1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

School project, need help!

Discussion in 'Chit Chat' started by Fiender, Jan 20, 2009.

  1. Fiender

    Full Member

    Joined:
    Nov 26, 2007
    Messages:
    267
    Likes Received:
    0
    Location:
    Rochester, NY
    Gender:
    Male
    Sexual Orientation:
    Gay
    Out Status:
    Out to everyone
    I'm writing my term paper about negative ways in which religion affects society.
    Among the subjects I'm going to touch upon is abortion versus adoption.

    I'm trying to find out how many children are placed into adoption agencies, orphanages, foster care (essentially any facility that isn't a permanent family), compared to how many of those children are actually adopted into a family.

    I've tried googling for statistics but I've gotten nothing.
     
  2. riddlerno1

    Full Member

    Joined:
    Aug 2, 2008
    Messages:
    1,068
    Likes Received:
    0
    Location:
    London
    Gender:
    Male
    Sexual Orientation:
    Gay
    Out Status:
    Out to everyone
    You could talk about how religion affects the homosexual society? Just an idea.....
     
  3. Fiender

    Full Member

    Joined:
    Nov 26, 2007
    Messages:
    267
    Likes Received:
    0
    Location:
    Rochester, NY
    Gender:
    Male
    Sexual Orientation:
    Gay
    Out Status:
    Out to everyone
    Well, I am doing that as well but one of the other topics I will be touching on s how religious ideals affect a person's view on abortion. One of the arguments against abortion is that the mother could simply give the child up for adoption and that's why I needed the adoption statistics.
     
  4. jotheoneandonly

    Joined:
    Jan 13, 2009
    Messages:
    147
    Likes Received:
    0
    Location:
    Cambridge
    i think that information is government protected... look what i found

    In September 30, 2001, according to the AFCARS - Adoption and Foster Care Analysis and Reporting System, 50,000 children were adopted from foster care in 2001. 59% of these children were adopted by their foster parents, 23 % by relatives, and 17% from non-relatives. The average age of these children was 7 years old.
     
  5. jotheoneandonly

    Joined:
    Jan 13, 2009
    Messages:
    147
    Likes Received:
    0
    Location:
    Cambridge
  6. Fiender

    Full Member

    Joined:
    Nov 26, 2007
    Messages:
    267
    Likes Received:
    0
    Location:
    Rochester, NY
    Gender:
    Male
    Sexual Orientation:
    Gay
    Out Status:
    Out to everyone
    This is for a term paper, so I need a website, or and official source or something (I think :icon_sad:slight_smile:
     
  7. jotheoneandonly

    Joined:
    Jan 13, 2009
    Messages:
    147
    Likes Received:
    0
    Location:
    Cambridge
  8. Fiender

    Full Member

    Joined:
    Nov 26, 2007
    Messages:
    267
    Likes Received:
    0
    Location:
    Rochester, NY
    Gender:
    Male
    Sexual Orientation:
    Gay
    Out Status:
    Out to everyone
  9. Tom

    Tom Guest

    Try something such as religion + world relations + hostilities, there must be some (un)surprising facts about how many times them three words crop up togther.
     
  10. Numfarh

    Numfarh Guest

    With the powers of the McGill Library Database in my hands, I will provide an abstract of such importance that you will demand to see the article!

    Predictors of children in foster care being adopted: A classification tree analysis

    Jessica Snowden, 1, a, , Scott Leona and Jeffrey Sierackia

    aLoyola University Chicago, Department of Psychology, Chicago, Illinois, United States


    Received 6 December 2007; revised 30 March 2008; accepted 30 March 2008. Available online 9 April 2008.

    I can't directly link you to the article, but if you are interested, I can send it in a PM. With all the important stuff like what journal.

    EDIT: I, for obvious reasons, cannot send you a PM. I could post it on your wall though.
     
  11. Fiender

    Full Member

    Joined:
    Nov 26, 2007
    Messages:
    267
    Likes Received:
    0
    Location:
    Rochester, NY
    Gender:
    Male
    Sexual Orientation:
    Gay
    Out Status:
    Out to everyone
    Obvious reasons, what do you mean?
    But eah, I'll take all the information I can get as long as It comes with a source that I can cite.
     
  12. Numfarh

    Numfarh Guest

    Well, you are not a Full Member so I can't PM you. I only noticed afterwards. Here's the article and one of the tables that I think is most pertinent.

    Children and Youth Services Review
    Volume 30, Issue 11, November 2008, Pages 1318-1327

    Predictors of children in foster care being adopted: A classification tree analysis


    1. Introduction
    Approximately 800,000 children are currently served by the child welfare system, with 500,000 of these children in the foster care system. Of these children, approximately 50,000 are adopted each year (U.S. Children's Bureau, 2007). Until 1997, the number of children in child welfare was even greater; recent decreases are due in part to the emphasis on family reunification for children in the system. Recognizing that children were literally growing up in foster care, the Adoption and Safe Families Act (ASFA) was enacted in 1997 by the federal government (Hannett, 2007). The primary goal of ASFA is to focus on permanency for children in foster care ([Gittens, 1994] and [Hannett, 2007]). When returning home is safe and appropriate for the child, this is, of course, the preferred method of achieving permanency. However, in many cases, family reunification is either not feasible or not in the best interest of the child. When returning home is not a viable option, adoption is the primary way that children can achieve permanency. Therefore, understanding the factors that impact adoption is critical to both evaluating and improving the child welfare system.

    Prior research has primarily studied “main effects” in regard to adoption rates and has not explored the ways in which variables (e.g., race/ethnicity, age, mental health status, state of adoption) might interact in complex ways to predict adoption. It is the primary hypothesis of this study that the probability of being adopted is determined by a multitude of interacting factors, creating multiple unique pathways to adoption. This study uses an exploratory multivariate analysis technique known as Optimal Data Analysis (ODA) to determine possible multiple pathways to adoption using Adoption and Foster Care Analysis and Reporting System (AFCARS) data.

    2. Predictors of adoption
    Prior research has uncovered numerous individual variables that predict adoption in “main effects” multivariate analyses. These factors include demographic variables, such as age and ethnicity, presence of mental illness/symptomology, and factors related to prior placement in the child welfare system, such as number of previous placements, type of placement, and state of placement.

    2.1. Demographic variables
    Age is amongst the most frequently studied variable in adoption research. Not surprisingly, previous research has found that the older children are the less likely to be adopted ([Courtney and Wong, 1996], [Earth, 1997] and [Kirton et al., 2006]). In general, infants are the most likely to be adopted, and the likelihood for being adopted decreases as age increases (Connell, Katz, Saunders, & Tebes, 2006). Although a specific cutoff age is difficult to ascertain, McMurtry and Lie (1992) found that children over the age of 8.6 were not likely to be adopted. Possible explanations for the increased probability of younger children being adopted included case worker concerns over the safety of young children and a greater interest in adopting younger children (Connell et al., 2006). In addition, older children are also at an increased risk for placement disruption after adoption, for this reason, they may be viewed as less attractive to potential families ([Fensbo, 2004], [Rosenthal et al., 1988] and [Rushton and Dance, 2006]).

    In addition to age, researchers have examined the relationship between other demographics and likelihood of adoption. Previous research has not linked gender to adoption status ([Brooks et al., 2002] and [Connell et al., 2006]); however, several racial differences in adoption rates have been documented. For instance, Earth (1997) found that after controlling for the age of the child, African American children are five times less likely than Caucasian children and two and a half times less likely than Latino children to be adopted. Other studies have also offered evidence that Caucasian children are more likely to be adopted than African American and Latino children (Brooks, James, & Barth, 2002; Courtney & Wong, 1996). However, the relationship between race/ethnicity and adoption is inconclusive; previous research using retrospective data has found no difference across ethnic groups on rates of exit into adoption (McMurtry & Lie, 1992) It may be that the inclusive results found in previous studies reflect the fact that researchers have primarily examined whether race or ethnicity impacts probability of adoption under specific conditions, such as adopting across racial/ethnic backgrounds.

    2.2. Mental and physical health status variables
    Children in foster care are up to nine times more likely to have a mental illness than children not in foster care (McIntyre & Keesler, 1986), and 50% to 80% of children in foster care are in need of mental health services ([Armsden et al., 2000], [Landsverk and Garland, 1999] and [Thompson and Fuhr, 1992]). In addition, children in foster care are at risk for a variety of problems, including health problems (25% have a chronic health problem and 30% have a disability) and school problems (Vandivere, Chalk, & Moore, 2003). Researchers have attempted to determine the relationship between mental and physical health variables and adoption, however, the results vary. While several studies have found that having a disability puts children at risk for not being adopted (e.g., McMurtry & Lie, 1992), others have found that this variable is unrelated to adoption status (Connell et al., 2006). Courtney and Wong (1996) found that children with health problems or mental disabilities were less likely to be adopted. When examining solely mental health disabilities, Connell et al. (2006) found having a diagnosed emotional or behavioral disorder to lead to a decrease in adoption.

    2.3. Family and placement variables
    Factors surrounding the child's birth family, reason for removal, and placements in foster care have also been examined as they relate to likelihood of adoption, although the research on these factors is scarcer than on demographic factors. Regarding the child's birth family, Courtney and Wong (1996) found that while the structure of the birth family was not related to adoption, increased levels of poverty of the birth family were associated with decreased likelihood of adoption. Connell et al. (2006) found that children who are removed due to sexual abuse, housing problems, parents' inability to cope with the child, or child abandonment are less likely to be adopted than children who are removed for other reasons. Although children who were removed for physical abuse, child behavior problems, or parental substance abuse did not have different outcomes than children who were removed due to neglect, children who were removed due to a history of sexual abuse were less likely to be adopted (Connell et al., 2006). On the other hand, Courtney and Wong (1996) found that physical abuse put children at risk for a longer period in the child welfare system until adoption.

    The child's current and prior placements in foster care have been linked to adoption. Placement in any setting other than a foster home (i.e., group homes, residential treatment, etc.) places children at less likelihood for adoption (Courtney & Wong, 1996). Similarly, other research has found that children in group homes have been found to be less likely than children in foster homes to be adopted (Connell et al., 2006). Foster parents who were specialized, meaning that they have received training in working with difficult child populations, were more likely to consider adoption than those who are not (Courtney & Wong, 1996). Moving beyond the type of the placement, the number of removals and length of care have been linked to adoption. Children are more likely to be adopted after they are removed from their birth home more than one time ([Connell et al., 2006] and [Courtney and Wong, 1996]). In addition, children are generally not adopted until after they have been in the system for at least 18 months (Connell et al., 2006). This is likely due to the legal processes of adoption and the need to move towards termination of parental rights (Courtney & Wong, 1996).

    2.4. Examining the child welfare system
    The opportunity to study policy issues related to the child welfare system in the United States has expanded greatly over the past fourteen years. The primary reason for this change is due to the creation of the Adoption and Foster Care Analysis and Reporting System (AFCARS), a database founded in 1994 as a result of an amendment to Title IV-B/E of the Social Security Act (Section 427). This act required that all states report data on youth in substitute care each year, or face financial penalty. The greatest advance brought by AFCARS was its requirement for full and complete data reporting across all states and its use of a uniform data collection across 89 data fields.

    The current study examines AFCARS data, with the primary goal of using a classification tree analysis data analytic strategy to understand adoption rates in the context of youth and family demographic and clinical variables. Our strategy, Optimal Data Analysis (ODA), allows the creation of a multivariable classification tree model for predicting decisions regarding adoption. Therefore, ODA allows for a deeper understanding of the influence of multiple factors related to adoption, under the over-arching assumption that adoption must be understood as an inherently contextual phenomenon. To our knowledge, adoption decisions have not been previously studied using this form of analysis. Previous studies that have assessed the influence of demographic and clinical variables on adoption have studied the relationships of specific “main effects” variables on adoption (e.g., race, age, disability). However, the present study expands upon previous research by using ODA in order to explore interactions across multiple demographic and clinical variables. More specifically, by allowing the variables and their interactions to guide the analyses, the current methodology provides a more accurate understanding of the contextual nature of decision making, thereby allowing clinicians and policy makers to examine the interactions that occur in decision making while emphasizing the fact that decisions do not occur in isolation.

    3. Methods
    3.1. AFCARS data
    Adoption and Foster Care Analysis and Reporting System (AFCARS) is an administrative database developed under federal mandate (Code of Federal Regulations, Title 45, Volume 4; 45CFR 1355.40); states report on all youth in their child welfare system on a yearly basis. The goal of AFCARS is to collect data on youth in the child welfare system across the United States and Puerto Rico as a means of aiding fiscal and policy considerations at both the state and federal levels.

    Beginning in 1995, states were required to submit uniform data encompassing 89 fields. As of fiscal year 1998, states received a financial penalty for failure to comply with AFACRS requirements. These data fields focus on several core categories: (a) youth, biological, and foster family demographics (e.g., child age, foster parent race), (b) youth physical and behavioral clinical status (e.g., whether there is an emotional or physical disability), (c) and funding sources (e.g., IV-E assistance, amount of monthly subsidies). More specifically, information regarding basic child demographics (e.g., current age, age at removal, race, disability status), information about the child's current and previous placements (e.g., foster parent race, foster parent age, foster care payments, the number of previous removals), information regarding the reason for the child's removal (e.g., behavior problems, neglect, physical abuse, sexual abuse, parent or child drug use), and information about the child's biological family (e.g., age of the caregiver from who they were removed, structure of the biological family) (see Table 1 for more information).

    Table 1.

    Descriptive statistics for the sample

    Variable Foster care Adoption
    Demographics
    Male 53.0% 50.6%
    Female 47.0% 49.4%
    African American 35.2% 33.7%
    Caucasian 38.7% 41.5%
    Multiracial 19.4% 18.0%
    Current age 10.2 (5.8) 7.2 (4.3)
    Age at removal 6.9 (5.3) 3.1 (3.5)
    Age at previous adoption 2.4 (.85) 1.9 (.9)


    Disability
    Mental retardation 3.1% 3.1%
    Other diagnosed condition 12.4% 17.9%
    Diagnosed disability 24.2% 27.8%
    Physical disability 1.3% 2.3%
    Emotionally disturbed 12.3% 10.5%
    Visual or hearing disability 2.7% 2.2%


    Current/previous placements
    Foster family — married 30.0% 51.5%
    Foster family — unmarried couple 1.5% 1.9%
    Foster family — single female 24.6% 22.3%
    Foster family — single male 3.1% 2.1%
    1st foster parent race — African American 17.6% 18.6%
    1st foster parent race — multiracial 54.4% 37.3%
    1st foster parent race — Caucasian 25.9% 42.1%
    Same race as both foster parents 29.5% 36.0%
    Same race as neither foster parent 32.0% 28.6%
    Monthly foster payment 1419 (4905) 703 (2697)
    Child adopted previously 1.5% 11%
    # of placement settings 3.2 (3.3) 3.0 (2.2)
    1st foster caregiver age 46.9 (11.4) 45.1 (10.1)
    Out of state placement 2.9% 9.1%
    # of previous removals 1.3 (.72) 1.2 (.57)


    Reason for removal
    Child behavior problems 14.5% 4.0%
    Neglect 51.0% 58.3%
    Parent drug 17.8% 24.1%
    Relinquishment 1.9% 5.9%
    Caretaker can't cope 19.8% 22.4%
    Inadequate housing 10.3% 12.5%
    Drug abuse, child 2.9% 4.5%
    Parent alcohol 7.8% 9.2%
    Parent death 2.2% 1.3%
    Parent jail 5.4% 6.1%
    Child disability 2.9% 3.4%
    Abandonment 6.4% 6.9%
    Physical abuse 15.8% 14.8%
    Sexual abuse 6.4% 6.0%
    Alcohol abuse child 1.1% 1.1%


    Biological family
    Age of caretaker removed from 35.4 (9.9) 33.3 (8.7)
    Birth family structure — married 15.4% 14.0%
    Birth family structure — unmarried 10.5% 11.8%
    Birth family structure — single mother 42.2% 46.1%
    Birth family structure — single father 4.2% 2.8%




    3.2. Participants
    A sub-sample of the AFCARS data from the year 2003 was utilized for the current study. The study was stratified according to adoptions status and 30,000 children that were adopted in FY 2003 and 30,000 children that remained in foster care in FY 2003 were randomly selected and, in order to maximize generalizability, all participants had an equal likelihood of inclusion in the final sample. In the final sample, 51.8% were males (n = 31,091) and 48.2% were female (n = 28,989). The average age of the sample at removal was 4.05, and the average current age was 7.95. Overall, 34.5% of the sample were African American (n = 20,670), 40% were Caucasian (n = 24,072), 18.7% were multiracial (n = 11,214) and 1% were Hispanic (n = 682) (see Table 1 for all descriptive statistics).

    3.3. Optimal Data Analysis
    In order to create a prediction model for adoption, Optimal Data Analysis (ODA) was used ([Soltysik and Yarnold, 1993] and [Yarnold and Soltysik, 2005]). ODA is a non-parametric method of data analysis that maximizes the classification accuracy of the model that is created for a sample of data. For the current study, ODA software for Windows was used to aid in the selection of the optimal predictors for adoption.

    ODA's approach is superior to traditional statistical techniques for several reasons. First, ODA's approach to testing multivariate interactions is unique compared to other traditional techniques, such as ANOVA and regressions. Where traditional techniques require hypotheses about various interactions a priori and only allow for the testing of a limited number of interactions, ODA discovers the optimal interactions for the user, with no limit to the interactions that it can examine and no need to specify interactions a priori. In addition, other traditional techniques limit the number of individual predictors that can be tested for entrance into the model, whereas ODA is univariate and does not have issues surrounding multicolinearity. For this reason, each variable in the current study was entered without specifying a priori interactions or selecting only specific predictors. While some may argue that only variables that have been with justification from the literature should be entered, ODA allows the researcher to look at all possible variables surrounding the child without increasing error. Therefore, ODA also can help to further research by examining predictors that have not been explored before and seeing if these predictors are part of the optimal model of adoption. Lastly, traditional techniques assume that the predictors that are selected are important predictors for each member of the sample. ODA, on the other hand, allows for the variables to classify different parts of the overall sample. For instance, perhaps predictors vary by age ODA allows for the construction of a model that will provide the most important predictors for each of these segments of the sample (Yarnold & Soltysik, 2005).

    ODA allows for identification of both main effects and interactions. Main effects were identified through univariate ODA (UniODA) (Yarnold & Soltysik, 2005). First, UniODA was conducted for each variable, which provided information about which attributes are individually significant as a predictor for adoption. While this provides information about main effects (i.e., whether boys are more likely to be adopted), it does not provide information about other attributes that may be interacting with that variable in predicting classification. For example, being male in combination with being young may predict adoption, whereas either of these attributes alone may not predict adoption. Therefore, the next step was to construct a CTA that provides further information about predictors of being adopted.

    The first step in the CTA was to select the attribute from UniODA that has the greatest effect strength for sensitivity. This attribute is the optimal predictor of adoption. For this predictor, ODA will provide a decision rule. For instance, if age was the most optimal predictor, the decision rule that ODA created (perhaps that younger children are more likely to be adopted than older children) would be used to partition the sample. Then, ODA would be conducted using all attributes again, this time only for members of each section of the partitioned sample (e.g., individuals who are below a certain age), and the attribute that has the greatest effect strength for sensitivity will be selected, and the sample will be partitioned further. This continues for each “branch” of the CTA until the sample can be reliably subdivided no further. At this point, that branch will end, and the researcher continues to utilize ODA to complete the other branches of the tree (Yarnold & Soltysik, 2005).

    To decrease the chance that the model capitalizes on chance, nondirectional Fisher's exact probabilities were utilized for each attribute. In addition, a sequentially rejective Bonferroni procedure was used to ensure an experimentwise Type I error rate that increases the confidence that the results of the final tree model do not capitalize on chance. Attributes are removed (“pruned”) from CTA if their Type I error rates do not meet the Bonferroni criteria (Yarnold & Soltysik, 2005).

    A leave-one-out (LOO) analysis, which is conducted by ODA, also decreases the likelihood that the results of the CTA capitalize on chance. LOO analyses are performed on each attribute that was analyzed at each decision point. LOO analyses check the expected cross-sample stability of an attribute as a predictor by removing one case from the model and then using the other cases to determine the decision rule. For instance, if there were 5 cases, LOO would hold out case 1 and then use cases 2–5 to create the decision rule. Then, the created rule is used to classify the held out case (e.g., case 1) as either admitted or deflected. The LOO procedure is then repeated, holding out a different case each time until all cases have been held out and classified (e.g., would hold out case 2, and use cases 1, 3–5 to create a decision rule and then would use the rule to classify case 2). If the accuracy of the initial model is replicated when cumulated across the LOO analysis, the attribute is determined to be LOO stable and the conclusion is drawn that the attribute's predictive effect is likely to generalize in an independent sample (Yarnold & Soltysik, 2005). In sum, LOO decreases the chance that decision rules are created due to idiosyncrasies in the given sample; therefore, only attributes whose classification accuracy is stable in LOO analysis were included in the final CTA.

    After the CTA was constructed, post hoc analyses were conducted to assess the final classification performance of the model, examining overall classification accuracy, sensitivity, predictive value, and effect strength overall (Yarnold & Soltysik, 2005). Overall classification accuracy is the percentage of the total sample that the tree model correctly classified. The effect strength sensitivity (i.e., the percentage of the actual members of a given category that the model correctly classified) and effect strength specificity (i.e., a prognostic index that indicates the percentage of the predicted classifications into a given category that were correct) will also be examined. Lastly, effect strength overall is simply the mean of the effect strengths for sensitivity and for predictive value, and reflects the model's performance from both descriptive and prognostic perspectives.

    4. Results
    4.1. Main effects
    By examining the effect strengths, a child's age at removal was determined to be the strongest predictor of whether or not the child was adopted. When looking at main effects, children who were removed at age 5 or under were significantly more likely to be adopted than children who were removed after the age of 5. Similarly, the second largest predictor of whether or not a child was adopted was the child's current age. Children currently over the age of 11.7 were much less likely to be adopted than children under the age of 11.7. In addition to age, the third strongest predictor of adoption was the structure of the foster family, with both married and unmarried couples being predictive of adoption. Several other demographic, disability, family, and placement related factors were found to be predictive of being adopted. For example, both being female and being Asian, Hawaiian/Pacific Islander, or Caucasian increased the rates of adoption.

    Children who are classified as the most likely to be adopted are in Group H (93.73%). These children were between the ages of 5 and 11.7 at removal, have married foster parents, and were previously adopted before the age of 2 or over the age of 5. The group that was classified as the second likeliest to be adopted was Group O (74.85%). These children were under the age of 5 at their current removal, resided with a foster family that was either an unmarried or married couple, and had more than 1 placement during their current removal.

    Overall, 32,490 children were classified in the final classification tree model. The reasons for the decline in the overall number of children from the original sample are two-fold. First, not every child was classified in the final tree model because some children did not meet the criteria of any of the final classification groups. Secondly, missing data accounts for some of the decline in the number of children in the final model. Missing data is common in the AFCARS database for some of the non-required fields, such as foster family structure. Variables such as this entered the final tree model, suggesting that they are critical for understanding the interactions of variables that impact adoption, but then also reduce the number of children that are classified.

    The ODA model correctly classified 23,245 (71.55%) of the 32,490 children. This represents an overall classification accuracy with an absolute ES of 43.10% (where 0 = chance accuracy and 100% = perfect accuracy) and is of moderate effect size (Yarnold & Soltysik, 2005). In addition, 79.30% of children predicted as not being adopted were actually not adopted, representing a strong effect strength of 58.60%, and 67.43% of children who were predicted as being adopted were actually adopted, representing a moderate effect strength of 34.86%. The mean sensitivity for both those who were and were not adopted was 71.18%, representing a moderate ES of 42.36%. Combining measures of sensitivity and predictive value into a single index of classification performance, the tree model had an overall ES of 44.56%, also a moderate effect size.

    5. Discussion
    The purpose of this study was to evaluate the contribution of demographic and clinical variables to likelihood of adoption in a nationwide child welfare sample using a decision tree model. In addition to univariate predictors of adoption, the model also combines multiple predictors in order to assess multivariate indicators and determine interactions among variables.

    5.1. Univariate predictors
    The pattern of results found in this study is consistent with previous “main effects” adoption research. For example, being White or Asian American/Pacific Islander was a significant predictor of being adopted, while being African American, Hispanic, multiracial, or Native American was significantly associated with remaining in foster care (Courtney and Wong, 1996 M. Courtney and Y. Wong, Comparing the timing of exits from substitute care, Children and Youth Services Review 18 (1996), pp. 307–334. Abstract | PDF (1591 K) | View Record in Scopus | Cited By in Scopus (39)[Courtney and Wong, 1996], [Earth, 1997] and [McMurtry and Lie, 1992]). Age, another consistent predictor in the adoption literature, (Courtney & Wong, 1996), was the most robust predictor of adoption in the current study. The optimal cutoff point for maximizing predictability of adoption status was 11.7, slightly higher than has been found in prior research (McMurtry & Lie, 1992). However, as the discussion of the multivariate results below suggests, these univariate results are limited in their policy application value because their effect is moderated by other variables in the study.

    Children with a diagnosed physical disability were more likely to be adopted, while children with emotional disturbances were less likely to be adopted. The findings regarding emotional disturbances have been observed in previous studies ([Connell et al., 2006] and [Courtney and Wong, 1996]). Interestingly, while the presence of a physical disability predicted adoption, the presence of a vision or hearing disability predicted a lower relative likelihood of adoption. This finding may suggest that child welfare agencies and departments can do more to find adults in the community willing to support youth with sensory disabilities; one example might be the adult visual and hearing impaired communities.

    5.2. Multivariate predictors
    The classification tree presented in Fig. 1 represents the primary goal of this study, to explore the possibility that probability of adoption can be a complex, contextual phenomenon. The unique pattern of findings across the various endpoints of the classification tree appears to support this contextual perspective. It is beyond the scope of this study to tease apart each of the 15 branches of the classification tree and their component variables to offer distinct policy applications and hypotheses for future research. However, two variables of particular policy interest, state and foster family structure, are discussed below.

    5.3. The role of state
    The role of state of placement in the multivariate analyses (see Fig. 1) illustrates the contextual basis of adoption likelihoods. State appears twice in the classification tree analysis and is involved in the endpoints “I” through “N”. For youth who were five years old or younger at removal from home and had a single foster parent (see figure note 1), the state of their placement impacted probability of adoption (scenario 1). Similarly, state of placement played a role in adoption for youth who were five years older or younger, whose foster parents were in a relationship or married, and for youth with zero or one previous placement (scenario 2).

    Interestingly, the following 6 states were associated with lower relative levels of adoption in scenario 1 but higher relative levels of adoption in scenario 2: California, Colorado, Minnesota, North Dakota, Rhode Island, South Dakota, and Wyoming. Further, the following 7 states were associated with higher relative levels of adoption in scenario 1 but lower relative levels of adoption in scenario 2: Hawaii, Indiana, Michigan, New Hampshire, North Carolina, Ohio, Oregon, and Tennessee. It is unclear why these states varied in their predictive direction depending on the different contexts studied here. Scenario 1 might be considered a relatively more challenging and unconventional adoption environment (single foster parent, 2 or more previous placements), and scenario 2 may be a relatively more straightforward and conventional adoption environment (younger child, married/coupled foster family, fewer previous placements).

    Future research should seek to first replicate these findings and continue to search for reliable adoption typologies that vary by state (e.g., the scenario 1 and scenario 2 results). Then, future research can explore potential policy explanations for state differences in adoption likelihoods across the various adoption typologies. For example, it may be that the states that are more successful with the straightforward adoptions have more effective adoption incentive policies (e.g., performance-based contracting), while states that are better at more complex adoptions may have policies that better screen and educate prospective adoptive parents in preparation for an unconventional adoption. Naturally, this type of research should have an expansive scope and study the full range of policies that might impact differences in relative adoption successes across states. At a more basic policy level, the results here support concerns in the literature that simply examining state adoption rates in terms of pure, uncorrected percentages is unwarranted (e.g., Woodruff, 2006).

    5.4. The role of foster family structure
    The variable of foster family structure also illustrates the contextual nature of adoption predictors. The univariate ODA results indicate that the structure of the foster family alone is the third strongest predictor of adoption, with married and unmarried couples having a higher likelihood of adoption. However, when looking at the classification tree model, the structure of the foster family enters the tree at 3 different points, each with a different predictive direction, many of which are different from the results of the main effect analysis. For example, even though unmarried couples in general are more likely to adopt, children who were over the age of 11.7 at the time of removal and lived with a Hispanic or multiracial unmarried foster parent were the most likely to remain in foster care (see Fig. 1, Group A).

    Therefore, the Group A result suggests that the interaction of youth demographic and family demographic and structure variables is crucial to understanding adoption likelihoods. A youth was 98% likely not to be adopted when he or she was over the age of 11, and came from a Hispanic or multiracial foster family where the primary caregivers were an unmarried couple (see Group A). This finding may offer several hypotheses for future research. For example, research has shown that Hispanics in general tend to have a more positive view of marriage (Oropesa, 1996) compared to African Americans and non-Latino Whites. The concept of Familsimo has evolved in this literature to reflect a greater reliance on extended family and marriage and a disdain for divorce (e.g., [Coohey, 2001], [McLoyd et al., 2000] and [Vega, 1990]). As a result, Hispanics tend to view cohabitation more positively as long as it involves future plans to marry, therefore seeing it as an intermediate step in a long-term plan (Oropesa, 1996). Therefore, it may be harder for unmarried Hispanic foster parents to see an older foster child fitting into a long-range plan if the youth will not be in the family for long. Further, older youth are more aware of the traditional family structure; as a result, many Hispanics may not want to undermine the concept of marriage by sending the wrong message to a youth by adopting him or her into a family with an unmarried couple.

    Naturally, this sort of hypothesis is speculative and should only serve to offer possibilities for future research. However, if future research supports this hypothesis, it would have broad policy implications for how child welfare entities work with Latino populations. For example, most state child welfare agencies have an office of multicultural affairs or diversity. Research supporting the above hypothesis could inform the work of these departments. For example, research supporting the above hypothesis could support the multicultural training of caseworkers and caseworker agencies by showing them how to better work with Latino families, such as the importance of considering their unique constructions of family.

    5.5. Limitations and directions for future research
    Despite its strengths, AFCARS has received a number of criticisms; the two primary criticisms are discussed here ([Smith, 2003] and [Wulczyn, 1996]). First, data are reported by states annually, but no linkage exists between any youth across years. This prevents the longitudinal study of youth beyond one year, a particular problem for questions related to the exploration of adoption and reunification patterns over time. Second, the data fields frequently only offer binary choices (e.g., presence or absence of emotional disorders), thus limiting our ability to measure the developmental, social, and political richness that is likely associated with a behavior as complex as the decision to adopt. These two primary limitations combine to form significant problems for the scientific and policy application of AFCARS. This problem is most pronounced in the study of differences in reunification and adoption performance comparisons at the state level, a primary purpose of the AFCARS system. Thus, the ability to look at adoption decisions from a longitudinal perspective, and examining symptoms on a continuum should be a goal of future research.

    In addition, our results support concerns in the literature that simply examining state adoption rates in terms of pure percentages is unwarranted (e.g., Woodruff, 2006). Clearly, state represents a complex social, political, and policy environment that must be properly taken into consideration before AFCARS data can be used as an accountability tool. Taking this into consideration, future research might seek to unearth possible reasons for the state effects found in the current study.

    5.6. Conclusions
    As previously discussed, understanding the factors that influence adoption is critical for both evaluating and improving the child welfare system. While the majority of the previous research has examined individual variables, as illustrated in the current study, factors that influence adoption do not occur in isolation and their impact often varies depending on their context. Therefore, examining adoptions from a contextual perspective is critical for better understanding why some children are adopted and others are not, with the ultimate goal of improving permanency for these children.
     
    #12 Numfarh, Jan 20, 2009
    Last edited by a moderator: Jan 20, 2009
  13. Johnviolist

    Regular Member

    Joined:
    Jul 31, 2013
    Messages:
    23
    Likes Received:
    0
    Location:
    Montana
    Gender:
    Male
    Sexual Orientation:
    Gay
    Out Status:
    Out to everyone
    What a great topic! I'm very anti theistic.
     
  14. Safekeeping

    Regular Member

    Joined:
    Jul 19, 2013
    Messages:
    70
    Likes Received:
    0
    Location:
    Fairfield, CT
    Gender:
    Male
    Sexual Orientation:
    Bisexual
    Out Status:
    Some people
    Alternatively, you could look at survey data and examine how religious people measure up in various activities related to social engagement (voting, volunteering, charitable donations, etc.).