Professional Documents
Culture Documents
The Effect of Universal Child Benefits On Female Labor Supply and Household Expenditure: Case of Mongolia
The Effect of Universal Child Benefits On Female Labor Supply and Household Expenditure: Case of Mongolia
The Effect of Universal Child Benefits On Female Labor Supply and Household Expenditure: Case of Mongolia
[1]
Khishigt Dandarchuluun, [2]Hyung Jai Choi
[1]
Associate Professor, [2]Professor,
[1]
Department of Economics, School of Arts and Sciences,
National University of Mongolia, [2] Department of Economics and Statistics,
Korea University, Sejong Campus
Abstract: This study investigates the causal impact of Mongolia’s universal child benefit program on
female labor supply and household expenditure using Difference-In-Difference-In-Differences
approach. In addition, we focused on the how households spend their benefit income. To examine the
relationship between benefit income and household expenditures, we use Seemingly Unrelated
Regression model. We use cross-sectional data from the Integrated Household Income and Expenditure
Survey with Living Standards Measurement Survey of 2002/2003 and Household Socio-Economic
Survey of 2007/2008 from the National Statistical Office of Mongolia. We find that between 2002/2003
and 2007/2008, women with children decreased their relative labor force participation by 12.27
percentage points. The estimated adverse effects of child benefit program on labor force participation
are proved by linear regression of annual hours of worked. When we estimate the effects of UCBP on
worked hours conditional on hours exceeding zero, women with children decreased their relative
working hours conditional on working by 155.32 hours. Moreover, we find that the effects of UCBP on
household expenditures depending on household behavior. For households involved in agriculture,
households receiving benefit income increased their expenditures on non-durable goods than food and
durable goods. Contrary, households not involved in agriculture, they increased their expenditures on
food items than durable and non-durable goods. For the low and middle-income group, there are
positive UCBP effects on the broad variety of food and non-durable expenditure categories. However,
there have been almost no significant effects of UCBP on the expenditures for the high-income group.
In addition, we found that the income transfers are shared between children and their parents.
I. INTRODUCTION
Child benefit program (sometimes called family allowance) is one of the important tools of social
protection and poverty reduction strategies. Governments of many countries (e.g. OECD countries,
most of EU countries) offer a cash benefit to families with children. In most cases, cash benefits have
targeted at low-income families and provide families with children based on the family net income, and
presence, number, age, and sometimes the birth order of child. Cash benefits can be unconditional,
conditional on child’s school enrollment, immunization, and nutrition, or else conditional to the parental
work requirement. The primary goals of these programs typically include encouraging fertility,
improving the wellbeing and long-term opportunities for children, and reducing poverty. However, it is
hard to evaluate the policies achieve their goals and whether there are any potential side effects.
Especially, consequences regarding labor market outcomes have been mostly ignored.The child benefit
program of Mongolia, which we focused on this study, is a universal cash benefit paid to a child under
the age of 18, regardless of family income. There is no requirement that certain amount must be spent
on particular services or goods. The stated original goals of the program were to reduce poverty,
increase children’s education, health, and nutrition, and redistribute mining revenues to Mongolian
families living below the poverty line. In January 2005, the Mongolian government introduced the
conditional child benefit (CCB) and expanded it into universal child benefit (UCB) in July 2006. In this
study, we examined the effects of UCBP on female labor supply and household expenditure using
Difference-In-Differences (DD) approach with Propensity Score Matching (PSM) method. In addition,
we focused on the how households spend their benefit income. In order to examine the relationship
between benefit income and household expenditures, we use Seemingly Unrelated Regression (SUR)
model.
supply. Nada Eissa and Jeffery B. Liebman (1996) have examined the labor supply response to the
earned income tax credit. They conclude that after the Tax Reform Act, single women with children
increased their relative labor force participation by up to 2.8 percentage points. Alexis Leon (2003) find
that statistically insignificant negative effect of family allowance on female labor force participation
using British Household Panel Data. Ghazala Naz (2003) evaluates the impact of Norwegian family
policy reform and find that negative effect of reform on women’s labor force participation. Highly
educated mothers decreased more than that of less educated mothers. Robert McNown and Cristobal
Ridao-Cano (2004) studied the effects of Canadian child benefit policies on fertility and female labor
supply. They highlighted that child benefit policies are expected to have positive effects on fertility,
with reverse effect on female labor supply. More recently, Tammy Schirly (2014) studied the universal
child-related income transfer on the labor supply of married individuals. She concludes that the
Canadian Universal Child Benefit has significant negative effects on labor supply. Lower-educated
mothers reduced their labor force participation 3.2 percentage points when receiving the benefit.
related everyday items. In addition, some families reduce spending on risky behavior items such as
alcohol and tobacco. David S. Johnson, Jonathan A. Parker, and Nicholas S. Souleles (2006) estimate
the change in consumption expenditures caused by the 2001 federal income tax rebates. They find that
households spent 20% to 40% of their rebates on nondurable goods shortly after they arrived rebates.
The expenditure responses were relatively large for low-income households. Kooreman (2000) studied
the effect of universal child benefits in the Netherlands and found that child benefit payments were
spent disproportionately on child-related goods. Furthermore, there is evidence that the child benefit
program did not meet its goals. Laura Blow, Ian Walker, and Yu Zhu (2005) examined household
expenditure patterns affected by universal child benefit using data from the Family Expenditure Survey
of the UK from 1980 to 2000. They find that households spend their child benefit income differently
than other income. Surprisingly, the child benefit is disproportionately spent on adult goods such as
alcohol. They find that this effect is much bigger for couples.
where lfp is a dummy equal to one if a woman who has worked during the last 12 months or looking
for a job last 3 months, zero otherwise. The vector 𝑍𝑖𝑡 represents the baseline set of controls we used.
In our baseline specification, 𝑍𝑖𝑡 includes age, age squared, marital status, household size, household
assets, asset squared, household income, the number of children, the number of preschool children,
Agriculture dummy, and Bachelor dummy variable, etc. 𝑍𝑖𝑡 also includes regional dummy variables.
𝐶ℎ𝑖𝑙𝑑𝑖 is a dummy equal to one for a woman who has a child under the age of 18, zero otherwise. We
expect 𝛾0 to be positive, which means women with children have higher participation rates than those
women without children. Because bringing up children may lead to an additional income requirement.
𝑃𝑜𝑠𝑡𝑡 is a dummy equal to one for after the UCBP, zero otherwise. 𝛾1 measures the average changes in
labor force participation for both women with children and without children between 2003 and 2008.
Our key coefficient is 𝛾2 (coefficient on the interaction Child*Post). We expect 𝛾2 to be negative,
which means that UCBP reduces the labor force participation of women with children.
Our OLS regression for worked hours is as follows:
variable, etc. 𝑋𝑖𝑡 also includes the regional dummy variables. 𝐶ℎ𝑖𝑙𝑑𝑖 is a dummy equal to one if a
household has a child under the age of 18, zero otherwise. 𝑃𝑜𝑠𝑡𝑡 is a dummy equal to one for after the
UCCBP, zero otherwise. Our key coefficient is 𝛾 (coefficient on the interaction Child*Post), and it
measures an average causal effect of UCBP on household expenditure. We estimate equation (3) by a
system of equations using Seemingly Unrelated Regression (SUR), which is more efficient than
estimates that each spending equation separately through OLS, because the SUR model considers the
variance-covariance structure of error terms in each equation. In addition, SUR allows us to perform
Wald test of joint significance of the treatment coefficient.Moreover, we estimate the change in
expenditure structure of household caused by the UCBP. Does UCBP affect the household expenditure
structure? If the UCBP affects household expenditure structure, whether there are similar effects on the
different behavior of household is also interesting. We use the share of each spending category as a
dependent variable in the equation (3) and run all share equations simultaneously through SUR model.
Our set of M linear equations as follows:
12,851 observations.For the impact of UCBP on household expenditure which provided the fifth
section, we exclude households that did not have complete data, and the resulting sample size is 14,333
observations. The “Integrated HIES with LSMS 2002/2003” and “HSES 2007/2008” data provides
information on an extensive set of spending categories. However, some spending categories were not
directly comparable, so we grouped the categories broad areas. Moreover, for the purpose of this study,
we attempt to group household spending categories into child-related spending and adult-related
spending.
Table:
Preschool
- - 0.47 0.50
children*
Number of
- - 2.00 1.05
children
Urban area* 0.65 0.48 0.56 0.50
Note: 1) * denotes a dummy variable 2) Labor force participation equal to one if a woman who has
worked (not include work on an own account, in the household farm, and herd) during the last 12 months
or looking for a job last 3 months, zero otherwise.
We estimate the effect of UCBP on FLFP using DD probit model. We use the PSM method with our
probit regression, to ensure that we are estimating the actual effect of UCBP on FLFP. The PSM method
will reduce the imbalance in covariates between treated and control groups. We use one-to-one, nearest-
neighbor propensity score matching to find the control individual with the propensity score closest in
value to the treatment individual. After the matching, we discard 5,601 treated observations and the
resulting sample size is 7250 observations.
Each column of Table 2 presents the separate probit results. Column (1) reports the result of DD
model and column (2) the results of matched DD model. In Table 2, expected participation response
was negative and statistically significant at the 1% level. The Married dummy, Agriculture dummy, the
interaction between Child and Post variable, log(Income), the number of children, the number of
preschool children, and Unemployment rate variables are negative; the Bachelor dummy, Age,
log(Asset), household size, Child dummy, and Post dummy variables positively correlated with labor
force participation. There is a coefficient of Child (γ0 ) is positive. It means that woman with children
has tended to be higher probabilities of participation than the participation of those women without
children. Also, γ1 is positive which suggesting that there is an overall positive trend in average
participation in the two groups. The key coefficient γ2 is a negative sign and the other coefficients on
the demographic characteristics, and regional variables are all economically (all have expected signs)
and statistically significant.
Note: 1) The dependent variable is labor force participation. Labor force participation equal to one for
a woman who has worked (not include work on an own account, on the household farm, and herd)
during the last 12 months or looking for a job last 3 months, zero otherwise. 2) *, **, and *** indicate
significance at 10%, 5% and 1% level, respectively. 3) The estimated coefficients on Region dummy
variables are not reported. The probit is a nonlinear model. Therefore, the coefficients cannot be used
directly as marginal effects. Also, in a nonlinear model, the cross difference does not represent the
treatment effect. Patrick A. Puhani (2008) shows that, in a nonlinear “difference-in-differences” model,
the treatment effect is the cross difference of the conditional expectation of the observed outcome minus
the cross difference of the conditional expectation of the potential outcome without treatment. Estimated
average participation responses of women with children are reported in the last row of Table 2. Since
PSM reduces the imbalance in covariates between the treated and control groups, we will explain the
results of matched DD model. When we add all variables in our probit model, we find that women with
children decreased their relative labor force participation by statistically significant 12.88 percentage
points.
In order to check the Hypothesis 2, we use income interaction variables in our probit model. We
estimate the following Difference-In-Difference-In-Differences (DDD) regression model:
(lfpit = 1) = Φ(α + βZit + γo Childi + γ1 Post t + γ2 (Child ∗ Post)it + γ3 LowInci
+ γ4 (LowInc ∗ Child)i + γ5 (LowInc ∗ Post)it
+ γ6 (LowInc ∗ Child ∗ Post)it ) (5)
where LowInc is a dummy equal to one for a woman who lives under the MSL, zero otherwise. If γ6 is
negative, it indicates that the UCBP has had more impact on the low-income group. We expect γ6 to be
negative. The results are reported in Table 3. The DDD and Matched DDD results are much similar
again. The coefficient of γ6 has negative, however, it has statistically insignificant. In other words, we
have not found any significant results for whether the UCBP has more impact on the low-income group.
Note: 1) The dependent variable is labor force participation. Labor force participation equal to one for
a woman who has worked (not include work on an own account, on the household farm, and herd)
during the last 12 months or looking for a job last 3 months, zero otherwise. 2) *, **, and *** indicate
significance at 10%, 5% and 1% level, respectively. 3) The estimated coefficients on Region dummy
variables are not reported.
Theoretically, an individual has a pure income effect when receiving the benefits. It means that
individual receiving the benefit are expected to purchase more goods that are normal. If the leisure is a
normal good, individuals to purchase more leisure (reduce their hours of work) when receiving the
benefits. Our next estimate is the effect of UCBP on worked hours.
When we estimate the effects of UCBP on annual worked hours conditional on hours exceeding zero,
women with children decreased their relative working hours conditional on working by 155.32 hours
(for the full sample 157.51 hours). That result reported in Table 4. The coefficient estimate of γ6 is
again statistically insignificant. It means that there is no significant difference in the response of worked
hours between low and high-income groups. Altogether the UCBP has negative effects on the annual
working hours of women with children.
Table 4. DDD OLS Results: The Effects of UCBP on Annual Worked Hours, Hours>0
Matched
DDD
DDD
Estimated coefficients
Child (γ0 ) 163.81** 160.73*
Post (γ1 ) -29.11 -42.92
Child*Post (γ2 ) -157.51** -155.32*
LowInc (γ3 ) 291.64** 276.09**
LowInc* Post (γ4 ) -364.34** -347.94**
LowInc*Child (γ5 ) -341.02** -393.41**
LowInc*Child* Post
317.2 327.82
(γ6 )
Age 48.92*** 42.65***
Age_sq -0.63*** -0.56***
Note: 1) The dependent variable is annual hours of worked. 2) *, **, and *** indicate significance at
10%, 5% and 1% level, respectively. 3) The estimated coefficients on Region dummy variables are not
reported.
B. Child benefits and household expenditure
The results that reported in Table 5 support our research hypothesis one, two, and three. For the low
and middle-income group, there are positive effects on the food and non-durable expenditure categories.
However, there have been almost no significant effects of UCBP on the expenditures for the high-
income group. In addition, the pattern of household spending affected by the UCBP depends on whether
households are engaged in agriculture or not. After the UCBP, households involved in agriculture
increased more their non-durable expenditure while households not involved in agriculture increased
more their food expenditure.
Table 5. Matched SUR Results: Effects of UCBP on Household Expenditure
Dependent Variables (MNT)
Durable Non-Durable Food
Full Sample
95954.47 225607.516** 280705.716***
(Obs=7268)
Low-Income
-13092.3 162622.1 242256.349**
(Obs=2183)
Middle-
Income 122269.3 186514.321* 288213.248*
(Obs=2905)
High-Income
-108911 76065.58 201848.5
(Obs=2180)
H’s Involved
in
112138.9 263743.342*** 259533.6
Agriculture
(Obs=2717)
H’s Not
100701.8 197392.4 288344.295***
Involved in
Agriculture
(Obs=4551)
Note: 1) We estimate each spending equation separately through OLS, and then estimate equation by
a system of equation using SUR. OLS and SUR results have been much similar to each other,
so we reported only SUR results because the SUR model considers the variance-covariance
structure of error terms in each equation. In addition, SUR allows us to perform Wald test of
joint significance of the treatment coefficient. The model includes the full set of variables such
as age, age_sq, married, bachelor, child_18, adults (or hhsize), agriculture, income, asset, urban,
region2, region3, region4, region5, Child, Post, and Child*Post. Households are classified
income groups based on the household per consumption (Lowest 30%, Middle 40%, and
Highest 30%).
2) Each cell of the Table corresponds the coefficient estimate of
Child*Post interaction which is our key treatment coefficient.
3) *, **, and *** indicate significance at 10%, 5% and 1% level,
respectively.
Table 6 correspond the coefficient estimates of benefit income variable by different subgroups of
households. For low and high-income households, an additional MNT of annual benefit income
increases household spending on food by MNT 0.436 and MNT 0.948 respectively. For household not
involved in agriculture, an additional MNT of benefit income significantly increased household food
expenditure by MNT 0.935. For household involved in agriculture, however, we did not find any
significant effects on durable, non-durable, and food expenditure. It is not surprising results because
households engaged in agriculture provide their necessities of some foods from their own enterprises.
For the more detailed categories of household expenditure which reported in panel (B), (C), and (D)
of Table 6, low and high-income households spend their benefit income for some risky behavior like
tobacco and cigarettes. The magnitude of this response is much bigger for high-income households.
Households involved in agriculture tend to spend their benefit income for non-durable items, while
households not involved in agriculture tend to spend their benefit income for food items.
Note: 1) We estimate each spending equation separately through OLS, and then estimate equation by
a system of equation using SUR. OLS and SUR results have been very close to each other, so
we reported only SUR results because the SUR model considers the variance-covariance
structure of error terms in each equation. The model includes the full set of variables Benefit,
age, age_sq, married, bachelor, adults (or hhsize), agriculture, enterprise, income, asset, urban,
region2, region3, region4, and region5. Households are classified income groups based on the
household per consumption (Lowest 30%, Middle 40%, and Highest 30%).
2) Each cell corresponds the coefficient estimate of Benefit variable.
3) *, **, and *** indicate significance at 10%, 5% and 1% level, respectively.
HSES provides more detailed data on expenditures which related to education. Table 7 shows the
estimated coefficients of benefit income variable according to the education-related spending category.
In panel (A) of Table 7, there are significant positive relationship between benefit income and spending
of “Transport”, “Books and Supplies”, “Uniform”, and “Others”; significant negative relationship
between benefit income and spending of “Dormitory”, “Tuition”, and “Food bought in the canteen or
other places”. As discussed before, these negative relationships can be attributed to Education sector
reform and National School Lunch Program. When we considering education-related expenses by
child’s education level, there is a significant positive response to the spending of general education and
vocational education. For spending on College and University, there are significant negative effects. It
is not surprising because the total amount of child benefit depends on the number of children under age
18. That is the reason of the negative relationship between benefit income and spending on tertiary
education. The results are robust across different subgroups of the households.
Table 7. SUR Results: Estimated Coefficients on Benefit Income Variable, by Education-Related
Category
Note: 1) The model includes the full set of variables Benefit, age, age_sq, married, bachelor, adults
(or hhsize), agriculture, enterprise, income, asset, urban, region2, region3, region4, and region5.
Households are classified income groups based on the household per consumption (Lowest
30%, Middle 40%, and Highest 30%).
2) Each cell corresponds the coefficient estimate of Benefit variable.
3) *, **, and *** indicate significance at 10%, 5% and 1% level, respectively.
We exclude all households that did not receive child benefit from our sample, and then estimate the
SUR model again. Table 8 shows the estimated coefficient of benefit income variable that is statistically
significant (statistically insignificant results are not reported). Households receiving benefit income
tend to spend more their benefit income for non-durable and food items.
Table 9 corresponds the estimated benefit coefficients on household expenditure share. An additional
MNT 100,000 of benefit income increases household expenditure share of food by 0.237%. In
concluding these results, the households spend their benefit income for both child and adult-related
goods and services. We found that there is a significant positive relationship between benefit income
and child-related expenses such as “Child clothing”, “Transport”, “Books and Supplies”, “Uniform”,
and “Others”. In addition, there is a significant positive relationship between benefit income and adult-
related some expenses such as “Women’s clothing”, and “tobacco and cigarettes”. On the other hand,
households spend their benefit income mostly for value-added productive items that are produced by
private producers. It means that public expenditures can be productive in the aggregate and have
positive effects on the economy.
Table 8. SUR Results: Estimated Coefficients on Benefit Income Variable, Benefit>0
Note: 1) The model includes the full set of variables Benefit, age, age_sq, married, bachelor, adults
(or hhsize), agriculture, enterprise, income, asset, urban, region2, region3, region4, and region5.
Households are classified income groups based on the household per consumption (Lowest
30%, Middle 40%, and Highest 30%).
2) Each cell corresponds the coefficient estimate of Benefit variable.
3) *, **, and *** indicate significance at 10%, 5% and 1% level, respectively.
Table 9. SUR Results: Estimated Benefit Coefficients on Household Expenditure Share, Benefit>0
CONCLUSION
In this study, we examined the effect of UCBP on female labor supply and household expenditure
using PSM with DDD approach. In addition, we focused on the how households spend their benefit
income. We use cross-sectional data from the Integrated Household Income and Expenditure Survey
with Living Standards Measurement Survey of 2002/2003 and Household Socio-Economic Survey of
2007/2008 from the NSO of Mongolia. In order to examine the relationship between benefit income
and household expenditures, we have used SUR model. There is no comprehensive impact evaluation
of the UCBP of Mongolia yet. Therefore, this study will be an important alternative evidence that
examined the impact of universal child benefits on female labor supply and household expenditure
because Mongolia’s child benefit system is universal and not targeting the low-income households. In
addition, this study can contribute to improving the effectiveness of such welfare policy.
We find that after the UCBP, women with children decreased their relative labor force participation
by 12.27 percentage points. When we estimate the effects of UCBP on worked hours conditional on
hours exceeding zero, women with children decreased their relative working hours conditional on
working by 155.32 hours. Our findings are suggesting that the UCBP reduces the labor force
REFERENCES
1) Alexander Bick. (2015). The Quantitative Role of Child Care for Female Labor Force
Participation and Fertility. Journal of the European Economic Association. Vol. 14: pp. 639–
668.
2) Alexis León. (2003). Family Allowances and Female Labor Force Participation.
Massachusetts Institute of Technology.
3) Angus Deaton, Javier Ruiz-Castill, and Duncan Thomas. (1989). The Influence of Household
Composition on Household Expenditure Patterns: Theory and Spanish Evidence. Journal of
Political Economy, vol. 97, no. I.
4) Angus Deaton. (1992). Understanding Consumption. Oxford: Clarendon Press.
5) Anthony Hodges et al. (2007). Child Benefits and Poverty Reduction: Evidence from
Mongolia’s Child Money Programme. Working paper of Division of Policy and Planning,
UNICEF.
6) Arnold Zellner. (1962). An Efficient Method of Estimating Seemingly Unrelated Regression
Equations and Tests of Aggregation Bias. Journal of the American Statistical Association, 57,
500-509.
7) Bernice Madison. (1964). Canadian Family Allowances and Their Major Social Implications.
Journal of Marriage and Family, Vol. 26, No. 2, pp. 134-141.
8) Charles Michalopoulos, Philip Robins, and Irwin Garfinkel. (1992). A Structural Model of
Labor Supply and Child Care Demand. University of Wisconsin-Madison, Institute for
Research on Poverty.
9) David Blau and Philip K. Robins. (1988). Child-Care Costs and Family Labor Supply. The
Review of Economics and Statistics, Vol. 70, No. 3, pp. 374-381.
10) David Coady et al. (2004). Targeting of Transfers in Developing Countries: Review of
Lessons and Experience. The World Bank, Washington, D.C.
11) David Johnson, Jonathan A. Parker, and Nicholas S. Souleles. (2006). Household Expenditure
and the Income Tax Rebates of 2001. The American Economic Review, Vol. 96, No. 5, pp.
1589-1610.
12) David Ribar. (1995). A Structural Model of Child Care and the Labor Supply of Married
Women. Journal of Labor Economics, Vol. 13, No. 3, pp. 558-597.
13) David Ribar. (1992). Child Care and the Labor Supply of Married Women. The Journal of
Human Resources, Vol. 27, No. 1, Special Issue on Child Care, pp. 134-165.
14) Duncan Thomas. (1993). The Distribution of Income and Expenditure within the Household.
Annales d’Économie et de Statistique, No. 29, Progrès récents en théorie du consommateur /
Recent Advances in Economic Theory, pp. 109-135.
15) Frances Woolley et al. (1996). The Case of Child Benefits. Canadian Public Policy, Vol. 22,
No. 1, pp. 24-39.
16) Ghazala Naz. (2004). The impact of cash-benefit reform on parents’ labour force participation.
Journal of Population Economics, Vol. 17, Issue 2, pp. 369-383.
17) Goodman Bacon et al. (2008). How to EITC recipients spend their refunds? Federal Reserve
Bank of Chicago. Economic perspectives, pp. 17-32.
18) Hal Varian. 2006. Intermediate Micro Economics.
19) Huck-ju Kwon and Woo-rim Kim. (2015). The Evolution of Cash Transfer in Indonesia:
Policy Transfer and National Adaptation. Asia and the Pacific Policy Studies, Vol. 2, No. 2,
pp. 425-440.
20) James Poerba. (1988). Are Consumers Forward Looking? Evidence from Fiscal Experiments.
The American Economic Review, Vol. 78, No. 2, Papers and Proceedings of the One
Hundredth Annual Meeting of the American Economic Association, pp. 413-418.
21) Jane Hopkins, Carol Levin, and Lawrence Haddad. (1994). Women’s Income and Household
Expenditure Patterns: Gender or Flow? Evidence from Niger. American Journal of
Agricultural Economics, Vol. 76, No. 5, Proceedings Issue, pp. 1219-1225.
22) Janet Currie (1994). Welfare and the Well-Being of Children: The Relative Effectiveness of
Cash vs. In-Kind Transfers. Tax Policy and the Economy, National Bureau of Economic
Research and MIT Press.
23) John Baumann et al. (1988). A choice-theoretical labour-market model: Empirical tests at the
mesolevel. Environment and Planning A, No. 20, pp. 1085-1102.
24) John Hoddinott and Lawrence Haddad. (1995). Does Female Income Share Influence
Household Expenditure? Evidence from Côte D’Ivoire. Oxford Bulletin of Economics and
Statistics, 57, 1, 0305-9049.
25) John Muellbauer. (1974). Household Production Theory, Quality, and the “Hedonic
Technique”. The American Economic Review, Vol. 64, No. 6, pp. 977-994.
26) Johannes Haushofer and Jeremy Shaprio. (2013). Household Response to Income Changes:
Evidence from an Unconditional Cash Transfer Program in Kenya. Working Paper.
27) Kanishka Misra and Paolo Surico. (2011). Heterogeneous Responses and Aggregate Impact
of the 2001 Income Tax Rebates. NBER Working Paper No. 16684.
28) Kelvin Lancaster. (1975). The Theory of Household Behavior: Some Foundations. Annals of
Economic and Social Measurement, Volume 4, number 1.
29) Laura Blow, Ian Walker, and Yu Zhu. (2006). Who benefits from Child Benefit? The
Warwick Economics Research Paper Series (TWERPS) 749, University of Warwick,
Department of Economics.
30) Lauren Jones, Kevin S. Milligan, and Mark Stabile. (2015). Child Cash Benefits and Family
Expenditures: Evidence from The National Child Benefit. NBER Working Paper No. 21101.
31) Libertad González. (2011). The Effects of a Universal Child Benefit. Barcelona GSE Working
Paper Series, No 574.
32) Marcia Meyers, Theresa Heintze and Douglas Wolf. (2002). Child Care Subsidies and the
Employment of Welfare Recipients. Demography, Vol. 39, No. 1, pp. 165-179.
33) Mark Berger and Dan Black. (1992). Child Care Subsidies, Quality of Care, and the Labor
Supply of Low-Income, Single Mothers. The Review of Economics and Statistics, Vol. 74,
No. 4, pp. 635-642.
34) Martin Browning and Dolores Collado. (2001). The Response of Expenditures to Anticipated
Income Changes: Panel Data Estimates. The American Economic Review Vol. 91, No. 3, pp.
681-692.
35) Michael Baker et al. (2008). Universal Child Care, Maternal Labor Supply, and Family Well‐
Being. Journal of Political Economy, Vol. 116, No. 4, pp. 709-745.
36) Michael Lokshin. (2000). The Effect of Early Childhood Development Programs on Women’s
Labor Force Participation and Older Children’s Schooling in Kenya. Policy Research Report
on Gender and Development Working Paper. No. 15.
37) Michael Lokshin et al. (2000). Effects of Child Care Prices on Women’s Labor Force
Participation in Russia. Policy Research Report on Gender and Development Working Paper.
No. 10.
38) Nada Eissa and Jeffrey Leibman. (1996). Labor Supply Response to the Earned Income Tax
Credit. The Quarterly Journal of Economics. Vol. 111, No. 2, pp. 605-637.
39) Nicholas Souleles. (1999). The Response of Household Consumption to Income Tax Refunds.
The American Economic Review, Vol. 89, No. 4, pp. 947-958.
40) Patrick Puhani. (2008). The Treatment Effect, the Cross Difference, and the Interaction Term
in Nonlinear Difference-In-Differences Models. Discission Paper, IZA DP No.3478.
41) Paul Gregg, Jane Waldfogel, and Elizabeth Washbrook. (2005). Expenditure Patterns Post-
Welfare Reform in the UK: Are low-income families starting to catch up? CASE Working
paper.
42) Paul Rosenbaum and Donald B. Rubin. (1983). The Central Role of the Propensity Score in
Observational Studies for CausalEffects. Biometrika, Vol. 70, No. 1. pp. 41-55.
43) Peter Kooreman. (2000). The Labeling Effect of a Child Benefit System. The American
Economic Review, VOL. 90 NO. 3.
44) Rachel Connelly. (1992). The Effect of Child Care Costs on Married Women’s Labor Force
Participation. The Review of Economics and Statistics, Vol. 74, No. 1, pp. 83-90.
45) Reagan Baughman et al. (2009). The Earned Income Tax Credit and Fertility. Journal of
Population Economics, 22, pp. 537-563.
46) Richard Nelson and Davide Consoli. (2010). An evolutionary theory of household
consumption behavior. Journal of Evolutionary Economics, Volume 20, Issue 5, pp 665–687.
47) Robert McNown and Cristobal Ridao-Cano. (2002). The Effect of Child Benefit Policies on
Fertility and Female Labor Force Participation in Canada. Discussion Papers in Economics,
Center for Economic Analysis, Working Paper No. 02-05.
48) Robert McNown and Cristobal Ridao-Cano. (2004). The Effect of Child Benefit Policies on
Fertility and Female Labor Force Participation in Canada. Review of Economics of the
Household 2. pp. 237-254.
49) Tammy Schirle. (2015). The effect of universal child benefits on labor supply. Canadian
Journal of Economics, Volume 48, Issue 2, pp. 437–463.
50) Ying Yeung and Stephen Howes. (2015). Resources-to-Cash: A Cautionary Tale from
Mongolia. Australian National University, Action Research Report.
51) Virendra Srivastava and David Giles. 1987. Seemingly Unrelated Regression Equations
Models: Estimation and Inference. Marcel Dekker, Inc.
52) William Greene. 2003. Econometric Analysis. Pearson Education