Haves, Halves, and Have-Nots: School Libraries and Student Achievement in California Page: 86

previous studies suggest that the level of combined staffing-both classified and certificated
may be related to the level of services available and to student achievement (Aaron, 1981;
Beaird, 1999; Farwell, 1998; Lance et al., 1999; Lance, Rodney, & Hamilton-Pennell, 2000b;
Lance, Rodney, & Hamilton-Pennell, 2001; Lance, Rodney, & Hamilton-Pennell, 2002; Lance,
Welborn, & Hamilton-Pennell, 1993; Loertscher et al., 1987; McCracken, 2001; McIntosh,
1994; Rojtas-Milliner, 2006; Underwood, 2004).
The third limitation to the Sinclair-Tarr and Tarr (2004) study is its use of the School
Characteristics Index (SCI) as the variable to control for school and community factors. The SCI
is based on a regression model derived not from the individual test scores used in the Sinclair-
Tarr and Tarr study (2004), but on the base Academic Progress Index (API) score for the
previous year, a weighted combination of overall school performance scores that include test
results across all grade levels in English/language arts, mathematics, science, social science, and
the California High School Exit Examination (CAHSEE) (California Department of Education,
2007; see also Technical Design Group, 2000); in several instances, the percentage of eligible
students taking the test influences the overall score, as students not taking the test are assigned a
minimal score. In a paper critical of the API and the ranking system derived from it, Russell
(2002) points out that "aggregating scores at the school level masks the successes and failures at
the grade and classroom levels" (ix). Russell concludes that aggregation of scores at the grade or
classroom level might "promote a closer examination of practices and issues within these smaller
operational units" (x), precisely because whole-school scores such as the API do not account for
variances in scores within an individual test. In a paper about the reliability of overall school
scores, Hill and DePascale (2002) additionally found that the average variance of student scores
"were substantially different for different grade levels" (2002, p.4). While the merits of the SCI

86