Findings from the Self-Sufficiency Project: effects on children and adolescents of a program that...

Applied Developmental Psychology 24 (2003) 201–239

Findings from the Self-Sufficiency Project:

effects on children and adolescents of a program that

increased employment and income

Pamela Morris*, Charles Michalopoulos

Manpower Demonstration Research Corporation (MDRC), 16 East 34th Street, New York, NY 10016, USA

Accepted 1 April 2003

Abstract

This paper examines the effects on children of an antipoverty employment program for Canadian

welfare recipients called the Self-Sufficiency Project (SSP). The SSP made work pay better than

welfare by offering a temporary, but generous, earnings supplement to single parents who left welfare

for full-time employment. The SSP was tested using a rigorous random assignment research design.

While the SSP was found to increase employment and income for parents of children in every age

group, the effects of the program on the children themselves differed with their age. For very young

children, the SSP had no effect on children’s outcomes. For children in the middle childhood period at

follow-up, the SSP increased children’s cognitive functioning and health outcomes, but had no benefits

on their social behavior. For adolescents, the SSP increased minor delinquency and substance use. The

results are discussed in terms of their contribution to research and policy.

D 2003 Elsevier Science Inc. All rights reserved.

Keywords: Maternal employment; Income; Welfare; Poverty; Experiments

1. Introduction

A number of studies have examined the effects of welfare and employment policies on

the economic status of low-income parents. Less is known, however, about the effect on

0193-3973/03/$ – see front matter D 2003 Elsevier Science Inc. All rights reserved.

doi:10.1016/S0193-3973(03)00045-5

* Corresponding author. Tel.: +1-212-340-8880; fax: +1-212-684-0832.

E-mail address: [email protected] (P. Morris).

P. Morris, C. Michalopoulos / Applied Developmental Psychology 24 (2003) 201–239202

children of policies that encourage welfare recipients to work. The few studies that do exist

focus on elementary school-age children, leaving us relatively uninformed about the effects

of such policies on both adolescent and very young children (Gennetian & Miller, 2002;

Huston et al., 2001; McGroder, Zaslow, Moore, & LeMenestrel, 2000). This paper seeks to

add to this literature by looking at evidence from a demonstration project called the Self-

Sufficiency Project (SSP), a policy innovation that made work pay better than welfare by

offering a temporary, but generous, earnings supplement to selected single-parent welfare

recipients in Canada. This research informs both developmental research and policy, and is

especially timely given the policy context of changes at the federal and state level in

the area of welfare reform. While the SSP tested a policy change in the context of two

Canadian provinces, the similarity with the few most generous U.S. policies, along with the

similarity in welfare and economic context between the United States and Canada, provides

an appropriate backdrop for considerations regarding U.S. policy as well as developmental

theory.

Three features combine to make this study unique. First, the intervention studied in the

SSP had substantial effects on the employment—particularly full-time employment—

earnings, and income of single-parent welfare recipients who were studied (Michalopoulos,

Card, Gennetian, Harknett, & Robins, 2000). Second, the SSP was studied using random

assignment, in which individuals were assigned in a lottery-like process to either a

program group, which was offered the SSP supplement, or a control group, which was

not. Random assignment is often considered the most reliable means of determining the

effects of a new policy. Moreover, the random assignment design allows definitive causal

inferences about the effects of an incentive to work—and more tenuous statements about

the resulting effects of increases in maternal employment and income—on children than

has been typically addressed in nonexperimental work. Third, the SSP collected extensive

information on preschool children, school-age children, and adolescents. The SSP therefore

offers a rare opportunity to inform our understanding of how programs that increase

employment and income may affect children at different developmental periods. Although

other studies have had one or two of these features, the SSP is the only study to have all

three.

In this study, we examine the effects of the SSP program on children’s developmental

outcomes. While we do examine the effects of the SSP program on potential mediators of the

program, including both direct targets of the program such as employment and income, and

more intermediate mechanisms such as child care and parenting behavior, we do not directly

examine the pathways by which the effects occurred here. As discussed later, by focusing on

the direct impacts of the SSP program, we preserve the integrity of the experimental design of

the study and can confidently attribute any effects on children, or mediators examined, to the

program itself.

1.1. The effects of welfare and employment programs on children and families

The effects of welfare, employment, and antipoverty programs on children have been

studied in several random assignment studies. The Negative Income Tax (NIT), which was

P. Morris, C. Michalopoulos / Applied Developmental Psychology 24 (2003) 201–239 203

studied experimentally in four cities in the United States from 1968 to 1982, provided a

guaranteed level of income for working poor families, although receipt of the NIT was not

contingent on work effort (Munnel, 1986; Office of Income Security Policy, 1983). The NIT

maintained or increased income, and it decreased parents’ work effort. For children, the NIT

was found to improve nutrition, early school achievement, and high school completion

(Mallar & Maynard, 1981; Salkind & Haskins, 1982).

In several recent studies of welfare reform programs in the United States, programs that

increased both employment and income also benefited children who were elementary school

age 2 to 3 years after the start of the programs (Morris, Huston, Duncan, Crosby, & Bos,

2001). The New Hope program increased employment and income among low-income

families in Milwaukee, WI by offering parents an earnings supplement, child care assistance,

and health insurance if they worked at least 30 hours per week. Teachers reported that boys in

the New Hope program were performing better academically and behaviorally in school than

comparable boys in a control group, although parent reports did not reveal significant

differences in program and control group children (and positive effects were not found for

girls; Bos et al., 1999; Huston et al., 2001).

The Minnesota Family Investment Program (MFIP; Knox, Miller, & Gennetian, 2000)

also increased both employment and income among welfare recipients in urban counties in

Minnesota in part by offering them financial incentives to work. Results from the MFIP

suggest that elementary school-age children benefit when their parents work part time and

their income increases (Gennetian & Miller, 2002). But, the duration of income improve-

ment may make a difference. Comparisons between the MFIP and a program in

Connecticut that combined a supplement with a time limit suggests that the MFIP, with

more sustained income increases, had more pervasive effects on multiple domains of

children’s well-being than Connecticut’s program, which curtailed income gains with a time

limit (Gennetian & Morris, 2003). And, in contrast to the New Hope and MFIP findings,

results from the evaluation of the National Evaluation of Welfare-to-Work Strategies

(NEWWS) suggest that programs that increase employment but not income do not have

consistent positive effects on early school-age children (Hamilton, Freedman, & McGroder,

2000; McGroder et al., 2000).

Data on older children are much weaker than for elementary school children because

most studies were not designed to assess the effects on older children. However, a

systematic meta-analysis of programs aimed to increase parental employment did find

small negative effects on adolescent school progress (Gennetian, Duncan, Knox, Clark-

Kauffman, & Vargas, 2002). These negative effects appear to be related to the increases in

responsibilities in caring for younger siblings that occurred when adolescents’ single parents

went to work.

1.2. Theoretical framework

Theories developed from sociology, psychology, and economics suggest that there are

several pathways by which children may be affected by welfare and employment programs.

Economic theory emphasizes the notion that employment and income can affect the resources


that families can provide for their children, which in turn can influence children’s

development (Becker, 1981; Coleman, 1988). Increased resources include both material

resources (that is, goods such as food or books) and nonmaterial resources (such as social

interactions with others, including teachers and peers) that parents provide for their children.

In this way, children are thought to benefit from parents’ increased income because this

increased income allows parents to invest more in their children’s development.

Psychology has emphasized instead the role of family functioning and parenting practices

that affect children’s well-being. Increased income may influence children’s development, for

example, by reducing parental stress and thereby changing the parent–child relationship

(Mcloyd, 1990; McLoyd, Jayartne, Ceballo, & Borquez, 1994).

1.2.1. Maternal employment

Maternal employment has been hypothesized to have both positive and negative effects on

children (Hoffman, 1989). In general, however, maternal employment has been found to have

few negative effects. There are two exceptions: maternal employment in the child’s first year

of life has been found to have negative effects, and boys have been found to be negatively

affected by their mothers’ employment (Baydar & Brooks-Gunn, 1991; Belsky & Rovine,

1988; Bronfenbrenner & Crouter, 1982; Brooks-Gunn, Han, & Waldfogel, 2002). In several

prior studies (Bronfenbrenner & Crouter, 1982; Desai, Chase-Lansdale, & Michael, 1989),

differences in the effects of maternal employment by social class have been observed, with

more negative effects of maternal employment for middle class as compared with lower

income families. Low-income children and children of single mothers have been found to

benefit from their mothers’ employment (Harvey, 1999; Moore & Driscoll, 1997; Vandell &

Ramanan, 1992; Zaslow & Emig, 1997). However, these effects may not be causal, as

differences between children of employed and nonemployed mothers appear to be due as

much to the differences in characteristics between employed and unemployed mothers as to

parents’ work status (Zaslow, McGroder, Cave, & Mariner, 1999).

While maternal employment seems to benefit low-income children, some factors may

reduce this effect. First, a high number of hours of employment early in a child’s life may be

associated with negative cognitive and achievement outcomes (Harvey, 1999). In a related

matter, extensive increases in employment without increases in income in the context of time-

limited welfare have been found to result in negative effects on school-age children (Morris,

Bloom, Kemple, & Hendra, in press). Moreover, maternal employment may be associated

with positive child outcomes only when women want to work (Alvarez, 1985; Farel, 1980).

In addition, research suggests that high-level, complex jobs may have positive outcomes for

children, while lower levels of employment may translate into negative outcomes (Parcel &

Menaghan, 1994, 1997). These studies raise the question whether moving mothers into full-

time, low-wage service employment will have positive effects on children.

1.2.2. Family income

In addition to increasing employment, the SSP was intended to increase family income,

with the long-term goal of moving families out of poverty. Research suggests that poverty can

negatively affect children’s functioning, especially if it occurs in early childhood and is


persistent (Duncan & Brooks-Gunn, 1997; Duncan, Brooks-Gunn, & Klebanov, 1994). The

negative influence of poverty may be more concentrated in children’s achievement and

academic functioning, rather than their behavior and health outcomes (Duncan & Brooks-

Gunn, 1997). As an antipoverty program, the SSP may therefore exert its strongest influence

on children’s achievement and school functioning and only secondarily affect the domains of

children’s behavior problems and health.

Both the resources provided to children and changes in family interactions may play a role

in the association between family income and children’s development, as poverty is

associated with fewer learning materials in the home and greater parental stress and more

insensitive parenting practices (Bradley & Caldwell, 1984; McLoyd et al., 1994; Smith,

Brooks-Gunn, Lee, & Klebanov, 1997; Sugland et al., 1995).

1.2.3. Mediators of employment and income: Child care, parenting, and parents’ emotional

well-being

Two sets of mediators have been identified in the literature as key intervening mechanisms

between changes in parents’ employment and family income and outcomes for children.

These include changes that may occur outside the child’s home—child care and children’s

activities—that parents’ use to structure children’s time as they engage in employment and to

pay for educational activities; and changes that occur within the parent–child relationship,

including parents’ emotional well-being and parenting behavior. Prior research on both

pathways is discussed below.

1.2.3.1. Child carze as a mediating factor. One pathway by which changes in employment

and income may affect children is through the use of child care and children’s participation in

after-school activities. For young children, compensatory education programs have been

found to benefit low-income preschool children, at least in the short term (Lazar &

Darlington, 1982; Lee, Brooks-Gunn, & Shnur, 1988; McKey et al., 1985). There are

benefits in cognitive and language skill development for children in low-income families

from high-quality care as compared with low-quality care (Burchinal et al., 2000; NICHD

Early Child Care Research Network, 2000; Peisner-Feinberg et al., 1999; Ramey et al., 2000;

Shonkoff & Phillips, 2000). In addition, formal, center-based child care is more beneficial to

cognitive development than home-based care when the two are of comparable quality

(NICHD Early Child Care Research Network, 2000). With regard to health outcomes, infants

who are placed in group child care arrangements have higher rates of contagious illnesses

than do infants who are cared for at home (NICHD Early Child Care Research Network,

2001, in press). The effects of child care (whether of high or low quality) on children’s social

behavior, however, are much less consistently positive or negative (NICHD Early Child Care

Research Network, 1998; Peisner-Feinberg et al., 1999).

For older children, emerging research suggests that formal after-school activities are

associated with positive outcomes for low-income children and children living in unsafe

environments (Pettit, Bates, Dodge & Meece, 1999; Posner & Vandell, 1994, 1999). It is

believed that these activities only help children when they provide a stimulating academically

focused environment, as well as protection against a deviant peer group.

1.2.3.2. Parents’ emotional well-being and parenting behavior . Family income and maternal

employment may also affect children through changes in parents’ emotional well-being and

parenting behavior. Financial strain has been found to negatively affect children through

marital conflict and parenting in two-parent families (Conger, Conger, & Elder, 1997) and

through depression and negative disciplinary practices in single-mother families (McLoyd et

al., 1994). Elder’s seminal work on economic hardship during the Great Depression identified

a link between job and financial loss and punitive, inconsistent parenting behavior (Elder,

1974, 1979; Elder, Liker, & Cross, 1984). Later work provided evidence that the relation

between economic changes and parenting were mediated by changes in parents’ emotional

well-being (like depression; see McLoyd, 1998 for a review). McLoyd’s work further

expanded this model for relevance to Black families (McLoyd, 1990), and provided a test

of this model among single-parent families, with findings suggesting that job loss, through

financial strain and psychological distress, affected parenting behavior and, ultimately,

adolescent well-being (McLoyd et al., 1994).

Notably, previous research on changes in income and employment has focused on the

negative impact of unemployment, job loss, and financial insecurity rather than whether

increases in income and entry into work can cause gains in children’s development in the

same way that job loss seems to undermine children’s development. However, the research

suggests that interventions aimed at removing the financial strain on the family may improve

parenting and children’s functioning (McLoyd, 1998).

1.2.4. Developmental differences

Child’s age may moderate the effects of maternal employment on children. As discussed

earlier, employment in the first year of a child’s life may be associated with more negative

child outcomes (Baydar & Brooks-Gunn, 1991; Brooks-Gunn et al., 2002), especially if it

involves many hours of work (Harvey, 1999). Spending long hours in nonmaternal care is

also associated with negative behavioral outcomes for young children (Currie & Thomas,

1995; Lamb, 1998; McLoyd, 1998), but high-quality care may help to deter some of these

negative effects (see Lamb, 1998).

Adolescent children and their younger school-age peers may be affected differently by

maternal employment. After-school arrangements are associated with positive outcomes

for both preadolescent and adolescent children, keeping them in structured care and away

from peers (Eccles & Gootman, 2002; Pettit et al., 1999; Posner & Vandell, 1994, 1999).

By contrast, adolescents may have difficulties if left alone after school and into the

evening hours as mothers take on off-hour and shift work, and maternal employment has

been found to be associated with reduced parental supervision and increased adolescent

delinquency among low-income families (Sampson & Laub, 1994). Adolescent children

may also be asked to take on greater household responsibilities as well as work when

their single mothers move into employment, and a high level of employment during

adolescence (particularly more than 20 hours of employment) has been linked with

children’s difficulties in school and increased drug and alcohol use (Burton, Brooks, &

Clark, 2002; Mortimer, Finch, Ryu, Shanahan, & Call, 1996; Steinberg & Dornbusch,

1991).



Child age has also been found to moderate the effects of family income on children.

Associations between poverty and child outcomes are stronger for preschool than for early-

school-age or older children. A longitudinal analysis has suggested that early childhood

poverty (when the child is age 0–5) is associated negatively with children’s completion of

schooling in adolescence, but that poverty from ages 6 to 15 is not (Duncan & Brooks-Gunn,

1997). This finding suggests that increasing income during the preschool years may be more

important for children’s outcomes than increasing income in adolescence.

1.2.5. Developmental differences in the combined effects of maternal employment and income

To our knowledge, there has been no research that attempts to examine the combined effects

of increasing maternal employment and family income across differing developmental periods.

While research has examined the moderating role of child age in the effects of maternal

employment and in the effects of family income, studies have not examined the effects of the

combined influence of the two, and how together they may affect younger, middle childhood

and adolescent children differently. The current study adds to our knowledge of developmental

processes by examining the effects of a program that increases both maternal employment and

family income on children of differing ages among long-term welfare recipient families.

Notably, as discussed in more detail in the Measures section, this work is complicated by the

fact that for differing ages of children, differing measures are needed to capture development-

ally appropriate constructs. While an analysis of the moderating role of child age would be

simplified by similar constructs collected across age groups, a recognition of developmental

change requires that measures change across the childhood age span.

1.3. The present study

The SSPwas a demonstration program designed to make work a viable alternative to welfare

for low-income parents, whose skills and experience would likely relegate them to low-paying

jobs. A group of about 6000 single parents in British Columbia and New Brunswick who had

been on welfare for at least a year were selected at random from the welfare rolls between

November 1992 andMarch 1995. In British Columbia, the SSP operated in the lower mainland,

which includes the Vancouver metropolitan area as well as neighboring areas to the north,

south, and east. In New Brunswick, the program operated in a region covering roughly the

lower third of the province, including the cities of Saint John, Moncton, and Fredericton.

Families who agreed to participate in the study were randomly assigned to either the program

group, which was offered the SSP supplement, or to a control group, which was not.

SSP’s financial supplement paid parents who left welfare and worked at least 30 hours per

week half the difference between their actual earnings and a target level of earnings.1 The target

earnings were set at Can$30,000 in New Brunswick and Can$37,000 in British Columbia a

year, although theywere adjusted slightly over time to reflect changes in the cost of living and in

1 Program group members could not qualify for the earnings supplement with jobs that were 100%

government-subsidized. Positions that were partially subsidized by the federal government or the province of New

Brunswick, however, were permitted.


the generosity of welfare benefits. When the study began, the target earnings levels corre-

sponded roughly to U.S.$22,500 and U.S.$27,750, respectively. The SSP supplement varied

with individual earnings, rather than family income, and was therefore unaffected by family

composition, other family members’ earnings, or unearned income.2 Finally, supplement

payments were available for a maximum of 3 years, and only to sample members who initiated

SSP payments by finding full-time work within 12 months of entering the study.

In this analysis, we examine the effects of the SSP on children ages 3–18 at the time of a

survey administered about 3 years after parents entered the study (the ‘‘3-year follow-up

survey’’). Children are divided into three distinct developmental periods: (1) those who were

infants and toddlers at random assignment (and thus preschoolers by the end of the follow-up

period); (2) those who were preschool and early school age at random assignment (and thus

were in elementary school, i.e., middle childhood, by the end of the follow-up); and (3) those

who were preadolescent and adolescent at random assignment (and thus in adolescence by the

end of the follow-up period).

Random assignment reduced the risk that program and control groups differed system-

atically at the start of the study; therefore any systematic differences in children’s

functioning can be directly attributed to the program and not to any unmeasured

characteristic. Although any differences in children’s outcomes between the program and

the control groups are likely to be related to the increased employment and income

stemming from the program, the experimental research design is fully maintained only by

examining the effect of the program on each mediator or outcome (and not by examining

the relationships between mediators and outcomes). As in previous research on the effects of

experimental welfare and employment programs on children (Gennetian & Miller, 2002;

Huston et al., 2001), we focus here on the experimental effects of the offer of the SSP

supplement on the direct targets of the intervention (employment and income), on the key

mediating processes (child care, parenting, and parents’ emotional well-being), and on

children’s outcomes.

More specifically, we address the following questions: (1) How did the SSP affect parents’

employment, income, and material hardship (the primary targets of the SSP)? (2) How did the

SSP affect children’s cognitive, behavior, and health outcomes? (3) How did the SSP affect

children’s interactions and experiences in and outside the home (including child care,

children’s activities, their parents’ emotional well-being, and parenting behavior—potential

mediators of the effects of income and employment)?

1.4. The Canadian context

How relevant is the SSP to our understanding of the effects of welfare and employment

programs in the United States? Under the policy innovation tested in the SSP, the income of a

2 Thus, the SSP supplement formula does not penalize single parents who receive child support, marry, or find

a partner. However, because benefits from SSP do not increase with family size (and Income Assistance [IA]

does), SSP is relatively less generous than IA for larger families.


typical parent with two children increased by about Can$400 per month, net of the welfare

benefits they would have forgone to receive the program’s supplement. This is very generous,

and more generous than many policies available to low-income families in either the United

States or Canada. However, there are a few policies that are similarly generous to the policy

tested under the SSP. The federal Earned Income Credit is very generous for certain

families—for example, it currently provides nearly U.S.$4000 to a parent with two children

who works full time at a minimum wage job. In addition, many states have implemented an

‘‘enhanced earnings disregard’’ as part of their welfare reform strategy, allowing welfare

recipients who work to keep more of their welfare dollars as their earnings increase and

disregarding a greater proportion of a welfare recipients’ earnings when calculating their

welfare benefit. In a few states, such as Connecticut and California, these earnings disregards

are now as generous or more generous than the SSP earnings supplement. In other states, low

welfare grants limit the income gains from earnings disregards and other policies, such as

time limits, and minimize a family’s ability to take advantage of them by discouraging

families from retaining a proportion of their welfare check as a supplement to their earnings

as they work. Finally, in some cases, a variety of benefits are available to families, but they

are administered from separate programs (like state EITC, food stamps, and child care

subsidies, for example), reducing the likelihood that any one family will take advantage of all

of the benefits available to them.

The Canadian context may affect the behavior of both the program and control groups in

the SSP—and is the background in which the experiment was tested. Canadian and U.S.

economic policies and outcomes have many similarities. The two countries offer similar types

of child care subsidies, and working mothers of young children make similar choices

regarding who cares for their children (Michalopoulos & Robins, 2000). Most single mothers

in both countries now work (Dooley, 1994; Meyer & Rosenbaum, 2000), and women in the

two countries earn similar amounts (Blackburn & Bloom, 1993). Federal governments in both

countries passed legislation that turned welfare from an entitlement program to a federally

funded block grant with considerable local discretion, although benefit levels in most

Canadian provinces tend to be higher than in much of the United States (Michalopoulos &

Robins, 2002). Although many U.S. and Canadian policies are similar, there are at least two

important differences. For example, Canadian safety net programs are more likely to be

universal rather than means-tested (Blank & Hanratty, 1993), and Canada has national health

insurance.

Because the SSP was conducted in both British Columbia and New Brunswick, the

SSP study has a distinct advantage that allows us to investigate how increases in employment

and income affect children in two different economic and cultural contexts. While the

Vancouver, British Columbia area (near Seattle) is an urban location with high average

wages, a high cost of living, and a high number of Asian immigrants, New Brunswick is

more heavily Catholic, more rural, and, as Canada’s only official bilingual province, has a

mix of English- and French-speaking citizens (Lin, Robins, Card, Harknett, & Lui-Gurr,

1998). The SSP has increased employment, earnings, and income by similar amounts in the

two provinces (Michalopoulos et al., 2000). If it has also affected children in similar ways,

this allows for the generalization of the results to a more diverse set of contexts.


2. Method

2.1. Participants

The sample consisted of 3259 families and 5078 children (ages 3–18) who were in the

home at baseline and for whom data on children were collected at the 36-month follow-up.3

This sample included 1654 families and 2582 children in the program group and 1605

families and 2496 children in the control group. There were three age groups: (1)

preschoolers (ages 3–5 at 36 months after random assignment; n = 1043 children); (2)

children in middle childhood (ages 6–11 at 36 months after random assignment, n = 2158

children); and (3) adolescents (ages 12–18 at 36 months after random assignment, n = 1417

children). Baseline demographic information of the parents of children in the program and

control groups, and the full sample of parents of each age group of children are presented in

Table 1.

2.2. Procedures

Data were collected from a variety of sources. A baseline survey administered at the time

of random assignment provided background information on the families. Administrative

records provided information on recipients’ use of the welfare program and receipt of the

earnings supplement. Follow-up surveys at 18 and 36 months after random assignment were

used to gather information on recipients’ employment, earnings, income, hardship, and

expenditures. The 18-month survey was completed by approximately 93% of the research

sample, and the 36-month survey was completed by approximately 90% of the research

sample. There were three primary sources of information on family and child outcomes, all of

which were collected 36 months after random assignment: (1) parent surveys on all children

in the household, (2) language tests conducted with children between 4 and 7 years old at the

time of the 36-month interview and math tests conducted with children between 7 and 15

years old at that time, and (3) surveys conducted with children who were then age 10 and

older.

Notably, differing assessments of children’s outcomes were collected across the differing

child age groups. Measures were collected to assess key developmental constructs at differing

ages—for example, vocabulary acquisition for younger children, externalizing and intern-

alizing behavior problems in preschool and middle childhood children, and delinquency

behavior in adolescence. Across all child assessments, the response rate was 81% of the

eligible child study sample. However, the response rate was much lower for direct assess-

ments with the older children (64%), largely because older children were more likely to be

difficult to contact and more likely to refuse to participate in the study. Response rates were

3 Data were collected on children who were not in the home at baseline but who were there at the 36-month

follow-up. These children were excluded from the analysis sample, since those in the program group had not

received the full ‘‘treatment.’’

Table 1

Selected baseline characteristics of parents, by child age

Characteristic Middle Full sample

Preschool

Children

Childhood

Children

Adolescents Program

group

Control

group

Difference

Gender (%)

Female 99.31 97.88 95.85 97.10 97.38 � 0.03

Age (%)

19–24 50.50 20.74 0.24 23.12 23.60 � 0.05

25–29 28.51 31.00 11.52 23.49 23.91 � 0.42

30–39 19.70 40.99 64.40 40.86 40.57 0.29

40–49 1.29 7.04 21.36 11.32 10.99 0.33

50 or older 0.00 0.22 2.48 1.21 0.94 0.27

Marital status (%)

Never married 71.12 51.85 31.47 51.94 52.52 � 0.59

Divorced, separated,

or widowed

26.31 45.70 66.85 46.13 45.23 0.90

Completed education (%)

Less than high school 51.73 52.18 56.45 1.94 53.83 � 1.90

High school 39.76 36.42 32.03 37.07 34.83 2.24

Some postsecondary 8.51 11.40 11.53 10.90 11.34 � 0.44

In school at baseline 17.43 14.39 9.83 14.76 14.21 0.55

Recent welfare history

Number of months on IA in prior 3 years (%)

10–23 34.12 17.48 17.33 21.10 23.36 � 2.26

24–35 40.45 35.35 30.27 34.70 35.20 � 0.50

All 36 25.42 47.17 52.40 44.20 41.43 2.76

Average IA payment in

prior month (Can$)

871.49 876.10 869.38 854.58 851.41 3.17

Work history and labor force status

Ever had paid job (%) 93.37 93.36 95.45 94.62 93.33 1.29

Average years worked 4.69 6.31 8.70 6.79 6.73 0.06

Labor force status at baseline (%)

Employed 30 h/week 4.77 6.61 8.50 6.57 7.43 � 0.86

Employed < 30 h/week 8.45 13.77 13.79 12.96 13.37 � 0.40

Looking for work 20.08 21.37 22.61 21.24 22.99 � 1.74

Not employed/looking for work 66.70 58.25 55.09 59.22 56.21 3.01y

Activity-limiting conditions (%)

Physical problem 16.83 20.93 30.54 23.28 23.80 � 0.52

Emotional problem 4.67 6.32 9.75 7.92 6.11 1.81*

(continued on next page)


Characteristic Middle Full sample

Preschool

Children

Childhood

Children

Adolescents Program

group

Control

group

Difference

Emotional well-being (%)

At risk for depression 24.95 25.82 28.82 26.24 26.29 � 0.05

Children

Number of children under age 19 (%)

1 50.65 37.62 32.16 48.21 49.37 � 1.17

2 35.12 42.06 40.42 36.13 34.73 1.40

3 or more 14.23 20.32 27.43 15.67 15.89 � 0.23

Ethnic background (%)

First Nations ancestry 9.73 8.12 9.62 8.78 8.32 0.46

Asian ancestry 4.77 5.02 5.45 4.79 5.07 � 0.28

French-speaking 13.66 11.93 13.34 13.91 13.40 0.52

Immigration (%)

Not born in Canada 10.78 12.97 16.08 12.70 13.85 � 1.15

Immigrated last 5 years 3.57 2.07 3.13 2.54 3.25 � 0.71

Period of intake (%)

11/1992–10/1993 35.31 34.86 33.07 34.04 34.08 � 0.04

1/1994–3/1995 64.69 65.14 66.93 65.96 65.92 0.04

Province (%)

British Columbia 48.27 50.11 48.40 48.25 48.91 � 0.66

New Brunswick 51.73 49.89 51.60 51.75 51.09 0.66

Sample size

(total = 3259 families)

1011a 1836a 1252a 1654 1605 3259

Source: Calculations from the baseline survey and IA administrative records.

Sample sizes vary for individual measures because of missing values.a Because children of more than one age group may come from the same family, the total number of families

(3259) across the three age groups is smaller than the sum of the families in each of the age groups of children.y p<.10, two tailed test for differences between the outcomes for the program and control groups.

*p<.05, two tailed test for differences between the outcomes for the program and control groups.

Table 1 (continued)


similar in the program and control groups, providing a measure of confidence in the estimates

of program effects. Extensive nonresponse bias analyses suggested that survey respondents

and nonrespondents had somewhat different family characteristics. However, there was little

evidence that impacts of the program based on information about survey respondents (that

were available for both respondents and nonrespondents to the child surveys like data on

employment, welfare receipt, and income) were different than they would have been if all

families had responded to the survey (Morris & Michalopoulos, 2000).


2.3. Measures

2.3.1. Parental employment, income, and material hardship

2.3.1.1. Employment. Parental reports of employment were obtained from the 18- and 36-

month follow-up surveys. Because some interviews were conducted in the 35th month after

random assignment, complete data are available only for 34 months following random

assignment. In each interview, respondents were asked for earnings, hours worked per week,

and weeks worked per month for each job they had held since the prior interview. In tying the

incentive to full-time (30 hours of work per week), the program was intended to increase full-

time, and not part-time, employment. For this reason, program impacts on full-time

employment only are examined here, in addition to impacts on earnings. A parent was

considered to be working full time in a month if responses indicated working at least 30 hours

per week for at least one week in the month. Average monthly earnings were computed from

reports on hourly wages, hours worked per week, and weeks worked per month.

2.3.1.2. Income. Monthly income from welfare and SSP supplement payments were obtained

from administrative records from the Ministry of Human Resources in British Columbia, and

Human Resources Development—NewBrunswick, and SSP’s ProgramManagement Informa-

tion System (PMIS). Income was computed as the sum of earnings, welfare payments, and SSP

supplement payments. All dollar amounts reported are in Canadian dollars.

2.3.1.3. Material hardship. The survey included information on several aspects of material

hardship: (1) Food hardship: two questions addressed parents’ difficulties with obtaining food

in the last 3 months: (a) whether the parent had ever used a food bank to get food and (b)

whether there was a time the parent could not afford food; (2) Household/structural problems:

three questions addressing (a) whether the gas/hydro had been turned off in the last 3 months

because of an inability to pay the bills, (b) whether the parent currently had any structural

problems with the home (e.g., leaky roofs and broken windows), and (c) whether the parent

currently had any problems with things not working in their home (e.g., appliances, heating,

electricity, and plumbing); (3) Health care problems included two questions about whether

there was a time in the last 6 months when (a) a member of the family needed to go to the

doctor but could not afford to go, and when (b) a member of the family needed medication

but could not afford it.

2.3.2. Children’s cognitive performance

Two tests were administered to children, but the tests varied depending on the child’s age.

For both age groups, well-established measures available in both official Canadian languages

(English and French) were administered.4 For younger children, it was believed that changes

in income and employment might be reflected in children’s early vocabulary, and therefore

4 All surveys administered in Canada must be available in both English and French, limiting the choice of tests

available to administer to children.


these children were administered a receptive vocabulary test. Older school-age children were

assessed with regard to their math skills, as no reading measure was available in both English

and French for administration to the school-age children (and it was believed that vocabulary

acquisition may be less sensitive to changes in employment and income than reading and

math skills for this age group). In addition, parents and children responded to several

questions about the child’s school performance.

2.3.2.1. PPVT-R score. Children ages 4–7 years were administered the Peabody Picture

Vocabulary Test—Revised (PPVT-R), a test of children’s receptive language ability.5 Raw

scores were converted to standard scores according to standards provided by the American

Guidance Service.

2.3.2.2. Math score. Amath skills test was administered to children ages 7–14 in Grades 2–7.

The test, which varied by the child’s grade level, consisted of 26 math problems for those in the

second grade and 34 items for those in Grades 3 and above. The test included a subset of items

from the Canadian Achievement Tests, Second Edition (CAT/2), a mathematics test developed

by the Canadian Test Centre that is administered annually in all provinces to approximately

300,000 students from Grades 2 up to the end of secondary school and college. The proportion

of correct items completed out of the total number of test items was computed for each child.

2.3.2.3. Academic achievement. Children’s functioning in school was assessed by means of

parent and child surveys for children of school age (6–18 years). Parents were asked about

children’s functioning in reading, writing, and math (for 6–15 year olds) or writing, math,

and science (for 15–18 year olds). Children ages 10–18 were asked about English, math, and

science. For each subject, functioning was ranked on a 5-point scale ranging from ‘‘not very

well at all’’ to ‘‘very well.’’ A measure of children’s average achievement was computed as

the average score across the three academic subjects. In addition, a dichotomous measure was

computed indicating whether the child scored below 3 (‘‘average’’) on any one of the

academic subjects to assess distributional as well as average effects, i.e., to determine if

average effects masked movement at the lower, more critical, portion of the distribution (cf.,

Gennetian & Miller, 2002).

2.3.3. Children’s behavior and emotional well-being

2.3.3.1. Behavior problems.Parents reported on children’s behavior for children up to age 11

using a scale developed for use in the National Longitudinal Survey of Children and Youth in

5 French-speaking children were administered the Echelle de vocabulaire en Images Peabody (EVIP), a test

comparable to the PPVT. Scores on the EVIP are not equivalent to scores on the PPVT (Marta Fahrenz, American

Guidance Service, personal communication, May 1999); therefore, the French- and English-speaking children

were analyzed separately. Unfortunately, the sample of children who chose to answer the test in French was too

small to allow an analysis of the impact of the SSP on children’s functioning on the EVIP. A total of 214 children

took the EVIP. These children were part of both the young preschool and middle childhood age groups.


Canada (NLSCY). The NLSCY is a unique survey of Canadians from birth to adulthood, and

the measure of behavior problems has items similar to the Behavior Problems Index (a 28-

item scale; Peterson & Zill, 1986) used in many U.S. studies of the effects of welfare reform

programs (Morris et al., 2001). A slightly different set of items was asked of parents of 3- to

5-year-old children than of parents of children ages 6–11, but internal reliability was high for

both age groups (a = .87 for the 21 items for the 3–5 year olds and a = .92 for 27 items for

the 6–11 year olds). An average score was computed across the items (which were coded on

a 3-point scale ranging from ‘‘never/not true’’ to ‘‘often/very true’’). Items included both

internalizing and externalizing aspects of children’s behavior. Sample items included, ‘‘My

child is too fearful or anxious,’’ ‘‘My child cries a lot,’’ ‘‘My child steals at home,’’ and ‘‘My

child gets into many fights.’’

2.3.3.2. Positive social behavior. Parents reported on children’s positive social behavior for

children up to age 11 using a 5-item scale developed for the NLSCY that are similar to items

in the Positive Behavior Scale used in other welfare demonstration studies (Polit, 1996). As

with the behavior problems scale, a slightly different set of items was asked of the parents for

their 3–5 year old children than for their 6–11 year old children, but internal reliability was

good for scales for both age groups (a = .75 for the 3–5 year olds and a = .80 for the 6–11

year olds). Sample items include: ‘‘My child tries to help someone who is hurt’’ and ‘‘My

child comforts a crying child.’’ Scores on the total positive social behavior scale ranged from

1 (never) to 3 (often). The average score across the five items tapping children’s social

interactions with peers was computed.

2.3.3.3. School behavior problems. Parents of children in school (ages 6–18) were asked how

often in the past school year they were contacted by the school about their children’s behavior

problems in school. This item has been used in a number of recent welfare and work

demonstration studies (e.g., Gennetian & Miller, 2002). Scores ranged from 1 (none or one

time) to 3 (four or more times).

2.3.3.4. Frequency of delinquent activity. Children 10–18 years old were asked about the

frequency of a number of delinquent acts, using a subset of items from LeBlanc and

Tremblay (1988) used in the NLSCY. Example items included ‘‘Did you skip a day of

school without permission?’’ ‘‘Did you get drunk?’’ ‘‘Did you run away from home?’’

Items referred to the last 6 months, except for one item that asked about skipping school

in the last month. Children ages 10–14 were asked seven items (a = 0.78), while children

ages 15–18 were asked about a wider range of delinquent activity across 14 times,

including theft, starting fires, and using weapons (a = .72). Each was scored on a scale

of 1 (never) to 4 (5 or more times). The average frequency across seven items for the

10- to 14-year-olds and 14 items for the 15- to 18-years-old was computed for each

child.

2.3.3.5. Substance use. Children ages 12–18 were asked whether they currently smoked,

drank alcohol weekly, and had used drugs in the last 6 months (not including inhalants). Each


measure was binary (yes or no), and children who had engaged in the activity were scored as

100, while children who had not were scored as 0.

2.3.3.6. At risk for depression. Children ages 15–18 were asked the number of days in the

past week they had experienced each of 11 depressive symptoms taken from the Center for

Epidemiological Studies Depression (CES-D) scale (a = .80; Radloff, 1977). Given the high

internal reliability of the original scale, this shortened version was believed to provide a close

approximation of scores on the total scale. The items were scored on a scale ranging from 0

(rarely/never) to 3 (5–7 days) and were summed to obtain a total score, as in the original

development of the scale (Radloff, 1977). However, one item was deleted because it was

negatively correlated with the other items on the scale, leaving a total scale of 10 items.6

Because this scale has been validated as a dichotomous and not a continuous measure

(Radloff, 1977), a clinical cutoff was used. Previous research has identified a threshold (a

score of 16 out of 60 using the original 20 items) at or above which scores may be indicative

of clinical depression. A corresponding threshold (a score of 8 out of 30) was used for this

smaller subset of items. This scale demonstrated good internal validity, with a = .80.

2.3.4. Children’s health

2.3.4.1. General health. Four items from the NLSCY addressed parental reports of children’s

health for all children in the sample (ages 3–18). Sample items included: ‘‘He doesn’t get sick

often’’ and ‘‘She seems less healthy than other children you know.’’ Items were asked on a 5-

point scale ranging from 1 (false) to 5 (true), with 3 being ‘‘sometimes false/sometimes true.’’

Items were rescored such that high scores indicated better health. An average score across the

four items was constructed to measure children’s general health. This scale had a moderate

internal reliability, with a = .62.

2.3.4.2. Long-term health problems. Parents were asked to report whether their children had

any long-term health problems that had lasted or would last 6 months or more, that had been

diagnosed by a health professional, and that limited the child’s participation in activities.

These problems included allergies, asthma, bronchitis, kidney problems, heart diseases,

epilepsy, cerebral palsy, mental handicap, and learning and emotional problems.

2.3.5. Parental emotional well-being

2.3.5.1. At risk for depression. As with the adolescent children, parents were asked about the

number of days in the last week they had experienced each of 11 depressive symptoms from

6 While children were asked about 11 depressive symptoms, one item detracted from the internal consistency

of the scale. The item ‘‘Everything I did was an effort’’ was negatively correlated with the total score and the other

items. It was suspected that some children interpreted this item to mean, ‘‘I put effort into everything I do’’ and

thus scored it differently than would be expected. Therefore, only 10 items were included in the score of

depression.


the widely CES-D scale (Radloff, 1977). Again, the high internal reliability of the original

scale made this reduced number of items appear sufficient to assess depressive symptoms.

The items were scored on a scale ranging from 0 (rarely/never) to 3 (5–7 days) and summed

to obtain a total score (a = .83), as in the original scoring (Radloff, 1977). As discussed

earlier, previous research has identified a threshold (a score of 16 out of 60 on the 20-item

scale) at or above which scores may be indicative of clinical depression. A corresponding

threshold of 9 (out of 33) was computed to measure parental risk for depression for this

smaller set of 11 items.

2.3.5.2. Parenting behavior. Parenting behavior was assessed via both parental and child

reports using scales developed for use in the NLSCY. Parents were asked one set of questions

about each of their 3- to 14-year-old children, and a different set of questions about each of

their 15- to 18-year-old children, to address the differing parenting challenges of younger and

older children. Children who were 12–18 years old were asked to assess their parents’

behavior using the set of questions asked of parents of 15–18 year old children. All items

were answered on a 5-point scale ranging from ‘‘never’’ to ‘‘many times each day.’’

For questions asked of parents of children ages 3–14, a principal components factor

analysis with varimax rotation suggested that the items tapped three distinct factors: (1) warm

parenting (a = .77), (2) negative parenting (a = .80), and (3) consistent discipline (a = .65;

see Table 2). For the items reported by parents of 15- to 18-year-olds and for the 12- to 18-

year-old child report items, two distinct factors: (1) authoritative parenting (a combination of

warmth, democratic, and limit-setting parenting; Baumrind, 1971; Maccoby & Martin, 1983;

a = .86 for the parent report measure and a = .88 for the child report measure) and (2)

negative parenting (a = .77 for the parent report measure and a = .68 for the child report

measure; see Table 3). Items assessing parental monitoring were also administered but the

internal consistency of the scale was so low (a = .36 for the parent report monitoring scale

and .47 for the child report monitoring scale) that it was not used in later analyses. In a small

number of cases, items double-loaded on the factors. In general, items were considered as part

of the factor on which the item more highly loaded. In a few rare cases, items that loaded on a

particular factor were not included as part of the scale for theoretical reasons. For each of the

scales, total scores were computed by averaging the items on the scale.

2.3.6. Child care and children’s activities

2.3.6.1. Child care use. Parents were asked in the 36-month survey about the regular child

care arrangements they had used for the youngest child in the household over the past 18

months. Child care arrangements include before- and after-school programs, day care centers,

care by a relative in or out of the child’s home, and care by a nonrelative in or out of the

child’s home. Parents could indicate that their children had multiple care arrangements.

Separate variables were constructed for children’s participation in (1) formal care (that

included center care and before and after-school programs) and (2) informal care arrange-

ments (that included child care by a relative or nonrelative in or out of the child’s home).

Also, parents were asked the number of times the primary child care arrangement had

Table 2

Factor analysis results of parenting by parents of children aged 3–14 years

Item Warm

parenting

Negative

parenting

Consistent

discipline

Praise child .72

Talk/play together .76

Laugh together .73

Do something special together .67

Play sports together .66

Proportion of time praise child .49

Get annoyed when child disobeys .69

Tell child he is not as good as others .42

Proportion of time express disapproval .56

Get angry when punish child .60

Punishments depend on mood .43

Have trouble managing child .74

Punish for the same things .72

How well gotten along in past 12 monthsa .68

Child ignores punishments .57 � .44

Child gets out of punishments .35 � .65

Child gets away with things .33 � .67

Make sure child obeys .64

Punish when child does not stop behavior .77

Alpha for scale .77 .80 .65

Source: Calculations from the 36-month follow-up parent survey.

Only factor items with loadings greater than or equal to |.30| are shown. Bolded factor loadings indicate items that

were used to create the respective scales. Except as otherwise noted, items were included on the factors on which

they most highly loaded. Responses for items ranged from 1 (never) to 5 (all the time). Exceptions are otherwise

noted.a Responses for this item ranged from 1 (very well, no problems) to 5 (not well at all, constant problems).


changed in the last 6 months (excluding periods of care by themselves), which provided a

measure of frequent child care changes for children who changed child care arrangements two

or more times in the last 6 months.

2.3.6.2. Children’s after-school activities. Parents of children ages 6–14 were asked about

their children’s participation in activities in the past year. Children ages 12–18 were also

asked about their own participation in activities. Four different activities were addressed: (1)

sports involving teaching or instruction (apart from physical education in school); (2) sports

without a coach or instructor; (3) lessons in music, art, or other nonsport activities (outside of

school); and (4) clubs, groups, or community programs with adult leadership. Responses were

scored on a 1–4 scale ranging from ‘‘about every day’’ to ‘‘rarely.’’

2.3.6.3. Frequency of doing household chores.Children ages 12–18 were asked the frequency

with which they did regular family chores or carried out responsibilities, such as doing

housework and caring for younger siblings. Responses ranged from 1 (never) to 5 (about every

day).

Table 3

Parenting factor analysis results of parenting for children aged 12–18 years

Parent behavior Parents of children 15–18 years old Children 12–18 years old

Authoritative Negative Monitoringa Authoritative Negative Monitoringa

Smile at child .73 .74

Praise child .76 .70

Listen to child .65 .76

Solve problems with child .56 .70

Show appreciate child .77 .77

Speak well of what

child does

.79 .78

Proud of what child does .76 .77

Forget rules .50 .51

Nag about small things .65 .67

Keep rules when

suits myself

.41 .47

Get angry and yell .66 .63

Threaten punishment .71 .62

Punish depending

on mood

.61 .66

Hit or threaten to hit .52 .45

Know what child

is doing

.39 .44 .44 � .35

Tell child when to

be home

.69 .34 � .59

Find out misbehaviorb .40 .26 .33 � .45

Let child go out � .62 .77

Repeated punishmentc .58 .38 N/A N/A N/A

Get along with childc,d .47 N/A N/A N/A

Alpha for scale .86 .77 .36 .88 .68 .47

Source: Calculations from the 36-month follow-up parent survey and the 36-month follow-up child survey.

Only factor items with loadings greater than or equal to |.30| are shown. Bolded factor loadings indicate items used

to create the respective scales. Items were included on the factors on which they most highly loaded. Responses

for items ranged from 1 (never) to 5 (very often) unless noted otherwise.a Monitoring scale was not internally consistent and was dropped from analyses.b Item not included on the negative parenting scale to be consistent with a priori theory.c Items used only in the 36-month follow-up parent survey; not included in the child surveys.d Responses ranged from 1 (very well, no problems) to 5 (not well at all, constant problems).


2.3.6.4. Employment.Children ages 15–18were askedwhether they were working for pay and,

if so, the number of hours per week. These items were used to determine the proportion of

children who are working, and the proportion of children workingmore than 20 hours per week.

2.4. Analysis strategy

Impacts of the program were estimated as the difference in average outcomes for the

program and control groups. Because families were randomly assigned to program and control


groups, any difference between the two groups can be reliably attributed to assignment to the

program. Sample attrition did not make the program and control groups dissimilar on baseline

characteristics (see Table 1).7 The standard errors on the measures analyzed at the level of the

child (rather than the family) were adjusted for the shared variance between siblings in the

same family, using the standard errors developed by White (1980). This adjustment allows for

nonindependence among sample members that may arise for any reason, including that which

arises because children come from the same family, neighborhood, or province. Two-tailed

tests with an alpha of .10 were used to test for the significance of program impacts. This alpha

is equivalent to a one-tailed test at p< .05, which is appropriate for assessing program effects.

This is the procedure used in previous studies of the effects of welfare and employment

programs on children (Gennetian & Miller, 2002; Huston et al., 2001).

Tables of the effects of the program contain the unadjusted means in the program and

control groups, the difference between the mean levels in the two groups, and the effect size of

the difference. Effect sizes were computed by dividing the difference between the program and

control groups by the standard deviation in the control group under consideration. As indicated

earlier, because the experimental design of the study is only maintained by examining each

outcome in a separate analysis of treatment effects, mediating analyses are not conducted here.

3. Results

3.1. Receipt of the supplement and impacts on employment, Income Assistance (IA) receipt,

and income8

As shown in Table 4, just over one-third of the program group members found full-time

employment during the year after random assignment and received at least one supplement

payment. By 12 months after random assignment, the SSP had doubled full-time employ-

ment, from about 14% of the control group to about 28% of the program group (for a

difference of 14 percentage points between program and control groups; p< .01). In the

second half of the follow-up period, the difference between the program and control group

levels on full-time employment diminished somewhat to about 10 percentage points, but was

still significant (p< .01). This reduction in program impact was largely due to increases in

employment by the control group, rather than declines in the employment rates of the

program group. Over the 36-month follow-up period, 51.6% of parents in the program group

7 Analyses were conducted controlling for a small set of baseline characteristics in estimating the effects of the

program. The impact estimates and standard error of the estimates did not change appreciably; therefore, the

unadjusted impact estimates are presented here.8 For a greater discussion of the effects of the SSP on adults’ economic outcomes, see Michalopoulos et al.

(2000). Results presented in Table 4 differ from those presented by Michalopoulos et al., because some sample

members did not complete any of the child components and therefore were not included in the tabulations shown in

Table 4.


worked full-time at some point, compared to 38.8% of parents in the control group, a

significant effect (p < .01).

Combined earnings and supplement payments produced much higher income on average

for families in the program group than families in the control group. As is shown in Table 4,

parents in the program group earned about Can$90 more per month than parents in the

control group (p< .01). Although they received about Can$80 per month less in income

from IA (p< .01), they received Can$150 per month in SSP supplement payments (p< .01).

The program impacts on the percentage of parents employed full time and income persisted

into the 6-month period prior to the interview (see Table 4 for impacts on employment and

income 6 months prior to interview for total sample; differences between program groups,

ps < .01).

In terms of program impacts on hardship, the SSP significantly decreased the proportion

of families using food banks or reporting that they could not afford food at the time of the

36-month interview (p < .05). However, there were no significant differences in the

proportion of families reporting household or structural problems or problems with health

care.

Only minor differences in impacts on adult economic outcomes were found for parents of

the three age groups of children. For parents of children in all three groups, about one-third of

families in the SSP group took up the supplement. And for all three groups, the SSP

significantly increased employment, earnings, and income (p < .01).

3.2. Impacts on children9

3.2.1. Impacts for preschool children

Impacts on younger children’s developmental outcomes are presented in Table 5. Children

in the control group had an average score of 91 on the PPVT-R test, which corresponds to a

percentile score of 27. There were no significant program impacts on PPVT-R scores for these

children.

Parents were asked about their children’s social behavior, emotional well-being, and health.

As can be seen in Table 5, there were no significant impacts in any of these areas. Analyses

were conducted to examine the impact of the SSP separately on internalizing and externalizing

dimensions of behavior. No significant effects of the program were found for either of these

two subscales of children’s behavior.

9 Given previous research on the differing effects of welfare and employment programs on boys and girls

(Huston et al., 2001), analyses were conducted to examine whether there were any significant differences in

impacts on child outcomes for boys and girls. While in some cases effects on girls were more pronounced than for

boys, differences in impacts were generally not statistically significant. Therefore, differences by gender were not

discussed in this paper. Differences in impacts for each of the three age groups was also examined for the two

provinces examined in this study: New Brunswick and British Columbia. Differences in impacts for the two

provinces were extremely rare, suggesting that the program had a similar effect on children’s outcomes in the two

different contexts.

Table 4

SSP summary of impacts on economic outcomes for families over the 36-month follow-up period

All families Families with preschool children

Program

group

Control

group

Difference

(impact)

Effect

size

Program

group

Control

group

Difference

(impact)

Effect

size

Received SSP

supplement (%)

35.91 36.11

Full-time employed,

Month 6

17.78 11.03 6.75** 0.22

Full-time employed,

Month 12

27.80 13.5 14.2** 0.42

Full-time employed,

Month 34a28.48 18.44 10.04** 0.26

Employment, earnings, and income, Months 1 to 34a

Ever full-time

employedb (%)

51.57 38.75 12.82** 0.26 53.14 36.49 16.66** 0.35

Monthly earnings

(Can$)

310.44 219.38 91.06** 0.23 298.78 181.78 117.01** 0.33

Monthly IA

income (Can$)

645.43 726.14 � 80.70** � 0.24 666.99 741.17 � 74.18** � 0.22

Monthly SSP

income (Can$)

152.14 0.00 152.14** 0.00 146.00 0.00 146.00** 0.00

Total income (Can$) 1113.22 957.33 155.89** 0.43 1119.82 928.84 190.98** 0.55

Employment and income 6 months prior to interview

Full-time employed (%) 33.98 23.86 10.12** 0.24

Monthly pre-tax

income (Can$)

1619.53 1443.03 176.50** 0.21

Expenditures and hardship, at 36 months

Food bank/not

afford (%)

35.27 40.76 � 5.49** � 0.11

Household

problems (%)

20.87 22.90 � 2.02 � 0.05

Health care

problems (%)

31.82 33.06 � 1.24 � 0.03

Sample size 1654 1605 3259 493 518 1011

Families with middle childhood children Families with adolescent children

Program

group

Control

group

Difference

(impact)

Effect

size

Program

group

Control

group

Difference

(impact)

Effect

size

Received SSP

supplement (%)

36.85 34.47


Ever full-time

employedb (%)

52.93 39.35 13.58** 0.28 48.07 35.87 12.20** 0.25


Families with middle childhood children Families with adolescent children

Program

group

Control

group

Difference

(impact)

Effect

size

Program

group

Control

group

Difference

(impact)

Effect

size


Monthly earnings

(Can$)

320.15 219.38 100.77** 0.24 296.78 218.94 77.84** 0.19

Monthly IA

income (Can$)

650.80 745.41 � 94.61** � 0.27 1103.48 741.53 � 82.71** � 0.23

Monthly SSP

income (Can$)

158.37 0.00 158.37** 0.00 144.12 0.00 144.12** 0.00

Total income (Can$) 1133.97 974.85 159.12** 0.41 1103.48 979.44 124.04** 0.35

Sample size 939 897 1836 647 605 1252

Source: Calculations from the baseline survey, IA administrative records, the 18-month follow-up core survey, and

the 36-month follow-up core survey.

Rounding might cause slight discrepancies in sums and differences. Sample sizes may vary for individual items

because of missing values.a Although information on employment comes from the 36-month follow-up core survey, some sample

members were interviewed as early as month 35, so that the valid information on employment and earnings is

available through month 34 only. Therefore, results related to employment and earnings are shown only through

34 months.b Full-time employment is defined as working 30 hours or more per week in at least one week during the

month.

**p < .01 for program versus control group.

Table 4 (continued)

Table 5

SSP impacts on child outcomes for the preschoolers at the 36-month follow-up

Outcome Program group Control group Difference (impact)a Effect size

Cognitive functioning

PPVT-R score 92.18 91.32 0.86 0.05

Sample size 403 425

Behavior and emotional well-being

Behavior problems 1.48 1.48 0.00 0.01

Positive social behavior 2.51 2.53 � 0.03 � 0.06

Sample size 497 535

Health

Average health 4.01 4.05 � 0.04 � 0.05

Any long-term problems (%) 25.60 27.43 � 1.83 � 0.04

Sample size 503 540

Source: Calculations from the baseline survey, the 36-month follow-up parent survey, and the PPVT-R.

Rounding may cause slight discrepancies in sums and differences. Preschool children were ages 3–5 at the 36-

month follow-up. Standard errors were adjusted to account for shared variance between siblings.a Differences were tested but all were nonsignificant.



3.2.2. Impacts for middle childhood children

For the middle childhood children, effects on cognitive outcomes were consistent across

parents’ reports and tests, but small (see Table 6). Children in the program group scored

significantly higher on a math test than their peers in the control group (with an effect

size = 0.14), and mothers in the program group rated their children higher on average

academic performance in school than did mothers in the control group (effect size = 0.11).

There were no significant program impacts on any of the measures of children’s behavior

(see Table 6). Analyses were conducted to examine the impact of the SSP separately on

internalizing and externalizing dimensions of behavior. No significant effects of the program

were found for either of these two subscales of children’s behavior. However, according to

parents’ reports, children in the program group were in better health (effect size = 0.11) and

were slightly less likely to have long-term health problems than children in the control group

(effect size =� 0.09). For example, 37% of children in the control group were reported to

have long-term health problems compared to 32.4% of children in the program group.

3.2.3. Impacts for adolescents

Table 7 presents the impacts on outcomes for the adolescents. Small significant unfavor-

able effects for the older children were found in two of the four measures of children’s school

functioning. Mothers in the program group reported significantly lower average school

Table 6

SSP impacts on child outcomes for the middle childhood children at the 36-month follow-up

Outcome Program group Control group Difference (impact) Effect size

Cognitive functioning

PPVT-R score (ages 6–7) 93.21 90.78 2.43 0.13

Sample size 293 292

Math score (ages 7–11) 0.56 0.52 0.04* 0.14

Sample size 699 622

Average achievement 3.71 3.61 0.10* 0.11

Below-average, any subject (%) 22.84 25.65 � 2.81 � 0.06

Sample size 1015 982


Behavior problems 1.42 1.43 � 0.01 � 0.03

Positive social behavior 2.58 2.59 � 0.01 � 0.02


Health

Average health 4.11 4.02 0.09* 0.11

Any long-term problems (%) 32.43 36.98 � 4.55* � 0.09


Source: Calculations from the baseline survey, the 36-month follow-up parent survey, the PPVT-R, the math skills

test.

Middle childhood children were ages 6–11 at the 36-month follow-up. Standard errors were adjusted to account

for shared variance between siblings.

*p < .05, two tailed test for differences between the outcomes for the program and control groups.


achievement for their children than did mothers in the control group, although there were no

significant program–control differences in the proportion of parents reporting that their

children were performing below average in school (see upper portion of Table 7). Nearly 19%

of children in the program group said they were below average in at least one subject in

school, compared to about 14% of the control group (p< .05). On the other hand, there were

no significant differences between the two groups on self-reported average achievement or on

the math skills test, the one objective measure of their academic performance.

Results on older children’s problem behaviors were more consistent with expectations.

While there were no differences in adolescents’ risk of depression, the SSP significantly

Table 7

SSP impacts on child outcomes for adolescents at the 36-month follow-up

Outcome Program group Control group Difference (impact) Effect size

Cognitive/academic functioning

Math score (ages 12–14) 0.45 0.46 � 0.01 � 0.03

Sample size 280 281

Average achievement 3.43 3.54 � 0.11y � 0.11

Below-average, any subject (%) 32.61 32.39 0.22 0.00

Sample size 726 673

Average achievement 3.50 3.57 � 0.07 � 0.09

Below-average, any subject (%) 18.91 14.26 4.65* 0.13

Sample size 512 470


Parental report

School behavior problems 1.40 1.34 0.06y 0.09

Sample size 740 677

Adolescent report

Delinquent activity, 12–14 years 1.35 1.38 � 0.03 � 0.06

Delinquent activity, 15–18 years 1.40 1.34 0.07* 0.21

Smoking (%) 26.52 22.13 4.39y 0.11

Drinks weekly or more (%) 8.91 4.65 4.27** 0.20

Drug use (%) 18.63 14.34 4.29y 0.12

Risk for depression, 15–18 years (%) 45.74 47.14 � 1.39 � 0.03

Sample size 568 509

Health

Average health 4.10 4.13 � 0.04 � 0.05

Any long-term problems (%) 38.99 38.11 0.88 0.02

Sample size 576 530

Source: Calculations from the baseline survey, the 36-month follow-up parent survey, the math skills test, and the

36-month follow-up child survey.

Adolescents were ages 12–18 at the 36-month follow-up. Rounding may cause slight discrepancies in sums and

differences. Standard errors were adjusted to account for shared variance between siblings.y p < .10, two-tailed tests for differences between the outcomes for the program and control groups.

* p < .05, two-tailed tests for differences between the outcomes for the program and control groups.

* * p < .01, two-tailed tests for differences between the outcomes for the program and control groups.


increased the use of tobacco, alcohol, and drugs by about 4 percentage points (for smoking,

p < .10; for weekly alcohol use, p< .01; for drug use, p< .10, all based on two-tailed tests).

The SSP also increased the frequency of delinquent activity for older children (ages 15–18;

p < .05) but not for their younger peers (ages 12–14). Post hoc analyses conducted at the item

level suggested that the SSP significantly increased involvement in minor delinquent activity,

such as staying out late or all night, but not major delinquent activity, such as stealing,

carrying weapons, and involvement with police. There were no significant differences

between program and control groups on measures of older children’s health.

3.3. Impacts on child care and children’s activities

Findings on child care and after-school activities are presented in Table 8. Recall that the

36-month follow-up interview collected information on child care arrangements for only the

youngest child in the family for the period between the 18-month and 36-month interviews.

Therefore, results reflect impacts of the program for children in the three age groups who

were also the youngest children in the family. Information on children’s activities, on the

other hand, was gathered for all children in the household who were age 6 and older, so all

children were included in the analyses on these measures.

As indicated in Table 8, mothers of the preschool and middle childhood children in the

program group reported greater use of child care than mothers in the control group. For the

preschoolers, the SSP significantly, albeit modestly, increased both formal care (p< .01)

and informal (home-based) child care arrangements (p< .05). However, the SSP also

slightly increased the instability of child care (p< .10). For children in middle childhood,

mothers in the program group reported slightly more use of informal child care (p < .10)

and after-school activities (p < .10) than did mothers in the control group. However, there

were no program–control differences in the proportion of children having used formal care

arrangements.

For the adolescents, there were no significant differences between the program and control

groups in children’s after-school activities, as reported by both parents and adolescents, or in

Notes to Table 8:

Source: Calculations from the 18-month follow-up core survey, the 36-month follow-up core survey, and the 36-

month follow-up parent and child surveys.

Rounding may cause slight discrepancies in sums and differences. Sample sizes may vary for individual items

because of missing values.a The child care participation data are for children at the 36-month follow-up who were also the youngest

children in the family. All child care participation data are for the 18 months prior to the 36-month follow-up

interview.b These measures were assessed separately for each child in the family. Standard errors were adjusted to

account for shared variance between siblings.y p < .10, two-tailed tests applied to differences between the outcomes for the program and control groups.

* p < .05, two-tailed tests applied to differences between the outcomes for the program and control groups.

** p < .01, two-tailed tests applied to differences between the outcomes for the program and control groups.


child care arrangements. However, the SSP increased children’s involvement in household

chores (p< .10). While there was no significant program impact on the proportion of

adolescents working for pay at the time of the 36 month follow-up interview, the SSP

Table 8

SSP impacts on child care and children’s activities

Preschoolers Middle childhood children

Program

group

Control

group

Difference

(impact)

Effect

size

Program

group

Control

group

Difference

(impact)

Effect

size

Child care for youngest childa

Any child care (%) 62.70 50.94 11.76** 0.23 45.43 39.97 5.46y 0.11

Any center care (%) 29.46 21.56 7.90* 0.19 8.52 7.03 1.49 0.06

Any informal child

care (%)

43.51 36.12 7.40* 0.15 36.91 31.73 5.18y 0.11

Changed care 2+

times (%)

5.95 3.24 2.71y 0.15 3.31 2.06 1.25 0.09

Sample size 370 371 634 583

Children’s activities (parental report)b

Any activity at least

once per week (%)

96.95 95.42 1.53y 0.07


Adolescents

Program

group

Control

group

Difference

(impact)

Effect

size

Child care for youngest childa

Any informal child

care (%)

3.69 5.51 � 1.81 � 0.08

Sample size 379 345

Children’s activities (parental report) (ages 12–14)b


once per week

91.53 92.35 � 0.82 � 0.03

Sample size 365 353

Children’s activities (adolescent report)b


once per week

86.29 85.69 0.61 0.02

Frequency household

chores

4.03 3.92 0.11y 0.10

Sample size 568 510

Employment (adolescent report) (ages 15–18)b

Working (%) 35.49 35.17 0.33 0.01

Work 20 or more

hours per week (%)

15.04 8.13 6.90* 0.25

Sample size 292 236

Table 9

SSP impacts on impacts on family functioning

Preschoolers Middle childhood children

Program

group

Control

group

Difference

(impact)

Effect

size

Program

group

Control

group

Difference

(impact)

Effect

size

Parental functioning

At risk for depression (%) 33.18 36.76 � 3.58 � 0.07 42.08 36.82 5.26* 0.11

Sample size 435 461 819 795

Parenting behavior (parental report)a

Warm parentingb 3.94 4.00 � 0.06y � 0.12 3.63 3.64 � 0.01 � 0.02

Negative parentingb 2.14 2.15 � 0.01 � 0.02 2.10 2.09 0.01 0.01

Consistent disciplineb 3.88 3.85 0.03 0.04 3.81 3.83 � 0.02 � 0.03

Sample size 497 527 1119 1049

Adolescents

Program

group

Control

group

Difference

(impact)

Effect

size

Parental functioning

At risk for depression (%) 44.62 43.86 0.77 0.02

Sample size 573 539

Parenting behavior (parental report)a

Warm parenting,

12–14 yearsb3.42 3.39 0.03 0.04

Negative parenting,

12–14 yearsb2.08 2.07 0.01 0.02

Consistent parenting,

12–14 yearsb3.67 3.67 0.00 0.00

Authoritative parenting,

15–18 yearsb4.05 4.09 � 0.04 � 0.06

Negative parenting,

15–18 yearsb2.24 2.16 0.08y 0.14

Sample size 382 360

Parenting behavior (adolescent report)a

Authoritative parentingb 3.74 3.73 0.01 0.01

Negative/inconsistentb 2.34 2.39 � 0.05 � 0.07

Sample size 550 494

Source: Calculations from the 36-month follow-up parent survey and the 36-month follow-up child survey.

Rounding may cause slight discrepancies in sums and differences. Sample sizes may vary for individual items

because of missing values.a These measures were assessed separately for each child in the family. Standard errors were adjusted to

account for shared variance between siblings.b These items are rated on a scale of 1 (never) to 5 (many times each day).y p < .10, based on two-tailed tests applied to differences between the outcomes for the program and control

groups.

* p < .05, based on two-tailed tests applied to differences between the outcomes for the program and control

groups.



increased the proportion of children working over 20 hours per week by 7 percentage points

(p< .05).

3.4. Impacts on parental emotional well-being and parenting behavior

Impacts on parental depression and parenting behavior for the three age groups of children

are presented in Table 9. For the preschoolers, approximately one-third of parents in both

program and control groups reported depressive symptoms that put them at risk for clinical

depression. Impacts on reported parenting behaviors were not common for parents of children

in this age group. Program group parents reported slightly less warm parenting than control

group parents (p < .10), but there were no significant differences on measures of negative

parenting and consistent discipline.

For middle childhood participants, the SSP increased the proportion of parents at risk for

depression by 5 percentage points (p < .05). Despite the small increase in parental depression

due to the SSP, the program did not significantly affect parental behavior as reported by

parents.

For parents of adolescents, program and control groups did not differ in their risk of

depression. In terms of reports of parenting behavior, program impacts were extremely rare.

The only difference that was statistically significant indicated more negative parenting among

program group than control group parents, as reported by parents of 15- to 18-year-old

children (p < .10). According to all the other parental and child report measures, however, the

SSP did not significantly influence parenting behavior among parents of adolescents.

4. Discussion

Although the SSP increased full-time employment and income for parents by a similar

amount for all three age groups of children, the pattern of significant effects of the SSP

on children varied considerably with the children’s age group.10 These findings suggest

that theoretical models of the effects of increases in maternal employment and family

income on children should include the moderating role of the developmental period of the

child. Moreover, the findings suggest that positive effects of increases in income and

employment appear to affect children through changes in the resources parents can

purchase for their children in terms of out-of-home environments rather than through

10 Post hoc analyses were conducted to address the extent to which differences in the effects of the programs

were due to differences in the background characteristics of families rather than the developmental challenges of

their children, as families with younger children have different background characteristics than families with older

children. While these analyses were hampered by the fact that different measures were collected for each of the

three age groups of children, they are suggestive. Using standardized child outcome measures created for each

domain of child functioning (cognitive/academic; behavior; health), multivariate analyses suggested that

developmental differences and not differences in family background characteristics accounted for the differences

in impacts between the age groups. Further information on this analysis is available from the authors.


changes in parenting and psychological distress. These findings are discussed in greater

detail below.

4.1. Youngest children

The lack of program impacts on the cognitive, behavioral, or health outcomes of very

young children suggests a model of the offsetting effects of employment and income on this

age group. Given the strength of prior work on the positive effects of income for very young

children (Duncan & Brooks-Gunn, 1997), it seems most likely that the effects of employment

may indeed have been negative and offset by the benefits of income—resulting in neutral

effects overall for the youngest children. Some research has suggested negative effects of

maternal employment occur with increases in employment during the child’s first year of life

(Baydar & Brooks-Gunn, 1991; Belsky & Rovine, 1988; Bronfenbrenner & Crouter, 1982;

Brooks-Gunn et al., 2002), but other research has pointed to the positive effects of

employment for low-income children in particular (as compared to their middle income

peers; Desai et al., 1989). However, none of this research has examined the combined effects

of increases in income and full-time employment on young children’s development. Our

analyses of potential mediators of the effects of employment and income suggest that the

increased instability of care and reduced parental warmth may have resulted in mitigating any

positive effects of income on children. Alternatively, of course, neither the increase in income

nor employment may have affected these young children given the effects of poverty on

young children’s well-being (Duncan & Brooks-Gunn, 1997). However, this hypothesis has

less support from prior research.

Considering how young they were at the start of the program, and the fact that young

children’s attachment relationships may be forming during this period (Ainsworth, Blehar,

Waters, & Wall, 1978), it is noteworthy that the increases in full-time maternal

employment did not result in negative impacts on the youngest group of children when

accompanied by large increases in income, at least given the measures utilized in this

study. Of course, the fact that only a small proportion of the program group took up the

supplement (about one-third) means that any effects observed for this group of families

would have needed to be very large to offset the likely null findings for the remainder of

the program group and result in any average effects to be observed when the full program

group is examined.

4.2. Middle childhood children

For middle childhood children, the SSP had small positive impacts on cognitive and health

outcomes but not on behavioral outcomes. The benefits to health outcomes are more

surprising than those to cognitive outcomes given the availability of health insurance in

Canada. Since these findings could not be verified with assessments by a health professional,

we view them somewhat cautiously.

The positive effects on cognitive outcomes are consistent with research that suggests a

stronger association between poverty and children’s outcomes for cognitive outcomes than


for behavioral outcomes (Duncan & Brooks-Gunn, 1997). Since these children ranged from

preschool to early school age (3–8 years) at the beginning of the study (although they were

in middle childhood by the end of the study), these findings are also consistent with

research suggesting that income may play a stronger role for preschool than for older

school-age children (Duncan & Brooks-Gunn, 1997). In fact, post hoc analyses support this

conclusion in that significant increases in academic achievement for the younger children

rather than for the older children in this group were observed. Because the SSP increased

voluntary and not mandatory maternal employment, and such employment has been found

to benefit children (Alvarez, 1985; Farel, 1980), maternal employment may not have

resulted in negative effects for these children. For these children, who may be less sensitive

to maternal separations than their younger peers, the negative effects of maternal

employment may not have offset the positive effects of the increased income, and this

combination may have resulted in the observed benefits to children’s cognitive devel-

opment.

The findings shed some light on the pathways that may have led to the favorable impacts

for middle childhood children and suggest a stronger role for child care, as opposed to the

parenting practices, in impacts. Parents in the program group relied on after-school activities

to a greater extent than parents in the control group because they engaged in longer hours of

employment. The greater participation in such activities may explain, at least in part, the

benefits to academic and cognitive performance found for children in the program group

(Posner & Vandell, 1994, 1999). In contrast, the SSP program did not affect parental

functioning, suggesting that parenting was not the cause of the favorable impacts for middle

children, despite theoretical and empirical work to suggest that increases in income and

employment should benefit parenting behavior (McLoyd, 1990, 1994). Because the measures

were reported by parents rather than outside observers, however, it is unclear whether no

change in parenting behavior occurred or whether the measures were not sensitive enough to

capture such changes. Future research is needed to examine whether observational measures

of parenting yield the same result.

The SSP’s effects for middle childhood children are consistent with the random assignment

evaluations of New Hope (Bos et al., 1999; Huston et al., 2001) and MFIP (Gennetian &

Miller, 2002). Both of these programs increased maternal employment and family income,

and both improved academic functioning for children in the middle childhood group

(although New Hope’s effects were concentrated only among boys and not girls). These

programs did not find effects for self-reported parenting behavior for parents of middle

childhood children either. In the New Hope evaluation, the researchers suggested that the

increased after-school activities might explain some of the benefits to boys that were found

(Huston et al., 2001).11 In contrast, programs studied in the random assignment NEWWS that

11 In New Hope, sex differences emerged for control group outcomes as well as experimental program

impacts—with boys at higher risk of negative outcomes and for whom positive impacts of the program were

found. No such gender differences in control group outcomes were found in this study. Notably, no sex differences

in effects emerged in the evaluation of the MFIP either (Gennetian & Miller, 2000).


required parents to look for work or enroll in education programs in order to receive welfare

increased parents’ employment but did not increase their income. The programs generally

produced neither systematically positive nor systematically negative changes for elementary

school-age children (McGroder et al., 2000). The concordance of findings in the SSP program

and in the U.S. New Hope and MFIP studies yields greater confidence that programs that

increase voluntary employment and income benefit children in the middle childhood age

range. They further suggest that the programs benefit children by increasing children’s care

experiences rather than through changes in parenting practices.

4.3. Adolescent children

Prior research has focused primarily on the negative effects of maternal separations for

young children, but the findings reported here suggest that welfare and employment programs

might have their most negative effects on older children’s functioning. The findings are

consistent with those found across welfare and employment programs, on a limited set of

measures of school progress (Gennetian et al., 2002).

The results may suggest the importance of monitoring adolescents as parents move from

welfare into work. First, the SSP significantly increased parents’ full-time employment, but

did not increase children’s time in after-school arrangements. Moreover, the SSP increased

the type of delinquent behaviors (for example, staying out late rather than conduct problems)

that are more closely associated with lack of parental supervision (Pettit et al., 1999) than

with negative parent–child interactions (Dodge, Pettit, Bates, & Valente, 1995; Patterson,

1995; Patterson, DeBaryshe, & Ramsey, 1989). An ethnographic study conducted with

parents in the SSP found that many chose not to take up the supplement because of concern

with their adolescent children (Bancroft & Vernon, 1995). These parents may have foreseen

what others in the program group did not. Results in the SSP are also consistent with research

suggesting that a high level of employment in adolescence is associated with greater alcohol

use and delinquency (Mortimer et al., 1996; Steinberg & Dornbusch, 1991). In the SSP, more

older children in the program group than in the control group reported taking on greater

family responsibilities by doing chores, and they also engaged in higher levels of employ-

ment. Researchers have suggested that associating with young adults and taking on adult

responsibilities may encourage early adultification of adolescents, and, in some cases,

encourage them to engage in adult risk-taking behavior earlier (Burton et al., 2002; Mortimer

et al., 1996; Steinberg & Dornbusch, 1991).

These results suggest a very different model for the effects of the SSP on adolescent

children than for their younger peers. First, given nonexperimental research that finds no

effects of income in adolescence on children (Duncan & Brooks-Gunn, 1997), we suspect

that increases in family income might not benefit older children as much as younger

children, although it is unlikely that increases in income would play out negatively for

adolescents. Second, we suggest that maternal employment may have negative effects on

children in this age group, with three possible pathways suggested by the data: (1) maternal

employment may reduce parental supervision and, in turn, increase adolescent delinquency

among low-income families; (2) maternal employment may lead to greater household roles


and responsibilities when single parents go to work, and these responsibilities may

overburden adolescents in low-income families; and (3) maternal employment may encour-

age links with the work force that may introduce these children to adult modes of behaving.

Exploratory correlational analyses reveal that maternal employment is indeed positively

related to adolescent delinquency (controlling for income), while the frequency of conduct-

ing chores is negatively associated with delinquency. Also, adolescent employment is not

significantly associated with adolescent delinquency. These findings provide initial support

for the first explanation. However, much more work is needed before definitive causal

conclusions can be drawn.

4.4. Limitations

Several limitations of the current study bear mention. First, while the random assignment

design allows us to confidently attribute any significant effects to the SSP program, analysis

of the processes by which such effects are occurring is strictly nonexperimental. While

analyses of the effects of the SSP on potential mediators of the relation between the SSP and

measures of children’s well-being are suggestive, they cannot directly assess the extent to

which changes in these potential mediators caused the changes outcomes for children

observed. It is not clear, for example, whether income and employment had offsetting effects

for the youngest children or both had neutral effects; whether the benefits for the middle

childhood children are due to benefits of increases in employment or income or both; and

whether the negative effects for the oldest children were a result of increased employment and

reduced supervision or increased employment and increased household and employment

responsibilities of their own. This study is powerful in attributing any effects on children to

the SSP program and to the combined changes in employment, income, and other mediators

that resulted from the program, but any conclusions that try to tease out the precise nature of

those pathways are much more tenuous.

Second, while it was necessary to vary measures by age group to address differing

developmental constructs at varying ages, it limits our ability to directly compare effects from

one age group to another. Without a set of measures that address identical constructs across

age groups, it is impossible to know whether differing results are due to differences in the age

of the children examined or the differences in the measures collected.

Third, it is important to recall that the response rates were lowest, at 64%, for the oldest

group of children. While no evidence of response bias was found (see Morris & Michalo-

poulos, 2000 for more detail on these results), such low response rates do lend more caution

to the interpretation of the findings. It is not clear whether the same effects would have been

found if more children had participated in the study.

Finally, for all three age groups of children, any effects of the SSP on developmental

outcomes are small, and only on selected measures of children’s functioning. Assuming

children scored around the 25th percentile on average on a standardized test in the control

group (that is close to the level observed, for example, for the PPVT scores in this study), this

effect size change would move children to the 30th percentile, which is an increase of only 5

percentile points. While the employment and income impacts of the SSP are relatively large,


these did not translate into big changes in children’s home and child care environments, or in

children’s developmental outcomes. While it is important to find that welfare and employ-

ment programs can have an effect on children, it is also noteworthy that these programs are

not a panacea. From a policy perspective, we still have a long way to go to help low-income

children. From a research perspective, the small and limited impacts make further studies

critical in enabling researchers to draw more definitive conclusions about the effects of such

programs on children.

5. Conclusion

The well-being of children in low-income families has been relatively understudied in the

move to increase the self-sufficiency of low-income, single parents. The SSP is one of a small

set of random assignment studies recently conducted on the effects of welfare and

employment programs on children. Together with these other studies, the SSP can dramat-

ically increase our understanding about how programs that increase employment and income

among single mothers may affect children, pointing to the importance of recognizing the

moderating role of children’s developmental stage. Moreover, the findings underline the

importance of examining differences in impacts on children across differing developmental

periods, and including adolescents in studies of the effects of increases in employment and

income on children.

Acknowledgements

This paper summarizes an evaluation managed by the Social Research and Demonstration

Corporation (SRDC) and evaluated by SRDC in collaboration with the Manpower

Demonstration Research Corporation (MDRC). The study was funded and conceived by

Human Resources Development Canada (HRDC).

Special thanks to Jean-Pierre Voyer and Allen Zeesman of HRDC’s Applied Research

Branch and to John Greenwood and Saul Schwartz at SRDC. The authors also thank Richard

Veevers, Ann Brown, June Lavergne, Scott Meyer, Charlene Walker, and Joan Conway at

Statistics Canada; Mary Pichette at HRDC; Sharon Manson Singer at British Columbia’s

Ministry of Social Development and Economic Security; Karen Mann and Gary Baird at

Human Resources Development, New Brunswick; Melony McGuire and Trudy Megeny at

EDS Systemhouse, in Nova Scotia; Betty Tully and Elizabeth Dunn at Bernard C. Vinge and

Associates in British Columbia; Shelly Price and Linda Nelson at Family Services Saint John,

in New Brunswick; Martey Dodoo, Tracey Hoy, and Wanda Vargas at MDRC for their

assistance in data collection, management, and analysis. Thanks also to Gordon Berlin,

Robert Granger, Lisa Gennetian, and Judith Gueron at MDRC; Dan Offord, Michael Boyle,

and Ellen Lipman at McMaster University; Ariel Kalil at the University of Chicago; members

of MDRC’s Income Support Studies Committee and the comments of two anonymous

reviewers for their thoughtful comments.


References

Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, S. (1978). Patterns of attachment: A psychological study

of the strange situation. Hillsdale, NJ: Erlbaum.

Alvarez, W. (1985). The meaning of maternal employment for mothers and the perceptions of their three-year-old

children. Child Development, 56, 350–360.

Bancroft, W., & Vernon, S. C. (1995). The struggle for self-sufficiency: Participants in the Self-Sufficiency Project

talk about work, welfare and their futures. Ottawa: Social Research and Demonstration Corporation.

Baumrind, D. (1971). Current patterns of parental authority. Developmental Psychology Monographs, 4 (Whole

No. 1), 1–103.

Baydar, N., & Brooks-Gunn, J. (1991). Effects of maternal employment and child-care arrangements in infancy on

preschoolers’ cognitive and behavioral outcomes, evidence from the children of the NSLY. Developmental

Psychology, 27(6), 932–945.

Becker, G. (1981). A treatise on the family. Cambridge, MA: Harvard University Press.

Belsky, J., & Rovine, M. J. (1988). Nonmaternal care in the first year of life and the security of infant–parent

attachment. Child Development, 59, 157–167.

Blackburn, M. L., & Bloom, D. E. (1993). The distribution of family income: Measuring and explaining changes

in the 1980s for Canada and the United States. In D. Card, & R. B. Freeman (Eds.), Small differences that

matter: Labor markets and income maintenance in Canada and the United States (pp. 233–266). Chicago:

University of Chicago Press.

Blank, R. M., & Hanratty, M. A. (1993). Responding to need: A comparison of social safety nets in Canada and

the United States. In D. Card, & R. B. Freeman (Eds.), Small differences that matter: Labor markets and

income maintenance in Canada and the United States (pp. 191–232). Chicago: University of Chicago Press.

Bos, J., Huston, A., Granger, R., Duncan, G., Brock, T., & McLoyd, V. (1999). New Hope for people with low

incomes: Two-year results of a program to reduce poverty and reform welfare. New York: Manpower Dem-

onstration Research Corporation.

Bradley, R. H., & Caldwell, B. M. (1984). 174 children: A study of the relation between the home environment

and early cognitive development in the first 5 years. In A. Gottfried (Ed.), The home environment and early

cognitive development (pp. 5–56). Orlando, FL: Academic Press.

Bronfenbrenner, U., & Crouter, A. C. (1982). Work and family through time and space. In S. Kamerman, & C. D.

Hayes (Eds.), Children in a changing world (pp. 39–83). Washington, DC: National Academy Press.

Brooks-Gunn, J., Han, W., & Waldfogel, J. (2002). Maternal employment and child cognitive outcomes in the first

three years of life: The NICHD study of early child care. Child Development, 73(4), 1052–1072.

Burchinal, M. R., Roberts, J. E., Riggins Jr., R., Zeisel, S. A., Neebe, E., & Bryant, D. (2000). Relating quality of

center-based child care to early cognitive and language development longitudinally. Child Development, 71,

339–357.

Burton, L., Brooks, J., & Clark, J. (2002). Adultification in childhood and adolescence: A conceptual model.

Working paper. College Park, PA: Penn State Center for Human Development and Health Research in Diverse

Contexts.

Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal of Sociology, 94,

S95–S120.

Conger, R. D., Conger, K. J., & Elder Jr., G. H. (1997). Family economic hardship and adolescent adjustment:

Mediating and moderating processes. In G. J. Duncan, & J. Brooks-Gunn (Eds.), Consequences of growing up

poor (pp. 288–310). New York: Russell Sage Foundation.

Currie, J., & Thomas, D. (1995). Medical care for children, public insurance, private insurance, and racial

difference in utilization. Journal of Human Resources, 30(1), 135–163.

Desai, S., Chase-Lansdale, L., & Michael, R. (1989). Mother or market? Effects of maternal employment on the

intellectual ability of 4-year old children. Demography, 26(4), 545–561.

Dodge, K., Pettit, G., Bates, J., & Valente, E. (1995). Social information processing patterns partially mediate the

effects of early physical abuse on later conduct problems. Journal of Abnormal Psychology, 104(4), 543–632.


Dooley, M. D. (1994). The converging market work patterns of married mothers and lone mothers in Canada.

Journal of Human Resources, 29(2), 600–620.

Duncan, G. J., & Brooks-Gunn, J. (Eds.) (1997). Consequences of growing up poor. New York: Russell Sage

Foundation.

Duncan, G. J., Brooks-Gunn, J., & Klebanov, P. K. (1994). Economic deprivation and early childhood develop-

ment. Child Development, 65, 296–318.

Eccles, J., & Gootman, J. A. (Eds.) (2002). Community programs to promote youth development. Washington,

DC: National Academy Press.

Elder, G. (1974). Children of the Great Depression. Chicago: University of Chicago Press.

Elder, G. (1979). Historical change in life patterns and personality. In P. Bates, & O. Brim (Eds.), Life span

development and behavior (pp. 117–159). New York: Academic Press.

Elder, G., Liker, J., & Cross, C. (1984). Parent–child behavior in the Great Depression: Life course and inter-

generational influences. In P. Baltes, & O. Brim (Eds.), Life span development and behavior (pp. 109–158).

Orlando, FL: Academic Press.

Farel, A. M. (1980). Effects of preferred maternal roles, maternal employment, and sociodemographic status on

school adjustment and competence. Child Development, 51, 1179–1196.

Gennetian, L., Duncan, G., Knox, V., Clark-Kauffman, B., & Vargas, W. (2002). How welfare and work policies

for parents affect adolescents: A synthesis of research. New York: Manpower Demonstration Research Cor-

poration.

Gennetian, L., & Miller, C. (2000). Reforming and rewarding work: Final report on the Minnesota Family

Investment Program: Vol. 2. Effects on children (pp. 1–212). New York: Manpower Demonstration Research

Corporation.

Gennetian, L., & Miller, C. (2002). Children and welfare reform: A view from an experimental welfare program in

Minnesota. Child Development, 73(2), 601–620.

Gennetian, L. A., & Morris, P. (2003). The effects of time limits and make work pay strategies on the well-being

of children: Experimental evidence from two welfare reform programs. Children and Youth Services Review,

25(1/2), 17–54.

Hamilton, G., Freedman, S., & McGroder, S. (2000). Do mandatory welfare-to-work programs affect the well-

being of children: A synthesis of child research conducted as part of the National Evaluation of Welfare-to-

Work Strategies. Washington, DC: U.S. Department of Health and Human Services, Administration for

Children and Families and Office of Assistant Secretary for Planning and Evaluation; and U.S. Department

of Education.

Harvey, E. (1999). Short-term and long-term effects of early parental employment on children of the National

Longitudinal Survey of Youth. Developmental Psychology, 35(2), 445–459.

Hoffman, L. W. (1989). Effects of maternal employment in the two parent family. American Psychologist, 44,

283–292.

Huston, A., Duncan, G., Granger, R., Bos, H., McLoyd, V., Mistry, R., Crosby, D., Gibson, C., Magnusson, K.,

Romich, J., & Ventura, A. (2001). Work-based anti-poverty programs for parents can enhance the school

performance and social behavior of children. Child Development, 72(1), 318–337.

Knox, V., Miller, C., & Gennetian, L. (2000). Reforming welfare and rewarding work: A summary of the final

report on the Minnesota Family Investment Program. New York: Manpower Demonstration Research

Corporation.

Lamb, M. (1998). Nonparental child care, context, quality, correlates and consequences. In I. E. Sigel, & K. A.

Renninger (Eds.), Handbook of child psychology (4th ed., pp. 73–134). New York: Wiley.

Lazar, I., & Darlington, R. B. (1982). Lasting effects of early education: A report of the consortium for

longitudinal studies. Monographs of the Society for Research in Child Development, 47(2–3), 1–151

(Serial No. 195).

LeBlanc, M., & Tremblay, R. E. (1988). A study of factors associated with the stability of hidden delinquency.

International Journal of Adolescence and Youth, 1, 269–291.

Lee, V. E., Brooks-Gunn, J., & Shnur, E. (1988). Does Head Start work? A one-year follow-up comparison of


disadvantaged children attending Head Start, no preschool and other preschool programs. Developmental

Psychology, 24(2), 210–222.

Lin, W., Robins, P. K., Card, D., Harknett, K., & Lui-Gurr, S. (1998). When financial incentives encourage work:

Complete 18-month findings from the Self-Sufficiency Project. Ottawa: Social Research and Demonstration

Corporation.

Maccoby, E. E., & Martin, J. A. (1983). Socialization in the context of the family: Parent–child interaction.

In P. Mussen (Ed.), Handbook of child psychology: Vol. 4. Socialization, personality and social development

(4th ed., pp. 1–101). New York: Wiley.

Mallar, C. D., & Maynard, R. A. (1981). The effects of income maintenance on school performance and educa-

tional attainment. Research in Human Capital and Development, 2, 121–141.

McGroder, S., Zaslow, M., Moore, K., & LeMenestrel, S. (2000). Impacts on young children and their families

two years after enrollment: Findings from the Child Outcomes Study. Washington, DC: U.S. Department of

Health and Human Services Administration for Children and Families and Office of Assistant Secretary for

Planning and Evaluation, and U.S. Department of Education.

McKey, R. H., Condelli, L., Gransom, H., Barrett, B., McConkey, C., & Plantz, M. (1985). The impact of Head

Start on children, families and communities. Final report of the Head Start Evaluation, Synthesis and Uti-

lization Project. Washington, DC: Department of Health and Human Services, Head Start Bureau.

McLoyd, V. (1990). The impact of economic hardship on black families and children: Psychological distress,

parenting and socioemotional development. Child Development, 61, 311–346.

McLoyd, V. (1998). Children in poverty, development, public policy, and practice. In I. E. Sigel, & K. A.

Renninger (Eds.), Handbook of child psychology (4th ed., pp. 135–210). New York: Wiley.

McLoyd, V. C., Jayartne, T. E., Ceballo, R., & Borquez, J. (1994). Unemployment and work interruption among

African-American single mothers, effects on parenting and adolescent socioemotional functioning. Child

Development, 65, 562–589.

Meyer, B. D. & Rosenbaum, D. T. (2000). Making single mothers work: Recent tax and welfare policy and its

effects. Cambridge, MA: National Bureau of Economic Research Working Paper 7491.

Michalopoulos, C., Card, D., Gennetian, L. A., Harknett, K. A., & Robins, P. K. (2000). The Self-Sufficiency

Project at 36 months: Effects of a financial work incentive on employment and income. Ottawa: Social

Research and Demonstration Corporation.

Michalopoulos, C., & Robins, P. K. (2000). Employment and child-care choices in Canada and the United States.

Canadian Journal of Economics, 33(2), 435–470.

Michalopoulos, C., & Robins, P. K. (2002). Employment and child-care choices of single-parent families in

Canada and the United States. Journal of Population Economics, 15(3), 465–493.

Moore, K., & Driscoll, A. (1997). Low-wage maternal employment and outcomes for children: A study. Future of

Children, 7(1), 122–127.

Morris, P., Bloom, D., Kemple, J., & Hendra, R. (in press). The effects of a time limited welfare program on

children: The moderating role of parents’ risk of welfare dependency. Child Development.

Morris, P., Huston, A., Duncan, G., Crosby, D., & Bos, H. (2001). How welfare and work policies affect children:

A synthesis of research. New York: Manpower Demonstration Research Corporation.

Morris, P., & Michalopoulos, C. (2000). The Self-Sufficiency Project at 36 months: Effects on children of a

program that increased employment and income. Ottawa: Social Research and Demonstration Corpora-

tion.

Mortimer, J. T., Finch, M. D., Ryu, S., Shanahan, M. J., & Call, K. T. (1996). The effects of work intensity on

adolescent mental health, achievement, and behavioral adjustment: New evidence from a prospective study.

Child Development, 67, 1243–1261.

Munnel, A. H. (1986, September). Lessons from the income maintenance experiments. Proceedings of a

conference sponsored by the Federal Reserve Bank of Boston and the Brookings Institution, Melvin

Village, NH.

NICHD Early Child Care Research Network (1998). Early child care and compliance, cooperation, defiance and

problem behavior at 24 and 36 months. Child Development, 46, 1145–1170.


NICHD Early Child Care Research Network (2000). The relation of child care to cognitive and language develop-

ment. Child Development, 71, 960–980.

NICHD Early Child Care Research Network (2001). Child care and common communicable illnesses:

Results from the NICHD study of early child care. Archives of Pediatrics and Adolescent Medicine,

155, 481–488.

Office of Income Security Policy, Department of Health and Human Services (1983). Overview of the Seattle–

Denver income maintenance experiment final report. Washington, DC: U.S. Government Printing Office.

Parcel, T. L., & Menaghan, E. G. (1994). Parent’s jobs and children’s lives. New York: Aldine de Gruyter.

Parcel, T. L., & Menaghan, E. G. (1997). Effects of low-wage employment on family well-being. Future of

Children, 7(1), 116–121.

Patterson, G. (1995). Coercion as a basis for early age of onset for arrest. In J. McCord (Ed.), Coercion and

punishment in long term perspectives (pp. 81–105). Eugene, OR: Oregon Social Learning Center.

Patterson, G., DeBaryshe, B., & Ramsey, E. (1989). A developmental perspective on antisocial behavior. Amer-

ican Psychologist, 44(2), 329–335.

Peisner-Feinberg, E. S., Burchinal, M. R., Clifford, R. M., Culkin, M. L., Howes, C., Kagan, S. L., Yazsjian, N.,

Byler, P., Rustici, J., & Zelazo, J. (1999). The children of the cost, quality, and outcomes study go to school.

Chapel Hill, NC: University of North Carolina (Technical Report).

Peterson, J. L., & Zill, N. (1986). Marital disruption, parent–child relationships, and behavior problems in

children. Journal of Marriage and the Family, 48, 295–307.

Pettit, G. S., Bates, J. E., Dodge, K. A., & Meece, D. W. (1999). The impact of after-school peer contact on early

adolescent externalizing problems is moderated by parental monitoring, perceived neighborhood safety and

prior adjustment. Child Development, 70(3), 768–778.

Polit, D. F. (1996). Self administered teacher questionnaire in the New Chance 42-month survey. New York:

Manpower Demonstration Research Corporation.

Posner, J. K., & Vandell, D. L. (1994). Low-income children’s after-school care: Are there beneficial effects of

after-school programs? Child Development, 65, 440–456.

Posner, J. K., & Vandell, D. L. (1999). After-school activities and the development of low-income urban children:

A longitudinal study. Developmental Psychology, 35(3), 868–879.

Radloff, L. S. (1977). The CES-D Scale: A self-report depression scale for research in the general population.

Applied Psychological Measurement, 1(3), 385–401.

Ramey, C. T., Campbell, F. A., Burchinal, M. R., Skinner, M. L., Gardner, D. M., & Ramey, S. L. (2000).

Persistent effects of early childhood education on high-risk children and their mothers. Journal of Applied

Developmental Science, 4, 2–14.

Salkind, N. J., & Haskins, R. (1982). Negative income tax: The impact on children from low-income families.

Journal of Family Issues, 3(2), 165–180.

Sampson, R. J., & Laub, J. H. (1994). Urban poverty and the family context of delinquency: A new look at

structure and process in a classic study. Child Development, 65, 523–540.

Shonkoff, J. P., & Phillips, D. A. (2000). From neurons to neighborhoods: The science of early childhood

development. Washington, DC: National Academy Press.

Smith, J. R., Brooks-Gunn, J., Lee, K., & Klebanov, P. K. (1997). Consequences of living in poverty for young

children’s cognitive and verbal ability and early school achievement. In G. J. Duncan, & J. Brooks-Gunn

(Eds.), Consequences of growing up poor (pp. 132–189). New York: Russell Sage Foundation.

Steinberg, L., & Dornbusch, S. M. (1991). Negative correlates of part-time employment during adolescence,

replication and elaboration. Developmental Psychology, 27(2), 304–313.

Sugland, B. W., Zaslow, M., Smith, J. R., Brooks-Gunn, J., Moore, K. A., Blumenthal, C., Griffin, T., & Bradley,

R. (1995). The early childhood HOME inventory and HOME short form in differing sociocultural groups: Are

there differences in underlying structure, internal consistency of subscales, and patterns of prediction? Journal

of Family Issues, 16(5), 632–663.

Vandell, D. L., & Ramanan, J. (1992). Effects of early and recent maternal employment on children from low-

income families. Child Development, 63(4), 938–949.


White, H. (1980). A heteroscedasticity-consistent covariance matrix estimator and a direct test for heteroscedas-

ticity. Econometrika, 48, 817–838.

Zaslow, M. J., & Emig, C. A. (1997). When low-income mothers go to work: Implications for children. Future of

Children, 7(1), 110–115.

Zaslow, M. J., McGroder, S., Cave, G., & Mariner, C. (1999). Maternal employment and meausres of children’s

health and development among families with some history of welfare receipt. Research in the Sociology of

Work, 7, 233–259.

Findings from the Self-Sufficiency Project: effects on children and adolescents of a program that...

Documents

Transcript of Findings from the Self-Sufficiency Project: effects on children and adolescents of a program that...