Since the 1960s, early childhood development (ECD) programs have been planned, implemented, and evaluated for preschool children and their families, particularly those living in socio-economically disadvantaged communities. Recently, ECD programs have grown in popularity because of their potential to reduce the incidence of serious psychosocial problems and save costs in government spending (Brooks-Gunn 2003; Camilli et al. 2010; Karoly et al. 2005; Nelson et al. 2003; Reynolds and Temple 2008). It is thought that if children’s competence and well-being can be promoted in their early years that this will prevent negative outcomes for children in the long-term (Nelson et al. 2009). Potential positive long-term outcomes that have been examined include increased rates of school completion and employment and decreased rates of grade retention, placement in special education, welfare participation, pregnancy during the teen years, and involvement in child welfare, criminal justice, and mental health systems.

While useful and important, an exclusive reliance on these outcomes may not do justice to the impacts that may result from disadvantaged children’s participation in ECD programs. For example, consider the outcome of teen-age pregnancy. While pregnancy may result in negative sequalae for teen-aged women who have limited resources and supports, the birth of a child may be experienced as joyous, life-changing event in a young woman’s life. If the woman has access to resources, having a child may be experienced as meaningful, growth-promoting, and positive. In this paper, we argue for the utility of adopting a narrative approach in the evaluation of long-term outcomes of ECD programs to tap the subjective experiences of youth. The purpose of the paper is to report on the long-term narrative outcomes of a quasi-experimental evaluation of Better Beginnings, Better Futures, a multi-site, multi-component, universal, research demonstration prevention program in Ontario, Canada. Children began participating in Better Beginnings at ages 3–4, and this article reports on narrative outcomes for the youth who are now ages 18–19.

Literature Review

The review begins with an examination of literature on the long-term preventive impacts of ECD programs and then considers research that uses narratives, particularly turning point stories, as a potentially valuable strategy for understanding the preventive impacts of ECD programs.

Long-Term Effects of Early Childhood Development (ECD) Prevention Programs

Several recent meta-analytic or systematic reviews have demonstrated the positive long-term impacts of ECD programs. Nelson et al. (2003) reviewed 10 controlled studies that examined the preventive impacts of ECD programs at high school and beyond and found positive impacts of these programs on social-emotional outcomes for youth. Farrington and Welsh (2003) found that four studies of home visitation or preschool education ECD programs that followed children into adolescence or early adult years had substantial impacts youths’ rates of criminal offending. In their review of five studies that examined adult outcomes, Karoly et al. (2005) found that ECD programs had significant positive impacts on several outcome domains: high school completion, adult crime and delinquency, employment and earnings, and social service use. Reynolds and Temple (2008) located 13 controlled evaluations of ECD programs that followed children up to age 13 or older. They found significant impacts of these programs on social and emotional development, school achievement, delinquency and crime, grade retention, special education participation, school completion, and employment and earnings. Similarly, in a review of 17 ECD programs on outcomes in adolescence, Manning et al. (2010) found significantly positive impacts on educational success, social deviance, social participation, cognitive development, criminal justice involvement, family well-being, and social–emotional development.

Two specific ECD programs, the Perry Preschool Project and the Chicago Child-Parent Center program, are particularly noteworthy because they are somewhat similar to Better Beginnings, and they have long-term follow-up research findings. Schweinhart et al. (2005) reported the impacts of a randomized trial of the Perry Preschool Project when participants were 40 years old. The Perry Preschool Project is an ECD program that was created to address the problem of poor academic performance in students from low-income, African-American neighborhoods in the Ypsilanti, Michigan school district. The Perry Preschool included both center-based programs and home visitation/family support. Participants in the Perry Preschool program significantly outperformed the control group on educational achievement and financial earnings at the age of 40, and they had significantly less involvement with crime and more personal and family success than those in the control group.

The Child-Parent Center program was designed to provide educational enrichment in preschool, school, and family services to children through grade 3 (Reynolds et al. 2007). The effects of this program were examined on the health and well-being of a group of 1,539 young adults, aged 24, in Chicago, Illinois, using a quasi-experimental design. Participants who had been enrolled in the Child-Parent Center program had significantly higher rates of school attendance and completion, and employment, as well as significantly lower rates of arrest, depressive symptoms, and out-of-home placements than comparison participants.

It is possible that the impacts of programs like the Perry Preschool and the Child-Parent Centers extend to the subjective experiences and life stories of the youth who participated in these initiatives. In the next section, we review research on a narrative approach to understanding the well-being of youth.

Narratives

Narratives are a way of understanding human experience. Bauer et al. (2008) argue that people make sense of their lives in terms of stories, which consist of characters, plots, themes, and emotional tone. Moreover, narratives can highlight a sense of unity, purpose, meaning, and coherence in a person’s life, even when an individual has undergone a traumatic experience (Bauer et al. 2008; Crossley 2003; Pals and McAdams 2004). Crossley (2003) and McAdams et al. (2006) indicate that constructing and telling a personal narrative enables an individual to develop a sense of identity, which is particularly important in late adolescence and early adulthood (McLean et al. 2007).

Turning Point Stories

Turning points, or events that represent an important change in one’s life or one’s outlook on the world, contribute to and generate information about identity, problem-solving and coping strategies, future outlook, the development and maintenance of interpersonal relationships, and a concern for future generations (Frensch et al. 2007; McAdams et al. 2001a). Turning points stories can describe a wide variety of life experiences, including drug addiction, diagnosis of HIV, and going away to university (McAdams et al. 2001a, b; Pillemer 1998).

Narrative Dimensions

A number of dimensions have been examined in the analysis of turning point stories and other narratives. These dimensions may be useful outcome measures for long-term evaluations of prevention-oriented ECD programs. Blagov and Singer (2004) have argued that specificity is an important narrative dimension, because people who are able to tell stories in a vivid and detailed way (i.e., with specificity) are less defensive than people who provide more vague, global accounts in their stories. Another important dimension is the resolution of a story (Pals 2006). Dumas et al. (2009) defined story resolution as “the ability to conclude low point life narratives with a positive tone and with evidence of coherence and emotional resolution” (p. 1532). In their review, Bauer et al. (2008) reported that people who are able to highlight personal growth in their life stories, and look at difficult life experiences as transformative in that new insight was gained about the self, show higher levels of happiness than those who do not show personal growth (Bauer and McAdams 2004). In their meta-analysis of 87 cross-sectional studies, Helgeson et al. (2006) found that the ability to find benefit in having experienced a traumatic event was related to more positive well-being and less depression. However, Park and Helgeson (2006) caution that self-reported growth is not always related to actual change in well-being.

Meaning-making refers to an individual’s ability to draw lessons and insights about life events and episodes and is important for maturity and a positive outlook for the future (McLean and Pratt 2006). As Adler and Poulin (2009) state: “Making meaning out of negative experiences is one of the primary psychological challenges in the wake of adversity” (p. 903). Tebes et al. (2004), in a study of 35 recently bereaved young adults, found that the ability to find meaning in an adverse experience was positively related to better grief resolution, better trauma resolution, and fewer psychiatric symptoms.

The coherence of stories is another dimension that has been studied. Stories can range from those that are very disjointed or incoherent to those that are well-articulated and understandable. As children move into adolescence and early adulthood, they become more capable of telling coherent stories (Bohn and Bernsten 2008; Habermas and de Silveira 2008). Turning point stories can also be examined in terms of affect transformation, or changes in emotion that occur between the beginning and the end of a turning point. Tebes et al. (2004) define cognitive transformation as realizing that an adverse experience has resulted in new opportunities, and re-interpreting that experience from one that is primarily traumatic to one that promotes growth. Affect transformation taps the pattern of redemption in a life story that McAdams and colleagues argue is important for personal growth and well-being (McAdams 2006; McAdams et al. 2001a, b; McLean and Pratt 2006).

Turning Point Stories and Well-Being

Prilleltensky and Prilleltensky (2006) have conceptualized well-being in terms of multiple ecological levels: personal, relational, and collective. Thus, well-being consists not just of the personal dimensions of subjective well-being and personal growth, but also of being embedded in a network of positive and supportive relationships and participating freely in social, community, and political life.

One important indicator of personal well-being is self-esteem. In a study of 74 mid-life adults, McAdams et al. (2001b) found that self-esteem was significantly positively correlated with affect transformation in participants’ narrative accounts of high point, low point, and turning point episodes in their lives. Similar results were found by McAdams et al. (2001b) in a study of 125 university students. Self-esteem could function as an important stress-meeting resource that helps young people successfully cope with stressful life events.

Family functioning is an indicator of relational well-being. In a longitudinal study of 100 individuals at ages 17 and 26 years, Dumas et al. (2009) found that scores on the Family Assessment Device (FAD), a measure of family functioning, at age 17 predicted positive resolution of low point stories at age 26. As well, positive resolution was related to indices of identity maturity and emotional adjustment at age 26. Like self-esteem, having the support of one’s family may be an important factor in enabling young people to successfully navigate low points in their lives.

Community involvement can be conceptualized as an important component of a person’s community well-being. In a study of 32 young adults aged 20 years, Frensch et al. (2007) found that the amount of their community involvement was significantly positively correlated with themes of generativity and caring in their turning point stories. In a study of young adults aged 26 years, Hasford et al. (2008) found that community involvement was positively correlated with several dimensions of stories about their community involvement: specificity, prosocial content, and personal impact. Community involvement can provide more opportunities for youth to tell stories about important events in their lives.

The Better Beginnings, Better Futures Project

Better Beginnings, Better Futures is a 25-year research demonstration project funded by the Government of Ontario to prevent social, emotional, behavioral, physical and cognitive problems in young children, promote the development of these children in those same areas, and enhance the ability of disadvantaged families and communities to provide for their children (Peters et al. 2003, 2010). Three Better Beginnings sites focusing on children aged 4–8 were established in 1991. These sites are in Sudbury, Cornwall, and Highfield (a neighborhood in the Greater Toronto Area). Sudbury is the largest city in the northern part of the province; its Better Beginnings project is centered in two ethnically diverse neighborhoods which include a significant proportion of native, Anglophone and Francophone families, as well as other groups. Cornwall is a medium-sized city in eastern Ontario; its Better Beginnings site was designed to serve the large French-speaking community in the neighborhoods within which it was based. The Highfield community is an ethnically diverse community with a large proportion of newcomers to Canada, situated in the northwestern part of Toronto, the largest city in the province. All three communities are socioeconomically disadvantaged, with household incomes well below the provincial average. In addition, two other Ontario communities, in Ottawa-Vanier and Etobicoke (Toronto), which were similar in economic and community characteristics to the communities receiving the intervention, were designated as comparison sites. In previous research, we have found positive impacts of Better Beginnings on child, parent, family, and community outcomes at grades 1, 2, 3, 6, and 9 (Peters et al. 2003, 2010).

The key features of the Better Beginnings model for primary prevention were established by a 25-member advisory group including program directors and researchers from across Ontario. The resulting program model was ecological, recognizing that strategies must seek to impact not only children themselves, but also the families and communities in which they exist. The model was also universal, in that, regardless of individual risk factors, all children and families living in a Better Beginnings community were eligible to take part in any of the programs. Requirements for comprehensive and integrated programming sought to ensure that Better Beginnings programs addressed multiple areas of need for families, children, and communities, while coordinating with existing service-providers and local institutions. Additionally, the model called for programs tailored to meet local needs and involving local residents in program planning and delivery.

Although programming in each site was designed to meet the same major project goals, the exact mix of programs offered in each site, as well as the methods in which programs were developed and delivered, varied considerably across sites. Each site developed an array of programs that were suited to the needs and desires of the community as articulated by community members. Some commonalities in activities did emerge across sites—for example, all three of the communities employed home visitors to provide parents with information, support, and assistance, worked with teachers and children to enhance children’s in-school experiences, and provided nutrition programs for children. In their analysis of multi-site, complex, community interventions, Hawe et al. (2004) have made a distinction between the form and function of interventions that is useful for understanding Better Beginnings programs. They argue that while the form or particular programs in communities may differ because of the need to tailor programs to the local context, that the intended function of the programs should be the same. While Better Beginnings programs differed somewhat from site to site, all sites were mandated by the government funders to have programs designed to have impacts on children, families, and communities.

Research Objectives and Hypotheses

The objectives of this research are to examine the impacts of Better Beginnings on narrative dimensions and to determine if the narrative dimensions are related to standardized measures of personal, family, and community well-being. More specifically, we tested the following hypotheses.

  1. 1.

    It is predicted that youth who participated in the three Better Beginnings sites will score significantly higher than youth from the comparison sites on several dimensions of turning point stories: story specificity, ending resolution, personal growth, meaning-making, coherence, and affect transformation.

  2. 2.

    It is predicted that standardized measures of individual (self-esteem), family (positive family functioning), and community (involvement in community activities) well-being will be directly correlated with dimensions of turning point stories: story specificity, ending resolution, personal growth, meaning-making, coherence, and affect transformation.

Methodology

Design

A randomized controlled trial design was not feasible for the evaluation of Better Beginnings in view of the government process for the selection of sites. Thus, a quasi-experimental longitudinal comparison site design was used, wherein cohorts of children and their families were recruited in the three project sites (Cornwall, Sudbury, Highfield) and two sociodemographically-matched comparison neighborhoods (Etobicoke and Ottawa-Vanier). Longitudinal follow-up analyses contrasted outcomes for youth aged 18–19 and their families from the Better Beginnings sites relative to youth and their families in the comparison sites.

Sample

Larger Sample

The sample for the present study was drawn from a larger sample of the Better Beginnings longitudinal research that began in 1993 when the children enrolled in Junior Kindergarten at 3–4 years of age and continued with data collection at Senior Kindergarten, grades 1, 2, 3, 6, 9, and the most recent data collection at grade 12 in 2007. Children and their parents from three Better Beginnings sites and two socio-demographically similar comparison sites participated in the research at each of these data collection periods. By grade 3, a total of 959 children and families had been recruited into the study through the children’s schools. In the Better Beginnings and comparison sites, the sampling strategy was to invite all families to participate. The resulting self-selected sample for Better Beginnings (n = 601) and comparison sites (n = 358) represented 50–60% of the entire birth cohort in their respective neighborhoods, based on school records. Sampling bias was tested using four indicators of behavior and social skills as assessed by teachers, and there were no significant differences between children who participated in the research and children who did not participate in the research during the initial waves of data collection (1993–1994 and 1997–1998).

Sample Attrition

Sample attrition (when no data were gathered from a family in a particular data collection period) occurred at an average of 10% every three years. From grade 3 to grade 6 assessments, 9% of Better Beginnings students and 15% of comparison students were not retained in the study. Then, between grades 6 and 9 there was a 7% decrease in sample size from the Better Beginnings sites and a 10% decrease in comparison sites. The retention rate from recruitment through to grade 12 assessments was 66.7% for Better Beginnings participants and 62.8% for comparison participants, which was not significantly different, χ2(1, N = 959) = 1.49, ns. Reasons for smaller sample sizes included being unreachable/not located, declining without explanation, or a child’s death. The total sample retention rate of 65.3% is similar to the 65% retention rate reported in the Statistics Canada’s National Longitudinal Survey of Children and Youth (NLSCY; Statistics Canada 2007).

Differences between those who remained in the study and those who dropped out between the last two data collection periods (grades 9 and 12) were examined for 19 sociodemographic variables,Footnote 1 using logistic regression to predict dropout from each of these variables. Two of the variables were significant at p < .05 when predicting dropout between the two time periods: parent’s year of birth and female labour force status. These two variables, along with others to be described, were used as covariates in the analyses.

Sub-Sample for Study of Narratives

In 2007–2008, a stratified random sampling strategy was used to recruit participants from the larger Better Beginnings sample into the Better Beginning narrative research. Stratification was by site and gender with roughly equal numbers of males and females drawn from each of the Better Beginnings and comparison community sites. For inclusion into the study, participants in both Better Beginnings sites and comparison sites were to have continuously lived in their community since they were in elementary school. This inclusion criterion was stipulated because youth were also asked questions about experiences in their communities both when they were young and at the time of the interview.

The narrative sub-sample size was n = 96 (for n = 51 males and n = 45 females). Roughly equal numbers of males and females were drawn from each of the Better Beginnings and comparison communities: Cornwall (10 males and 10 females), Highfield (10 males and 9 females), Sudbury (13 males and 10 females), Etobicoke (10 males and 8 females) and Ottawa Vanier (8 males and 8 females). The narrative sub-sample did not differ from the larger Better Beginnings sample on several demographic variables: sex of respondent, immigrant status, or cultural identification. However, on average participants in the narrative sub-sample had a lower mean monthly household income (M = $3496.75) than participants in the larger Better Beginnings sample (M = $3880.84), t(458) = 2.48, p < .05. In addition, the samples differed in length of youths’ residence in neighborhood (in years), t(772) = 3.97, p < .05. Participants in the narrative sub-sample had lived in their neighborhood for longer (M = 5.33 years) than participants in the larger Better Beginnings sample (M = 4.00 years). Lastly, in comparison to the larger Better Beginnings sample, more participants in the narrative sub-sample still lived in Better Beginnings neighborhoods at the time of interviews, χ2(1, N = 543) = 50.35, p < .01.

There are numerous reasons why the samples may have differed on these measures. First, we aimed to recruit participants for the narrative sub-sample who had lived in Better Beginnings communities consistently over time. Therefore the difference in length of youths’ residence in a Better Beginnings neighborhood is expected. In addition, there may be a reciprocal relationship between the length of a youth’s residence in their neighborhood, their current residence in a Better Beginnings neighborhood, and the difference found in monthly household income. Better Beginnings programs are located in lower than average socio-economic communities. Families that continuously reside in those communities may do so because of a lack of financial opportunities that would cause them to relocate. Also of importance, those participants who lived in Better Beginnings communities and participated in Better Beginnings programs for a greater length of time may have had a greater desire to give back to the program by participating in this study.

Since this is a quasi-experiment and the sites could differ in terms of demographic characteristics, the data were weighted to remove mean differences in covariates between Better Beginnings and comparison sites (Friedman 2001; Ridgeway et al. 2006). This procedure, generalized boosted regression, provides a robust estimate of the probability that a case lies in one group or the other, given its scores on a set of variables which predict group membership. These probabilities are created from the regression coefficients of many bootstrapped samples (30,000 samples in this case), which are used to create a final set of regression weights for the predictors. Based on these and a participant’s scores on the predictors, each case in the comparison group is assigned a weight. When applied, these weights produced a comparison sample closely resembling the Better Beginnings sample, so that the two groups did not differ on the following variables: birth year and gender of the interview respondent (typically the mother), gender of child, marital status, single parent status, respondents’ education, employment status, family income, cultural category (Anglophone or Francophone), and immigrant status.

Measures

All the measures were collected at grade 12 when the youth were ages 18–19.

Narratives

Interviews

This study focused on youths’ turning point stories as one part of a larger narrative study. The entire interview contained four parts: (a) the youth’s life story—a descriptive account from pre-school through high school, (b) a turning point story, (c) participation in Better Beginnings and other community programs, and (d) stories about their communities. For the turning point story, youth were given the following instructions by the interviewer.

“In looking back on your life, you may be able to identify certain key ‘turning points’—episodes through which you experienced an important change in your life. Please choose one key turning point scene and describe it in detail. If you feel your life story contains no clear turning points, then describe a particular episode in your life that comes closer than any other to qualifying as a turning point—a scene where you changed in some way. Please describe what led up to the event, what happened in the event, where and when it happened, who was involved, what you were thinking and feeling, and so on. Also, please tell me how you think you changed as a result of this event and why you consider this event to be a turning point in your life.”

Participants were encouraged to describe their turning points in terms of a narrative story. Expectations were left open so that any experience that was deemed an important life change or turning point for the participant was acceptable as an answer to this interview question. The interviews were audio-recorded and transcribed verbatim.

Coding

All of the turning point stories were coded on several dimensions: specificity (1 = not specific to 5 = very specific), ending resolution (1 = very unresolved to 4 = very resolved), personal growth (0 = no personal growth to 2 = event explicitly promoted personal growth), meaning-making (0 = no meaning reported to 3 = insight [explicit transformation of understanding of self or the world]), coherence (0 = no story or story is too incoherent to understand to 3 = classic pattern in which story builds to high point, dwells on it, then resolves it), and early affect and later affect (1 = very negative affect to 7 = very positive affect). Affect transformation was calculated for each turning point story by subtracting the scale value of early affect from the value of later affect. Affect transformation is an indicator of a redemption pattern in which a turning point story shifts from something that is initially challenging or negative to a positive outcome in the end (McAdams 2006). We also coded as present or absent the type of story that the participants told (e.g., relationship, autonomy).

Here, we give an example of part of a turning point story to illustrate how it was coded for personal growth, meaning-making, and ending resolution. The following narrative of a Better Beginnings youth shows a high level of personal growth, meaning-making, and ending resolution.

In high school things changed completely… when I was 16 I got pregnant… and after I had her I stopped going to school. I just stayed home to take care of her… I just wanted to finish my high school ‘cause I felt like I’m just at home doing nothing…and then I saw this commercial that said I could finish my high school, so now I am finishing school at home, on the computer… and I’m doing good now…this will help me because I want to be a fashion designer…and it makes me feel better and lets my mom be proud of me… so I’m doing everything I can to get that. I will make that goal so I can move a step forward instead of a step back.

This story was rated a 2 for personal growth (on a scale of 0–2). It is clear that this event explicitly promoted positive development in the participant’s life because she seems to have learned a lot from her current situation and is prepared to meet her goals. Although she previously dropped out of school due to her pregnancy, this experience has taught her that she wants to be in school so she can achieve her career goals and be proud of herself. In this way, she has learned something valuable from this experience and derived significant meaning from it. The insight she demonstrates and transformation of her understanding of herself and her world make this story a 3 (on a scale of 0–3) in terms of meaning-making. It was rated a 4 for ending resolution (on a scale of 1 to 4) because it can be seen that the participant has come to terms with her situation as a new mother and is happy with her life and where it is going.

Inter-Rater Reliability

All of the turning point stories were coded by one rater, while a second rater coded a sample of 23 stories. The primary rater was blind as to whether the youth had participated in Better Beginnings. To determine inter-rater reliability, Pearson correlation coefficients were calculated for each of the variables. Estimates of inter-rater reliability were quite high: specificity (r = .70), ending resolution (r = .92), personal growth (r = .96), meaning-making (r = .75), coherence (r = .71), early affect (r = .87), and later affect (r = .90). The kappa statistic was used to test the reliability of the coding for the different types of stories: relationship (κ = .91), achievement (κ = .80), autonomy (κ = .72), mortality (κ = .83), and other story (κ = .78).

Standardized Measures of Well-Being

Self-Esteem

Youths’ self-esteem was assessed with a four-item, youth-rated scale from the Canadian National Longitudinal Study on Children and Youth (NLSCY). Each item was rated on a five-point Likert scale ranging from “false” to “true.” Items include statements such as “A lot of things about me are good.” Scores range from 4 to 20, with higher scores indicating higher self-esteem (α = .60). The reliability of this scale was lower at this data collection point than in previous data collection points (α = .75 at grade 6 and α = .85 at grade 9).

Family Functioning

The quality of family functioning was assessed using seven parent-rated items from the General Functioning Scale of the Family Assessment Device (FAD) (Epstein et al. 1983) (α = .75). The items included statements such as “in times of crisis we can turn to each other for support” and “we express feelings to each other,” which were rated on a four-point scale (1 = strongly disagree to 4 = strongly agree). Scores ranged from 4 to 28, with a higher scored indicating better family functioning. The reliability of this scale with the current sample was consistent with that found by the scale developers (α = .78).

Community Involvement Index

The level of youths’ community involvement was assessed with a youth-rated index that was created using seven items from the NLSCY. The dichotomous items involved yes/no responses about youths’ involvement in various activities within the past 12 months (e.g., fund raising, supporting a cause, helping neighbors or relatives). Scores ranged from 0 to 7, with higher scores indicating more involvement.

Procedure

The procedure for the larger study and this one, which is a subset of it, is the same for participants in the intervention sites and participants in the comparison sites. In the larger study, site researchers collected quantitative data through in-person interviews with children and parents. For this part of the study, all the parent (e.g., family functioning) and youth (e.g., self-esteem) measures were obtained from both the intervention and comparison samples as part of the grade 12 data collection. Additional narrative data were collected from a sub-sample of youth shortly after they had completed the quantitative interview protocol. The site researchers who interviewed the youth were trained in an 8-h session that addressed ethical issues, the nature of qualitative data, and tips for how to ask open-ended questions and probes, with particular attention to discussing personal matters with youth. This session included conducting mock interviews with feedback provided by the lead researchers.

The same recruitment procedures were used for both Better Beginnings and comparison youth. Site researchers contacted potential participants and invited them to be a part of the study. For those who consented, the site researchers scheduled and conducted a semi-structured individual interview that was audio recorded. All youth participants were provided an honorarium of $25 for their participation. The interviews were approximately 45 min in length and were done in privacy at a convenient place for youth, typically in the youths’ homes. After an interview was completed site researchers uploaded the digital audio recording to a secured website. The interview was later transcribed verbatim by hired research assistants. These transcripts were then coded as was described earlier. The lead researchers listened to the audio-recordings of a sample of the narrative interviews and provided feedback to the site researchers on their interviewing. Also, monthly teleconferences were held with all the site researchers and lead researchers to review progress, share experiences, and troubleshoot problems.

Statistical Analyses

The first hypothesis, that the stories of youth involved in the Better Beginnings sites would show more enhanced development than the stories of youth in the comparison communities, was tested by comparing stories from the Better Beginnings sites and comparison sites with respect to each of the story indices (e.g., specificity, ending resolution). Because cases were clustered within Better Beginnings and comparison sites, and sites were chosen non-randomly, standard errors had to be computed with due allowance for clustering, and on the understanding that inferences should not be drawn beyond the set of sites from which data were available. STATA (Stata Corporation 2005) was used to obtain appropriate standard errors for the influence of covariates. Covariate-adjusted residuals were then used to compare Better Beginnings and comparison sites. Tests for differences between the combined Better Beginnings sites and combined comparison sites were tested using a Z statistic, which was computed by dividing the regression coefficient by the bootstrapped standard error, again in STATA. Two-tailed tests of significance (p < .05) were employed for all analyses. A standard set of covariates was used in all analyses, including and birth year and gender of the interview respondent (typically the mother), gender of child, marital status, single parent status, respondents’ education, employment status, family income, cultural category (Anglophone or Francophone), and immigrant status.

To test the second hypothesis that standardized measures of individual (self-esteem), family (positive family functioning), and community (involvement in community activities) well-being will be directly correlated the dimensions of turning point stories, two different methods of analysis were used. First, Pearson correlations were computed between the standardized measures of well-being and the turning point story dimensions. Second, for those standardized measures of well-being that were found to be significantly correlated with at least one of the turning point dimensions, simultaneous multiple regression analyses were computed. The standardized measures of well-being were entered as the independent variables, while the turning point dimensions were the dependent variables. Two-tailed tests of significance (p < .05) were employed for all the correlational and regression analyses.

Results

Preliminary Analyses

The percentages of participants who told different types of stories were as follows: autonomy (49%), relationship (48%), achievement (23%), mortality (16%), and other (15%). Note that since the story categories were not mutually exclusive, the percentages do not sum to 100%. Also, the types of story did not differ significantly for Better Beginnings and comparison participants.

The means, standard deviations, and ranges for the narrative dimensions and the standardized measures of well-being are presented in Table 1. Mean scores were in the middle to high end of the range for most narrative dimensions. Means for standardized measures of well-being were moderately high for self-esteem and family functioning, while the mean for community involvement was moderately low. Table 1 also shows the intercorrelations of all the measures. All intercorrelations of the turning point story dimensions were significant, with correlations ranging from .41 to .76. For the standardized measures, self-esteem was not significantly correlated with the other two measures, but family functioning and community involvement were significantly correlated with one another, r = .32, p < .01.

Table 1 Sample sizes, means, standard deviations, ranges, and inter-correlations of all measures
Full size table

Narrative Outcomes

Analysis of the differences in the narratives of the Better Beginnings and comparison site youth (presented in Table 2) showed strong support for the hypothesis that Better Beginnings youth would demonstrate more enhanced development than the comparison site youth. The overall analyses indicated that the stories of Better Beginnings youth had more positive ending resolutions, described more personal growth, showed more meaning-making, were more coherent, and showed a higher level of affect transformation than were the stories of youth from the comparison communities. All of these differences were statistically significant and the effect sizes for the overall comparisons were moderate to large (.37–.76), suggesting that participation in Better Beginnings programming had an impact on the youths’ narratives. The most pronounced and consistent effects were with regard to arguably the most important of the indices—those relating to personal growth and meaning-making. The only index on which Better Beginnings and comparison site youth did not differ significantly was the specificity index, p < .07.

Table 2 Covariate adjusted means, standard deviations, Z test, and effect sizes for narrative dimensions by Better Beginnings project sites and comparison sites
Full size table

Relationship of Turning Point Story Dimensions to Standardized Outcome Measures

Table 1 shows the correlations between turning point story dimensions and standardized measures of well-being. We found that self-esteem was significantly correlated with ending resolution, r = .28, p < .01, personal growth, r = .23, p < .05, and meaning-making, r = .23, p < .05. Community involvement was significantly correlated with specificity, r = .25, p < .05, meaning-making, r = .27, p < .01, and coherence, r = .24, p < .05. There were no significant correlations between family functioning and any of the narrative dimensions.

Simultaneous linear regressions were run using self-esteem and community involvement as predictors of each narrative dimension (see Table 3). Since no significant correlations were found with family functioning, it was not included as a predictor in the regression analyses. Similar to the correlations, we found that self-esteem was a significant predictor of ending resolution, β = .26, p < .01, and personal growth, β = .21, p < .05. The significant correlation between self-esteem and meaning-making was not sustained in the regression. The results of the community involvement regressions were also consistent with the correlations, as the community involvement index significantly predicted specificity, β = .27, p < .01, meaning-making, β = .24, p < .05, and coherence, β = .22, p < .05. Overall, each model was significant at p < .05, except for affect transformation.

Table 3 Simultaneous regression predicting turning point dimensions from self-esteem and community involvement
Full size table

Discussion

Narrative Outcomes

The findings from this project indicate that a narrative approach can have significant value in evaluating long-term prevention programs. Narratives, and in particular, turning point stories, have not previously been used as a way of examining long-term program outcomes. Previous research had shown Better Beginnings, Better Futures to be beneficial for families and communities in the short- and medium-term in terms of various quantitative outcome measures, including improved social, emotional, and physical health, increased positive parenting behaviors, and improved neighborhood and school characteristics (Peters et al. 2003, 2010). However, this quantitative evaluation did little to allow an understanding of the real differences in the lived experiences of youth who had participated in Better Beginnings versus those who had not.

The stories of youth from Better Beginnings sites were significantly more developed than the stories of youth from the comparison sites on a number of important dimensions (e.g., personal growth). Since we know that these narrative dimensions are correlated with indicators of well-being at multiple levels, including individual happiness (Bauer et al. 2008), family functioning (Dumas et al. 2009), and community involvement (Frensch et al. 2007; Hasford et al. 2008), examining the youths’ turning point stories allows us to conclude that youth from Better Beginnings communities experienced higher levels of well-being than youth from the comparison communities. Moreover, the ESs of the intervention on the narrative dimensions were of greater magnitude (.45–.76) than those typically reported in the literature on the effects of ECD programs on more traditional outcome measures.

Our experience suggests that narrative analysis can produce significant added value in understanding and articulating programmatic outcomes, particularly in the case of long-term prevention-oriented programs such as Better Beginnings. We encourage other researchers engaged in evaluating similar efforts to consider the addition of a narrative analysis component to allow for richer depth of understanding of programmatic outcomes and greater trustworthiness of evaluation results.

Relationship of Turning Point Story Dimensions to Standardized Measures of Well-being

The second hypothesis concerned the relationship between turning point story dimensions and standardized measures of well-being: self-esteem, family functioning, and community involvement. Two of these three measures (self-esteem and community involvement) were related to various turning point story dimensions. Self-esteem was found to be positively correlated with the dimensions of ending resolution, personal growth, and meaning-making. These findings are consistent with those of McAdams et al. (2001b), who found that self-esteem was significantly positively correlated with the dimension of redemption in mid-life adults’ and college students’ narrative accounts of high point, low point, and turning point episodes in their lives. Community involvement was also found to be related to various turning point story dimensions, in particular, the dimensions of specificity, meaning-making, and coherence. These results are consistent with other studies that have also found a relationship between community involvement and the stories that young people tell about turning points and other aspects of their lives (Frensch et al. 2007; Hasford et al. 2008).

The relationship of self-esteem and community involvement with young people’s narrative accounts of turning points in their lives suggests that these variables may serve as mechanisms through which involvement in Better Beginnings produces enhanced well-being and a more positive sense of identity. Participation in Better Beginnings allows children to engage in new activities under the supervision of trained staff. This helps them build skills and experience positive outcomes, contributing to a sense of mastery and self-esteem. These feelings of mastery and self-esteem, and the skills on which they are based, can help them deal with adverse experiences and develop a positive sense of identity, which is reflected in their narratives.

The enhanced opportunities for community involvement provided by Better Beginnings can also lead to enhanced well-being and identity development. Hasford et al. (2008), as well as Prilleltensky and Prilleltensky (2006), suggest that the community is an important context for learning and that participating freely in social, community, and political life promotes individuals to gain maturity and communication skills, and to bring an optimistic outlook to situations. Similarly, individuals who are meaningfully engaged in community activities will have encountered experiences and developed skills that would also help deal with adverse experiences and develop a positive sense of identity, which is reflected in their turning point stories. Future studies would benefit from the inclusion of more comprehensive measurement of skills, self-esteem, and community involvement, to establish the mechanisms through which participation in ECD programs produces the kinds of positive outcomes evident in youths’ narratives.

Based on past research (e.g., Dumas et al. 2009), it seems reasonable to expect that characteristics such as family cohesion, communication, and support would be positively associated with various dimensions of a turning point story. However, of the three standardized measures of well-being used in this study, family functioning was the only measure not found to be positively associated with any of the turning point story dimensions. It is important to note that whereas the measures of self-esteem and community involvement were completed by the youth, the family functioning measure was completed by parents. Parent and youth perceptions of family functioning may differ and it is possible that the parent-completed measure did not capture the relationship between family functioning and dimensions of turning point stories told by youth. The lack of relationship found in the current study, then, does not necessarily mean that there is no relationship between family functioning and dimensions of turning point stories. To better understand this relationship, future work will need to examine family functioning in a more comprehensive manner (and perhaps longitudinally).

Limitations

There are several limitations to this study. First, we relied solely on youth-rated and parent-rated outcome measures. Future research could also incorporate teacher-rated measures and archival measures (e.g., crime in the community). Second, the reliability of the measure of self-esteem was relatively low (α = .60) compared to the reliability estimates that we found at Grade 6 (α = .75) and Grade 9 (α = .85). Third, there is the inherent limitation of determining causality in longitudinal, community-based intervention research that uses quasi-experimental designs. The extended time lapse since the youth participated in the program and the complexities of local contexts make it difficult to attribute the findings solely to the intervention. Fourth, the cross-sectional analyses restrict the ability to determine the direction of causality between narrative dimensions and other outcomes (e.g., whether narrative dimensions factors shape self-esteem or vice versa) and the mechanisms of change (e.g., how intervention conditions may have produced changes in particpants’ narratives). These methodological limitations, in addition to inconsistent findings in the relationships between narrative dimensions and outcomes, suggest that future studies should emphasize specification and testing of theory, more comprehensive measurement of outcomes, and larger sample sizes, in order to better understand the mechanisms of change.

Despite these limitations, the relationship between measures of self-esteem and community involvement and various dimensions of turning point stories highlights the importance of these different types of measures of well-being of youth and suggests the utility of a narrative approach for the evaluation of the long-term outcomes of ECD programs.