Table IIC.I presents a logistic regression model for the likelihood of reunification within three years of entry and also a parallel model for the likelihood of a family exit within three years of entry. The units of analysis are all spells that started during the years 1990-94 in the nine study states. Because the spells were observed through 1997, there are no censored observations in these data.
Summary of Logistic Reunification Analysis.
|Age||Reunification is least likely for children under one year of age at entry, and also for children over twelve years of age at entry. The largest difference is between the ratio for infants and that for 9-11 year olds, as 9-11 year olds are 2.56 times more likely to reunify than infants.8|
|Gender||Gender has little apparent effect on the likelihood of reunification, with males estimated as being 1 percent more likely than females to exit via reunification.|
|Race||The odds of reunification for white and Hispanic children are even. The odds of reunification are 54 percent [(1.00/0.65) -1] more likely for children from either of these groups than for African American children in foster care.|
|Region||Reunification is 54 percent [(1.00/0.65) -1] more likely for children living in those portions of these states that are not in the primary urban counties. 9
The main urban counties have lower reunification rates.
|Sequence||Reunification is 39 percent [(1.00/0.72) -1] more likely for children in their first spell in care than it is for children in reentry foster care spells.|
|Placement||Reunification is less likely for children in family foster care arrangements (relative or nonrelative) than for children in institutions or group homes.10 This result is weakened by the fact that both kinship and nonrelative foster care are included in the foster care category.|
|Stability||Reunification is 150 percent more likely [(1.00/0.40) - 1] from spells where the child stays in one single placement than in spells where the child moves at least one time between placement settings.|
|Year:||Reunification became less likely for each successive annual cohort between 1990 and 1994. The odds for the reunification of 1990 entrants were 22 percent greater [(1.12/0.92)-1) for 1990 entrants than they were for 1994 entrants.|
|State:||Reunification is most likely in Wisconsin and New Mexico and least likely in Alabama. The largest difference in odds is between Wisconsin and Alabama: entrants in Wisconsin are over 4.0 times as likely [1.45 0.36] as entrants in Alabama to exit foster care via reunification.|
It is worth restating that these statistics were computed simultaneously in a single multivariate model. Thus, unlike the bivariate results shown earlier, each of these relationships is observed with the influence of all of the other variables in the model being controlled. It is important to note that, for the most part, the basic direction of most of the bivariate results observed in Table IIA. I are largely preserved in the results produced under multivariate controls. Thus, the independent variables we have discussed each appear to contribute independently to the prediction of reunification levels.
8. The following are equivalent expressions: 9-11 year olds are 2.56 times more likely to reunify than infants. Infants are 156 percent less likely to reunify than 9-11 year olds. The odds ratio of reunification between infants and 9-11 year olds is .39 to 1.
9. Primary urban counties have been as follows: AL (Jefferson), CA (Los Angeles), IL (Cook), MD (Baltimore City), MI (Wayne), MO (Jackson and St. Louis City), NM (Bernalillo), NY (New York City), and WI (Milwaukee).
10. Because kinship care cannot be identified in all states, relative and nonrelative foster care are combined into one category for the nine-state analysis tables that follow, a category for kinship care will be reported separately. For states where kinship cannot be identified, relative and nonrelative care will continue to be combined, and the value for kinship care is reported as "missing."