Behavior Problems Index (BPI)

Description of the Behavior Problems Index (BPI)

The Behavior Problems Index was created by Nicholas Zill and James Peterson to measure the frequency, range, and type of childhood behavior problems for children age four and over (Peterson and Zill 1986). Many items were derived from the Achenbach Behavior Problems Checklist (Achenbach and Edelbrock 1981) and other child behavior scales (Graham and Rutter 1968; Kellam et al., 1975; Rutter, Tizard and Whitmore 1970). In 1981 the National Health Interview Survey (NHIS) Child Supplement collected data on a nationwide probability sample of 15,416 children 17 years of age and under. [1] Mothers of children 4-17were asked a series of structured questions concerning their child's problem behaviors and use of mental health services (NCHS 1982: 100-102). The specific questions varied somewhat depending on the age of the child. The behavior problem items utilized in the NLSY79 Child surveys were developed from these items.

The NLSY79 Child BPI

The Behavior Problems Index (BPI) used in the NLSY79 Child includes 28 questions administered to mothers in the Mother Supplement (items 1-26, 31, and 32)*. These questions ask mothers about specific behaviors that their children ages four and over may have exhibited in the previous three months. Three response categories are used: (1) "often true," (2) "sometimes true," and (3) "not true."

* Items 27 through 30 are not part of the Behavior Problems scale. They were added by CHRR to tap dimensions that are particularly relevant for older children. 

The BPI items in the NLSY79 Child are used to derive two types of summary "total" scores, as well as a set of distinct clusters of items, representing common syndromes of problem behavior found in children and adolescents. Principal components analysis with varimax rotation was used to verify that the BPI contained separable dimensions similar to those that were hypothesized to occur, and that the groups of items used to compute the subscale scores did hang together as expected (Zill, 1985). The two types of overall "total" BPI scores as well as related subscales are described below.

Scores Based on Dichotomized Items. The NLSY79 Child database provides one overall BPI score and six subscores based on the dichcotomized recoding of the original items. The six subscores are produced by summing across subsets of dichotomized responses to items in the following domains: (1) antisocial behavior, (2) anxiousness/depression, (3) headstrongness, (4) hyperactivity, (5) immature dependency, and (6) peer conflict/social withdrawal. Appendix D in the Codebook Supplement displays the components of these subscales.

In computing scores for the overall Behavior Problems Index, and the set of six subscales defined above, responses to the individual items are first dichotomized and then summed. In this recoding process, each item answered "often" or "sometimes true" is given a value of one. Each item answered "not true" is assigned a value of zero. Higher scores therefore represent a greater level of behavior problems.  Users should note that two of the items (Q.31 Is disobedient at school and Q.32 Has trouble getting along with teachers) are appropriate only for children who have attended school.  Only the overall score and the antisocial subscore use these two items. Thus, for the overall BPI score and antisocial subscore, parallel raw scores are computed for children in school and children not yet in school. For children ages 4-5, not in school, the total dichotomized-based BPI score is comprised of 26 items. For children ages 4-11 in school, the total is based on 28 items and for children ages 12 and older, the BPI total includes 23 items. Available as raw, percentile, and standardized scores, these dichotomized-based scales are identified in the NLSY79 Child database according to the following sets of question names: BPIyyyy, ANTIyyyy, ANXyyyy, DEPyyyy, HEADyyyy, HPRyyyy, PEERyyyy. The composition of each of these scores can be found in Appendix D, Parts 1 and 2.

Scores Based on Trichotomous Item Recodes. In addition to the dichotomous-based scales described above, three additional scales have also been created based on trichotomous recodes of the original three answer responses. CHRR has prepared a set of BPI scores that measure overall problems based on trichotomous items, and two scores that measure a child's tendency to internalize problems or externalize behaviors. The overall trichotomous-based BPI score is based on a summation of 26 items for children ages 4-5, 28 items for children age 6-11, and 23 items for children age 12 and older. The internalizing scores for children age 4-11 include 10 items and 6 items for children age 12 and older. The externalizing score is based on 18 items for children age 4-5, 20 items for children age 6-11, and 19 items for children age 12 and older. Available as raw, percentile, and standardized scores, these trichotomous-based scales are assigned the following question names: BPTOTRyyyy, BPTOTPyyyy, BPTOTZyyyy, BPEXTRyyyy, BPEXTPyyyy, BPEXTZyyyy, BPINTRyyyy, BPINTPyyyy, BPINTZyyyy. The composition of each of these scores can be found in Appendix D, Parts 1 and 3.

In constructing the trichotomous-based scores that reflect a child's tendency to internalize or externalize behaviors, the individual items are not dichotomized.  Rather, these scores are summed from trichotomous inputs that are first recoded from the original three responses: 1 (Often True), 2 (Sometimes true) and 3 (Not true) into 0 (Not true), 1 (Sometimes true) and 2 (Often true). The composition of these trichotomous-based overall total score and the externalizing and internalizing scales can be found in Appendix D.

Imputation of BPI scores. Since 1992, imputed values have been assigned for the overall dichotomous raw score for all children for whom one item only was missing. Norms are also provided for all those children. Similarly, scores have been imputed for the trichotomous-based overall, internal, and external subscales where only one item is missing. Starting in 2000, the total trichotomous score (BPTOTR) was no longer imputed. Imputation flags (assigned to the ASSESSMENT area of interest) are available that identify those cases that have been prorated (see question names for these variables listed below). 

BPMISSyyyy. BEHAVIOR PROBLEMS INDEX: TOTAL RAW SCORE - PRORATION FLAG
BPMIS2yyyy. BEHAVIOR PROBLEMS INDEX: EXTERNAL RAW SCORE TRICHOTOMOUS ITEMS - PRORATION FLAG
BPMIS3yyyy. BEHAVIOR PROBLEMS INDEX: INTERNAL RAW SCORE TRICHOTOMOUS ITEMS - PRORATION FLAG

Score Formats/Codebook Conventions. The overall dichotomous raw score includes one extra digit, with the final digit representing one implied decimal place.  The external and internal raw scores have been rounded to the nearest full digit. 

Age eligibility for the BPI. The Behavior Problems Index (BPI) is administered to NLSY79 mothers for each of her children ages 4-14.

Norms for the BPI. Single-year age norms were developed from the 1981 National Health Interview Survey administration for all children and for males and females separately. With the exception of the non-dichotomous scores (based on trichotomous items), overall as well as "same-gender" normed scores have been created based on data from the 1981 National Health Interview Survey. Girls are systematically more likely to be reported as exhibiting "better" behavior on most of these scales.  Normed scores for the BPI include both percentile and standard scores (with a national mean of 100 and a standard deviation of 15) and are based on single year of age data. For children below the age of six, separate norms are created for children in school and out of school.

Nationally normed percentile and standard scores are provided for the three trichotomous scores (total, internal, and external), but normed "same-gender" scores are not available. 

As with the other mother-report assessments, a user who wishes to select a sample of children of a particular age should use MSAGEyyyy, the child age variable based on the date the Mother Supplement was administered. Users will find the Behavior Problems scores for the most recent survey year documented in Table 1 in the Child Assessments--Introduction section of the User's Guide.  The components of the BPI scores are listed in Appendix D, as are the BPI norming tables.

Completion Rates for the BPI. Through 2000, the overall completion rate for the Behavior Problems scale was about 93 percent, with somewhat lower levels of completion by Hispanic children in the sample. With the introduction of CAPI administration of the Mother Supplement in 2002, overall completion rates for the BPI have increased to over 95%.

Validity and Reliability of the BPI. The Behavior Problems Index is among the most frequently used of the NLSY79 child assessments, both as an outcome in its own right and as a robust predictor of a wide range of child attitudes and behaviors.

A fully representative sample of children would be expected to have a mean standard BPI score of 100, however mean scores for the NLSY79 child sample in the early survey rounds are one to two points above this level. In more recent survey rounds, scores on the BPI closely approximate the norming. While the age-specific BPI scores vary over time, the evidence suggests that the sample of NLSY79 children may have normed scores not substantially different from the overall 1981 national norming sample.

Researchers continue to frequently use the BPI assessment in studies based on the NLSY79 child data. The overall scale, typically used more often that its subscales, is often seen as an outcome predicted by family circumstances and parental characteristics. The overall index has also been used to test the reliability and validity of other scales, such as the temperament scales (Baydar 1995). Original or modified versions of the internalizing and externalizing subscales have been used as determinants and outcomes in a number of studies (Chase-Lansdale and Gordon 1996; McLeod and Shanahan 1993; McLeod and Edwards 1995; Mott, Kowaleski-Jones, and Menaghan 1997; Guttmannova, Szanyi, and Cali 2007; McCarty, Zimmerman, Diguiseppe, and Christakis 2005).

The six NLSY79 Child behavior problem subscales (antisocial, anxious/depressed, headstrong, hyperactive, dependency, and peer conflict) are often used as child outcomes of interest within a single study. The antisocial and anxious/depressed subscales are also studied separately in some cases. Several researchers have created their own subscales from the items in the BPI child assessment to use as child outcomes. The most frequently studied outcome of this type is oppositional action, a fifteen-item scale that represents "acting out" behaviors (Cooksey, Menaghan and Jekielek 1997).

The NLSY79 Child Handbook: 1986-1990 discusses additional literature on this assessment. The NLSY Children, 1992: Description and Evaluation includes a discussion of the reliability and validity of the assessment based on the earlier waves of child data. Both of these documents are available on the Research/Technical Reports page. Users are encouraged to review an annotated listing of research in which the BPI scales are used by accessing the NLS online bibliography.

NOTE: Several scales and items related to behavior problems are administered as part of the NLSY79 Young Adult survey once child respondents reach young adult age. For example, the following 6-point Schieman Anger scale is administered to young adults starting with the 2008 survey: