Unlike Standard & Poor's, the Regents, in determining that a higher poverty weighting was required, had focused on the specific circumstances of New York City schools, including an especially heavy concentration of high-needs students, very low graduation rates, large classes and a disproportionate number of schools in need of improvement, and thereby determined that the appropriate low-income weighting for New York City was 1.8. With respect to the state as a whole, the Regents recommended weightings for low-income students ranging from 1.5 to 2.0, depending on the concentration of poverty in the district. Because the Standard & Poor's weighting of 1.35 for low-income students was not focused on the specific circumstances of New York City schools, its use to determine the actual cost of providing a sound basic education to economically disadvantaged New York City students was irrational, as the referees found. The referees thus properly determined that a poverty weighting of at least 1.5the lower end of the range proposed by the Regentsmust instead be used. B. The 50% Cost Filter Finally, as endorsed by the Zarb Commission, Standard & Poor's applied a "cost effectiveness filter" of 50% in an alleged effort to screen out successful school districts that either spent money inefficiently or spent more than was necessary to provide the opportunity for a sound basic education. Under that approach, Standard & Poor's considered the average expenditures of only the lower-spending half of successful school districts in New York State, thereby excluding from the analysis 140 of the 281 successful school districts meeting the Regents Criteria standard. Multiple witnesses and amici heavily criticized this filter. Indeed, several testifying witnesses criticized the use of any cost reduction filter, and others would have used an approach substantially different from the one adopted by defendants, such as simply eliminating the highest- and lowest-spending 5% {**8 NY3d at 41}of districts as "outliers." Defendants' own expert, Dr. Robert M. Palaich, testified that his own firm would not use the 50% filter, and there was no evidence offered that this filter is generally accepted by experts in educational finance or, more fundamentally, that the higher-spending districts that were excluded from defendants' analysis by the cost filter were in fact inefficient. Inasmuch as defendants made no attempt to determine why [*18]some successful schools spent less per pupil than others, the assumption that this must have been because the lower-spending schools were more efficient is utterly speculative. Indeed, certain expert amici posited that the lower spending in the selected schools might instead have been due to low wage costs and a low concentration of disadvantaged students, not to efficiency. Only one decisionmaker anywhere in the countrythe New Hampshire Legislaturehas ever implemented a 50% cost reduction filter. But as even defendants acknowledge, "in New Hampshire it appears that the efficiency factor was selected to drive costs down" to a predetermined amount; it was not based on the expertise of any education finance experts. The 50% number not only is wholly arbitrary, but also has the effect of eliminating most of the