Adopted Chinese Girls’ Character Strengths from Preschool to School Age: A Longitudinal Exploratory Study

Sample

Internet discussion groups have become a popular forum for adoptive families to communicate on all aspects of adoption-related topics. In this study, participants were recruited through these discussion groups, as well as major adoption agencies. In early 2005, through one group moderator, a recruitment letter, with an introduction of the research project by this moderator, was posted on the message board of the internet moderators’ group. The other moderators were asked to disseminate the recruitment letter to members of their respective groups. At the same time, the recruitment letter, together with the same introduction, was mailed to the directors of 10 adoption agencies in the U.S. (e.g., Chinese Children’s Adoption International, China Adoption With Love, Inc., and Alliance for Children).

Overall, the study was endorsed by 120 internet discussion groups and 6 adoption agencies. The groups included organizations associated with Chinese adoptions in general (e.g., Families with Children from China; Raising China Children), as well as groups with a more specific focus. The latter included (a) groups for families of children adopted from certain regions of China and (b) groups organized around specific developmental issues and topics, such as attachment, special needs, identity, and general post-adoption adjustment. As most families belong to more than one organization, some received information about the study simultaneously from multiple sources. Parents interested in participating in the study contacted the research program directly with information about the number of children they had adopted from China, number of biological children, age of each child, and a regular mailing address.

A total of 1,001 families from the U.S. and 91 families from other countries (e.g., Canada, Australia, and the U.K.) requested surveys. The U.S. families were from 49 states. A packet that included study instructions, consent form, demographic survey, survey on the adopted child’s pre-adoption experience, and the Child Behavior Checklist for 1.5-5 years (CBCL/1.5-5) (Achenbach & Rescorla, 2000) or CBCL for 6-18 years (CBCL/6-18; Achenbach & Rescorla, 2001) was mailed to the adoptive parents via regular mail within two days of receiving the parents’ request. An email confirming the mailing of surveys was sent to the adoptive parents within a week thereafter. For the returned surveys, an email thanking the family was sent and for surveys that were not returned within three weeks, an email reminder was sent to the parents. A total of 852 families (about 78.1%) returned the surveys. The total number of children within these families was 1,193, of whom 1,122 (94%) were adopted from China; the rest were the biological children of the adoptive families.

Two years later in the spring of 2007 (Time 2), 780 of the 853 families (for 72 families, contact information was no longer valid) were successfully contacted for the second phase of the study (one family declined to participate without explanation). Using a similar method from Time 1, a packet containing the CBCL was sent to the 780 adoptive families. Data from 675 families (86.5%) with 1,111 children were obtained. These families had 882 adopted Chinese children and the rest were the biological children of the adoptive parents.

Data Source for the Current Analysis

In both phases of the study, the adoptive mothers were asked to “Please describe the best things about your child”. Their responses were used to identify character strengths of the adopted Chinese children. Our data gathering approach was similar to the study of young children’s personality, in which parents responded to one open-ended question about their young children (Kohnstamm, Halverson, Mervield, & Havill, 1998). Additionally, Glascoe and Dworkin (1995) suggested that parents are often the most reliable source of their children’s behaviors. In a recent study on children adopted from China, Tan (2010) successfully identified behaviors that concerned the adoptive parents most using a similar approach. Equally importantly, our method closely resembles the method used in the only other study of the VIA Classification of Character Strengths among young children by Park and Peterson (2006b).

Demographic variables were gathered using a child and family background survey at Time 1. Parents were asked to report on current marital status, presence of other children living in the home, household income, parents’ age, child’s date of birth, age at adoption, and date of data collection. Children’s chronological ages in months were calculated from their date of birth and date of data collection. They were used as predictors for the logistic regression analyses in the current analysis.

Selecting Criterion for Children in the Current Analysis

In response to the need for longitudinal studies on young children’s character strength development, we chose to study a sub-sample of adopted Chinese girls who were 4-5 years of age at Time 1 and were 6-7 years at Time 2. We were interested in the transition from preschool to school age because developmentally this is a period for adopted children to start understanding the meaning and implications of adoption (Brodzinsky, Singer, & Braff, 1984), and socially this is also a period during which children’s social context expands significantly (Selman, 1980).

From the dataset, we identified 184 children who met the age criterion. After excluding seven boys and 10 girls with missing Time 2 data, 162 girls from 157 families (i.e., five parents had two girls) aged 4-5 years (i.e., preschool ages) and 6-7 years of age (i.e., early school ages) at Time 2 were included for this analysis. The attrition appeared to be random as there was no difference in any area between the children who dropped out and those who remained. At Time 1, they were 4.9 years (SD = 0.56) on average and were 6.8 years (SD = 0.55) on average at Time 2. These girls’ average age at adoption was 15.1 months (SD= 5.2, Range= 7-55). The adoptive mothers’ average age at Time 1 was 45.2 years (SD = 5.2). Over 90% of the mothers completed at least college, and nearly 60% had a masters degree or higher. The adoptive families’ median annual household income in 2004 was $80,000 to $89,999, with only 10% of families making less than $50,000 per year. Sixty-four percent of mothers were married or living with a spouse or partner, and 69% of children had other adoptive siblings in the home.

Data Coding Procedures

Paralleling the coding procedures used in Park and Peterson’s (2006b), we developed a codebook based on the VIA Classification’s 24 character strengths. We adopted the age-appropriate manifestations and behavioral indicators described in Park and Peterson’s study, along with the description of each strength (Peterson & Seligman, 2004) and additional relevant example descriptions based on initial readings of the data. An abbreviated version of the codebook is presented in Table 1.

While conventionally two raters are involved in coding qualitative data, as was the case in Park and Peterson’s content analysis, a more stringent approach with three raters (i.e., first and second author, and an advanced graduate student) was used for the current analysis to help further validate the themes emerged from the data and ease the limitation of having relatively brief responses.

During the training session, the first author provided didactic explanations of each individual character strength code as well as group discussions to clarify meanings. All three raters then coded 15 cases for practice, followed by discussing each rater’s line of reasoning to build initial consensus. This process facilitated refining the codes and establishing reliability, particularly for the responses that were somewhat ambiguous. For example, one mother described her child as ‘charming,’ and this resulted in discrepant independent coding. Because ‘charming’ is associated with pleasing others, the codes of Kindness and Social-Intelligence were discussed as possible codes. Through group discussion, it was agreed upon that “charming” implied the ability of knowing how to please others in different situations and that charming behaviors can be displayed for purposes other than trying to be kind; thus, ‘charming’ was considered a descriptor of Social Intelligence.

Table 1. Abbreviated Character Strength Codebook

Following this training session, the three raters independently coded 50 randomly selected cases. A total of 181 codes (M = 3.62 per child) were endorsed by at least one rater. Agreement on each code between any two of the three raters was fairly high (i.e., 75.1%), although the three-rater agreements were understandably lower (i.e., 58.0%). Instances in which there were discrepancies across raters were then discussed, similar to the initial try-out discussions described above, which refined the codes. The raters discussed the codes until consensus across the three raters was achieved.

Character strength codes were then grouped into the six virtues from the VIA Classification. This was done in an effort to reduce the likelihood for Type I errors during statistical analyses as it resulted in fewer numbers of statistical tests to be conducted. Specifically, character strengths data were grouped into the VIA Classification’s core virtues based on the presence or absence of any one or more character strengths associated with a virtue. Thus, similar to the character strengths, virtues were dichotomous variables, not summative scores. Specifically, the 24 character strengths were grouped in accordance with the VIA Classification’s six virtues-Wisdom and Knowledge, Courage, Humanity, Justice, Temperance, and Transcendence (see Table 1 or Table 3 for information on which character strengths comprise each of these virtues). Character strengths data were re-coded for each virtue based on the presence or absence of any one or more character strengths associated with a virtue. When none of the character strengths within a virtue category was coded for a child, the virtue was coded 0 (not present), but when one or more character strengths within a virtue category were coded for a child, the virtue category was coded a 1 (present). For example, if a child received codes for Kindness and Love, that child received a code of 1 for Humanity. A child who received a code for Love only was given a code of 1 for Humanity. Similar to the character strengths, virtues became dichotomous variables, not summative scores.

Data Analysis Plan

Frequency counts of the coded character strengths were used to determine the prevalence rates of each character strength in the sample at Time 1 and Time 2. Character strength prevalence rates at Time 1 and Time 2 were each rank ordered from lowest (1) to highest (24) and the Spearman ρ correlation analysis was used to compare the rank order profile of Time 1 to Time 2. Nonparametric statistical testing was also used to test for significant differences between prevalence rates of character strengths over time from preschool (Time 1) to school age (Time 2). Specifically, the McNemar test of marginal homogeneity was used since this test is appropriate for nominal data with matched pairs of subjects. The degree of similarity between Time 1 and Time 2 for each character strength was also analyzed through correlation analyses. Recognizing that 24 tests for the character strengths were conducted, more stringent significance levels were examined, including the use of Bonferroni corrections (.05 divided by 24). Each of the analyses described above were then also calculated for virtues.

To further reduce the likelihood of Type I errors, only the 6 broader virtues were used for further analyses to determine significant predictive relationships with additional child, parent, and/or family variables. Logistic regression rather than multiple regression was used with the virtues due to the fact that dichotomous criterion variables (i.e. presence or absence of virtues) were analyzed rather than continuous variables. To reflect several child, parent, and family variables, child’s age (at Time 1), age at adoption, household income, single or two parent household, mothers’ age, and presence or absence siblings served as the predictor variables. As a check for multicolinearity among these variables, inter-item correlations among these variables at Time 1 were found to be low, ranging from -.21 to .34, indicating sufficient independence for logistic regression.

Descriptive Statistics

At Time 1, parent descriptions of the child’s strengths ranged from 2 words (e.g., “affectionate, caring”) to 101 words (M = 21.2). After each description was coded, the average number of character strengths per child was 2.93 (SD = 1.62, Range = 0-8 strengths). At Time 2, the parent descriptions range was from 3 to 96 words (M = 17.7). At Time 2, the average number of character strengths for a child was 2.94 (SD = 1.27; Range = 1-6).

Table 2 summarizes the frequency of character strengths at Time 1 and Time 2, with the character strengths organized in ascending order based on the prevalence of the character strengths at Time 1. Considerable variability existed in the prevalence of each character strength in the sample. Thirteen of the 24 character strengths (e.g., Gratitude) at both times were found in less than 5% of the children. The remaining 11 character strengths ranged in prevalence from 10.5% (Bravery) to 58.0% (Love) at Time 1 and 8.6% (Love of Learning) to 48.8% (Kindness) at Time 2.

Although our sample includes only girls who were younger than Park and Peterson’s (2006b) sample, we included in the table the results of Park and Peterson for a rough comparison. Both similarities and differences were noted. Nonetheless, the considerable variability in frequencies at which the more prevalent character strengths were described makes the similarities and differences between samples in the rank ordering and actual frequency of character strengths more interpretable than within the rarely described character strengths that varied much less. The top two strengths at both time points, Love and Kindness, were the top two strengths among those reported in Park and Peterson’s sample. Humor was the third and fourth most prevalent character strength among the current sample and Park and Peterson’s sample, respectively. Character strengths of the Wisdom and Knowledge virtue (i.e., Love of Learning, Curiosity, and Creativity) were prevalent at lower rates among Chinese adoptees when compared to the non-adopted children, though the order of prevalence within both samples was the same (i.e., Love of Learning was less frequent than Curiosity, which was less frequent than Creativity). The most dramatic difference appeared with respect to Zest and Social Intelligence, which were much more prevalent among the adopted sample (i.e., 32.1% and 23.5% at Time 1, respectively) compared to Park and Peterson’s sample (i.e., 9.9% and 12.1%, respectively). Zest and Social Intelligence were among the top five most prevalent character strengths at both time points among the adopted sample, while only the ninth and eleventh most prevalent among Park and Peterson’s sample.

Table 2. Frequency of Character Strengths at Time 1 and Time 2 (N = 162)

As seen in Table 3, even greater variability existed in the prevalence of each virtue in the sample, ranging from 4.3% (Justice, Time 2) to 77.8% (Humanity, Time 1). We did not include the results from Park and Peterson (2006b) in the table since they did not provide the statistics in their study.

Change and Continuity in Prevalence from Preschool to School Age

When comparing the ranked character strengths from lowest (1) to highest (24) from both time points, the Spearman ρ correlation was .94 (p < .001), indicating very high similarity in the sample’s character strength profile from Time 1 to Time 2.

Table 3. Frequency of Virtues at Time 1 and Time 2 (N = 162)

Further analysis through McNemar tests of homogeneity showed that Love was the only character strength with a significant change (< .002) in prevalence over time. It decreased from being represented among 58.0 % of children at Time 1 to 40.7% of children at Time 2, χ2 (1, N = 162) = 10.72, p < .001. An increase in Kindness from 39.5% at Time 1 to 48.8% at Time 2 also approach the less stringent statistical significance (not applying a Bonferroni correction), χ 2(1, N = 162) = 3.70, p = .054. In terms of virtues, none were significant using a Bonferroni correction significance level (p < .008). Using the less stringent significance level, Courage was the only virtue that increased in prevalence from 43.2% of children at Time 1 to 53.1% of children at Time 2, χ2(1, N = 162) = 3.88, p < .05.

To determine if the same character strengths were described for the same children at both time points, we focused on the 11 character strengths that occurred within 10% or more of the sample (as there is too little variability in the rarely described character strengths).

As shown in Table 2, Kindness, Humor, and Perseverance were the only strengths that were correlated significantly when the most stringent p value was applied. With less stringent p values, there were five character strengths with moderately strong correlations (r = .19 to .29). The correlations between time points for Kindness and Humor were fairly strong (r = .35 and .41, respectively). More specifically, 64% of children with Humor as character strength at Time 1 continued to be described similarly at Time 2, and 70% of children with Kindness at Time 1 continued to be described similarly at Time 2. These results suggest both change and continuity in the children’s character strength development.

Child, Parent, and Family Variables Related to Virtues

To test relationships between the (presence or absence of) virtues and child, parent, and family predictors, logistic regressions for Time 1 and Time 2 were performed with several predictors, including child’s age, age at adoption, family income, single or two parent household, mother’s age, and whether the child was the only child in the household.

As shown in Table 4, most of the variables did not have a significant impact on the odds of the virtues being present with a few exceptions. For instance, during preschool years (Time 1), the odds were higher for older children as well as children who were older at the time of adoption to show character strengths that indicated Courage. Interestingly, at Time 2, children from single parents household were about 3 times more likely than their peers from two-parent households to show character strengths that were indicative of Humanity.

Table 4. Logistic Regression of Virtues at Time 1(T1) and Time 2 (T 2)

Discussion

While deficit-based approaches have provided important insights in the adopted children’s post-adoption problematic adjustment, there is a need for strength-based approach in understanding adopted children’s post-adoption resilience and flourishing. To the best of our knowledge, the current study serves as the first longitudinal study of the VIA Classification among young children and the second study on internationally adopted children to primarily assess a broad spectrum of strengths.

Our study showed that at both times, the adopted Chinese girls were similar to the non-adopted children in the study by Park and Peterson (2006b) in average number of identified strengths. The adopted Chinese girls at both times also share a similar character strength profile to the children from the general population in Park and Peterson’s (2006b) study. For instance, among the 13 infrequent character strengths in our sample, 12 overlap with what was reported in Park and Peterson’s sample. At the same time, 10 of the more frequent strengths overlap between the two studies, suggesting that the adopted Chinese children were similar to the non-adopted children. Given that the adopted Chinese girls had spent some time in the orphanages, this finding is impressive and highlights the resilience of child development. Future research is needed to tap into specific mechanisms within the child and adoptive family that foster the adopted children’s resilience from pre-adoption orphanage deprivation.

However, there are indeed some differences in the prevalence of specific character strength. For instance, over 30% of the adopted Chinese girls were reported to have Zest, as compared to under 10% of the sample in the study by Park and Peterson (2006b). In an earlier study by Tan and Marfo (2006), they had speculated that a selection process might have been used by the Chinese child welfare authorities to identify babies that were mostly likely to thrive in adoptive environment. Children who exhibited characteristics that might signal Zest may have been deemed most suitable for international adoption by the Chinese authorities. Of course, more studies are warranted to help determine whether the differences might have resulted from differences in the sample’s age and gender differences or differences in earlier psychosocial experiences between our study and theirs.

Change and Continuity in Character Strengths

The longitudinal design provides good insights into both change and continuity of the adopted Chinese girls’ character strength and virtue from preschool ages to early school-ages. Results showed that most character strengths and virtues remained stable over a two-year span. There was, however, a significant decrease of Love and a significant increase of Courage. In terms of Love, it should be noted that while it decreased in frequency, it was still the second most prevalent character strength when the children were in school ages. One interpretation for this decrease could be related to parents’ greater sensitivity to establishing a loving relationship with an adopted child early in their post-adoption lives. A concern for the development of an attachment disorder, characterized as an inability to give or receive love and affection (e.g., Levy & Orlans, 2000), might initially lead adoptive parents to be particularly vigilant about their child’s expression of love. As children get older, that heightened vigilance may fade. The decrease in Love could also make sense from a developmental perspective, as the expression of love through affection for a caregiver is something that can be seen from infancy.

With increasing age and cognitive maturation, other developmentally salient strengths (e.g., learning in school) might become the focus of the adoptive parents’ descriptions. Courage, a virtue comprised of Authenticity, Bravery, Perseverance, and Zest and described as an “exercise of will to accomplish goals in the face of opposition, either external or internal” (Peterson & Seligman, 2004, p. 199), appears to become more prevalent over the course of the transition from preschool to school ages. Given that these children were racially different from their parents, future research on their children’s ethnic identity formation might help shed some light into whether “Americanization” plays a role. On the other hand, virtues of Justice, Temperance, and Transcendence were endorsed at low rates across this transition, indicating they would not increase significantly until older ages.

At the individual level, some character strengths were more consistent over time than others. Humor and Kindness were the most consistent and highly correlated over time (of the 11 more prevalent strengths for which the frequency was high enough to be meaningfully interpreted), indicating a higher likelihood that humorous or kind girls in their preschool years will continue to exhibit those attributes in their early elementary school years. Most of the character strengths were either not significantly correlated or had weak positive correlations from Time 1 to Time 2, which is interesting given that the rank ordered character strength profile for the sample was so highly correlated. One speculation is that while there appear to be clear distinctions between the types of character strengths that are displayed frequently during this age range, individual children for the most part have not yet developed stable signature strengths at this young age. More research is needed to explore the malleability of one’s character strengths throughout childhood.

Another key variable to consider in adoption research is the age at which the child is adopted. Among Romanian adoptees, children adopted after 24 months were at greater risk for higher levels of problem behaviors (Meese, 2005) and lower levels of behavioral and emotional strengths (Pearlmutter et al., 2008). Nevertheless, age at adoption in this sample was only predictive for Courage, with only a relatively trivial increase in likelihood of occurrence. This finding of a lack of significant or strong associations with age at adoption was consistent with existing research (e.g., Tan & Marfo, 2006). This finding may have been influenced by the relatively young ages at adoption. Only 13% of the sample was adopted at 24 months or older.

Finally, children in single–mother households appeared to score higher on the virtue of Humanity at Time 2 than their peers in dual-parent households. According to Groze and Rosenthal (1991), adopted children may benefit from single-parent families’ relatively simple relationship structure. More research is needed to understand how family structure affects adopted children’s character strength development.

Limitations

Our study focused on trans-racially adopted Chinese girls. In the U.S., the stereotypes associated with children of Asian descent (e.g., being model minorities) has been frequently reported (e.g., Lee, 1996). Thus, our sample’s gender and racial background might have played a role in our results. It is beyond the scope of the current study to determine to what extent the very fact that these children were Chinese girls might influence the adoptive parenting, yet this issue deserves attention in future research. The results from our study might not generalize to trans-racial adoptees from a different racial background or from a domestic adoption, in-racial adoptees, or male adoptees.

Using one open-end question to gather data was a notable limitation of our study. However, similar approaches have been successfully used in many studies (e.g., Park & Peterson, 2006a, 2006b). Nevertheless, research using a more systematic approach could help learn more about the development of character strength in young children and help validate the method used in the current study. Potential biases associated with non-random, convenience sampling procedures should also be considered when evaluating the findings of the current study, despite a 78% response rate at Time 1 and 87% response rate at Time 2. These considerably high response rates for mail-based survey research may have helped mitigate biasing effects of self-selection.

Implications

Overall our study showed that the character strength profile among the current sample of adoptees remained relatively stable after transitioning from preschool age to school ages, and the profile was similar to that of a previously reported US sample of non-adopted young children. Based on this longitudinal data and comparison to the US non-adopted sample, certain character strengths appear to be ones that develop at younger ages (e.g., Love, Kindness, Humor, Zest, Social Intelligence), while others do not appear to manifest at young ages (e.g., Gratitude, Forgiveness, Modesty, Prudence).

Transitioning to school can present challenges to both adopted and non-adopted children and their parents. Examining positive characteristics of internationally adopted Chinese girls during the transition can broaden our understanding of their post-adoption development. The similarities help support some generalizable understanding of character development in young children. The differences, however, could provide some targets for building upon unique strengths. For instance, given their noticeably high prevalence of Zest and Social Intelligence, the adopted Chinese girls may benefit from educational activities that tap into these strengths.

Conclusion

China has been the leading country of the US international adoptions for a decade (US Department of State, 2011). Professionals involved with children and families will inevitably encounter many of these adoptees in their respective fields. It could prove useful to understand which strengths to expect based on a child’s developmental stage and capitalizing on their unique strengths in any efforts aimed at fostering their resilience and ability to thrive.