalities is around 16 % making the loss from income taxes 0: = 352 euros per mother per month. The total cost-bene t calculation per mother with assumed characteristics shows = 32 euros loss. We ended up with a negative number, but not very large. The deciding factor here is clearly the loss in tax income. If the same calculation was made from the public sector as a whole point of view, it would have been much more negative, since the average tax rate is higher in central government taxation and people who do not participate tend to get some other means-tested bene ts from the central government. This di erence between local government and central government cost bene t analysis points why it could be tempting for the municipalities to have the supplement policy, although the total welfare e ect was more negative. 21

22 7 Conclusion In this paper we have estimated how much child care bene ts a ect maternal labour supply. The Finnish municipal supplement to child home care allowance provides exogenous variation to parents labour supply. To be entitled to municipal supplement a parent needs to stay at home taking care of children, to have children of certain age and live in municipality that has the policy in place. Because all municipalities do not have supplement policy, we are able to compare parents with similar observable characteristics in control group to those who are entitled to the supplement. We have variation over time as municipalities changed their policies many times during our estimation period. The municipal supplement to home care allowance has a negative e ect on mother s labour supply. Our main estimate indicates that increasing the municipal supplement by 100 euros per month causes 10 % fewer women to participate. We did not nd any signi cant e ect to fathers labour supply. After estimating the direct e ect of supplement policy to mothers participation we estimated the elasticity over the extensive margin of labour supply. We used the supplement policy as an instrument to the increase in income arising from participation. From the second stage we could calculate an estimate for participation elasticity, that was 2.6 for mothers in couples were father is participating. Our estimates seem large compared to many earlier papers. These result, on the other hand, are in line with aggregate statistics (OECD 2007) showing low mothers employment rate when the youngest child is under 2 years old, but reaching the high levels of other Nordic countries when the youngest child is little older. This paper provides a reason for these statistics, generous child home care bene ts, and estimates the participation elasticity of these bene ts in population of mothers. Our results also show that when the home care allowance period ends, the mothers return to employment. Thus, there does not seem to be delayed e ects of these policies. We did not estimate the e ect of these policies on fertility with microdata, but gure 3 indicates that these policies did not have large impact on fertility rates. The mother who had a job prior to maternity leave, can return to the same job after the home care allowance period, provided the same job still exists. The home care allowance system is di erent from unemployment bene t system, where people do not have any job to take up whenever they feel like it. Also the value of staying at home for a mother of young child is larger than for an unemployed women. Thus, these features of home care allowance system enhance the incentives for mothers to take up the bene t, and consequently we estimated a large participation elasticity. 22

23 To be able to assess optimality of public day care policy as a total in Finland, we should have some idea of the e ect these policies have on children themselves. This remains a question for future research for now. References [1] Baker, Gruber and Milligan, 2005: Universal Childcare, Maternal Labour Supply and Family Well-being. Working Paper 11832, National Bureau of Economic Research, [2] Bertrand, Du o and Mullainathan, 2004: How Much Should We Trust Di erences-in-di erences Estimates? Quarterly Journal of Economics February 2004, Vol. 119, No. 1: [3] Blau, 2003: Child care subsidy programs. Working Paper no. 7806, National Bureau of Economic Research, Cambridge, MA. Reprinted in Robert Mo tt, ed.: Means-tested transfer programs in the United States, Chicago: University of Chicago Press. [4] Blau and Robins, 1988: Child-Care Costs and Family Labour Supply. The Review of Economics and Statistics 70: [5] Blundell, Brewer and Shephard, 2005: Evaluating the labour market impact of Working Families Tax Credit using di erence-in-di erences. Institute for Fiscal Studies, Externally published reports, June [6] Blundell and Macurdy, 1999: Labor supply: A review of alternative approaches. Handbook of Labor Economics, in: O. Ashenfelter & D. Card (ed.), Handbook of Labor Economics, edition 1, volume 3, chapter 27, pages Elsevier. [7] Card and Levine, 2000: Extended bene ts and the duration of UI spells: evidence from the New Jersey extended bene t program. Journal of Public Economics 78, pages [8] European Comission, Ten years of the European Employment Strategy (EES). Directorate-General for Employment, Social A airs and Equal Opportunities, Unit D.2. [9] Eissa and Liebman, 1996: Labor Supply Response to the Earned Income Tax Credit. The Quarterly Journal of Economics, Vol. 111, No. 2 (May, 1996), pp

24 [10] Laine and Uusitalo, 2001: Kannustinloukku-uudistuksen vaikutukset työvoiman tarjontaan. VATT Research Reports 74, Helsinki. [11] Lalive and Zweimuller, 2004: Bene t entitlement and unemployment duration. The role of policy endogeneity. Journal of Public Economics 88, pages [12] Lefebvre and Merrigan, 2008: Child-Care Policy and the Labour Supply of Mothers with Young Children: A Natural Experiment from Canada. Journal of Labour Economics, 2008, vol. 26, no.3, The University of Chicago. [13] Lundin, Mörk and Öckert, 2008: How far can reduced childcare prices push female labour supply? Labour Economics, forthcoming in 2008, Elsevier B.V. [14] Milligan & Stabile, The integration of child tax credits and welfare: Evidence from the Canadian National Child Bene t program. Journal of Public Economics, Elsevier, vol. 91(1-2), pages , February. [15] OECD, 2005: Babies and Bosses - Reconciling Work and Family Life (Vol. 4): Canada, Finland, Sweden and the United Kingdom (2005). [16] OECD, 2005: Society at a Glance: OECD Social Indicators 2005 Edition. OECD, Volume 2005, Number 2, March [17] OECD, 2007: Babies and bosses: Reconciling work and family life. OECD, [18] Saez, 2002: Optimal Income Transfer Programs: Intensive versus Extensive Labor Supply Responses. The Quarterly Journal of Economics, Vol. 117, No. 3, (Aug., 2002), pp , MIT Press. [19] Schone, 2004: Labour supply e ects of a cash-for-care subsidy. Journal of Population Economics, 2004, vol. 17, Springer-Verlag. [20] Viitanen, 2007: Childcare voucher and labour market behaviour: Experimental evidence from Finland. She eld Economic Research Paper Series, SERP Number: , United Kingdom. 24

25 Appendix One potential (and even desired) e ect of larger home care allowance is that women are more fertile. Based on Figure 3 below this does not seem to be the case here. The gure describes ratio of all living births in a municipality to number of women in same municipality. It is di cult to see any signi cant change in trend due to supplement policy reform in Supplement group. Ratio of Births to Women Year Sd*(+/-0.5) Supplement Mean Supplement Sd*(+/-0.5) No Supplement Mean No Supplement Figure 3: Supplement group contains municipalities that have the supplement policy in some or all of the years and No Supplement group contains the rest of municipalities. 25

26 ,03,04,05,06,07,001,002,003,004 Ratio of migrants to inhabitants Baby-migrants to inhabitants Year Year Sd*(+/-0.5) No Supplement Mean Supplement Sd*(+/-0.5) Supplement Mean No Supplement 26