Lifetime Impact of Cash Transfer on Fertility

In most OECD countries, fertility level is below the natural generation replacement level. A number of undesirable social and economic effects stem from decreasing population and rising age composition of population. To counter decreasing fertility, many OECD countries implement pro-natal policies including direct cash transfer schemes. Impact of cash schemes are estimated in many countries including Germany, Sweden, France and Israel. However evaluations of long run impact of such policies are surprisingly rare. We investigate whether the cash transfer increases completed fertility, exploiting quasi-experiment arising from a pro-natal cash transfer called ANC (Allowance for Newborn Children). We first devise a measure of ANC impact for different birth cohorts using age-specific fertility rates because the policy lasted for a decade and it affected cohorts with different intensities at different ages. The results show that ANC impact on fertility has little permanent component.


Introduction
Fertility levels of most OECD countries started to fall rapidly in 1970s and their fertility levels were below replacement level by 1983 (See Figure 1).By 2002, total fertility levels of all OECD countries except Mexico and Turkey were below replacement level (OECD 2005).Assuming no external migration, population level will decrease in those countries and their population structure would be adversely affected.Decreasing population lowers the growth of real GDP, per capita GDP and domestic savings and changing population structure creates pressure on government budget as smaller number of working population will have to support greater number of pensioners.All OECD countries have tried to instigate birth by policy instruments including family cash benefits, childcare support programs, increased duration of maternity and parental leave and general social protection system.Cash transfer is one of the most popular pronatal policies.As of 1998, almost eighteen OECD countries were spending more than one percent of total GDP on family cash benefits to instigate birth including Australia, France and the United Kingdom (Sleebos 2003).Buttner and Luz (1990) and Hoem (1990) find that a set of pro-natal policies mainly aimed to reduce child rearing cost to instigate births in East Germany and Sweden were effective in increasing fertility.Recent evidence from France and Israel indicates that such policies are also effective in increasing fertility (Salanie andLaroque 2005, Cohen et al. 2007).Zhang et al. (1994) show that all of the fiscal incentives for giving births such as personal tax exemption for children, child tax credit, family allowances, and maternity leave benefits in Canada all have statistically significant and positive effect on fertility, using vital statistics from 1921 to 1988. 1 In Canada, provincial government of Quebec implemented a universal cash transfer scheme called Allowance for Newborn Children (ANC) from May 1988 to September 1997 to all legal residents of Quebec. 2 Milligan (2005) use household level Census data and find that implied 1 Personal tax exemption giving tax deduction to dependents was first introduced in 1918 in Canada.In 1945, the family allowance benefit to all families with dependent children was introduced and maternity leave benefits were available starting in 1971 as part of Unemployment Insurance Program.Later, child tax credit was introduced in 1977.
2 Duclos, Lefebvre, and Merrigan (2001) found a positive effect of the ANC on fertility using data obtained only from vital statistics.percentage increase in probability of having a child is approximately 12 percent as summarized in Table 1. 3 However the evaluation of long-run policy impact is surprisingly rare.
In this paper, we first devise a measure of policy impact on different birth cohorts, because ANC lasted for a decade and each birth year cohort of women were affected by the policy at different ages.Then we estimate effect of ANC on completed fertility level, the measure of lifetime fertility across time for each given cohort.We exploit quasi-experimental nature of the universal cash transfer program that lasted for a decade in only a part of the country to investigate if the policy has statistically significant impact on completed fertility.The results indicate that ANC had little impact on increasing completed fertility of Canada.

Empirical Strategy
Before we run our main analysis, we first replicate Milligan (2005) but we add age group dummy variable.The difference-in-differences method is used for the analysis of age-specific impact of the ANC to verify that the policy impact varies by age group.If different age groups are affected differently by ANC, some of the age-specific impact variation may be due to the shift in birth timing and long run ANC impact may only be estimated by using completed fertility level.The dependent variable is whether the household had at least one child during the five year census window.The linear probability method is used for the analysis.The equation to be estimated takes the following form: where i indexes individuals, j indexes provinces, t indexes time k indexes age groups and m indexes individual traits.We used all the individual traits used in Milligan (2005).X variables include the number of children who were born before the census window in each household; the 3 See Table 1 for summary of past studies.Buttner and Luz (1990) studies strong pro-natal policy implemented in East Germany in 1972.They show that age specific fertility levels increases in response to the policy.Hoem (1990) shows that increasing financial incentive to birth increased fertility rates in Sweden.Salanie and Laroque (2005) uses microsimulation model using individual level data to show that financial incentives have significant impact on increasing fertility in France.ages of the female and male members of each household as they were reported as one of four age groups, 15-24, 25-34, 35-44, and 44+; immigrant

Data
Before ANC, provincial government of Quebec and the federal government already had various policy measures related to child rearing before ANC. 4 In May 1988, provincial government of Quebec enacted a universal cash transfer system called ANC for all legal residence of the Province.A household with a newborn baby received C$500 for the first and the second child and C$8,000 for the third or higher parity child in cash.The amount gradually increased to C$500 for the first child, C$1,000 for the second child and C$8,000 for the third or higher parity child in cash by May 1992 (See Table 2).The subsidy amount rose over the years and observing this, the expectation of households may have been to shift births to later in their lives because delaying the timing of the births could result in a higher amount of cash transfer, based on the trend.Unfortunately, it is impossible to test the impact of the year-to-year revelation of transfer amount increase on timing of fertility, because the census is not available year-to-year but is only available every five years (See Figure 2).The program lasted for a decade until May 1997 until it was replaced by other programs.5 For age-specific impact analysis in equation ( 1

Results and Discussions
Birth subsidy may only cause a temporal shift in the timing of childbirths.Milligan (2005) argued that ANC impact was permanent rather than transitory because the difference in the average number of childbirth between Quebec and the rest of Canada rose between 1991 and 1996 (p.550).However, ANC was enacted in 1988 and it is not likely that the transitory impact would be completed in 8 years.Therefore we directly compare the completed fertility levels of Quebec and the rest of Canada from 1934 to 1962 birth cohorts. 9We first test if ANC impact varies by age group using household level Census data (See Table 3).We find strong support for age-specific responses to ANC and the need to generate age-specific measure of ANC impact.
8 1944 birth cohort was under the influence of ANC while on average 5.22 children are born out of 1,000 women of the age group, the total average number of children born between the ages of 44 and 53 in Quebec.Likewise, 1963 birth cohort was between 25 and 34 under ANC and on average the total number of children born between the age range was 859 per 1,000 women.9 We selected 1962 birth cohort as the upper limit because completed fertility levels of 1963 and younger cohorts were not actual results but projections.
Then we use completed fertility level data to estimate the permanent impact of fertility (See Figure 3 and Table 4-6).Results indicate that there is little ANC impact on completed fertility.We include completed fertility level of rest of Canada as independent variable as a control for overall trend in Canadian completed fertility level that may be incidentally correlated with ANC impact.As seen in Figure 3, analysis summarized in panel (A) indicates that ANC does not have significant correlation with completed fertility level of Quebec.
We then tested if the model is robust to the measure of ANC impact.Instead of using ageadjusted exposure to ANC as a proxy for ANC impact, we tried the number of childbearing years each cohort spent while ANC is implemented.Childbearing age used is 15-44, as total fertility rate of a given year is measured as the total number of children that would be borne by a typical woman in a country if she is to speed through age 15 to 44 during the year.Panel (B) in Table 4 shows that the use of alternative measure as ANC impact has little impact on the result.In all three models, completed fertility level of Quebec is the dependent variable, and they progressively take number of childbearing years spent under ANC, completed fertility of rest of Canada, cohort birth year and its quadratic term as independent variables.As in panel (A), ANC impact does not have a significant impact on completed fertility.It is only significant in model (A) but as we introduce cohort birth year as time trend, the significance disappears.Therefore it seems that the correlation is robust to the use of ANC impact measure.
We  3), the interaction terms between Quebec dummy variable with linear and quadratic cohort year variables and the significance disappears.Thus, panel (C) shows that age-adjusted exposure to ANC does not have statistically significant impact on fertility level of Quebec.
In theory, costly cash transfer schemes that reduces childrearing cost may have a positive impact on fertility (Becker 1960).However, if the subsidy amount is substantial enough to have permanent impact on fertility is an empirical question.It depends on the number of households at the margin whose lifetime fertility is affected by the cash transfer amount.Assuming that the utility of a child is diminishing in the number of children and the cost of childrearing is constant over the parity of births, cash subsidy may not be enough to increase the total number of lifetime births.Instead, for households with borrowing constraints, cash subsidy may only be enough for them to have a child that they are saving up for earlier.Therefore, whether the subsidy is substantial enough to encourage households to increase the total number of children is an empirical question, as children are discrete and expensive.
The mechanism is formally illustrated in a simple two-period fertility decision model with liquidity constraint in Parent and Wang (2007).Assuming price of a child is identical over the period and also assuming that utility from the first child exceeds the price of a child, households bound by liquidity constraint in the first period would be able to have a child in the first period.
However, whether the household would have another child in the second period depend on if the size of child benefit is sufficient to make the utility from the second child greater than the price of a child.If the child benefit is insufficient to increase lifetime fertility, it merely has temporal effect in the timing of birth.Using difference-in-difference analysis on household level Census data, they also show that the impact of reformed Family Allowance program of Canada in the mid-1970s indicated in Zhang et al. (1994) is actually temporal, emphasizing that the lack of permanent effect is an empirical issue.This means that ANC may not be substantial enough to have a permanent impact on fertility either.
Analysis of the age-adjusted exposure to ANC and completed lifetime fertility level of Quebec shows that ANC impact is not permanent.ANC cash transfer amount did not affect significant number of households at the margin such that their lifetime fertility is affected by the subsidy amount.The evidence implies that significant number of households who are not at the margin shifted the timing of births in response to ANC.

Concluding Remarks
The Allowance for Newborn Children (ANC) from Quebec, Canada is a fiscal incentive scheme to stimulate birth; it has a positive and significant impact on fertility.However, ANC impact may be due mainly to the shift in the timing of childbirth and not to the increase in lifetime fertility level.The permanent impact of ANC is more accurately tested by using completed fertility level of female birth cohort.It may be more accurate to estimate the permanent impact of pro-natal cash transfer by analyzing the impact on completed fertility level rather than cross-sectional dataset as it was previously done, for the transfer may have statistically significant temporal shift in birth timing as well.Although there is mechanical link between fertility level and childrearing costs, the subsidy amount must be substantial enough to permanently influence enough households who are willing to increase their lifetime fertility because children are innately discrete and expensive.The ANC subsidy amount relative to the total childrearing cost seems insufficient to increase the total lifetime fertility of affected households and the cash transfer policy have little impact on the lifetime fertility increase.
Childrearing is costly hence cash transfer schemes to reduce childrearing cost are costly as well.
For example it costs almost C$154,000 to raise a child from birth to the age of 18 in Canada according to the Canadian Council of Social Development.Based on survey by the U.S.
Department of Agriculture using the data from the Consumer Expenditure Survey by the U.S.
Department of Labor, it costs from C$124,800 to C$250,260 to raise a child from birth to the age of 18, depending on the income level and the living arrangements of the parents.Using the method highlighted in Phipps (1998), ANC transfer amount accounts for 1.3 percent, 3.2 percent and 30.1 percent of direct cost of children (Duclos, Lefebvre, and Merrigan (2001)).Cost of children also has indirect economic costs.It affects savings, consumption and labor supply decision of a household (Browning 1992) hence ANC transfer amount may be smaller than the estimate if indirect economic costs are accounted for.This paper shows that cash transfer of such magnitude is not substantial enough to increase completed fertility level but it may only shift the timing of births of the affected population.
All OECD countries except for Turkey and Mexico has above replacement level of fertility (OECD 2005) and all OECD countries implement pro-natal policies in one form or another (Sleebos 2003).Cash transfer scheme is a popular policy choice in many OECD countries spanning over different continents, including Australia, Belgium, France, Germany, New Zealand, South Korea and Sweden, while long run impact of the cash transfer has not been fully tested.This may be due to ease of monitoring, ease of implementation and political decision making.However, if future studies with more completed fertility level data, as they become available, also reveal that long run impact of cash transfer is limited, immigration may be an alternative solution for policymakers genuinely concerned about decreasing national income, productivity and rising burden on government budget due to the current fertility transition of developed countries.2).For the analysis presented in panel (B), we replaced the measure of ANC impact with the number of years spent under ANC during birth years of 15-49.The analysis in panel (C) is identical to the model specified in equation (3).The time trend dummy variables in panel (C) are the cohort birth year and its quadratic term.Quebec specific time trend dummy variables are the interaction term between Quebec dummy variable and cohort birth year and its quadratic term.Standard errors are reported in parenthesis.*** indicates significance at 1% level, ** indicates significance at 5% level and * indicates significance at 10% level.

Figure 2 :Figure 3 :
Figure 2: Census and ANC Timing Cohen et al.uses child subsidy level variation in June 2003 in Israel, using household level data from 1999 to 2005 to show that reduction in child benefit has significant impact on fertility decision of Israeli household.Milligan uses Canadian Household Census data to show that ANC had significant impact on fertility.
status; mother tongue; and terminal level of schooling of the male and female spouse of each household; the marital status of the heads of trend and time trend dummy variable coefficients of cohort year and its quadratic term controls any region invariant time trend.The interaction term coefficient captures Quebec specific time trend.The coefficient of interest is , the measure of impact of ANC exposure on completed fertility.Again if there is statistically significant impact of ANC on completed fertility level the coefficient would be positive and significant.
), we selected data from 1991 and 1996 Canadian Census Public Use Microdata Files on Families to control for the individual household level traits such as income, education level, language, immigration status, and age groups.1986and2001censusfilesreported 345,351 families and 348,104 families, respectively.manychildrenarebornonaverage per 1,000 women per cohort when they were under the influence of ANC.Age-adjusted exposure to ANC was estimated as it follows.First, for each cohort the age at which they were under the influence of ANC was calculated, that is, the age of each cohort between 1988 and 1997.Then, the age-specific fertility of Quebec was summed up by the age at which each cohort was under the influence of ANC.For example, for 1944 birth cohort, as they were under the influence of ANC between 44 and 53, we summed up age-specific fertility level from 44 to 53.ANC impact, weighted by age-specific fertility level, is only 5.22 average births per 1,000 women for 1944 cohort.8Figure3showscompleted fertility levels of Quebec and the Rest of Canada along with age-adjusted exposure to ANC for birth cohorts 1934 to 1962.
6We comparedand 1997ale age group probability of fertility of Quebec households relative to rest of Canada for partially affected Census window (1991) versus completely affected Census window (1996) to test if subsidy amount was substantial enough to significantly affect fertilities of households likely to be facing terminal period in lifetime birth cycle.7Weusedcompletedfertilitylevel of Quebec and the rest of Canada for 1934 to 1962 cohorts to analyze the effect of ANC on completed fertility level, as in equation (2).However, measuring exposure to ANC per birth cohort is not straightforward.Fertility rate varies by age hence each cohort is affected differently by ANC.Age-specific fertility measures how many children are born per 1,000 women in the age group on average.We created a measure of ANC exposure called age-adjusted ANC exposure by incorporating age-specific fertility rate of Quebec.The measure uses the age of each cohort during the decade at which ANC was implemented and estimates howBetween 1988and 1997, the ANC policy window, 1944birth cohort was aged between 44 and 53 (See Figure3).Once weighted by the age-specific fertility level, 1944 cohort was the first cohort whose fertility level was significantly affected by ANC.Age-adjusted exposure to ANC gradually increases until 1962 cohort, who was aged between 24 and 33 when ANC was present, at 798 average births per 1,000 women.Between 1934 and 1963 cohort, the decade long ANC progressively affects age-adjusted exposure to ANC.

Table 3
presents regression results of equation (1).It reports the ANC impact coefficient of three sub samples by age group and the ANC coefficient of a pooled sample.We test if birth probability of different age group is differently affected by ANC.From Table3, ANC impact varies by age group.The birth subsidy impact on fertility is significant and positive for the two younger cohorts; however, it is not significant for the oldest cohort.In the pooled regression, the impact of ANC on the youngest cohort is 6.6 percent and it is 4.2 percent for the second youngest cohort.The impact is 1.1 percent for women older than 35.As ANC impact varies by age group and there is strong evidence that we need age-specific measure of ANC impact on fertility.
Completed fertility level of Quebec is consistently below completed fertility level of the rest of Canada.Completed fertility level of Quebec slowly and consistently declines from 3.02 in 1934 birth cohort to 1.61 in 1965 birth cohort.Completed fertility level of the rest of Canada also consistently declines from 2.38 in 1944 birth cohort to 1.63 in 1962 birth cohort.Completed fertility levels of both Quebec and rest of Canada seem to move concurrently with each other and age adjusted exposure to ANC does not seem to affect the pattern of Quebecois completed fertility level differently from completed fertility level of rest of Canada.Linear regression results from equation (2) are reported in Table4.Models in panel (A) take the completed fertility level of Quebec as the dependent variable and age-adjusted exposure to ANC, completed fertility level of rest of Canada, linear and squared cohort year variables as the independent variables.Age-adjusted exposure to ANC is statistically insignificant for all models.
estimate equation (3), taking completed fertility levels of both Quebec and the rest of Canada as dependent variables.The model takes age-adjusted exposure to ANC, cohort birth year, its quadratic term and Quebec indicator variable as independent variables.Age-adjusted exposure to ANC was assumed to be zero for the rest of Canada regardless of cohort birth year.From panel (C) in Table4presents the estimation result of equation (3).The results indicate that ageadjusted exposure to ANC is significant in model (2) only.However, model (2) implicitly assumes that Quebec and the rest of Canada have identical time trend.We introduce Quebec specific time trend dummy variable in model (

Figure 1: Total Fertility Rates of OECD Countries from 1980 to 2007
Asia includes Japan and South Korea.Southern Europe includes France, Italy, Portugal and Spain.Northern Europe includes Austria, Belgium, Denmark, Finland, Germany, Iceland, Ireland, the Netherlands, Norway, Sweden, Switzerland and the United Kingdom.Other OECD Europe includes Czech Republic, Estonia, Hungary, Poland, Slovak Republic and Slovenia.Oceania includes Australia and New Zealand Source: OECD Statistics (http://stats.oecd.org/Index.aspx).Note:

Table 1 : Summary of Short Term Policy Evaluation Studies
Table 102-4505) for the age at which each cohort was affected by ANC.The completed fertility of the rest of Canada was estimated by subtracting population-weighted completed fertility rate of Quebec from completed fertility rate of Canada per cohort (Source: Statistics Canada data table 91-209X for completed fertility of Canada (http://www.statcan.gc.ca/pub/91-209x/2011001/article/11513/figures/desc/desc05-eng.htm),Institut de la Statistique Quebec for completed fertility of Quebec, CANSIM database Table 051-0001 and Canada Yearbook for population of Canada and Quebec from 1949 to 2006, for every five years).Completed fertility levels of 1963-1978 cohorts were also available but they were projections hence we only used real data up to 1962 birth cohort.

Table 3 : Regression Results by Female Age Groups Dependent
Variable: Had at least one child during the 5-year Census window.Notes: This analysis is identical to the model specified in equation (1).The age cohort variables and important dummy variables are reported: '15-24 X Quebec X 1996 dummy' indicates the dummy variable for households with female age 15-24 in the presence of the ANC, '25-34 X Quebec X 1996 dummy' indicates the dummy variable for households with female age 25-34 in the presence of theANC and '35 and over X Quebec X 1996dummy' indicates the dummy variable for households with female age over 35 in the presence of the ANC.Standard errors are reported in parenthesis.*** indicates significance at 1% level, ** indicates significance at 5% level and * indicates significance at 10% level

Table 4 : Age-Adjusted ANC Exposure Impact on Completed Fertility
Dependent Variable -Completed Fertility Level of Quebec and the rest of Canada (pooled sample) The analysis in panel (A) is identical to the model specified in equation (