Child Care: Child Care Subsidies Increase Likelihood That Low-Income
Mothers Will Work (Letter Report, 12/30/94, GAO/HEHS-95-20).

Since 1988, Congress has created four child care programs for low-income
families.  Two of them subsidize child care for welfare recipients who
are trying to become self-sufficient through education, training, and
employment.  Two others provide child care subsidies to working poor
nonwelfare families.  GAO found that reducing child care costs increases
the likelihood that poor, near-poor, and nonpoor mothers will work.
This effect is strongest for the poor and near-poor mothers.  More
specifically, GAO predicts that providing a full subsidy to mothers who
pay for child care could boost the percentage of poor mothers who work
from 29 to 44 percent and that of near-poor mothers who work from 43 to
57 percent.  By comparison, the probability of nonpoor mothers working
could increase from 55 to 65 percent.  GAO concludes that among the
factors that encourage low-income mothers to seek and keep jobs--factors
such as more education, training, and transportation--affordable child
care is a decisive one.  Thus, any effort to move more low-income
mothers from welfare to work will need to take into account the
importance of child care subsidies to the likelihood of success.

--------------------------- Indexing Terms -----------------------------

 REPORTNUM:  HEHS-95-20
     TITLE:  Child Care: Child Care Subsidies Increase Likelihood That 
             Low-Income Mothers Will Work
      DATE:  12/30/94
   SUBJECT:  Disadvantaged persons
             Child care programs
             Income maintenance programs
             Welfare recipients
             Women
             Surveys
             Aid to families with dependent children
             Employment or training programs
             Subsidies
IDENTIFIER:  AFDC
             AFDC Transitional Child Care Program
             AFDC/JOBS Child Care Program
             Job Opportunities and Basic Skills Training Program
             JOBS Program
             Child Care and Development Block Grant
             HHS At-Risk Child Care Program
             Urban Institute National Child Care Survey of 1990
             
**************************************************************************
* This file contains an ASCII representation of the text of a GAO        *
* report.  Delineations within the text indicating chapter titles,       *
* headings, and bullets are preserved.  Major divisions and subdivisions *
* of the text, such as Chapters, Sections, and Appendixes, are           *
* identified by double and single lines.  The numbers on the right end   *
* of these lines indicate the position of each of the subsections in the *
* document outline.  These numbers do NOT correspond with the page       *
* numbers of the printed product.                                        *
*                                                                        *
* No attempt has been made to display graphic images, although figure    *
* captions are reproduced. Tables are included, but may not resemble     *
* those in the printed version.                                          *
*                                                                        *
* A printed copy of this report may be obtained from the GAO Document    *
* Distribution Facility by calling (202) 512-6000, by faxing your        *
* request to (301) 258-4066, or by writing to P.O. Box 6015,             *
* Gaithersburg, MD 20884-6015. We are unable to accept electronic orders *
* for printed documents at this time.                                    *
**************************************************************************


Cover
================================================================ COVER


Report to the Congressional Caucus for Women's Issues, House of
Representatives

December 1994

CHILD CARE - CHILD CARE SUBSIDIES
INCREASE LIKELIHOOD THAT
LOW-INCOME MOTHERS WILL WORK

GAO/HEHS-95-20

Child Care Subsidies


Abbreviations
=============================================================== ABBREV

  AFDC - Aid to Families with Dependent Children program
  CCDBG - Child Care and Development Block Grant
  FSA - Family Support Act of 1988
  IMR - Inverse Mills Ratio
  JOBS - Job Opportunities and Basic Skills Training program
  SIPP - Survey of Income and Program Participation
  TCC - Transitional Child Care program

Letter
=============================================================== LETTER


B-256750

December 30, 1994

The Honorable Patricia Schroeder
Co-Chair, Congressional Caucus for
 Women's Issues

The Honorable Olympia J.  Snowe
Co-Chair, Congressional Caucus for
 Women's Issues
House of Representatives

For many years, the Congress has been interested in encouraging
low-income mothers to seek employment as an alternative to receiving
welfare.  These mothers face many challenges in obtaining work and
establishing stable and successful work records.  They often must
acquire additional education and job skills, find a job that is
consistent with their skills, and have access to reliable
transportation.  Another major concern for mothers entering or
staying in the labor force is the availability of affordable child
care. 

Because low-income mothers often must pay for all or part of their
child care expenses, the cost of child care remains an employment
barrier to many of them.  Recognizing this need, the Congress has
created four child care programs for low-income families since 1988. 
Two of them subsidize the child care costs of welfare recipients who
are attempting to become self-sufficient through education, training,
and employment.  Two others provide child care subsidies\1 to working
poor nonwelfare families.\2 In addition to paid care, many low-income
mothers use informal care--care provided free of charge by relatives
or friends. 

To better understand the role that child care costs play in the
likelihood that low-income mothers will work, you asked us to examine
this relationship.  Specifically, we determined the probability of
poor and near-poor mothers working as their child care expenditures
change, as compared with nonpoor mothers.\3

To estimate the impact of child care expenditures on mothers'
decision to work, and to compare the impact among poor, near-poor,
and nonpoor mothers, we developed an empirical model using available
data from the Urban Institute's 1990 National Child Care Survey. 


--------------------
\1 A subsidy can be paid in the form of a voucher for parents to
purchase care or a contract with child care providers for a number of
slots. 

\2 The Aid to Families with Dependent Children (AFDC) Child Care and
Transitional Child Care (TCC) programs were included in the Family
Support Act of 1988.  The At-Risk Child Care Program and the Child
Care and Development Block Grant (CCDBG) were authorized by the
Omnibus Budget Reconciliation Act of 1990. 

\3 Poor is defined as at or below the 1989 federal poverty level,
near-poor is defined as above 100 percent and up to 185 percent of
the federal poverty level, the nonpoor is defined as over 185 percent
of the federal poverty level.  The federal poverty level for a family
of three, consisting of one adult and two children, in 1989 was
$9,890. 


   RESULTS IN BRIEF
------------------------------------------------------------ Letter :1

Our analysis predicts that reducing child care costs increases the
likelihood that poor, near-poor, and nonpoor mothers will work.  This
effect is strongest for the poor and near-poor mothers.  More
specifically, our model predicts that providing a full subsidy to
mothers who pay for child care could increase the proportion of poor
mothers who work from 29 to 44 percent, and that of near-poor mothers
who work from 43 to 57 percent.  By comparison, the probability of
nonpoor mothers working could increase from 55 to 65 percent. 

The results of our analysis suggest that among the factors that
encourage low-income mothers to seek and keep jobs--factors such as
more education, training, and transportation--affordable child care
is a decisive one.  Thus, any effort to move more low-income mothers
from welfare to work will need to take into account the importance of
child care subsidies to the likelihood of success. 


   BACKGROUND
------------------------------------------------------------ Letter :2

Child care costs are a significant portion of most low-income working
families' budgets.  They consumed on average 27 percent of monthly
income for families with incomes below poverty who paid for child
care in 1991, compared with an average of 7 percent for families with
incomes above poverty.\4 According to the Bureau of the Census, the
average weekly child care expenditure for all families who paid for
care was about $63 in 1991.  For families below poverty, that figure
was only slightly lower, at about $60.\5

The 1988 Family Support Act (FSA) legislation contained provisions
for child care assistance to help welfare recipients obtain
employment, leave welfare, and stay employed.  FSA requires state
agencies to guarantee child care to (1) employed AFDC recipients, (2)
participants in the Job Opportunity and Basic Skills (JOBS) Training
program, (3) other AFDC recipients in state-approved education and
training programs, and (4) AFDC recipients who leave the welfare
rolls as a result of increased earnings from employment. 

The Congress also recognized the importance of child care assistance
to working poor nonwelfare families.  The At-Risk Child Care program
targets subsidies to non-AFDC working families who would be at risk
of becoming eligible for AFDC if child care assistance were not
provided.  The CCDBG's major purpose is to provide subsidies to
low-income working families who need child care to work or to
participate in education and training. 

Not all families pay for child care.  In our sample, 55 percent of
poor families used informal child care arrangements.  By comparison,
the fraction of near-poor and nonpoor families using informal care
was 37 and 21 percent, respectively.  When paid child care is used,
most child care subsidy programs do not cover the full cost.  The TCC
and At-Risk Child Care programs both use a sliding scale to set their
subsidy rates, based on families' ability to pay.  The AFDC child
care and CCDBG programs authorize a maximum subsidy rate for child
care providers or slots in a particular locality.  Parents whose
costs are below the established rate may be fully subsidized, while
parents whose providers charge more than the rate authorized must pay
the difference themselves.\6

Because most mothers do pay for child care while they work, their
decision to work is dependent, at least in part, on how much they
will make after they have paid child care expenses.  Economic theory
would suggest that reduced child care expenditures will lead to an
increase in the probability that a woman will participate in the
labor force.  In general, previous studies have found a significant
positive effect of child care cost reductions on the labor force
participation of mothers.  (See app.  I for a discussion of these
studies.) While some researchers have looked specifically at
low-income mothers, there has been no effort to separately study poor
and near-poor mothers.\7


--------------------
\4 "Who's Minding the Kids?  Child Care Arrangements:  Fall 1991,"
Survey of Income and Program Participation, Current Population
Reports, P70-36, U.S.  Bureau of the Census (Washington, D.C.: 
1991), p.  21. 

\5 "Who's Minding the Kids?  Child Care Arrangements:  Fall 1991," p. 
21. 

\6 States are required to periodically conduct studies of the cost of
child care in each local community in their state.  These studies,
called market rate studies, are done for different types of care and
for different ages of children.  Based on these studies, both
programs authorize maximum subsidy rates up to the 75th percentile of
the price distribution of local child care providers or slots. 

\7 See Mark C.  Berger and Dan A.  Black, "Child Care Subsidies,
Quality of Care, and the Labor Supply of Low-Income, Single Mothers,"
Review of Economics and Statistics, Vol.  74, No.  4 (Nov.  1992),
pp.  635-642; and Paul Fronstin and Douglas Wissoker, "The Effects of
the Availability of Low-Cost Child Care on the Labor Supply of
Low-Income Women" (unpublished paper presented at the Annual Meeting
of the Population Association of America, Miami, Florida, May 5-7,
1994). 


   SCOPE AND METHODOLOGY
------------------------------------------------------------ Letter :3

For this study, we used data from a nationally representative sample
of households with children--the Urban Institute's 1990 National
Child Care Survey.  We also included data from the Survey's
Low-Income Substudy.  Because the survey collected detailed data on
child care use, family income, and employment histories of both
parents, it was possible to estimate the impact of child care
expenditures on mothers' decision to work. 

We developed measures of predicted wages and child care expenditures
that account for the fact that not all mothers were employed, used
child care, or paid for care.  We then separated the sample into
poor, near-poor, and nonpoor mothers, in order to test whether the
effect of child care costs on the decision to work differed across
these three groups.  We developed estimates under two scenarios--one
in which some mothers have access to informal child care and the
other in which mothers lose all access to that care.  The poor and
near-poor samples contain both welfare and nonwelfare families. 

As in all empirical analysis, these estimates are limited by the data
on which they are based.  Uncertainty results as well from the
necessity of predicting wages and child care expenditures for poor
women from a relatively small sample.  Our results, however, are
generally consistent with the work of other researchers.  Appendix I
contains a detailed discussion of this literature. 


   LARGEST IMPACT ON DECISION TO
   WORK COMES FROM SUBSIDIZING
   POOR AND NEAR-POOR MOTHERS
------------------------------------------------------------ Letter :4

Our analysis showed that subsidizing child care costs has the
greatest impact on poor and near-poor mothers' decision to work, as
compared with nonpoor mothers.  When we considered that some mothers
use informal care, our model predicted that a full child care subsidy
would result in a 15-percentage-point increase in the average
probability of poor mothers working.\8 That is, for every 100 poor
mothers, the approximate number who work would rise from 29 to 44. 
For near-poor mothers, our model predicted that a full subsidy of
child care costs would lead to a 14-percentage-point increase--a rise
in the approximate number who work from 43 to 57 of every 100.  For
nonpoor mothers, our model predicted that the same full subsidy would
increase the approximate number who work by 10 percentage points,
from 55 to 65 of every 100. 

Taking into account the fact that some mothers receive partial
subsidies, we also simulated the response of mothers in each income
group to child care subsidy rates of 10 percent, 25 percent, and 50
percent.  Figure 1 illustrates the average probability of working for
each income group for the different subsidy rates. 

   Figure 1:  Average Probability
   of Mothers Working, by Child
   Care Subsidy Level and Income
   Group

   (See figure in printed
   edition.)


--------------------
\8 These results are indicative only of how the labor supply of
mothers would change with a given child care subsidy rate, holding
all other variables constant.  They do not take into account labor
demand changes; short-term lags, gaps, or bottlenecks in the supply
of child care; or other changes in economic conditions. 


   SUPPLY OF INFORMAL CHILD CARE
   AFFECTS MOTHERS' DECISION TO
   WORK
------------------------------------------------------------ Letter :5

Fifty-five percent of poor mothers in our sample used informal care
for their children, compared with 37 percent of near-poor mothers and
21 percent of nonpoor mothers.  However, if more mothers who are
currently not working decided to enter the labor force or participate
in education or training programs, the need for child care would
increase.  At the same time, informal child care might become less
available. 

Various welfare reform proposals would encourage mothers to enter the
labor force.  This would result in an increased demand for child
care, causing the price to rise.  This, in turn, could draw informal
child care providers into the paid child care market, thereby
shrinking the supply of informal care.  Because mothers who are not
currently working may already be less likely to have a source of
informal child care, they could be in the position of having to pay
for child care if they work. 

While it is unlikely that all informal care will ever disappear
completely, a decrease in the supply of informal care has
implications for the effectiveness of child care subsidies.  To
demonstrate this point, we considered the extreme case of mothers
losing all access to informal care. 

When we assumed that mothers have to pay for child care, our model
predicted that the probability of poor mothers working would fall 23
percentage points to only 6 percent.  That is, when mothers face a
market in which there are no informal care options and no child care
subsidy, the model predicted approximately 6 of every 100 poor
mothers would continue to work.  Under the same conditions, our model
predicted that approximately 18 of every 100 near-poor mothers and 44
of every 100 nonpoor mothers would continue to work. 

With a child care subsidy, however, the number of mothers predicted
to work would rise until, with a full subsidy, the number reached 44
percent for poor mothers, 57 percent for near-poor mothers, and 65
percent for nonpoor mothers.  This reflects the fact that with a full
child care subsidy, some mothers are no longer dependent on informal
care to enter the work force. 


   CONCLUSIONS
------------------------------------------------------------ Letter :6

Child care subsidies are predicted to make a substantial difference
in the probability of poor and near-poor mothers working, especially
if informal child care becomes less available.  These findings are
important for understanding how to help both current welfare
recipients enter the labor force and working poor families remain off
the welfare rolls. 


---------------------------------------------------------- Letter :6.1

Our work was based on information we developed using available data
from the 1990 National Child Care Survey.  As a result, we did not
obtain agency comments on this report.  We are sending copies of this
report to the Secretary of Health and Human Services and to other
interested parties.  We will also make copies available to others on
request. 

Major contributors to this report are listed in appendix II.  If you
have any questions concerning this report or need additional
information, please call me on (202) 512-7215. 

Jane L.  Ross
Director, Income Security Issues


ECONOMIC ANALYSIS OF THE EFFECT OF
CHILD CARE SUBSIDIES ON MOTHERS'
DECISION TO WORK
=========================================================== Appendix I

To study how child care expenditures affect the probability of
mothers working, we adopted an economic model of the decision by
women with children to explicitly include child care costs.  From a
theoretical model of mothers' labor force participation, we derived
an empirical model and used data from the Urban Institute's National
Child Care Survey and Low-Income Substudy to test the effect of child
care expenditures on the probability of poor, near-poor, and nonpoor
mothers working.  We tested this empirical model using a two-stage
procedure to correct for possible sample selection bias in the
prediction of mothers' hourly wage and weekly child care
expenditures.  Our results indicated that predicted child care
expenditures have a negative and significant effect on the
probability of mothers working in each income group, and the effect
is largest for the poor and near-poor mothers.  This result is true
when we assumed that some mothers use informal care, and it became
even more substantial when we assumed that no mothers use informal
care. 


   THEORETICAL MODEL
--------------------------------------------------------- Appendix I:1

We adopted our model of the labor force participation decision for a
mother with children from work by Dr.  Rachel Connelly of Bowdoin
College.\9 In the Connelly model, the labor force participation
decision is assumed to be the outcome of a mother's maximization of
her utility over a composite market good, X; child quality, Q; and
leisure, tL; subject to a production function for Q; a money budget
constraint; and two time constraints, one for herself and one for the
children.  The production of child quality is a function of the
amount of time the mother spends with her child, tQ; the amount of
time spent in child care, tCC; and the quality of that care, q.  It
is also dependent on the number of children in the family, N, and the
ages of the children, A. 

The utility function, production function, and budget constraints are
specified in the following set of equations: 











where tm is time in market work, tL is leisure time, W is the market
wage, V is nonlabor income (including the husband's earnings), and
PCC is the hourly price of child care (PCC = P(q, N, A )). 

One of the first order conditions for the maximization problem is
given as



where PCC\* is the price of child care at the optimally chosen level
of quality, q\* .  The third expression is the net benefit of tQ,
which depends on the net benefit to Q of parental child care versus
nonparental child care time and the PCC\* ,which is the money savings
of an hour of tQ. 

Connelly's model predicts that for those women who participate in the
labor market, their market wage will be equal to both the value of
their leisure time and the value of the time they spend caring for
their children.  It is then possible to derive an equation for time
in market work for mothers that is a function of the mother's market
wage, the price of child care, and a set of individual and family
characteristics that affect the marginal rate of substitution between
goods and leisure and the production of child quality. 


--------------------
\9 Rachel Connelly, "The Effect of Child Care Costs on Married
Women's Labor Force Participation," Review of Economics and
Statistics, Vol.  74, No.  1 (Feb.  1992), pp.  83-90. 


   PREDICTING HOURLY WAGES AND
   WEEKLY CHILD CARE EXPENSES
--------------------------------------------------------- Appendix I:2

We estimated the parameters of the time in market work equation
described above using a limited dependent variable model, such as
probit, where I = 1 if tm >0, and 0 otherwise.  However, because some
mothers in the sample were not earning a wage at the time of the
survey, and some were either not using or not paying for child care,
it was necessary first to predict wages and child care expenditures
for these mothers.  We did this by using information about the
mothers in the sample who were earning a wage and those who were
paying for child care to predict wages and expenditures for those who
were not.  These predicted wages and expenditures could be biased,
however, if there were some unobserved differences between the
working and nonworking mothers or the paying and nonpaying mothers. 
This problem is known as sample selection bias.  We corrected for
possible sample selection bias using a standard two-stage Heckman
procedure.\10 We first estimated a reduced form probit equation to
calculate a statistical correction factor, known as the Inverse Mills
Ratio (IMR).  We estimated separately the reduced form probits of the
probability of working and the probability of paying for child care
and constructed an IMR from each set of reduced form coefficients. 
We then included the IMR as an additional variable in our estimates
of wages for those mothers who were working at the time of the
survey, and child care expenditures for those mothers who paid for
child care.  We used the coefficients from these estimates to
calculate an unbiased and consistent predicted wage and predicted
weekly child care expenditure for each mother in the sample. 


--------------------
\10 For a detailed explanation of this procedure see James J. 
Heckman, "Sample Selection Bias as a Specification Error,"
Econometrica, Vol.  47, No.  1 (Jan.  1979), pp.  153-161. 


      FIRST STAGE ESTIMATION OF
      HOURLY WAGE
------------------------------------------------------- Appendix I:2.1

In the first stage of estimation, we estimated a correction term to
correct for the possible bias introduced by the fact that hourly
wages are not observed for nonworking women.  The estimation of the
IMR involved first estimating a reduced form probit of the
probability that a woman was working in the week before the survey. 
That is, we estimated work status as a linear function of all of the
exogenous variables in the model. 

LFPART = fn (RACE,UND6,KID612,EDUC,WRKEXPER,EXPSQR,
MARRIED,SPDVWD,URBAN,SUBURB,NORCENTR,
NOREAST,WEST,KID1318,OTHADLT,UNEMADLT,OTHINCOM)

Table I.1 gives a complete definition of each variable.  The omitted
category for the marital status variable is the never married
category; for the variables measuring urbanicity, the omitted
category is the rural category; and for the regional variables, the
Southern region is the omitted category.  Table I.2 gives descriptive
statistics for mothers in each of the three income groups.  We
estimated the reduced form probit and subsequent wage over the entire
sample of poor, near-poor, and nonpoor women.  This is the same as
assuming that all of the women in the sample faced the same labor
market, regardless of their family income. 


      SECOND STAGE ESTIMATION OF
      HOURLY WAGE
------------------------------------------------------- Appendix I:2.2

We then calculated the IMR term using the estimated coefficient
vector from the probit equation on the probability of working.  We
included the IMR in the Ordinary Least Squares estimate of the
natural log of wages for the sample of women who were currently
working.  This is a standard wage equation, based on human capital
variables and a set of variables controlling for region and
urbanicity of residence. 

HREARN = fn (RACE,EDUC,WRKEXPER,EXPSQR,URBAN,SUBURB,
NORCENTR,NOREAST,WEST,IMR1)

The coefficients from this selection-corrected wage estimation were
then used to calculate predicted natural log wages for all women in
the sample.  For women who were not currently working, this predicted
wage can be thought of as the wage they would be expected to earn if
they were working. 

In the same way, weekly child care expenditures were estimated for
all women in the sample, correcting for the possible bias introduced
by the fact that not all women either used or paid for child care. 
The estimation involved estimating a reduced form probit of the
probability that a woman paid for child care in the week before the
survey.  We did this estimation separately for poor, near-poor, and
nonpoor women because we assumed that the child care market, which is
much more local than the labor market, may be very different for
women with different levels of family income. 


      FIRST STAGE ESTIMATION OF
      WEEKLY CHILD CARE
      EXPENDITURE
------------------------------------------------------- Appendix I:2.3

We assumed that a woman's decision to pay for child care is affected
by the number of children she has and their ages, her own education,
race, and marital status, the presence of other adults in the
household and their employment status, and the set of variables
controlling for region and urbanicity of residence. 

PAYCARE = fn (UND6,KID612,KID1318,EDUC,OTHINCOM,OTHADLT,
UNEMADLT,URBAN,SUBURB,NORCENTR,NOREAST,WEST,RACE,
MARRIED,SPDVWD)


      SECOND STAGE ESTIMATION OF
      WEEKLY CHILD CARE
      EXPENDITURE
------------------------------------------------------- Appendix I:2.4

Using the coefficients from this equation, we constructed an IMR term
and included it as an explanatory variable in an Ordinary Least
Squares estimation of weekly child care expenditures for the families
that paid for nonparental care.  Again, we made these estimates
separately for poor, near-poor, and nonpoor women because
expenditures on child care are expected to differ systematically
among the three income groups.  We assumed that the amount that women
must pay for child care is a function of the number and ages of the
children, the woman's own race and education level, the presence of
others in the household, and the set of variables controlling for
region and urbanicity of residence. 

WKEXPEND = fn (UND6,KID612,KID1318,RACE,EDUC,OTHINCOM,
OTHADLT,URBAN,SUBURB,NORCENTR,NOREAST,WEST,IMR2)

We did this selection correction because there may be unobserved
differences between mothers who pay for care and those who do not
that are not accounted for in the observed characteristics included
in the model.  The coefficients from the selection-corrected
expenditure estimation were then used to calculate predicted child
care expenditures for all women in the sample, conditional on the
mother actually using formal care (E(PCC|PCC >0)).  We multiplied
this conditional child care expenditure by the probability that a
mother would actually use formal care to arrive at the unconditional
predicted child care expenditure (E(PCC)).  This unconditional
predicted child care expenditure can be thought of as the price of
child care women face in considering employment, given the current
availability of informal child care possibilities. 


   ESTIMATING THE PROBABILITY OF
   WORKING
--------------------------------------------------------- Appendix I:3

In this stage, we estimated a structural probit of the probability
that a woman will work, including as independent variables the
predicted natural log wage and the predicted weekly child care
expenditure, as well as a set of variables that may affect the
decision to work separately from the wage and expenditure variables. 
These include variables such as race, marital status, number of
children under 6 and between 6 and 12, and urbanicity of residence. 

LFPART = fn (LNWGHAT,CCEXP,RACE,MARRIED,SPDVWD,UND6,
KID612,URBAN,SUBURB). 


      DATA SOURCE
------------------------------------------------------- Appendix I:3.1

We used data from the Urban Institute's 1990 National Child Care
Survey, a nationally representative sample of 4,392 households with
children.  We merged this main sample with the Urban Institute's Low
Income Substudy, a companion sample of 430 households with family
income below $15,000, in order to have enough cases to study poor and
near-poor mothers separately.  Our sample included all mothers
between ages 18 and 64 with at least one child under age 13 present
in the household.  Households without a mother present or missing
information about the mother's education, race, work experience, or
earnings were not included in the sample.  Our final sample of 3,930
mothers was further divided by total family income into three
groups--poor mothers, with family income at or below the 1989 federal
poverty level; near-poor mothers, with family income above 100 and up
to 185 percent of the federal poverty level; and nonpoor mothers,
with family income above 185 percent of the federal poverty level. 
Because women in different income groups may face different markets
for child care, it is important to estimate their child care
expenditure, and its impact on their decision to work, separately. 


   EMPIRICAL RESULTS
--------------------------------------------------------- Appendix I:4

Table I.7 presents the results of the structural probit of the
probability of working.  For all three income groups, we found that
the predicted child care expenditure (CCEXP) has a negative effect on
the mothers' probability of working.  This effect is statistically
significant for poor mothers at the .025 level, and for near-poor and
nonpoor mothers at the 0.01 level.\11 In addition, for all three
income groups, the natural log of hourly wage (LNWGHAT) has a
positive effect on the probability of mothers working, but this
effect is significant for the nonpoor mothers only, at the 0.10
level. 

Looking at the individual and family characteristic variables, we
found some similarities as well as some interesting differences
across the three income groups.  According to this model, race has no
significant impact on the probability of mothers working in any of
the three income groups.  For all three income groups, currently
married women (MARRIED) have a lower probability of working, and
separated, divorced, or widowed women (SPDVWD) have a higher
probability of working.  These results are significant at least at
the 0.05 level in every case except for the effect of being currently
married on the probability of poor women working. 

The number of children under age 6 (UND6) has a negative and
significant effect\12 on the probability of working for poor and
nonpoor women,\13 and a positive but insignificant effect on the
probability of working for near-poor women. 


--------------------
\11 An estimate is considered statistically significant if the
probability is low that the true value of the coefficient is zero. 
In the case of near-poor and nonpoor mothers, the probability of the
true coefficient being zero is no greater than 0.01. 

\12 The variable is significant at the 0.01 level for nonpoor women
and at the 0.10 level for poor women. 

\13 The variable is significant at the 0.01 level for nonpoor women
and at the 0.10 level for poor women. 


   CALCULATING AVERAGE PROBABILITY
   OF WORKING
--------------------------------------------------------- Appendix I:5

The coefficients from this structural probit allowed us to calculate
the average probability of working for each income group.  We call
this the baseline probability of working (see row 1 in table I.8). 
For poor mothers, our model predicted that this baseline probability
is 29 percent--that is, an estimated 29 of every 100 poor mothers are
working.  Near-poor mothers have an estimated 43 percent baseline
probability of working, and for nonpoor mothers the estimated
baseline probability of working is 55 percent. 

Calculating the elasticity of the probability of working due to a
change in child care expenditure gave us a measure of the sensitivity
of mothers' decision to work to child care expenditures.  This price
elasticity of employment is -0.50 for poor women, -0.34 for near-poor
women, and -0.19 for nonpoor women.  Thus, a 1 percent decrease in
child care expenditures results in a 0.50 percent increase in the
average probability of working for poor mothers.  Near-poor and
nonpoor mothers' responses are smaller. 


   CHILD CARE SUBSIDY SIMULATIONS
--------------------------------------------------------- Appendix I:6

Following the work of Connelly,\14 and as another measure of the
sensitivity of the probability of mothers working to changes in child
care expenditures, we simulated varying levels of child care
subsidies for the women in the samples and calculated how their
average probability of working would change.  We first multiplied the
predicted child care expenditure value for each woman (CCEXP), by
different subsidy rates (10-percent, 25-percent, 50-percent, and
100-percent subsidies).  We then recalculated the average probability
of working using the original beta coefficients from the structural
probits and the new child care expenditure values created by each
subsidy.  Rows 2 through 5 in table I.8 present the new average
probabilities of working predicted by our model, given the different
subsidy rates.  These values can be interpreted as the new average
probability that a woman in a specific income group will work, if all
other variables are held constant and only her child care
expenditures are changed. 


--------------------
\14 See Rachel Connelly, "The Effect of Child Care Costs on Married
Women's Labor Force Participation," pp.  83-90. 


   THE EFFECT OF WELFARE REFORM
   PROPOSALS ON INFORMAL CARE
--------------------------------------------------------- Appendix I:7

Legislative proposals for welfare reform have focused on requirements
for many more mothers on welfare to participate in education and
training; some of these proposals would also place a 2-year cap on
recipients' ability to receive welfare without working.  These new
requirements could mean an increase in the number of low-income
mothers in the workforce and, thus, an increased demand for child
care.  These new entrants to the workforce may be less likely to have
a source of informal care than the mothers who are already working. 
In addition, some of these mothers may be women who formerly provided
informal care to other low-income mothers.  Their entrance into the
workforce, therefore, both increases the demand for care and
decreases the supply of informal care at the same time.  Economic
theory suggests that as the demand for child care increases, the
price of formal care will rise.  This, in turn, will raise the
opportunity cost of informal care providers and draw more of them
into the formal child care market, further decreasing the supply of
informal care.  The end result for low-income mothers of an increase
in the demand for child care and a decrease in the supply of informal
care may be that more of them are forced to pay for child care while
they work. 


   CALCULATING AVERAGE PROBABILITY
   OF WORKING USING THE
   CONDITIONAL CHILD CARE
   EXPENDITURE
--------------------------------------------------------- Appendix I:8

To simulate the decision-to-work response of mothers in the three
income groups when they have no access to informal care, we
recalculated the average probability of working using the
coefficients from the estimated structural probit reported in table
I.7 and the conditional child care expenditure (E(PCC|PCC >0)).  Row
6 in table I.8 demonstrates how much the average probability of
working falls when mothers face the full cost of child care
expenditures with no access to informal care.  Once child care
subsidies are provided, however, the probability of working once
again begins to rise, until, with a 100-percent subsidy, it reaches
the same probability level as in the unconditional case, as can be
seen in row 7 of table I.8. 


   COMPARISON WITH THE LITERATURE
--------------------------------------------------------- Appendix I:9

If we compare the results of this analysis with those of other
researchers who have done similar work, we find that they are quite
consistent, allowing for the differences in the populations studied. 

Connelly calculated a price elasticity of employment of -0.20 for her
sample of married women, using data from the 1984 Panel of Survey and
Program Participation (SIPP).\15 She also simulated how a 100-percent
child care subsidy would affect the probability of employment for the
women in her sample.  She calculated an increase in probability of
employment from a mean of 58.8 percent to 68.7 if mothers received a
100-percent child care subsidy. 

Blau and Robins calculated a price elasticity of -0.38, also on a
sample of married women only, using data from a 1980 household survey
of the Employment Opportunity Pilot Projects.\16 Using the same 1984
SIPP data as Connelly, Ribar calculated a price elasticity for
married women of -0.74,\17 while Gustafsson and Stafford, using data
on married women in Sweden, calculated a price elasticity of
employment of -0.872.  None of these studies looked separately at
single mothers. 

A recent study by Kimmel examined single mothers and married mothers
separately.\18 Kimmel, using SIPP data from the 1987 panel,
calculated a price elasticity of employment of -0.521 for single
mothers, and -0.309 for married mothers.  Kimmel's results of a
higher price elasticity for single mothers than married mothers are
particularly relevant, since single mothers are more likely to be
low-income.\19 Kimmel also calculated mean predicted probability of
labor force participation for single mothers currently receiving AFDC
support.  She found that the baseline predicted probability of labor
for participation was 0.121 but that it rose to a predicted
probability of 0.462 when child care expenditures are subsidized at
100 percent.  This last result is very similar to the predicted
probability calculated for poor women in our sample under the same
conditions of a simulated 100-percent subsidy. 


--------------------
\15 See Rachel Connelly, "The Effect of Child Care Costs on Married
Women's Labor Force Participation," pp.  83-90. 

\16 David M.  Blau, and Philip K.  Robins, "Child Care Costs and
Family Labor Supply," Review of Economics and Statistics, Vol.  70,
No.  3 (1988), pp.  374-381. 

\17 David C.  Ribar, "Child Care and the Labor Supply of Married
Women." The Journal of Human Resources, Vol.  27, No.  1 (Winter
1992), pp.  134-165. 

\18 Jean Kimmel, "Child Care and the Employment Behavior of Single
and Married Mothers," Staff Working Paper #92-14, W.E.  Upjohn
Institute for Employment Research (1992). 

\19 In our sample, only 30 percent of poor women are married.  (See
app.  I, table I.2). 


   LIMITATIONS OF THE STUDY
-------------------------------------------------------- Appendix I:10

As in any empirical analysis, there is a level of uncertainty around
each of the estimates presented here.  The 1990 National Child Care
Survey has both strengths and limitations.  The detailed information
on characteristics of parents, including their education and
employment as well as the level of detail in the child care use data,
make this data set very useful for answering questions about child
care and employment decisions of mothers.  However, less specific
data on expenditures for all the children in the household,
especially in terms of identifying when there are multiple child care
arrangements, which arrangements were for formal care, and which were
for informal care, may mean that our predicted child care expenditure
variables could be biased.\20

In addition, our sample included relatively low numbers of poor women
who were employed and who used formal care for their children. 
Because predicted wages and predicted child care expenditures had to
be estimated for a large portion of the poor mothers' sample, the
estimates may be more imprecise.\21 However, the strong significance
of the predicted child care expenditure variable in the structural
probit on probability of working gave us some confidence in,
especially, the direction and relative magnitude of our findings in
terms of the impact of child care expenditures on mothers' employment
decisions. 



                                    Table I.1
                     
                               Variable Definitions

Variable            Definition          Variable            Definition
------------------  ------------------  ------------------  --------------------
UND6                Number of children  SPDVWD              Equals one if
                    under age 6 in the                      separated, divorced,
                    household                               or widowed

KID612              Number of children  KID1318             Equals one if there
                    aged 6-12 in the                        is a child aged 13-
                    household                               18 present in the
                                                            household

EDUC                Years of formal     OTHINCOM            Equals one if the
                    education                               household has income
                                                            from any source
                                                            other than
                                                            respondent's or
                                                            spouse's earnings

WRKEXPER            Years of work       OTHADLT             Equals one if there
                    experience,                             is another adult in
                    defined as age                          the household
                    minus years of                          besides the
                    education minus 6                       respondent (and
                                                            spouse, if married)

EXPSQR              Years of work       UNEMADLT            Equals one if there
                    experience squared                      is an unemployed
                                                            adult in the
                                                            household besides
                                                            the respondent (and
                                                            spouse, if married)

URBAN               Equals one if       LFPART              Equals one if
                    respondent lives                        respondent was
                    in an urban area                        employed and earning
                                                            a nonzero wage in
                                                            the previous week

SUBURB              Equals one if       PAYCARE             Equals one if the
                    respondent lives                        respondent paid for
                    in a suburban area                      child care in the
                                                            previous week

RACE                Equals one if       HREARN              Hourly wage of
                    respondent is                           respondent
                    white.

WEST                Equals one if       LNWGHAT             Predicted log hourly
                    respondent lives                        wage of respondent
                    in Western region

NORCENTR            Equals one if       WKEXPEND            Weekly child care
                    respondent lives                        expenditure for all
                    in North Central                        children
                    region

NOREAST             Equals one if       CCEXP               Predicted
                    respondent lives                        unconditional weekly
                    in Northeastern                         child care
                    region                                  expenditure

MARRIED             Equals one if       EXPHAT              Predicted
                    respondent is                           conditional weekly
                    currently married                       child care
                                                            expenditure
--------------------------------------------------------------------------------


                          Table I.2
           
            Descriptive Statistics for Poor, Near-
            poor, and Nonpoor Mothers (unweighted
                           samples)

                                               Near-  Nonpoo
                                                poor       r
                                        Poor  (n=585  (n=280
                                     (n=541)       )      4)
----------------------------------  --------  ------  ------
Age                                    30.10   31.00   33.20
Education (years)                      11.60   12.20   13.70
Work experience (years)                12.40   12.90   13.50
Number of children under age 6          0.85    0.79    0.69
Number of children aged 6-12            1.20    1.20   0.990
Percent white                          62.00   80.00   92.00
Percent married                        30.00   61.00   89.00
Percent separated, divorced, or        39.00   26.00    8.00
 widowed
Percent never married                  31.00   13.00    3.00
Percent in labor force                 29.00   43.00   55.00
Percent pay for child care             24.00   30.00   45.00
Percent urban                          37.00   33.00   38.00
Percent suburban                       23.00   29.00   37.00
Percent rural                          40.00   38.00   25.00
Percent in Western region              18.00   24.00   19.00
Percent in Southern region             45.00   38.00   32.00
Percent in North-Central region        23.00   25.00   28.00
Percent in Northeastern region         14.00   13.00   21.00
------------------------------------------------------------


                          Table I.3
           
              Reduced Form Probit of Labor Force
                    Participation Decision

                              Estima
                                 ted  Standa      T-
                              coeffi      rd  statis    Mean
Variable                       cient   error     tic   value
----------------------------  ------  ------  ------  ------
CONSTANT                           -   0.161       -    1.00
                               1.187           7.369
UND6                               -   0.033       -    0.72
                               0.244           7.377
KID612                             -   0.029       -    1.05
                               0.229           8.013
KID1318                            -   0.060       -    0.22
                               0.039           0.647
EDUC                           0.084   0.010   8.565   13.20
WRKEXPER                       0.062   0.011   5.452   13.24
EXPSQR                             -   0.000       -  216.26
                               0.002           4.374
URBAN                              -   0.053       -    0.37
                               0.073           1.381
SUBURB                             -   0.053       -    0.34
                               0.042           0.795
RACE                           0.015   0.066   0.225    0.86
WEST                               -   0.060       -    0.19
                               0.115           1.917
NORCENTR                           -   0.054       -    0.27
                               0.072           1.344
NOREAST                            -   0.060       -    0.19
                               0.087           1.438
MARRIED                        0.083   0.086   0.962    0.77
SPDVWD                         0.307   0.095   3.218    0.15
OTHINCOM                           -   0.042       -    0.47
                               0.071           1.687
OTHADLT                        0.272   0.091   2.989    0.15
UNEMADLT                           -   0.109       -    0.09
                               0.334           3.073
------------------------------------------------------------
N = 3,930.  Log likelihood ratio = 255.08. 



                          Table I.4
           
           Ordinary Least Squares Estimation of Log
           of Hourly Wage, Corrected for Selection
                             Bias

                                      Estima
                                         ted  Standa      T-
                                      coeffi      rd  statis
Variable                               cient   error     tic
------------------------------------  ------  ------  ------
CONSTANT                              0.0290  0.1840   0.156
RACE                                  0.0540  0.0390   1.404
EDUC                                  0.1300  0.0080  16.687
WRKEXPER                              0.0200  0.0080   2.649
EXPSQR                                     -  0.0002       -
                                      0.0003           1.502
URBAN                                 0.2670  0.0330   8.155
SUBURB                                0.1620  0.0320   5.023
WEST                                  0.0010  0.0380   0.032
NORCENTR                                   -  0.0340       -
                                      0.0570           1.702
NOREAST                               0.0670  0.0370   1.804
IMR1                                       -  0.0940       -
                                      0.0400           0.427
------------------------------------------------------------
N= 1,950.  R\2 = 0.249. 



                          Table I.5
           
           Probit Estimation of the Probability of
                Paying for Child Care Services

                                       Near-          Nonpoo
                        Poor            poor               r
                      estima          estima          estima
                         ted             ted             ted
                      coeffi          coeffi          coeffi
                       cient           cient           cient
                      (stand          (stand          (stand
                         ard             ard             ard
Variable              error)          error)          error)
--------------------  ------  ------  ------  ------  ------
CONSTANT                   -               -               -
                       1.249          1.384\          1.135\
                      (0.402          (0.402          (0.229
                           )               )               )
UND6                   0.054           0.141
                      (0.088          (0.086          0.085\
                           )               )          (0.041
                                                           )
KID612                     -           0.001               -
                       0.014          (0.069          0.079\
                      (0.070               )          (0.036
                           )                               )
KID1318                0.093               -               -
                      (0.164           0.191           0.092
                           )          (0.160          (0.070
                                           )               )
EDUC                   0.056           0.061           0.102
                      (0.029          (0.028          (0.012
                           )               )               )
OTHINCOM                   -                           0.009
                       0.051          0.26\2          (0.050
                      (0.126          (0.119               )
                           )               )
OTHADLT                    -           0.282           0.097
                       0.167          (0.209          (0.111
                      (0.262               )               )
                           )
UNEMADLT                   -               -               -
                       0.016           0.244           0.203
                      (0.288          (0.263          (0.139
                           )               )               )
URBAN                      -               -
                       0.180           0.174          0.15\6
                      (0.151          (0.143          (0.065
                           )               )               )
SUBURB                     -           0.061
                       0.141          (0.141          0.190\
                      (0.165               )          (0.064
                           )                               )
WEST                   0.012           0.116               -
                      (0.189          (0.157           0.158
                           )               )          (0.072
                                                           )
NORCENTR                   -           0.112               -
                       0.085          (0.151          0.168\
                      (0.170               )          (0.063
                           )                               )
NOREAST                0.087           0.059               -
                      (0.201          (0.188          0.301\
                           )               )          (0.069
                                                           )
RACE                       -           0.028           0.070
                       0.024          (0.160          (0.091
                      (0.156               )               )
                           )
MARRIED                    -               -               -
                      0.532\          0.451\           0.468
                      (0.198          (0.188          (0.149
                           )               )               )
SPDVWD                                 0.283           0.054
                      0.31\5          (0.188          (0.169
                      (0.162               )               )
                           )
N                        541             585            2804
\Log likelihood         38.5            55.1           169.4
 ratio
------------------------------------------------------------


                          Table I.6
           
             Ordinary Least Squares Estimation of
             Weekly Child Care Expenditures, With
                Correction for Selection Bias

                                       Near-          Nonpoo
                        Poor            poor               r
                      estima          estima          estima
                         ted             ted             ted
                      coeffi          coeffi          coeffi
                       cient           cient           cient
                      (stand          (stand          (stand
                         ard             ard             ard
Variable              error)          error)          error)
--------------------  ------  ------  ------  ------  ------
CONSTANT                               15.97           4.849
                      10.453          (25.96          (23.15
                      (41.57               )               )
                           )
UND6                   1.847
                      (6.256          18.236          17.432
                           )          (3.519          (2.388
                                           )               )
KID612                 6.021               -               -
                      (4.43)           1.612           1.220
                                      (2.937          (2.26)
                                           )
KID1318                    -               -               -
                       5.112           7.788           16.26
                      (9.977          (7.316          (4.151
                           )               )               )
RACE                   8.922               -           7.410
                      (9.067           1.001          (4.736
                           )          (5.993               )
                                           )
EDUC                   0.234                           2.688
                      (1.983          2.3187          (0.982
                           )          (1.334               )
                                           )
OTHINCOM               6.896               -           0.945
                      (8.143           5.727          (2.739
                           )          (5.692               )
                                           )
OTHADLT                    -               -           2.289
                       1.757           2.034          (4.606
                      (10.9)          (6.79)               )
URBAN
                      14.106          12.485          16.696
                      (10.12          (6.032          (4.038
                           )               )               )
SUBURB                                 0.961            6.39
                      12.044          (5.864          (4.021
                      (10.63               )               )
                          9)
WEST                   3.529               -           1.299
                      (12.27           3.484          (4.118
                          4)          (6.369               )
                                           )
NORCENTR                   -           0.269               -
                       3.275          (6.123           6.013
                      (11.58               )          (3.669
                           )                               )
NOREAST                                    -           2.980
                      12.818           3.120          (4.579
                      (12.84          (7.773               )
                          6)               )
IMR2                       -               -               -
                       2.901          16.044           16.97
                      (18.76          (11.15          (12.64
                          9)               )               )
N                        129             176            1262
R\2                    0.077           0.254           0.168
------------------------------------------------------------


                          Table I.7
           
            Structural Probit Estimation of Labor
           Force Participation Decision of Mothers
                     With Young Children

                                       Near-          Nonpoo
                        Poor            poor               r
                      estima          estima          estima
                         ted             ted             ted
                      coeffi          coeffi          coeffi
                       cient           cient           cient
                      (stand          (stand          (stand
                         ard             ard             ard
Variable              error)          error)          error)
--------------------  ------  ------  ------  ------  ------
CONSTANT                   -           0.078          0.606\
                      0.603\          (0.398               c
                           a               )          (0.240
                      (0.428                               )
                           )
LNWGHAT                0.292           0.260          0.175\
                      (0.249          (0.224               a
                           )               )          (0.116
                                                           )
CCEXP                      -               -               -
                      0.046\          0.022\          0.009\
                           c               d               d
                      (0.023          (0.009          (0.003
                           )               )               )
RACE                       -               -               -
                       0.070           0.094           0.078
                      (0.145          (0.143          (0.090
                           )               )               )
MARRIED                    -               -               -
                       0.155          0.520\          0.258\
                      (0.217               d               b
                           )          (0.196          (0.153
                                           )               )
SPDVWD                0.638\          0.298\          0.381\
                           d               a               c
                      (0.180          (0.185          (0.171
                           )               )               )
UND6                       -                               -
                      0.116\          0.0\82          0.137\
                           a          (0.111               d
                      (0.089               )          (0.052
                           )                               )
KID612                 0.009               -               -
                      (0.073           0.038          0.215\
                           )          (0.064               d
                                           )          (0.037
                                                           )
URBAN                      -               -               -
                      0.420\          0.232\           0.020
                           c               b          (0.071
                      (0.150          (0.141               )
                           )               )
SUBURB                     -               -               -
                       0.158           0.120           0.016
                      (0.157          (0.134          (0.066
                           )               )               )
N                        541             585            2804
Log likelihood ratio  32.998          32.954          102.83
------------------------------------------------------------
\a Statistically significant at the 0.100 level. 

\b Statistically significant at the 0.050 level. 

\c Statistically significant at the 0.025 level. 

\d Statistically significant at the 0.010 level. 



                          Table I.8
           
              Average Probability of Labor Force
            Participation for Poor, Near-Poor, and
                       Nonpoor Mothers

                                                      Nonpoo
                                               Near-       r
                                        Poor    poor  (perce
                                      (perce  (perce     nt)
                                         nt)     nt)  n=2,80
                                       n=541   n=585       4
------------------------------------  ------  ------  ------
Baseline probability                      29      43      55
10% subsidy                               30      44      56
25% subsidy                               32      46      58
50% subsidy                               36      50      60
100% subsidy                              44      57      65
Baseline probability with                  6      19      44
 conditional price
Conditional price with 100% subsidy       44      57      65
------------------------------------------------------------

--------------------
\20 If the number of hours that children are in care includes both
formal and informal care, predicted child care expenditures could be
underestimated. 

\21 This may be one reason why the effect of predicted wages on the
probability of working for poor women is not statistically
significant, as shown in table I.7. 


GAO CONTACTS AND ACKNOWLEDGMENTS
========================================================== Appendix II


   GAO CONTACTS
-------------------------------------------------------- Appendix II:1

Lynne Fender, Assistant Director, (202) 512-7229
Alicia Puente Cackley, Senior Economist, (202) 512-7022


   ACKNOWLEDGMENTS
-------------------------------------------------------- Appendix II:2

The following team members also contributed to this report:  Cynthia
A.  Bascetta and Sarah L.  Glavin. 

In addition, the following individuals acted as peer reviewers for
this report:  Mr.  Douglas Besharov, the American Enterprise
Institute; Dr.  Rachel Connelly, Bowdoin College; Dr.  Sandra
Hofferth, Institute for Social Research; and Dr.  Douglas Wissoker,
the Urban Institute. 


BIBLIOGRAPHY
========================================================== Appendix II

Beller, Andrea, and Karen Folk.  "Part-time Work and Child Care
Choices for Mothers of Preschool Children." Journal of Marriage and
Family, Vol.  55 (1993), pp.  146-157. 

Berger, Mark C., and Dan A.  Black.  "Child Care Subsidies, Quality
of Care, and the Labor Supply of Low-Income, Single Mothers." Review
of Economics and Statistics, Vol.  74, No.  4 (Nov.  1992), pp. 
635-642. 

Blau, David M., and Philip K.  Robins.  "Child Care Costs and Family
Labor Supply." Review of Economics and Statistics, Vol.  70, No.  3
(Aug.  1988), pp.  374-381. 

Brayfield, April.  "Child Care Costs as a Barrier to Women's
Employment." Final Report submitted to U.S.  Department of Labor's
Women's Bureau.  Washington, D.C., 1992. 

Connelly, Rachel.  "The Effect of Child Care Costs on Married Women's
Labor Force Participation." Review of Economics and Statistics, Vol. 
74, No.  1 (1992), pp.  83-90. 

Connelly, Rachel.  "The Effect of Child Care Costs on the Labor Force
Participation and AFDC Recipiency of Single Mothers." Discussion
Paper # 920-90, Institute for Research on Poverty, University of
Wisconsin-Madison, 1990. 

Fronstin, Paul, and Douglas Wissoker.  "The Effects of the
Availability of Low-Cost Child Care on the Labor Supply of Low-Income
Women." Unpublished paper presented at the Annual Meeting of the
Population Association of America, Miami, Florida, May 5-7, 1994. 

Gustafsson, Siv, and Frank Stafford.  "Child Care Subsidies and Labor
Supply in Sweden." The Journal of Human Resources, Vol.  27, No.  1
(Winter 1992), pp.  204-230. 

Heckman, James J.  "Sample Selection Bias as a Specification Error."
Econometrica, Vol.  47, No.  1 (Jan.  1979), pp.  153-161. 

Hofferth, S.L., A.  Brayfield, S.  Deich, and P.  Holcomb.  The
National Child Care Survey, 1990.  Washington, D.C.:  The Urban
Institute, 1991. 

Kimmel, Jean.  "Child Care and the Employment Behavior of Single and
Married Mothers." Staff Working Paper #92-14, W.E.  Upjohn Institute
for Employment Research, 1992. 

Ribar, David C.  "Child Care and the Labor Supply of Married Women."
The Journal of Human Resources, Vol.  27, No.  1 (Winter 1992), pp. 
134-165. 

"Who's Minding the Kids?  Child Care Arrangements:  Fall 1991."
Survey of Income and Program Participation, Current Population
Reports, P70-36, Washington, D.C.:  U.S.  Bureau of the Census, 1991. 

Child Care:  Working Poor and Welfare Recipients Face Service Gaps. 
(GAO/HEHS-94-87, May 13, 1994).  Washington, D.C.:  U.S.  General
Accounting Office, 1994. 

