POLICY RESEARCH WORKING PAPER 2206
Household Childcare Replacing family allowances
with childcare subsidies in
Choices and Women s Russia might have a strong
Work Behavior in Russia positive effect on women's
participation in the labor force
and thus could be effective in
Michael M. Lokshin reducing poverty
The World Bank
Development Research Group
Poverty and Human Resources U
October 1999
POLICY RESEARCH WORKING PAPER 2206
Summary findings
Lokshin models mothers' participation in the labor force, The results of this analysis indicate that the extent to
their working hours, and household demand for which mothers participate in the labor force, and for
childcare in Russia. The model estimates the effects of how many hours, depends on the costs of childcare and
the price of childcare, mothers' wages, and household on what level of hourly wage is available to them and to
income on household behavior and well-being. other members of the household.
The theoretical model yields several predictions. To Lokshin's siinulations show that family allowances -
test these, reduced-form equations of rhe discrete and intended to reduce poverty - do not significantly affect
continuous household choices are estimated jointly using the household choice of childcare arrangements.
the method of semi-parametric full information Replacing family allowances with childcare subsidies
maximum likelihood. This method controls for the might have a strong positive effect on women's
correlation of error terms across outcomes, and the participation in the labor force and thus could be
correlation of error terms that can result when panel effective in reducing poverty.
data are used.
This paper - a product of Poverty and Human Resources, Development Research Group - is part of a larger effort in the
group to understand the role of gender in the context of the household, institutions, and society. Copies of the paper are
available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Patricia Sader, room MC3-
632, telephone 202-473-3902, fax 202-522-1153, Internet address psader@worldbank.org. Policy Research Working
Papers are also posted on the Web at http://wwvw.worldbank.org/html/dec/Publications/Workpapers/home.html. The
author may be contacted at mlokshin@worldbank.org. October 1999. (35 pages)
The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about
development issues. An objective of the series is to get the findings out qeuickly, even if the presentations are less than fully polished. The
papers carry the names of the authors anzd should be cited accordingly. The findings, interpretations, and conclusions expressed in this
paper are entirely those of the authors. They do not necessarily represenzt the viewc of the World Bank, its Executive Directors, or the
countries they represent.
Produced by the Policy Research Dissemination Center
Household child care choices
and women's work behavior in Russia
Michael M. Lokshin'
JEL Classification: J21,015
Key words: Labor Supply, Child Care, Semi-Parametric Maximum Likelihood, Russia
1 Development Research Group, World Bank, 1818 H Street NW, Washington DC, 20433,
University of North Carolina at Chapel Hill, Department of Economnics, Chapel Hill NC, 27514. The
financial support of the World Bank's Gender Board is gratefully acknowledged. I thank Dr. David Blaa
(UNC-CH) and Dr. Martin Ravallion (The World Bank) for valuable comments.
1. Introduction
This paper examines the interdependency of the women's labor force activity and household child
care choices in Russia.
In the days of the Soviet Union, more women participated in the labor force than in almost
any other country in the industrialized world. In the 1980s about 90 percent of prime-age women
were either employed or went to school (Lapidus, 1985). Women in the Soviet Union worked
full-time the whole year round. There was very little part-time employment; less than one percent
of the work force was employed under such arrangements.
Soviet women could not have been involved in the economy to such an extent if a wide range
of government-subsidized child care programs, such as nurseries, preschool, kindergartens, and
after-school programs, had not existed. The number and variety of state-provided child care facilities
increased steadily throughout the Soviet era. By the late 1980s some 15 million children between one
and six years of age (70 percent of children from that age group) were registered in public child care
institutions (Matthews, 1986).
Reforms launched by the Russian government in early 1992 lead to a dramatic change in the
socioeconomic environment in Russia and put a great strain on the existing system of social
protection and state-subsidized institutions. A sharp decline in GNP in the 1990s resulted in an ever
widening budget deficit, shrinking government-funded programs, and a dramatic decline in the
number of state-run child care organizations. According to GosKomStat (the Russian government's
national statistics agency), the proportion of children in preschool facilities dropped by more than
50 percent (GosKomStat, 1995) between the mid-1980s and the mid-1990s.
Not only has the number of kindergartens and nurseries decreased sharply, but the cost of
sending children there has soared. In the days of the Soviet Union, families' child care costs were
partly or totally covered by subsidies from the federal and local governments and/or by funding from
employers. By now virtually all government child support programs in Russia have been eliminated
and only a handful of companies can afford to pay for the daycare services for their employees1
children. This affects all families with young children and low-income families in particular.
Thus, over the last decade, the situation in Russia has moved from one in which child care
was provided by the government and almost all households with children had access to affordable
and often free-of-charge child care facilities, to one in which the cost of day-care is becoming an
important determinant of household labor supply decisions.
The research described in this paper is prompted by the complexity of the problems faced by
families with children in the transitional economy and by the significant impact that the reform in
the child care system has had on the political and economic environment in Russia.
This study appears significant in light of many contemporary economic theories of
childbearing and female labor force participation. The availability and affordability of child care are
significant factors in a mother's decision about whether to enter, reenter, or remain in the labor force,
and on her consequent decisions about her fertility (Blau and Robins 1989; Leibowitz et al. 1988).
In addition, the availability, convenience, or costs of child care may prevent some women from
participating in the labor force or from getting a full-time job (Bloom and Steen 1990; Blau and
Robins 1989; Floge 1985). Such constraints may be especially acute among less-educated women
or unmarried mothers (Mincer and Ofek, 1982; Presser and Baldwin, 1980). To the extent that child
1
care represents yet another important mechanism in the process by which social inequality is
disproportionately experienced by women and children, a better understanding of the child care
choices and effects on children raised within different socio-economic groups will inform theoretical
and policy discussions about the contexts in which economic changes enhance or compromise
children's development.
In this paper, I model households' child care choices, the decisions that Russian mothers
make about whether to participate in the labor force, and the number of hours that they work. Using
a model of consumer demand for state-provided child care, I estimate how the price of child care,
mothers' wages, and household income all affect households' behavior and welfare.
The econometric model that I use is derived from a theory described in the literature on the
household decision-making about women's participation in the labor market and about child care.
The theory has several testable predictions. I test the hypotheses implied by this economic theory by
jointly estimating reduced-form models of both the discrete and continuous choices of households
using the method of Semi-Parametric Full Information Maximum Likelihood (SPFIML). This
method takes into account the error term correlations across outcomes, and correlation of the error
terms that can result when panel data are used.
The estimation reveals that the decision that mothers make about participating in the labor
force and about the number of hours they work are relatively sensitive to changes in hourly wages,
and, to a lesser degree, to changes in the cost of child care. To examine how household behavior is
affected by child care subsidies and hourly wage policies, I use simulations based on the paper's
estimates. These simulations indicate that the payment of family allowances to households with
children does not have a significant effect on whether or how much they use formnal child care or
whether and how much the mothers work. However, a decrease in child care cost has a strong
positive effect on the labor activity of women with children and on the use of formal child care.
This is the first study to analyze the mothers' employment, mode choice, and demand for
formal child care in Russia in a unified framework. I use this approach because of the simultaneity
of the household decisions about modes of care, employment and demand for child care. Analyzing
all these decisions together has enabled me to obtain consistent estimates of the effects of the child
care prices and of the level of wages on the discrete and continuous outcomes.
Until lately, there has been little research on the economics of child care outside of the United
States. Recently, a certain amount of research on child care has been conducted in the nations of
Western Europe where (as in the US) growing numbers of women with young children have been
entering the workforce (see, for example, Gustafsson and Stafford, 1992; Cleveland, Gunderson and
Hyatt, 1996; Van Den Brink and Groot,1997). To date, there is only a very limited amount of
research on child care and women's labor market activity in the developing countries. This includes
a paper by Wong and Levine (1992) that focuses on child care and household time use by analyzing
the effects of household composition on mothers' employment in Mexico. A paper by Connelly,
DeGraph and Levison (1996) examines the effect of child care arrangements on the rate of women's
participation in the labor force in Brasil. To my knowledge, there are no formal economic studies
of child care in Russia.
This paper is based on recent progress in the theory of demand for child care and women's
labor supply in the U.S. Methodologically, the paper follows the work of Blau and Robins (1988),
Ribar (1992 and 1995), Conelly (1992), Michalopoulos, Robins, Garfinkel (1992), Kimmel (1995,
2
1996), Averet, Peters and Waldman (1 997) all of whom jointly model households' decisions about
child care and mothers' decisions about entering the workforce.
3
2. Theoretical model
The analysis applies to households with children under seven years of age. There are three
forms of child care available to households in Russia: informal care provided by the mother, informal
child care provided by other household members, and formal child care. For households with
children and two parents, the husband is considered a potential provider of free child care. In a
household with a single mother who has no relatives living with her, it is assumed that any informal
child care is provided by children themselves or relatives who live outside the household.
The theoretical model used in this paper is based on the assumption that household members
make choices about their consumption of child care quality, of market goods, and of leisure. A
household's decisions as to the quality of child care it wishes to obtain, and about the amount of time
each member of the household can work are motivated by the desire to achieve the highest level of
household welfare. Household utility maximization occurs at the point where: (i) the ratio of the
marginal utility of the market good consumption to the marginal utility of the quality of child care
purchased on the market equals the ratio of their respective prices, and (ii) the loss in utility from an
additional hour of work is offset by the gain in utility from the additional consumption of market
goods that becomes possible due to the additional earnings and utility that are yielded when the
household substitutes formal child care for in-house care.
This theoretical framework makes it possible to model a wide range of household child care
choices and labor supply decisions that involve various combinations of formal and informal care
that are typical of the extended household structure that is prevalent in Russia.
The model is made tractable through a number of simplifying assumptions. First, it is
assumed that children require continuous care. Second, the household structure and the number of
children are assumed to be exogenous. It seems reasonable to make this assumption, even though
it has been shown (Blau, Robins, 1989) that household fertility decisions depend on the cost of child
care, because of a relatively short time span covered by the three rounds of the survey and rapid
changes that are occurring in the prices of child care. Third, the assumption is made that free child
care is available for a mother during the entire time she is at work. The assumption is important to
enable me to apply the same theoretical framework to households with different structure. Also, this
assumption reflects the fact that the availability of acceptable options is a major influence on the
mother's decision to work. Forth, it is assumed that household members derive utility from the
quality of child care they choose. This utility is represented by the discounted value of a potential
improvement in children resulting from a higher quality of child care or by the current utility of the
family knowing that their children are in competent hands. Fifth, it is assumed that mothers spend
all their free time on child care, i.e., that the mother's leisure time and time she spends caring for
children are perfect substitutes. This, of course, is a simplifying assumption, but Blau and Robins
have demonstrated (1989) that no new insights are achieved by separating the time that a mother
spent on child care from her leisure time. Sixth, it is assumed that households can use all three types
of child care at once, in which case the average quality of child care is the weighted sum of the
quality provided by the different sources. In other words, it is assumed that there is a perfect
substitution between the child care arrangements. In the model, the quality of child care provided
by the mother and/or by the other members of the household is considered to be exogenous.
However, the quality of market-provided care can be chosen by the household.
4
In the one-period utility maximization problem the household chooses its consumption of
a Hicksian composite good G, the average per hour quality of child care Q, the leisure time of the
mother Lm, and the leisure time of other household members L. subject to its budget and time
constraints. The household utility function is assumed to be twice-continuously differentiable and
quasi-concave:
MaxU =U(Lm, Lo, G, Q). (1.1)
The total quality of child care Q is the weighted sum of the exogenous quality of the child care
provided by the mother Qm, the quality of child care purchased on the market Qp, and the exogenous
quality of child care provided by relatives Qo:
Q =Q/Lm +Qp(Hm -T.)+Q.To (1.2)
The budget constraint includes total household expenditures on child care as a function of the
number of children in the household, of the per unit quality price of child care, of the quality of
formal care, and of the time spent by the children in care:
G = E +Wm}H[m +WOHo-NPqQp( H-To), (13)
where E is the exogenous non-wage household income, Hm is the mother's actual work time, Ho is
the other household members' actual work time, N is the number of children in the household, Pq is
the exogenous price of the unit of quality of formal child care, T. is the amount of time spent by
other household members on child care, Wm is the market wage available to the mother, and W0
indicates the market wage available to the other household members.
Finally, the model specifies - under the assumption that children require constant care- the
time constraints affecting the mother, the other household members, and the children:
Lm +Hm =L4 +1Ho +To =1 (1.4)
Ho -TO 20 (15)
0 i > 2e for any jq]
=Pr[e-ji,-er > X(,6ti, -/jjt) +Z(yit -Yjie) for any jqgl]
The demand functions of a mother's
hours at work and hours spent by children in formal care can be specified in a linear form as:
H,ik t herek= K (2.2)
Here, H,,k is the continuous dependent variable k associated with household i in state j at time t. In
the first continuous outcome equation, H2,1 is the number of hours that a mother supplies to the labor
market (if she works), and H1'2 is the number of hours spent by children in formal child care facilities
in those states where formal child care arrangements have been chosen. Xit and Z4 are the vectors of
the variables defined above, pk and ak are vectors of unknown parameters, and Cit is an error term
with mean zero.
The theoretical model assumes that a household makes simultaneous decisions about the
mode of child care it wishes to use, the labor supply of each of its members, the amount of time that
each family member spends on child care, and about the amount of time that their children spend in
formal care. All of these decisions are determined by the exogenous characteristics of the family and
individual family members, both observable and unobservable. It is possible to estimate discrete
choice equation (2.1) and continuous variable equations (2.2) jointly.
There are several estimation issues that need to be discussed. First, the error terms in the
discrete e and continuous Vk equations may be correlated across states and among each other. The
correlation across states is a correlation among disturbances in the state-specific indirect utility,
functions V1jt's. If, for instance, a mother's participation in the labor force is determined by, among
the other factors, some unobservable taste for work, it is going to hold true for all of the states in
which a mother works.
The same example of an unobservable preference for work can help to illustrate the
possibility of a correlation among the errors terms of equations (2.1) and (2.2). In the continuous
equation, the number of hours that a mother supplies on the labor market depends, among other
things, on her taste for work. Women with a high preference for work can be expected to work
longer hours and are more likely to be employed. This means that there may be a correlation between
the disturbance in the equation for a household's choice of a discrete state and the equation for the
amount of time that women supply on the labor market. Similar correlations can exist for the
equations that determine the following: (i) labor supply decisions of the other family members, (ii)
time spent on child care by the mother and the other family members, and (iii) household child care
arrangements.
In addition, because panel data are used in the model, there exists the possibility of a
correlation in the error terms among the multiple observations of the same family (correlation
between Eii,2 and ejjtl, t1tt2 and correlations between ,tlk and iC2k, t1#t2). Macro time effects (not
7
household-specific) in disturbances e and i can be controlled by the introduction of time-specific
dummy variables into equations (2.1) and (2.2).
If it is assumed that the errors are not correlated when in fact they are, then the point
estimates will be biased and inconsistent. To account for such possible error correlations in a
tractable way I impose a factor structure on the disturbances in equations (2.1) and (2.2):
-'i,, = 4j, + Pi 1VI i + Pi, 2 V2 i (3.1)
1= At + fjVIi + Z2V2U (3.2)
4' =t + l VIi + 5t2V2ti (3.3)
where pij, is an independent extreme value and , Xi,, and yi, are mutually-independent random
variables that are assumed to be independent of the regressors in the model. V1 is a permanent factor
(the factor that remains the same for the household at any time point and is similar to the
household-specific effect) while V2 is a transitory factor (within a single household, the factor will
be different at any two different points in time). These factors are unobservable variables that
influence the choices made by households and that have no correlation with explanatory variables.
p's, 's, and ;'s are factor loadings that represent the effect of a given factor in each equation.
We introduce a two-factor structure to account for the two major possible sources of
heterogeneity in the disturbances - the inter-equation correlation of the error terms and the panel
structure of data. The use of both time-invariant and transitory factors helps to take into account this
potentially complex form of correlation. These assumptions about the structure of the error terms in
the equations (2.1) and (2.2) are considerably more flexible than the common practice of imposing
a specific functional form of the distributions of v's (see for example, Gourieroux and Monfort
(1996) and Berry and Pakes (1990)).
The system of equations (2.1-2.2) with the error structure (3.1-3.3) can be estimated by the
Semi-Parametric Full Information Maximum Likelihood (SPFIML) method developed by Liard
(1978), and Heckman and Singer (1984), and applied to simultaneous equations by Mroz and
Guilkey (1992), and Mroz (1999).
Assuming the independent, extreme-value distribution of e, the contribution of the discrete
outcome part of the model to the likelihood function has the form of a multinominal logit,
conditional on V's. Assuming that the error terms of the continuos outcome equations are normally
and independently distributed, these equations contribute to the likelihood function in the form of
normal regressions. I assume that the distributions of the v's in equations (2.1-2.2) may be
approximated by the following step functions:
M
Pr(V, = vim) =Pm P > O and XPm = 1 (41)
m=4
K
Pr(V2 = v2k)7zi )> Oand YZ; =1 (4.2)
k --
where v's are the points of support in the distribution of the factors 1 and 2, P and Xt are the
probability that the factors take value v, and M and K are the numbers of points of support of the
distribution of each V.
8
The above specification implies that each household has an access to all possible child care
arrangements and all potential employment outcomes. However, a significant proportion of Russian
families have no access to formal child care facilities. Also, Russian legislation specifies that only
children older than 18 months of age can be accepted by formal care institutions. For these groups
of families the conditional contribution of the discrete outcome equation to the likelihood function
is calculated based on a restricted set of possible forms of care, i.e., that there are no formal care
arrangements available in the choice set of these households:
Households with access to all forms of care Households with no access to formal care
or households with children younger 18 months
e e,Xi, +Pj,VI.+P2V2, e6jx,+A1V1,+A2v2k
Pr(Yj, = jIVIm,V2k) = 4 for j=1.4 Pr(Y, = jilVlmV2k)NA =_2 for j=1,2
1 + A'1+A V1k+A12'2 1+ + t
k=4 k=4
pr(yi, =IV°lmIV2k)A 4 Pr(yi, = 0JIVI.mV2k)NA= 2 (5)
1 + YeA,9,,+"v 1 + A;eXl+flVl8+An2V2k
k=1 k=i
where Pr(Yit=jlvlm,v2k)A is the probability that household i (which has an access to formal child care
facilities) chooses state j at time t conditional on the realization of factors V, and V2, and
Pr(Yjt=jlv,m,v2k)NA is the same probability for those households with no access to the formal child
care.
For the continuous outcome equations (2.2), the conditional on the v's probabilities are:
1 H.' -xtd - Z,; - 'ZiVIm - '2V2k
Pr(H iIv,m,v2k)=- Z -) (6.1)
a it
1 .2 2a
Pr(H2 vl ,Hv2k )=t it ZitJ Y -1Vlm-2V2k) (6.2)
i,'l. 2k ki C2it
where Hi,, and Hi,2 are the dependent variables in the continuous outcome equations (2.2), 4 is the
probability density function of standard normal distribution, and a'it and a2i, are the square roots of
the variances of the error terms in equations (2.2).
Thus, the semi-parametric log-likelihood function for the system of equations (2.1-2.2) with
the error structure (3.1-3.3) is:
N M T K
3 =jn(,Pm[J7Ijzk Pr(YI, = jlvlmv2k ) Pr(Yt = jVIlmsV2k ) Pr(Hi,' vjm,V2k )Pr(Hj, IV,m,V2k )]) (7)
i4 m=lI t-1 k/ri
where N is a number of households in the sample, T is the number of times that a household appears
in the sample (the total number of observations in the sample is T*N), M is the number of points of
support for the first factor, Pm are probability weights associated with the first, time-invariant factor,
K is the number of points of support for the second, transitory factor, 7rk are probability weights
associated with the second factor, vim and v2k are the points of support of the distribution of factors
9
V, and V2.
Choosing a priori numbers of points of support M and K, the log-likelihood function q is
maximized over a's, n's, y's, 'Ts, c's, P's, i's, and v's. For identification purposes, the two points
of support for both factors are normalized to equal 0 and 1, respectively4. The number of points of
support is increased until the difference in the log-likelihoods of consequent maximizations satisfies
the convergence criteria.5
The joint distribution of the error terms (3.1-3.3) is unknown, so the sample statistics of the
estimates cannot be derived analytically. It is feasible to estimate the covariance matrix 0) of the
coefficients in the model (2.1-2.2) by inverting the Hessian matrix of the second derivatives of the
log-likelihood function W. However, the numerical approximation of the Hessian matrix can be
difficult to obtain when the values of certain estimates by far exceed the rest of the coefficients, or
when the function becomes flat near the optimum. In the paper I use method of bootstrapping to
determine the correct standard errors for the estimated coefficients.
Let S,=Si(Yit*,Hlit*,H 2 *,Xjt*,Zit*), i=l,...,n be a randomly drawn with replacement sample
from the empirical distribution. It follows from the model (2.1-2.2) that Hlit*, H2it*, Yit*, are
functions of Xi,* ,Zi,* and the error terms. Let ,Bj* denote a vector of the bootstrap estimates obtained
from a maximization of the log-likelihood function 9 on the sample S1. This process can be repeated
B times to yield bootstrap estimates , I *... PB*. The bootstrap estimator of the asymptotic covariance
matrix 0 is given then by
0= . I(-Il3 )(4-I3)' where 8 = B D.
The estimates of the base sample coefficients are used as starting points for the bootstrap
optimization in order to speed up the convergence. The covariance matrix El is estimated based on
B=500 bootstrap repetitions of the log-likelihood maximization, which should provide a suitable
level of accuracy for the estimated standard errors (D. Andres, M. Buchinsky).
4. Data and Variables
4The functional form for the normalization of probability weights, the points of support for the
likelihood faction (6) and the estimated parameters are given in the Appendix
5 As a convergence criteria I use Akaike's information criteria, recommended by Gritz (1993).
According to this method the full model is estimated with an increasing number of points of support,
until the improvement in the value of the log-likelihood function SE is less than the number of additional
parameters estimated. In the case of two-factor model, this method is applied first to determine the
number of points of support for the permanent factor V,. Then, keeping the number of points of support
of the permanent factor constant, the "optimal" number of points of support for the time-variant factor V2
is determined by increasing the number of points of support for that factor until additional point of
support fail to produce any significant improvement in the value of log-likelihood function.
10
This research is based on the data from the Russian Longitudinal Monitoring Survey (RLMS)6, the
first and only nationally representative sample of households in the Russian Federation. The survey
comprised seven rounds conducted in (D) September 1992, (11) February 1993, (1E1) August 1993, (LV)
November 1993, (V) December 1994, (VI) October 1995 and (VI) October 1996. Rounds I-IV
surveyed over six thousand households while Rounds V, VI and VII surveyed a different panel of
approximately four thousand households. The data were weighted across the rounds for
comparability and to ensure that the survey was representative on the national scale.
In this analysis, I use a pooled sample of households with children younger than 7 years old
based on the results of the last three rounds of the survey. The data on the first four rounds of the
survey (1992-93) gathered no information on the availability of child care facilities, which makes
it impossible to apply the model of child care choice and labor market behavior to the households
covered in the first four rounds.
The initial sample of households for rounds V, VI, and VII of the survey was identified from
a stratified three-stage cluster sample of residential addresses. Cities as well as urban and rural
portions of rayons (political and geographic units about the size of counties in the United States)
were the area units selected in the first stage. These 38 rayons were stratified by the eight regions
and by the percentage of the urban population within each region. Within each area chosen in the
first stage, a sample of voting districts (primary population points) was randomly chosen from a
geographically ordered list of voting districts falling in that area.
There are 1,262 households with children under the age of seven in the pooled sample of
rounds V, VI, and VII, and these households are represented by 2,162 observations (an average of
1.77 observations per household). The data set includes information on the individual members of
these households, household-specific information and data on the community level. It also contains
information on the modes of child care arrangements made for each child in the household, the
amount of time each child spent in formal and informal child care, and the amount of money paid
for formal child care during the week of the survey. The part of the questionnaire that was
administered to each individual household member yielded data on how much time each household
member spent looking after children and was active at the labor market, as well as information on
their monthly wages. The part of the questionnaire that was administered to one respondent per
household on matters that affected the household collectively yielded information on any non-wage
household income and on the household's composition. The part of the questionnaire that was
administered to a group of respondents who represented the whole local community yielded data
about the availability of different forms of child care, and this information was collected for each of
the 160 primary population points.
4.1 Dependent variables
6 The weights and a range of issues related to the sample design and collection of these data are
explained in depth in the documents that can be found in the home page of the RLMS. The data sets can
be obtained free through the home page: www.cpc.unc.edu/projects/rlms/rlms_home.html. Lokshin and
Popkin (1999), and Lokshin, Popkin, Harris (1999) give additional information on the sample and data
set.
11
The dependent variable for the discrete outcome equation is defined according to the possible
combinations of a mother's employment status and the mode of child care, which are shown in Table
1. These combinations are: (0) - the mother does not work and stays at home with her children; (1)
-the mother works, the other household members also work, informal child care arrangements are
used; (2) - the mother works, the other household members do not work, informal child care
arrangements are used; (3) -the mother works, the other household members work, formal child care
is used; (4) -the mother works, the other household members do not work, both formal and informal
care arrangements are employed. The distribution of households by the mothers' labor force
participation and by the mode of child care is presented in Table 2.
More than 45 percent of households with young children have non-working mothers. The
percentage of mothers who stayed at home with their children increased slightly from 1994 to 1996.
Among the households that use other types of care, the largest single group is formed by families
using household members other than the child's mother as child care providers. A third of the
households with a working mother use this type of care. A relatively high share of the families, 8
percent, used only formal facilities for child care. And a small minority of Russian households uses
both formal and informal child care.
Table 3 illustrates the distribution of the dependent variables for continuous outcome
equations, i.e., the time that mothers spent working and the time that children spent in formal care.
Both continuous outcomes are observed only among the sample of working mothers or on the sample
of children in formal care.
4.2 Explanatory variables
The definitions and descriptive statistics for the explanatory variables in the system of equations
(2.1-2.2) are presented in Table 3. Several key variables of interest are discussed in details below.
Price per quality unit of child care (P). In the RLMS, households reported their weekly
expenditures on child care and the time that their children spent in a formal child care facility during
the week of the survey. There is no direct way to relate such information to the quality of child care
provided as no data were collected on the regional characteristics of child care facilities (such as the
sizes of groups of children in pre-school establishments, quality of personnel, etc.). Like Blau and
Robins (1988), I assume that the quality of formal child care is uniform within a population point7
and I use the average local per hour price of care as a proxy for the child care price.
Mother's offered wage (W,, The wage rates available to each mother have been imputed
using Mincer's (1974) type earning function regression with a control for selectivity (standard
Heckman correction)8 run on a sub-sample of working women for whom hourly wage data were
available. The hourly wage has been calculated as a ratio of the women's monthly earning and the
7 The average prices of child care are calculated for 30 population points in the sample. Each
population point includes about 40 households with children.
8 Regression coefficients for the wage equations are shown in the Appendix 2. For identification
in the selection equation I use the standard set of household characteristics that can influence the
mother's labor force participation decision, but are uncorrelated with the potential wage rate.
12
total number of hours they worked during the month the survey has been administered. In the
absence of data on the total amount of time a mother had worked during the preceding month, the
imputations were made based on the number of hours worked during the week of the survey.
In the wage regression, the following explanatory variables have been used to predict
mothers'hourly earning - the mother's educational level, her age, details on the region she lived, the
type of settlement where she lived (urban-rural), the number of children she had (as a proxy for work
experience), her marital status, and the amount of time she had been in her current main job.
Imputations are made based on the women's predicted hourly wages with the job tenure of
non-working mothers being equal to zero. Here the offered wage is assumed to be a wage that a
mother could earn if she were to start a new job.
Ofered wages ofotherhousehold members (Wa:The wage rates available to otherhousehold
members are calculated in a similar way to the wage rates available to mothers. Different regressions
were run to predict wages for household members of different ages or genders. After the imputations
two methods were used to obtain the wage WO. Under the first specification the offered wage of other
household members is equal to the lowest wage earned by any household member except the mother.
The second specification uses the average wage of all working household members as an explanatory
variable in the model.
Non-wage household income (E): Non-wage income is measured as the household monthly
income from all sources other than the wage income. This may have included social security
transfers, private transfers, in-kind inconme, and income from home production. The structure of
household income changed over the rounds of the survey and certain adjustments were made to
ensure compatibility of the income data across all of the survey rounds..
Other explanatory variables include some individual characteristics of the mother such as her
age and level of education, household demographics and size, the number of children in the
household and their ages, the number of pensioners in the household, and the household's
geographical characteristics.
5. Results
5.1 Estimated coefficients
The results of the estimation of the system of simultaneous equations (2.1-2.2) are presented for the
specification with four points of support for the permanent factor V, and four points of support for
the transitory factor V2. A further increase in the numbers of points of support M and K failed to
result in a significant increase in the value of the likelihood function9. The estimated coefficients are
shown for the model estimated with (V,>O and V2>0) and without (V1=O and V2=O) adjusting for
unobservable heterogeneity. The estimation of the model without adjusting for a possible correlation
in the error terms between equations (2.1-2.2) is essentially a joint yet independent estimation of the
9 Using more than four points of support for each factor lead to a significant increase in the time
of convergence for the optimization procedure. The bootstrap technique that uses multiple optimizations
of the log-likelihood function would require a prohibitively long time for the estimation of the standard
errors in that case.
13
modified multinominal logit (MNL) of the form (4) for the discrete outcome and two ordinary least
square (OLS) estimations of continuous outcome equations.
According to the likelihood-ratio test criterion, the MNL'OLS specification is rejected in
favor of the SPFIML estimation. The log-likelihood value for the independent MNIVOLS estimates
is -9650.50 based on 164 parameters. The log-likelihood value for the SPFIML estimate is -9362.65
based on 182 parameters. This is an increase of 287.85 in the log-likelihood value for 18 additional
parameters. This means that using the model that does not control for any correlation in the error
terms may have biased the estimated coefficients and introduce inefficiencies in terms of the size of
the standard errors.
Tables 4a and 4b present the estimated coefficients of the discrete outcome equation for both
specifications. The effect of mothers'wage rates on the discrete outcomes was much stronger when
SPFIML was used than it would have been if it had been assumed that the error terms for all states
were independent. For example the coefficient on the log of mothers' wages for state I estimated by
MNL is one third of the value estimated under SPFIML specification. The effects of the child care
prices on household discrete outcome choices are also stronger if the model is estimated by SPFIML.
While both methods produce estimates consistent with the predictions of the economic theory, using
SPFIML yields more accurate estimates.
Estimates of the continuous outcome equations are shown in Table 5. The estimated
coefficients of the independent equation of the hours of work are consistent with the coefficients of
the model estimated with the selectivity bias correction. However, the latter yields significantly more
precise estimates. The standard errors of the SPFI4L estimation are on average 50 percent lower
than the standard errors obtained from the OLS estimation. Both estimations show that increases in
the prices of child care have a strong negative effect on the number of hours that the mothers work,
while increases in wages available to mothers have a significant positive effect on their participation
in the labor market.
The estimation of the equation of the demand for child care indicates that the correlation of
the error terms reduces the effects of mothers' offered wages and the price of child care on the
number of hours during which households use formal care facilities.
5.2 Simulations
To examine the effects of the estimates summarized above on the model (2.1-2.2) I simulate how
households would respond to changes in the specific parameters used in the model. In a given
simulation, the same values for all of parameters of interest are assigned to all the households in the
sample. The simulated probabilities for the discrete model outcomes and simulations for the
continuous models are generated for each household at every time point by integrating over the
estimated heterogeneity distribution and then averaging the probabilities across the sample.
5.2.a Discrete outcome model of the household's child care mode and the mother's labor supply
The simulated distributions of the probabilities for the discrete outcome equation are shown
in Table 6.
Price of child care: I estimate the impact that subsidizing the prices of child care would
14
have on the probability, that households choose a particular child care/labor supply mode, make a
particular decision about the labor supply of the mothers, and about the amount of time children stay
in formal care'". As predicted by the theoretical model, these simulations show that an increase in
the per hour price of care discourages households from choosing formal child care arrangements
(mode 3 and 4), and also discourages mothers from working.
If formal child care were fully subsidized (in other words, the price was zero) as opposed to
the current situation in which the average child care price is 5.6 rubles per hour, this would result
in an 1 1.3 percent increase in the rate of mothers' labor force participation. It also would result in a
10.5 percent increase in the use of formal care facilities. On the other hand, increasing the prices of
child care would discourage other household members from working. Doubling the cost of child care
would decrease by 9.6 percent the proportion of households where other members work. A negative
effect of child care prices on the states where only informal care is used can be attributed to the
existence of a shadow price of the time other household members spend taking care of children,
which can be correlated with the actual market price of child care. Alternatively, this negative effect
of child care prices can be attributed to monetary transactions between the members of a household.
Mothers 'offered wage rate: The potential wages that a mothercould earn if she works have
a strong and positive effect on the probability of her participation in the labor market. A higher wage
offer will increase the opportunity cost of her staying at home and, therefore, will increase mother's
propensity to work. Simulated changes in the probability that various types of child care
arrangements would be chosen and that mothers would participate in the labor force in response to
changes in the level of offered wages are presented in Table 6.
The doubling of a mother's offered wage rate (from 10 to 20 rubles per hour) would increase
the number of households with working mothers by 24 percent. At the same time, the proportion
of households with non-working mothers would decline from 50.5 percent of the sample to 40.7
percent. This change in the offered wage rate would also affect the distribution of households with
working mothers, particularly in those where other household members also have jobs. The
proportion of such households would decline by 17 percent relative to the initial state. The doubling
of the offered wage of the mother would also increase by 19.1 percent the use of paid care facilities.
Offered wage rate ofother household members: Anincrease in the wage rate of the otier
household members reduces the probability that the households will choose to keep members other
than the mother from working (states 2 and 4). The third row in Table 6 demonstrates the impact of
an increase in the wage rate available to other household members by 100 percent (from 10 to 20
rubles per hour") - a 14.6 percent increase in the proportion of households where other household
members work. The slight decrease in the number of non-working mothers can be attributed to
imperfect substitution between the maternal and other forms of care. An increase in the wage rates
10 The changes in the price of care have no effect on the behavior of those households that had no
access to the formal care facilities. These households, however, are included in the sample and all the
results are obtained by averaging the predicted probabilities over the whole sample.
"' The use 10 ruble wage subsidies and 5.6 ruble child care subsidies is arbitrary. Following Blau
(1999) 1 assume that the relative cost-effectiveness of the two types of subsidies is not very sensitive to
the specific magnitudes of the subsidies.
15
of other household members would increase total household income, thus (according to the
theoretical model) discouraging mothers from working. However, as household income grew and
formal care became more affordable, households would switch from the maternal to formal care
mode and the number of working women with children would increase.
Demographic and geographical variables: Table 7 presents simulations that summarize
the effects of several demographic variables on the discrete outcomes. Households with children
under the age of three are significantly less likely to choose to put themselves in states where the
mothers of these children work. The probability that those households with children younger than
18 months will have a working mother is 50.4 percent lower than the probability that households
with older children will have a working mother. Younger children require more intensive care and,
therefore, have a strongest impact on the likelihood that their mothers will be employed.
This estimation shows that household non-wage income has no significant effect on the
household's choice of a type of child care. However, the household's structure does appears to be
an important determinant of its choice of child care. The availability of grandparents and other family
members in households encourages these households to use informal types of child care and may
minimize their propensity to use formal care. The presence in the household of children between the
ages of seven and twelve years of age and of teenagers has a positive effect on mothers' labor force
participation in all states of child care arrangements. Households in the rural areas of Russia are
more likely to have working mothers than families in the urban or metropolitan areas of the country.
A mother's educational level can be regarded as a proxy for the quality of maternal child
care. Theory would suggest that the mother's educational level would have a negative effect on her
labor force participation, in part because better-educated mothers are likely to provide higher quality
care. However, the estimation fails to show any such pattern and instead indicates that mothers who
are high school graduates are 13.2 percent less likely to choose formal child care arrangements than
mothers with higher levels of education.
Maternal employment and the use of paid care are both higher in single-parent households
than in two-parent households. Such households are more likely to use formal child care, and single
mothers are 10.3 percent more likely to work than married women with children.
5.2.b Continuous outcome models for women's time spent working and children's time spent in
fornal care
Women's time spent working: A simulation of the effects of the exogenous variables
on the mothers'hours of work model confirms the predictions of the theoretical model. An increase
in mothers' wages has a positive effect on the number of hours that mothers spend at work, while an
increase in the costs of child care has a negative effect on their work hours. An increase in the cost
of child care lowers mothers"'effective" wages, which, in turn, makes them work shorter hours. The
effect of child care costs on labor hours is weaker than the effect of wages. The wage rate of other
household members seems to have no significant effect on mothers' work hours. The higher the
household's non-wage income, the less time mothers work. Of the demographic variables, the
presence of small children in a household decreases the mother's labor activity. On the average,
single mothers work 20 hours more per week than married women with children. Younger and
less-educated mothers tend to work longer hours. Mothers in the metropolitan areas of Russia tent
16
to work shorter hours than mothers in rural areas.
Children's time spent informal care: The theoretical model offers no predictions
about how the prices of child care affect the number of hours that children spend in formal care.
However, assuming that formal care is a normal good, a negative child care cost coefficient of the
hours of care regression could be expected. Indeed, simulations show that an increase in the average
price of child care within a given population point decreases the number of hours that children spend
in formal care. The mothers'wage rate has a positive and significant effect on the use of formal care,
while the presence in the household of children under the age of 18 months reduces the household's
use of formal care. The estimations fail to reveal any significant differences in children's attendance
at formal care facilities among different regions of Russia. Better-educated mothers use formal care
more often than less-educated mothers.
6. Robustness
The robustness of these results can be tested by evaluating some of the alternative specifications of
the model (2.1-2.2). Table A3 in the Appendix shows the SPFIML estimation coefficients for the
discrete and continuous outcomes equations where the regressor is the minimum wage rate of the
household members other than the mother. These estimations suggest that using this specification
makes little difference in the effects of the key policy variables compared to when the regressor is
the average wage rate of the household members other than the mother.
The "average wage" estimation results are more consistent with the prediction of the
theoretical model in terms of how the wage rates of the other household members affect their
employment. For those states with the working household members (other than the mother) (states
1 and 3), an increase in the "average wage" variable has a positive effect on their level of labor force
participation. However, in state I (households in which the mother and other members work but only
informal child care is used), the "minimum wage" specification results in a negative significant
coefficient for this parameter.
In the continuous outcome equations for the hours of mothers' employment and the demand
for formal care the estimation of "minimum wage" specification shows that the wage rates of other
household members have a weaker impact on the outcomes in both equations. The effects of the
other parameters are not different from the predictions of the "average wage" model estimation.
7. Policy implications
This analysis demonstrates that the price of child care and the wage rates of mothers and other
household members have a strong effect on the maternal employment and on the use of formal child
care. When allocating limited budget funds, policy-makers need to decide what policies are most
likely to improve the well-being of Russian families. The three key policies in this area are wage rate
subsidies, child care subsidies and direct government transfers to the households with children.
These instruments can affect the welfare of Russian households by increasing their income and/or
improving the developmental and educational outcomes of young children.
The discussion in this section provides the information that the Russian government needs
to determine the costs and benefits of accomplishing a given policy objective and to identify the
17
tradeoffs involved in attempting to achieve multiple objectives.
The empirical analysis of the model (2.1-2.2) indicates that wage and child care subsidies
have the biggest positive impact on the employment rate of women with children. Wage subsidies
make having a job more attractive than not working. Child care subsidies can affect the utility
derived from working and from using formal care. They tend to increase maternal employment by
inducing certain households to switch from a state where the mother stays at home with her children
to states where the mother works. Child care subsidies also affect households with working mothers
by prompting them to change from informal to formal child care. However, these subsidies do not
in any way affect the level of maternal employment in households with working mothers. Wage
subsidies are paid to all working women with children and thus the net increase in employment will
be reduced by the amount paid to women who already have a job.
The effectiveness and distributional impact of child care and wage subsidies can be
determined through simulations. Suppose, similar to the results in Table 7, the government were to
introduce child care subsidies that make the child care free. These subsidies are available only to the
families that use paid child care. The child care subsidies will induce some non-working mothers to
enter the labor market, and may induce households with working mothers to use more formal care.
The results of this simulation are presented in Table 8. The total government expenditure on fully
subsidized care would consist of its subsidies to those households that were using the formal care
before the new subsidy policy was implemented plus the subsidies to those households that switch
from using informal care to using formal care because of the new subsidized price. This policy
measure would result in a 11.2 percent growth in the number of working mothers"2 and would affect
471.9 households with children in the sample.
An alternative approach for the govermment to take would be to introduce wage subsidies for
the women with children. Suppose the Russian government wants to spend the same amount of
money as it spent on child care subsidies on wage subsidies. All households with working mothers
would be eligible for such subsidies. This increase in the hourly wage rate would induce some
mothers who previously did not work to enter the labor market. Also, those mothers who were
employed before the policy was implemented would work longer hours. The number of households
that use formal care would also go up. The magnitudes of these changes can be determined by
solving equation (8):
E =Nb - o+Na (Wb-+dv), (8)
where Nb is the number of the households where the mother works before the wage subsidies, Na is
the number of households where the mother enters the labor market after the increase in hourly
wages, aW is the change in wage rates due to the subsidies, Wb is the wage rates before the subsidies,
and E is the government expenditure. The fixed-point solution of equation (8) with respect to Nb, Na,
and 8W yields a 5.6 percent growth in women's labor force participation rate under this policy. If
the wage subsidy were equal to the amount of money that the govermment would otherwise have
spent on child care subsidies, this would increase mother's wage rates by 2.06 rubles per hour, and
12 When child care subsidies lead to a raise in the demand for formal care, this can also result in
the increase in the demand for the labor of prime age women, as most of the staff in the child care
facilities are women.
18
1044.1 households would benefit from this policy.
Child care subsidies would increase maternal employment by almost twice as much as wage
subsidies. Both measures would increase the total household incomes of the eligible households. In
the case of those households that were using formal child care before the child care subsidies were
introduced, this increase in income would be a result of a decrease in the amount of money that they
previously spent on child care and an increase in the mothers' wage income because she would be
working longer hours. In the case of those households that start using market care because of the
lower prices, and as a result of which the mother enters the labor market, the increase in their income
is generated by the wage income that the newly-employed mother brings into the household budget.
The wage subsidies would increase total household income by increasing the household's
earned incomes because of the additional time that mothers who were already in the labor force
would spend at work and because of the additional wage that mothers who had not previously been
working would bring into the household budget after they became employed.
A third policy measure that may influence the welfare of Russian households with children
is a family allowance transfer. Assume again that the government wants to spend the same amount
of money on family transfers as it spent on the first two policies. According to the model, this would
result in a slight drop in the level of labor force participation of mothers, but would increase the total
household income by increasing households' non-wage income. All households with children would
benefit from this policy.
These simulations indicate that child care subsidies are substantially more effective than
wage subsidies at increasing employment per ruble of government expenditure. They would also be
more effective in inducing households to use formal child care.
Comparing the effects of the above three policies on the income levels of Russian households
with children reveals that the child care subsidies would produce the largest increase in the family
income both for the beneficiaries, and, if averaged out, for the whole sample. Wage subsidies would
produce the next biggest increase, following by family allowances transfers. Households that use
payed care would experience on the average a 20 percent increase in their incomes as a result of the
fully subsidized child care. The effect of the wage subsidies would be significantly smaller (yielding
a 7.5 percent increase in total household income), although it would affects more families. The
uniform family allowances would increase the income level of all households with children by only
3.4 percent.
8. Conclusion
Estimating the joint model of households' child care choices, mothers' labor supply decision, and
household demand for formal child care confirm the predictions of the theoretical model developed
in this paper. The estimations indicate that economic incentives have a powerful effect on the work
behavior of women with children in Russia. The level of wages available to them and the costs of
child care can both be expected to affect women's labor force participation and labor supply
decisions. Child care costs affect which child care arrangement households choose. When the costs
of care are high, this discourages households from using formal child care and increases the number
of households that rely only on informal care.
Government subsidies on child care may increase the number of mothers who work, thus
19
increasing the incomes of poor households and lifting some families out of poverty. The simulations
in this paper have shown that measures such as subsidies aimed at reducing the costs of market child
care are more effective than measures that increase women's wages in increasing the number of
mothers who work and the number of hours that they work.
in reducing poverty.
A significant proportion of Russian households with children use a network of family
members to provide most of their child care. Those families in which some members do not work
are unlikely to use paid care as informal care is available from those family members. It would
appear that in Russia this kind of informal care has substituted for the care that used to be provided
by the Soviet government, which would explain the fact that there is still a relatively high level of
participation by women in the labor force despite the sharp drop in the number of kindergartens and
nurseries in the last 10 years.
Further research is needed in several areas. First, this paper has assumed the perfect elasticity
of the labor market with respect to an increase in women's labor supply. However, given Russia's
shrinking economy, it seems unlikely that the market could actually accept a significant influx of
women without any wage adjustments. An increase in the women's labor supply might lead to a drop
in real wages, which would mean that the actual effects of the policies simulated in this paper would
be quite different. This may also be true of the child care market where an increase in the demand
for the formal care might cause the market prices of child care to increase.
The next question that needs further research is the distributional impact of the various
potential government policies. Even if the child care subsidies produce, on average, the largest
increase in the household income (compared to subsidizing wage rates or having a system of family
allowances), it is unclear which households would benefit most from these subsidies. The poorest
households with children in Russia may not benefit from this kind of subsidies, which means that
other policy measures would be needed to improve the well-being of such families.
The serious limitation in the present analysis is the lack of direct information on the quality
attributes of care provided under different arrangements and at different facilities. While the data
used in this paper are averaged by population point, center-specific information would be more
appropriate. Further work should also consider marriage and fertility decisions of the households as
factors that may influence households' choices of child care arrangements and its members' labor
supply decisions.
20
References
Andrews, D., Buchinsky, M., (1999) "A three-step method for choosing the number of bootstrap
repetitions," forthcoming in Econometrica
Averett., S., Peters, H., Waldman, D., (1997) "Tax Credits, Labor Supply, and Child Care." The
Review of Economics and Statistics, 79, No. 1.
Berry, S., Pakes, A., (1990) 'The Performance of Alternative Simulation Estimators." Discussion
Paper, Yale University
Blau, D., Robins, L., (1988) "Child care cost and family labor supply." Research in Economics and
Statistics 70(3):374-81.
(1989) "Fertility, employment and child care cost." Demography 26(2): 287-99
(1991) "Child care demand and labor supply of young mothers over time." Demography
28(3): 333-351
. (1989) "The labor market effects of child care subsidies." Unpublished manuscript.
University of North Carolina at Chapel Hill, April.
(1990) "The economics of child care" New York: Russel Sage Foundation
(1993) "The Supply of Child Care Labor. Journal of Labor Economics, 1993, vol 11, no.2
Blau, D., Hagy, A., (1998) "The demand for quality in child care" Journal of Political Economy.
106(1): 104-139.
Bloom, D., Steen, T., (1990) "The labor force implications of the expanding child care industry."
Population Research and Policy Review 9:25-44.
Chaplin, Hofferth, D. and S., and Wissoker, D. (1995) "Price and Quality in Child Care Choice: A
revision." Journal of Human Resources.
Coe, R., (1978) "Dependency and Poverty in the Short and the Long Run," in Accounting for Race
and Sex Differences in Eamings and Other Analyses of the First Nine Years of the Panel
Study on Income Dynamics, eds. Greg J. Duncan and James N. Morgan, Vol. 6., Five
Thousand American Families--Patterns of Economic Progress, Ann Arbor: Institute for
Social Research, University of Michigan.
Connelly, R., (1992) 'The effect of Child care costs on married women's labor force participation."
Review of economics and statistics 74(1), 83-90
, (1992) "Self-employment and providing child care" Demography, Vol. 29, No. 1
,(1991) 'The importance of child care costs to women's decision making." In Blau, D.,
ed. "The economics of child care.", New York: Russell Sage Foundation, 87-117
, (1989) "Determinants of weekly child care expenditures: a comparison of married and
unmarried mothers." Mimeo, Bowdoin College
Connelly, R., DeGraff, D., (1996) "Women's employment and child care in Brasil" Economic
development and cultural change
Cleveland, G., Gunderson, M., Hyatt, D., (1996) "Child care costs and the employment decision
of women: Canadian evidence." Canadian Journal of Economics, XXIX, No. 1
Floge, L., (1985) "The dynamics of child-care use and some implications for women's
employment." Journal of Marriage and the Family 47:143-154.
Glinskaya, E., Mroz, T. (1998) 'The gender gap in wages in Russia from 1992 to 1995."
Forthcoming in Journal of Public Economics.
Gourieroux, C., Monfort, A., (1996) "Simulation-Based Econometric Methods." Core Lectures,
Oxford University Press, Oxfort
Gritz, R., (1993) "The impact of training on the frequency and duration of employment" Journal of
Econometrics, vol. 57, pp.21-51
Gustafsson, S., Stafford, F., (1992) "Child care subsidies and labor supply in Sweden" Journal of
21
Human Resources 27(1): 204-229.
Heckman, J., (1974) "Effects of child care programs on women's work effort." In Economic of the
family, ed. T. Schultz, 491-518. Chicago: University of Chicago Press.
,(1979) "Sample selection bias as a specification error." Econometrica 47: 153-61
Singer, E., (1984) "A method of minimizing the impact of distributional assumptions in
econometric models for duration data." Econometrica. 52: 271-320
Hofferth, S., (1987) "Child care in the United States." Statement before the House Select
Committee on Children, Youth and Families, July 1.
Kimmel, J.,(1995). 'The effectiveness of child care subsidies in encouraging the welfare-to-work
transition of low-income single mothers." American Economic Review. Papers and
Proceedings, May.
Laiard. N., (1978). "Nonparametric Maximum Likelihood estimation of a mixing distribution," Journal
of the American Statistical Association, 73(1978), 805-811
Lapidus, G., (1985). "The Soviet Union." In J. Farley, ed. Women Workers in Fifteen Countries.
Ithaca, NY: ILR Press.
Leibowitz, A., Waite, L., Witsberger C., (1988) "Child care for preschoolers: difference by child's
age." Demography 25(2):205-20.
Leibowitz, A., Klerman, J., Waite, L., (1992) "Employment of new mothers and child care choice:
difference by children's age." Journal of Human Resources 27(1): 112-33.
Lokshin, M. (1999) "Household child care choice and women's labor supply" Dissertation thesis.
University of North Carolina at Chapel Hill.
Lokshin, M., Harris, K., Popkin, B., (1999) "Single mothers and poverty in Russia", CPC Working
Paper (Chapel Hill: Carolina Population Center, The University of North Carolina).
Lokshin, M., Popkin, B., (1999) "The emerging underclass in the Russian Federation: Income
Dynamics 1992-1996." Economic Development and Cultural Change forthcoming in April
1999.
Maasen Van Den Brink, H., Groot, W., (1997) "A household model of paid labor, household work
and child care" De Economist, 145, No. 3.
Matthews, M., (1986). Poverty in Soviet Union. Cambridge: Cambridge University Press.
Michalopoulos, C., Robins, P., Garfinkel, I., (1992) "A structural model of labor supply and child
care demand." Journal of Human Resources 27(1) 166-203
Mincer, J., Ofek, H., (1982) "Interrupted work careers." Journal of Human Resources 17:3-24.
Mincer, J., Polachek, S., 1974. "Family investment in human capital: earnings of women," Journal
of Political Economy (Supplement), 82:S76-S108
Mroz, T., Guilkey, D., (1992) "Discrete factor approximation for use in simultaneous equation
models with both continuous and discrete endogenous variables" Working paper series.
Carolina Population Center. University of North Carolina at Chapel Hill.
, and Popkin, B., (1995) "Poverty and the Economic Transition in the Russian
Federation," Journal of Economic Development and Cultural Change 44: 1-31.
, Popkin, B., Mancini, D., Glinskaya, E., Lokshin, M. (1997) "Monitoring Economic
Condition in the Russian Federation: The Russia Longitudinal Monitoring Survey 1992-
1996." Report submitted to the U.S. Agency for International Development. Carolina
Population Center, University of North Carolina at Chapel Hill, North Carolina. (The reader
can go to the website for the RLMS for detailed data on the survey and instruments and
ways to obtain at no cost the data set. The world wide web address for the Russia
Longitudinal Monitoring Survey is: http.//www.cpc. unc. edu/Droiects/r/ms/r/ms home. html
, (1996) "Applications of discrete factor models in labor economics" Preliminary draft.
Department of Economics and Carolina Population Center, University of North Carolina at
22
Chapel Hill.
, (1999) "Discrete Factor Approximations in Simultaneous Equation Models:Estimating
the Impact of a Dummy Endogenous Variable on aContinuous Outcome." forthcoming in
Journal of Econometrics
Popkin B., Marina Mozhina, and Alexander K. Baturin, "The Development of a Subsistence Income
Level in the Russian Federation" (Chapel Hill: Carolina Population Center, The University
of North Carolina, 1992).
Presser, H., Baldwin, W., (1980) "Child care as a constraint on employment: prevalence,
correlates, and bearing on the work and fertility nexus." American Journal of Sociology
85:1202-1213
Prosser, W., (1986) "Day care centers 1976-1984: Has supply kept up with demand?" Technical
analysis paper. U.S. Department of Health and Human Services, May.
Ribar, D., (1995) "A structural model of child care and the labor supply of married women." Journal
of Labor Economics 13(3):558-97
. (1992) "Child care and the labor supply of married women: reduced form evidence."
Journal of Human Resources 27(1): 134-65
"Russia in numbers" (1996) State Statistical Agency (GosKomStat): Moscow
Wong, R., Levine, R., (1992) "The effect of household structure on women's economic activity and
fertility: evidence from recent mothers in urban Mexico." Economic Development and
Cultural Change, Vol. 41 No. 4: 89-102.
23
Appendix
In the SPFIML estimation the following functional forms were assumed in estimating the
probability weights ir and P, and the points of support V1 and V2:
Zr." exp(bmn), 1 IYm
mn = X ) n =l..., N-1 N4
1+ Yexp(bmn ) 1 + Eexp(bmn)
Vm 1 exp(amn) n = 2,..., N-I VmI = 0; VmN =1
mn + exp(amn )
Table A 1.1 Points of support, probability weights and factor loading for the SPFIML estimation with
3 points of support for both transitory and permanent factors
Permanent factor Transitory factor
Point of support
V(1) 0.0000 0.0000
V(2) 0.0453 0.3576
V(3) 1.0000 1.0000
Probabilities
P(1) 0.3244 0.4073
P(2) 0.6743 0.5772
P(3) 0.0012 0.0153
Factor loadings
Discrete outcome equation
p(l) 0.0000 0.0000
p(2) 90.6366 -10.2445
p(3) 107.6548 -77.1019
p(4) 169.2239 15.0137
p(5) 171.8540 10.0019
continuous outcome equation
Hours of work p 17.5526 9.6144
24
Table A2: Selectivity corrected wage regression estimation for different age/gender groups
Women < 25 Men < 25 Women 25-54 Men 25-59 Women >55 Men > 60
Coeff. Std. Err Coeff. Std. Err Coeff. Std. Err Coeff. Std. Err Coeff. Std. Err Coeff. Std. Err
Age 8.779 11.928 10.808 10.044 0.147 0.115 -0.168 0.123 1.658 2.924 2.726 2.399
Age squared -0.400 0.563 -0.510 0.474 -0.004 0.003 0.005 0.003 -0.029 0.045 -0.040 0.037
Age cubed 0.006 0.009 0.008 0.007 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Education
Hligh school Reference
Technical/Vocational 0.467 1.160 -1.084 1.061 -0.497 0.190 -0.189 0.198 2.209 1.559 -3.381 1.507
University 1.540 3.169 2.138 3.175 0.104 0.212 0.057 0.212 -0.220 1.457 -3.328 1.343
Location
Rural areas of Russia -0.736 0.576 -0.586 0.456 -0.112 0.177 -0.189 0.195 -0.524 0.446 -0.864 0.466
Urban areas of Russia -0.338 0.561 -0.133 0.461 0.384 0.179 0.414 0.195 -0.505 0.444 -0.581 0.467
Metropolitan areas Reference
Interactions
Age/vocational -0.033 0.053 0.050 0.047 0.013 0.004 0.005 0.004 -0.039 0.024 0.052 0.024
Age/university -0.059 0.139 -0.071 0.137 0.006 0.005 0.002 0.005 0.007 0.024 0.059 0.022
Rurual/vocational 0.493 0.349 -0.143 0.295 0.223 0.122 0.095 0.137 0.365 0.333 0.933 0.388
Rural/university 0.343 0.595 -1.312 0.974 0.366 0.141 0.323 0.152 0.358 0.287 -0.363 0.474
Other urban/vocational 0.437 0.279 -0.012 0.251 0.082 0.108 0.048 0.123 0.588 0.271 0.689 0.348
Other urban/university 0.085 0.368 -0.618 0.403 0.043 0.111 0.156 0.112 0.477 0.215 0.268 0.299
Time dummies
Round 5 Reference
Round 6 -0.374 0.108 -0.191 0.098 -0.184 0.033 -0.230 0.037 -0.165 0.085 -0.402 0.093
Round 7 -0.236 0.112 -0.091 0.104 -0.111 0.034 -0.200 0.038 -0.097 0.089 -0.255 0.096
PSU characteristics
Average wage 94/10000 0.109 0.028 0.019 0.017 0.022 0.006 0.021 0.007 -0.024 0.026 0.047 0.017
Average wage 95/10000 -0.069 0.030 0.020 0.018 0.016 0.007 0.014 0.008 0.054 0.026 -0.016 0.019
Average wage 96/10000 -0.363 0.489 0.794 0.436 -0.293 0.146 -0.612 0.165 -0.133 0.378 -0.247 0.411
entni94 0.007 0.899 -0.933 0.745 0.600 0.259 1.094 0.292 0.393 0.695 -0.476 0.655
25
Table A2: (Continues)
Women < 25 Men < 25 Women 25-54 Men 25-59 Women >55 Men > 60
Coeff. Std. Err Coeff. Std. Err Coeff. Std. Err Coeff. Std. Err Coeff. Std. Err Coeff. Std. Err
Unemployment rate 93 3.896 4.805 8.600 4.171 8.296 1.386 8.175 1.592 -1.077 3.800 10.508 3.865
Unemployment rate 94 -4.239 6.955 -3.457 5.802 -0.914 1.912 -1.744 2.131 -9.342 5.406 4.316 6.350
Unemployment rate 95 0.028 3.980 -0.805 3.589 -3.332 1.121 -4.129 1.232 6.254 3.417 -9.507 3.537
percent of loss making -0.520 0.800 -0.640 0.660 0.000 0.210 0.290 0.240 -0.700 0.560 -0.790 0.630
enterprises
percent of profit to previous year -0.059 0.066 0.007 0.058 0.008 0.020 0.059 0.022 0.055 0.052 0.076 0.055
Long term credit per worker 1.825 1.880 1.580 1.639 -0.505 0.538 -0.189 0.637 -1.203 1.662 -0.465 1.645
percent to previous year
Short term credit per worker 0.070 0.126 -0.082 0.102 -0.034 0.033 -0.066 0.038 0.199 0.113 -0.265 0.129
percent to previous year
Constant -62.097 83.817 -73.735 70.626 -0.523 1.496 3.544 1.580 -28.597 62.740 -56.721 52.225
26
Table A3: Discrete outcome SPFIML model estimation with heterogeneity, specification with the
minimum wage for the other household members.
Mode of child care and household labor supply arrangement.
Case when the mother does not work is used as a reference.
Other work Other do not work Other work Other do not work
Informal Informal Formal Informal & formal
child care child care child care child care
Coef. Std. Err. Coef. Std. Err. Coef. Std. Err. Coef. Std. Err.
Presence of children
younger than 18 months -2.52 0.48 -3.23 0.68 -5.38 1.00 -5.80 1.00
1.5-3 years old -1.30 0.42 -1.90 0.58 -3.98 0.82 -4.18 0.97
3-7 years old 0.33 0.43 0.28 0.61 -0.32 0.86 0.27 1.04
7-12 years old 0.99 0.37 0.79 0.52 0.77 0.48 0.14 0.59
12-18 years old 1.34 0.46 0.46 0.61 3.20 0.57 2.69 0.66
Family size 6.54 1.37 -2.60 1.94 3.86 1.73 -5.32 1.97
Number of children in the household -8.40 2.34 3.49 3.55 -16.23 2.56 -0.95 2.93
Number of pensioners -3.96 1.99 1.86 2.81 -14.36 2.39 -14.35 2.84
Single mother household -0.64 0.48 1.15 0.61 0.98 0.53 2.56 0.59
Households with two parents Reference
Household incomes
Household non-wage income -0.58 1.38 1.45 1.61 -0.57 1.53 0.15 1.74
Log wage rate of others -2.51 1.77 -8.10 2.00 0.72 2.49 -11.03 2.45
Log of average cost of child care -2.32 1.05 -2.49 1.39 -2.53 1.20 -4.04 1.54
Household regional dummies
Other urban areas of Russia -0.25 0.34 -0.10 0.45 -2.49 0.46 -2.16 0.52
Moscow and St. Petersburg -1.27 0.62 -1.41 0.84 -0.69 0.78 0.76 0.98
North and North-West -1.32 0.58 -2.13 0.76 3.05 0.62 4.21 0.76
Central and Central Chernozem 0.41 0.49 -0.22 0.68 2.74 0.58 3.39 0.73
Volgo-Vyatskiy -0.59 0.51 -1.55 0.69 2.56 0.55 2.81 0.73
North Caucasis 0.01 0.46 -0.72 0.68 1.41 0.60 1.70 0.78
Ural -0.18 0.50 -1.08 0.66 3.35 0.56 4.16 0.71
Western Siberia -0.18 0.51 -1.05 0.72 1.48 0.65 2.47 0.81
Eastern Siberia Reference
Time dummies
Round 5 0.09 0.30 -0.03 0.41 -0.76 0.37 -1.20 0.49
Round 6 0.27 0.31 0.12 0.43 -0.77 0.36 -0.85 0.46
Round 7 Reference
Characteristics of the mother
Mother's agein years 1.25 1.17 0.65 1.42 -1.79 1.65 -3.18 1.99
High school completed -0.51 0.39 -0.40 0.53 -1.35 0.49 -1.02 0.59
Technical/vocational school 0.23 0.36 -0.28 0.49 -0.02 0.46 0.08 0.55
University Reference
Log of mother's wage 12.75 3.15 10.80 4.35 19.69 3.55 21.72 3.03
Constant -4.83 1.17 -1.56 1.74 -31.08 2.14 -28.01 2.17
Table A4: Continuous outcome SPFIML model estimations with a control for heterogeneity
27
Hours of work Hours in formal care
Coef. Std. Err. Coef. Std. Err.
Presence of children
younger than 18 months -7.29 1.30 -13.19 2.09
1.5-3 years old -14.29 1.13 1.36 1.85
3-7yearsold -8.50 1.14 2.99 1.91
7-12 years old -4.43 1.08 -12.79 1.28
12-18 years old -6.19 1.02 -11.76 1.37
Family size/lO 43.18 1.39 -2.14 2.84
Number of children in the household/lo 8.92 1.11 120.11 5.84
Number of pensioners /10 9.11 1.15 0.01 2.49
Single mother household 19.86 1.05 0.17 1.36
Households with two parents Reference
Household incomes
Household non-wage income/10000 -40.70 1.41 -3.54 1.74
Log wage rate of others /10 7.52 2.43 4.70 3.94
Log of average cost of child care/1O -18.85 1.04 -3.94 1.43
Household regional dummnies
Other urban areas of Russia 0.16 1.08 1.04 1.01
Moscow and St. Petersburg -22.29 1.14 1.03 1.55
North and North-West -22.27 1.05 2.41 1.35
Central and Central Chernozem -15.92 1.09 -0.45 1.23
Volgo-Vyatskiy -2.46 1.02 -4.21 1.32
North Caucasis -7.45 1.21 -2.62 1.47
Ural -20.53 1.13 -0.25 1.24
Western Siberia -20.85 1.02 -0.88 1.41
Eastern Siberia Reference
Time dummies
RoundS -11.80 1.04 -0.12 1.06
Round 6 3.20 1.03 -1.64 1.06
Round 7 Reference
Characteristics of the mother
Mother's age in years /100 5.63 1.07 1.83 2.59
High school completed 14.52 1.03 -1.04 1.11
Technical/vocational school 18.53 1.03 -2.50 1.03
University Reference
Log of mother's wage/1O 58.79 2.34 -5.21 5.04
Constant 121.59 1.05 -3.38 3.12
28
Table 1: Solutions of the household utility maximization problem
Mother works Other household Informal child Formal child care
members work care
-Lm>>1 Ho>0 1-L,-To>O
TO>O
0 No
1 Yes Yes Yes No
2 Yes No Yes No
3 Yes Yes No Yes
4 Yes No Yes Yes
Table 2: Distribution of households with children 0-7 years old by the choice of child care arrangements
RLMS Rounds l-VII, 1992-96.
Rounds of survey
Child care mode V, 12/94 VI, 10/95 VII, 10/96
Mother does not work 45.0 46.2 47.3
Others do not work, informal care only 22.7 23.2 18.5
Others work, informal care only 8.0 8.6 8.8
Others work, formal care only 19.7 15.8 18.8
Others don't work, formal and informal care 4.5 6.2 6.6
Total number of households 796 695 670
29
Table 3: Summary statistics for the explanatory variables, the means and standard deviations. Pooled
sample, rounds V-VII, RLMS.
E,xplanatory variables Mean Standard Deviation
Log of mother's imputed hourly wage 2.43 0.42
Log of the other household members imputed hourly wage (average) 2.43 0.83
Log of the other household members imputed hourly wage (minimum) 2.34 0.85
Log of the average cost of child care 1.80 1.08
Mother's work hours per week 36.67 13.87
Hours per week children spend in formal child care facilities 41.09 22.28
Non-wage household income 3300.88 855.1
Presence of children 0-18 months 0.19
Presence of children 18 month - 3 years old 0.22
Presence of children 3 - 7 years old 0.67
Presence of children 7-12 years old 0.26
Presence of children 12-18 years old 0.17
Mother's age in months (years) 367.6 (30.6) 94.7 (7.9)
Total number of children 18 years old and younger 1.68 0.89
Number of pensioners in the household 0.17 0.42
Mother's years of education 12.3 3.6
Single parent family living alone indicator 0.05
Single parent living with grandparents indicator 0.06
Other households with single parent 0.04
Nuclear family living alone 0.54
Nuclear family living with grandparents 0.16
Other households with nuclear family 0.06
Other households with children 0.09
Metropolitan areas indicator 0.09
Other urban areas indicator 0.66
Rural areas indicator 0.25
Sample size 2169
30
Table 4a: Discrete outcome SPFIML model estimation with heterogeneity
Mode of child care and household labor supply arrangement.
Case when the mother does not work is used as a reference.
Other work Other do not work Other work Other do not work
Informal Informal Formal Informal & fonmal
child care child care child care child care
Coef. Std. Err. Coef. Std. Err. Coef. Std. Err. Coef. Std. Err.
Presence of children
younger than 18 months -3.12 0.49 -3.64 0.70 -5.38 1.00 -5.80 1.00
1.5-3 years old -1.45 0.45 -1.89 0.62 -4.13 0.91 -4.73 1.22
3-7 years old 0.55 0.45 0.73 0.64 -0.80 0.93 -0.90 1.30
7.12 years old 0.97 0.40 0.81 0.54 0.32 0.54 -0.16 0.67
12.18 years old 1.27 0.47 0.43 0.64 2.43 0.62 1.95 0.76
Family size 7.45 1.44 -0.40 1.90 2.31 2.04 -3.57 2.34
Number of children in the household -8.87 2.41 1.43 3.29 -10.98 3.77 -0.05 4.50
Number of pensioners -3.84 2.23 1.70 3.07 -12.01 2.88 -10.54 3.84
Single mother household -0.33 0.50 1.36 0.62 0.64 0.53 1.92 0.62
Households with two parents Reference
Household incomes
Household non-wage income -0.15 1.94 2.50 2.43 -0.96 2.10 -0.06 2.56
Log wage rate of others 0.25 2.21 -8.09 2.66 726 3.26 -6.36 3.09
Log of average cost of child care -3.19 1.16 -3.39 1.52 -2.93 1.36 -4.45 1.73
Household regional dummies
Other urban areas of Russia -0.36 0.36 -0.04 0.47 -2.23 0.50 -1.70 0.56
Moscow and St. Petersburg -1.54 0.67 -1.52 0.88 -1.15 0.87 0.24 1.19
North and North-West -1.29 0.63 -2.12 0.80 2.54 0.71 3.38 0.94
Central and Central Chernozem 0.65 0.53 -0.10 0.70 2.83 0.73 3.17 0.90
Volgo-Vyatskiy -0.44 0.56 -1.51 0.72 2.34 0.70 2.29 0.91
North Caucasis 0.40 0.52 -0.30 0.71 1.78 0.79 1.80 1.00
Ural -0.12 0.57 -1.13 0.73 3.05 0.72 3.49 0.90
Western Siberia -0.14 0.55 -1.10 0.73 1.30 0.75 2.11 0.96
Eastern Siberia Reference
Time dummies
Round 5 0.03 0.29 -0.06 0.39 -0.81 0.37 -1.19 0.50
Round 6 0.35 0.30 0.21 0.39 -0.63 0.37 -0.66 0.46
Round 7 Reference
Characteristics of the nother
Mother's age in years 1.60 1.35 1.05 1.86 -0.70 1.83 -1.49 2.10
High school completed -0.46 0.42 -0.28 0.55 -1.31 0.57 -1.08 0.69
Technical/vocational school 0.26 0.38 -0.19 0.50 -0.17 0.51 -0.17 0.63
University Reference
Log of mother's wage 14.18 4.37 13.24 6.11 17.98 6.15 19.67 6.02
Constant -6.82 1.44 -3.65 1.82 -14.86 2.40 -11.05 2.41
Table 4b: Discrete outcome SPFIML model estimation without heterogeneity
31
Mode of child care and household labor supply arrangement.
Case when the mother does not work is used as a reference.
Other work Other do not work Other work Other do not work
Infonnal Informal Fornal Informal & formal
child care child care child care child care
Coef. Std. Err. Coef. Std. Err. Coe. Std. Err. Coef. Std. Err.
Presence of children
younger than 18 months -1.74 0.27 -1.73 0.38 -5.38 1.40 -5.80 2.31
1.5-3 years old -0.87 0.25 -1.10 0.34 -2.01 0.49 -2.02 0.72
3-7 years old 0.06 0.25 0.45 0.35 -0.09 0.54 -0.05 0.81
7-12 years old 0.68 0.20 0.48 0.26 0.18 0.25 -0.08 0.34
12-18 years old 0.97 0.24 0.56 0.33 1.14 0.28 1.05 0.38
Family size 4.83 0.69 -2.53 0.95 2.27 0.92 -2.81 1.15
Number of children in the household -6.57 1.38 2.95 1.81 -6.09 1.76 1.96 1.32
Number of pensioners -2.05 1.12 2.60 1.32 -5.05 1.38 -3.08 1.41
Single mother household -0.27 0.26 1.37 0.25 0.28 0.31 1.54 0.35
Households with two parents Reference
Household incomes
Household non-wage income -0.15 0.96 1.66 1.09 -0.39 1.16 0.35 1.06
Log wage rate of others 2.28 1.23 -3.29 1.16 7.46 1.92 -3.37 1.49
Log of average cost of child care -1.27 0.64 -1.07 0.87 -1.31 0.77 -3.08 1.19
Household regional dummies
Other urban areas of Russia -0.18 0.17 0.02 0.23 -0.75 0.21 -0.29 0.34
Moscow and St. Petersburg -1.03 0.35 -0.61 0.45 -1.00 0.46 0.16 0.69
North and North-West -0.55 0.32 -0.85 0.43 0.70 0.38 1.10 0.55
Central and Central Chernozem 0.19 0.27 -0.13 0.35 0.90 0.36 0.67 0.56
Volgo-Vyatskiy -0.32 0.28 -0.72 0.35 0.78 0.37 0.02 0.57
North Caucasis -0.02 0.27 -0.29 0.35 0.34 0.39 0.00 0.60
Ural 0.03 0.29 -0.17 0.36 1.34 0.36 1.26 0.54
Western Siberia -0.32 0.28 -0.68 0.37 0.16 0.36 0.49 0.55
Eastern Siberia Reference
Time dummies
Round 5 0.14 0.16 0.03 0.21 -0.14 0.19 -0.35 0.32
Round 6 0.23 0.16 0.02 0.21 -0.02 0.19 0.04 0.30
Round 7 Reference
Characteristics of the mother
Mother's age in years 0.64 0.98 -1.80 0.98 -0.50 0.98 -1.32 0.98
High school completed -0.48 0.77 -0.59 0.82 -0.63 0.85 -0.77 0.97
Technical/vocational school 0.12 0.65 -0.48 0.84 -0.02 0.81 -0.28 0.94
University
Log of mother's wage 3.82 0.99 0.49 0.99 3.68 0.99 0.40 0.99
Constant -2.88 0.83 0.73 0.84 -2.46 0.84 0.72 0.86
Table 5: Continuous outcome SPFIML model estimations with and without heterogeneity
32
Hours of work Hours in formal care
No heterogeneity He snet No heterogeneity ,Hetomencty
Coef. Std. Err. C-eL Std.Err, Coef Std. Err. Coe- Std. Err.
Presence of children
younger than 18 months -9.02 10.51 -6:57 4.60 -25.02 3.71 -1390 2.44
1.5-3 years old -10.99 8.34 -14.03 4.06 0.28 3.29 1.93 2.31
3-7 years old -4.66 8.10 :4.15 4,27 6.83 3.59 0,61 2.44
7-12 years old -4.03 6.12 -40 3.04 -18.36 2.82 4 1 -3. 1.59
12-18 years old -4.59 7.08 -7.31 2.21 -12.63 3.23 -1 1.71 -170
Family size/10 42.9 2.51 40.86 3.2 -1.72 1.16 -3.59 4.72
Number of children in the household/lo 15.8 4.68 14.58 4.16 12.33 2.08 13.13 1.67
Number of pensioners 110 12.5 4.04 10.99 2.)) 1.32 1.94 -042, 5.69
Single mother household 20.72 6.75 22.27 2.53 0.86 3.22 -0,54 1.54--
Households with two parents Reference
Household incomes
Household non-wage income/10000 -39.39 31. 05 4- . 497 -1.96 1.29 -4.1O 6227
Log wage rate of others /10 25.92 31.6 25.12 3.,0 8.87 14.61 233 5.45,
Log of average cost of child carellO -22.71 177 -1956 1.44 -14.7 8.34 -3.53 3.95
Household regional dummies
Other urban areas of Russia 0.17 4.68 0.85 2,48 0.37 2.17 0.29 1.19
Moscow and St. Petersburg -20.80 9.86 -22.44 1. 70 1.43 4.48 -0.43 2,51
North and North-West -16.63 8.48 -i2.13 L,91 3.49 3.76 27: 1.97
Central and Central Chernozem -12.00 7.80 -16.08 332- 4.11 3.76 -028 2.3
Volgo-Vyatskiy 1.73 8.44 .2,19 2.35 4.14 3.99 -3.71, 2,07
North Caucasis 0.42 8.50 -7.20 2.68 1.73 4.10 -.93 2.23
Ural -15.46 7.96 -19.72 2.06 8.22 3.73 0.11 1.94
Western Siberia -12.38 7.94 -20.24 2.38 4.28 3.75 -1.62 2.01
Eastern Siberia Reference
Time dummies
RoundS -11.78 4.43 -11,66 -2.14 -0.03 2.02 4.08 1.18
Round 6 3.96 4.51 3.24 2,5 1.25 2.03 -2.03 - 1.24
Round 7 Reference
Characteristics of the mother
Mother's age in years /100 6.67 2.76 10.31 2.31 0.72 1.21 1.99 451
High school completed 10.24 6.22 13.93 1.0 3.08 2.84 -04.91 1.42
Technical/vocational school 14.34 5.36 -417.91 ,4 1.86 2.50 -L1.7 1.30
University Reference
Log of mother's wage/1i 46.6 80.2 33.18 3.20 55.2 34.8 1.89 8.97
Constant 116.55 26.23 121.55 4,32 -0.03 2.02 -2.22 4.00
33
Table 6: Simulation of the effects of various policies on the household choices of child care mode
and mother's labor supply
Mother does Mother works
not work
other work, other do not other work, other do not
informal care work, formal care work, formal
informal care and informal
care
Change in price of child -5.7 (11.9) 2.5 (12.8) 1.0 (10.1) 0.8 (4.9) 1.4 (29.2)
care from 5.6 rubles per
hour to O
Change in offered mother's -9.8 (24.0) 4.1 (17.3) 0.7 (7.5) 3.5 (18.4) 2.5 (21.2)
wage from 10 to 20 rubles
per hour
Change in offered other -0.6 (1.3) 1.8 (8.2) -2.8 (39.1) 4.2 (21.6) -2.5 (77.0)
household members' wage
from 10 to 20 rubles per
hour
Table 7: Effect of the demographic variables on the household choice of child care mode and
mother's labor supply
Mother does Mother works
not work
other work, other do not other work, other do not
informal care work, formal care work, formal
informal care and informal
care
Single mother relative to the -4.8 (10.8) -9.8 (83.0) 9.2 (53.1) -0.01 (5.9) 6.3 (59.0)
married mother
Moscow and St. Petersburg 10.8 (18.9) -7.1 (51.0) -2.0 (27.3) -5.8 (50.8) 3.9 (42.7)
relative to other regions of
Russia
Mother with a high school 13.2 (6.9) -1.5 (7.5) 0.3 (3.7) -5.2 (38.9) -0.6 (14.1)
education relative to the
other mothers
Presence of children 0-18 40.2(50.4) -11.6 (103.1) -6.1 (114.5) -16.7 (449.1) -5.7 (560.7)
months of age
Presence of children 1.5-3 25.2 (37.7) -2.3 (12.4) -3.5 (52.6) -14.3 (239.1) -5.1 (314.5)
years old
Presence of children 3-7 -0.5 (0.97) -3.4 (15.6) 1.5 (16.4) -3.07 (19.6) -1.5 (28.4)
years old
Presence of children 7-12 -5.9 (13.7) 6.1 (24.2) 0.3 (3.2) 0.8 (4.8) -1.23 (27.1)
years old
34
Table 8: Simulation of the three possible scenarios of the government supportfor the households
with children.
Percentage change Cost for Total Change Who is
the increase in total eligible,
govern in hhold # of
ment, income income, eligible
1992 from the per bholds
Policy instrument mother's mother's formal formal rubles change eligible
labor hours of care use care per in LFP hhold
supply work hours month and
hours of
work
Child care subsidy 11.9 2.6 10.1 2.2 396,702 198,555 1,261.2 Formal
5.6-0.0 rubles/hour child care
users
471.9
Wage subsidy 5.7 1.2 6.2 1.0 396,702 96,332 472.2 Working
10-12.06 mothers
rubles/hour 1044.1
Family allowances -0.2 -0.01 -0.01 -0.01 396,702 (0) 198.5 All
transfer households
1999
35
Policy Research Working Paper Series
Contact
Title Author Date for paper
WPS2187 Who Determines Mexican Trade Jean-Marie Grether September 1999 L. Tabada
Policy? Jaime de Melo 36896
WPS2188 Financial Liberalization and the Pedro Alba September 1999 R. Vo
Capital Account: Thailand, 1988-97 Leonardo Hernandez 33722
Daniela Klingebiel
WPS2189 Alternative Frameworks for Stijn Claessens September 1999 R. Vo
Providing Financial Services Daniela Klingebiel 33722
WPS2190 The Credit Channel at Work: Lessons Giovanni Ferri September 1999 K. Labrie
from the Republic of Korea's Financial Tae Soo Kang 31001
Crisis
WPS2191 Can No Antitrust Policy Be Better Aaditya Mattoo September 1999 L. Tabada
Than Some Antitrust Policy? 36896
WPS2192 Districts, Spillovers, and Government Reza Baqir September 1999 S. Devadas
Overspending 87891
WPS2193 Children's Growth and Poverty in Michele Gragnolati September 1999 M. Gragnolati
Rural Guatemala 85287
WPS2194 Does Democracy Facilitate the Jean-Jacques Dethier October 1999 H. Ghanem
Economic Transition? An Empirical Hafez Ghanem 85557
Study of Central and Eastern Europe Edda Zoli
And the Former Soviet Union
WPS2195 Aggregating Governance Indicators Daniel Kaufmann October 1999 D. Bouvet
Aart Kraay 35818
Pablo Zoido-Lobat6n
WPS2196 Governance Matters Daniel Kaufmann October 1999 D. Bouvet
Aart Kraay 35818
Pablo Zoido-Lobat6n
WPS2197 Production Sharing in East Asia: Francis Ng October 1999 L. Tabada
Who Does What for Whom, and Why? Alexander Yeats 36896
WPS2198 How the Chinese System of Charges Hua Wang October 1999 H. Wang
and Subsidies Affects Pollution Ming Chen 33255
Control Efforts by China's Top
Industrial Polluters
WPS2199 Managing Risks of Capital Mobility Mansoor Dailami October 1999 W. Nedrow
31585
WPS2200 The Role of Trust in Financial Sector Biagio Bossone October 1999 E. Mekhova
Development 85984
Policy Research Working Paper Series
Contact
Title Author Date for paper
WPS2201 Financial Development and Industrial Biagio Bossone October 1999 E, Mekhova
Capital Accumulation 85984
WPS2202 Specialization without Regret: Michael R. Carter October 1999 M. Fernandez
Transfer Rights, Agricultural Yang Yao 33766
Productivity, and Investment in an
Industrializing Economy
WPS2203 Market versus Administrative Michael R. Carter October 1999 M. Fernandez
Reallocation of Agricultural Land Yang Yao 33766
in a Period of Rapid Industrialization
WPS2204 Corruption under Moral Hazard Gunnar S. Eskeland October 1999 H. Sladovich
Henrik Thiele 37698
WPS2205 Foreign-Owned Capital and Marcelo Olarreaga October 1999 L. Tabada
Endogenous Tariffs 36896
(12|!