[Note: I have since done a more sophisticated analysis of this data. See: It could be culture.]
I looked at the 8th and 12th grade NAEP composite math and verbal scores by racial identification by year. I identified race through school report (variable: SDRACE) and student report (variable: DRACEM). I identified mixed race Black/White students through student report and, in the case of 8th graders, since sample sizes were permitting, through cross classification between student and school report. In all analyses, individuals identifying as Hispanic were excluded.
Student report data comes from a two-part ethnic/race question on the NAEP student questionnaire. (See here.) The question allows for multiple ethnic/race responses. The NAEP DRACE classifications — White, Black, Hispanic, Asian/Pacific Islander, American Indian/Alaska Native, More than one, unclassified — created from these responses are mutually exclusive. Non-hispanic, mixed race B/W students can be identified with NAEP explorer by cross tabing DRACEM with self reported “White” and “Black.”
School report data comes from official school reports. Schools are federally mandated to keep records of student racial classification. Student racial classification is based, in turn, on parental report. In some instances school reports do not match with student reports (e.g., the school report will indicate that an individual is Black while the individual will report that she is White). This allows for mixed race identification. The underlying assumption, in this case, is that unidentified B/W students are being identified as opposed to misclassified monoracial and/or otherwise mixed students. I used four cross classifications: School White, Student Black; School Black, Student White; School Mixed, Student Black; School Black, Student Mixed.
As can be seen below, for Blacks and Whites, school and student reported race scores closely match. Overall, identified B/W mixed race students perform intermediate to Whites and Blacks. The only pronounced deviation was with 8th grade students who were school identified as being Black and who self-identified as being White — but the standard errors in this case ranged from 3.4 to 5, so this anomaly is probably due to the small sample size.) It’s possible that the intermediate performance of the identified mixed race students is an artifact of my classification. If, for example, the identified students were in reality a composite of Blacks identifying as Whites and Whites identifying as Blacks, one would also get intermediate scores. The concordance of findings across classifications, though, argues against this.
The first graph shows math plus verbal scores by year by 8th grade racial identification based on school and self report for monoracials and self report for biracials. The second graph shows the same, replacing 8th with 12th grade.The third graph duplicates the first with the exception that the above discussed biracial cross-classifications are included.