[187 [187 Misc.2d Misc.2d 72] 72] scores to account for socioeconomic characteristics of at-risk students. This decision rests on the premise that was not established at trial: that at-risk students' educational potential is immutably shaped by their backgrounds. This is not the position of SED or the Regents, and it is contrary to the evidence at trial. Defendants argue that BOE has made similar adjustments in its official rankings of schools. However, BOE has done this primarily to compare schools with similar demographic characteristics; it has not taken the position that the demographic characteristics of a school's student body excuse poor performance. Dr. Armor's analyses of resources failed to track the effect of resources provided to individual students. For example, dollars spent on a reading recovery program in a given school would be attributed to a school's budget, but very few students would actually receive that benefit. Studies that examine over-all spending on student achievement are of limited probative value. Rather, an accurate measurement of spending effectiveness must examine the particular inputs upon which money is spent. Even if over-all spending were a probative measure, Dr. Armor's spending data were incomplete. The data failed to account for variance in school costs and resources depending on such factors as the size of the school, whether its space was owned or leased, transportation costs, and whether it received private funding. Finally, some of Dr. Armor's resource variables were not well chosen to measure the effect of inputs on student performance. For example, the evidence at trial indicated that teachers with two years' or less experience were particularly correlated with lower student outcomes. Dr. Armor measured the supposed effect of teachers with five years of experience or less. Additionally, although the percentage of teachers with a Master's degree is sometimes used by experts to measure teacher quality, the evidence indicated that it is not the strongest such measure. 2 . D r . H a n u s h e k ' s Te s t i m o n y Dr. Hanushek also testified concerning the alleged disconnect between student achievement and school resources. Dr. Hanushek began by examining spending and performance trends in the United States as a whole. He presented charts which demonstrated that nationwide student performance on the SAT tests declined in the mid-1960's to the early [187 [187 Misc.2d Misc.2d 73] 73] 1980's. After a slight increase thereafter, student performance has remained essentially flat since 1985. This decline and plateau persisted during this period despite decreases in teacher-student ratios, and steady increases in per pupil spending and teacher experience as measured by teachers with a MA and years teaching. Dr. Hanushek noted a similar trend for national assessment for educational progress tests, though this data does not reach back into the 1960's. Dr. Hanushek also presented data concerning nationwide labor market outcomes based on 1990 census data, in which he found that school funding levels do not generally have an effect on individuals' subsequent earnings. The one group for which there was a statistically significant positive correlation was black women. However, the effect observed was very small. By contrast, Dr. Hanushek found that the parental education level did have a strong effect on their children's adult earnings levels. Dr. Hanushek then sought to examine the effect of spending on New York City public schools. He first examined 619 New York school districts outside of the City to see if the number of students who received Regents, as opposed to local, diplomas, was affected by funding levels. He examined the number of students who received Regents diplomas for a single year, 1996-1997. 37 In his final analysis, he sought to account for students' socioeconomic status. After accounting for socioeconomic status, Dr. Hanushek found that there was in fact a negative relationship between spending and the number of students who received Regents diplomas. Dr. Hanushek then turned his attention to New York City to examine whether elementary students' test performances in 1997-1998 bore any relation to district spending. He isolated "high poverty" schools—those with over 90% of students receiving free lunch—and determined whether those schools scoring above the mean on nationally normed City tests (high-performing schools) spent more or less per pupil than schools [187 [187 Misc.2d Misc.2d 74] 74] scoring less than the mean (low-performing schools). 38 He found that the high-performing schools receive less funding than the low-performing schools. This result was statistically significant. He observed the same patterns for "moderate poverty schools" with 75% of their students receiving free lunch. Dr. Hanushek performed similar analyses for other inputs. His analysis of school facilities is discussed in section IV (F) above. He also determined that pupil-teacher ratio and "computer availability" are not related to student performance. Dr. Hanushek also performed parallel analyses for middle schools, again using the "high poverty" and "moderate poverty" and "high performing" and "low performing" typology. The results were essentially the same as those for elementary schools. Dr. Hanushek again found a statistically significant negative correlation between per pupil spending and student performance. He also found no correlation between student performance and "computer availability." The result was equivocal for the relationship between student performance and pupil-teacher ratios in high-poverty schools. Finally, Dr. Hanushek conducted regression analyses for all schools, looking at the effect, or lack thereof, of the inputs discussed above. He controlled for students' socioeconomic status. His results were consistent with those of the analyses discussed above. In summing up his testimony Dr. Hanushek stated that he believed that poor students can improve their academic performance, and that teacher quality, at least, can make a difference in student outcomes. However, he testified that he felt that school systems are not using correct criteria in identifying, or compensating, better teachers. He recommended tying teacher compensation to outcomes. He concluded that New York City has sufficient resources at present to improve student outcomes without any influx of new funds. Dr. Hanushek's analyses suffer from some of the same shortcomings as Dr. Armor's. Crucially, his State-specific analyses studying the effect of resources on students are snapshots of single years, and do not measure the effect of resources over time. Like Dr. Armor's, those of his analyses which focus only on gross per pupil spending are of limited value