Executive Summary State of Rhode Island Education Adequacy Study by R. C. Wood & Associates Joint Committee to Establish a Permanent Education Foundation Aid Formula for Rhode Island In April of 2006 the firm of R. C. Wood & Associates replied to a national request for proposals as presented by the state of Rhode Island. The firm was selected after review by the Joint Committee on Legislative Services. The firm made numerous visits to the state and conducted numerous in depth analyses to determine the adequacy of the funding for elementary and secondary schools in the state. For this research project the research team consisted of Dr. Craig Wood, University of Florida, Mr. Steve Smith, Education Finance, Policy Consultant, Denver Colorado, Dr. Bruce D. Baker, University of Kansas, Dr. Bruce Cooper, Fordham University, Dr. Ronald DiOrio, University of Rhode Island, Dr. Charles H. McLaughlin, Jr., Rhode Island College, and Dr. Robert Shaw, Brown University. The firm conducted four research methodologies in order to determine what fiscal adequacy ranges that the state of Rhode Island should consider in order to provide every child in the state with an adequate opportunity to meet high educational standards. Additionally, the firm offered an overall professional opinion that the state of Rhode Island should move from an appropriation model of distributing funds to school districts to a student need based driven model. The education finance distribution aid formula should have a base student allocation conceptually founded on one of the research methodologies or some combination thereof. The education finance distribution formula should reflect an adequate amount for the base student funding as well as reflect vertical equity adjustments for student needs as evidenced by weights and the cost of delivering educational services in the state. The student need weights as identified by the research team were: • Students in Poverty, • Students in English Language Programs, and • Students in Special Educational Programs. 1 In order to identify the adequacy target expenditure, four education finance models were conducted. The four models were: • Successful Schools Model • Advanced Statistical/Cost Function Model • Professional Judgment Model • Evidenced Based Model The Successful Schools Model is essentially the process of examining the expenditures of schools that are deemed “successful” as measured by state assessments. With adjustments created via discount rates that account for various school demographics, a model can be determined that yields a targeted expenditure equal to what successful schools are achieving in the state of Rhode Island. Depending upon the discount rate applied the increase expenditures required to reach targeted adequacy levels range from $ 56.7 to $128.3 Million. Based on the successful school approach, R.C. Wood & Associates propose a base cost of $9,200 to $9,250, along with special student population weights of 25 percent for free and reduced lunch students and English Language Learners, and 100 percent for special education students. The Advanced Statistical/Cost Function Model was also conducted. This model essentially creates a regression equation consisting of a host of variables to create a curve of best fit. Cost of education variables such as poverty, language proficiency, and disabilities as well as competitive wages and issues of scale were addressed. Based on the cost function model the research team estimated the targeted increase to be $ 42.4 to $46.4 Million. Based on the cost function approach, R.C. Wood & Associates propose a base cost of $9,150 to $9,200, along with special student population weights of 25 percent for free and reduced lunch students and English Language Learners, and 100 percent for special education students. The Professional Judgment Model was conducted with a statewide survey of every building principal and numerous focus group meetings with “expert educators” to estimate the adequacy levels for various prototype schools. Nine different prototype schools were created, reflecting small, medium, and large, elementary, middle, and high schools. Organizational and scale variables ranged from an increase of 1.8 percent to a high of 31.8 percent. Overall, this model produced an estimate of an additional $ 153.5 Million. Additionally, the expert panels determined that “Insufficient Progress” students would require an additional $51.3 Million for extended educational opportunities for a total of $ 204.8 Million of increased funding to meet adequacy targets. Based on the Professional Judgment approach, R.C. Wood & Associates propose a base cost of $10,112 The Evidenced Based Model is essentially built on the concept of identifying the costs of educational strategies and concepts that appear to be the most successful in improving student performance. Numerous examples of recently identified effective strategies that have met strict evaluation procedure were utilized. The research team also was concerned that the bulk of these strategies are virtually impossible to cost out and to 2 determine if they might be generalizable to the state of Rhode Island. Nonetheless, the professional opinion of the research team was that certain programs, e.g., full day kindergarten could be implemented and other pilot programs should be undertaken by the state along with the creation of a state of the art program evaluation system. The total costs associated with model were $53.35 Million. Additionally, the report contains a number of views on education finance aspects unique to Rhode Island conducted by notable Rhode Island educators. These viewpoints yield additional insights and thoughts for consideration beyond the four research methodologies presented. Thus, in final summary the four models and the target expenditures they generate were as follows: • Successful Schools Model $56.7 to $128.3 Million. • Production/Cost Production Model $42.4 to $46.4 Million • Professional Judgment Model $204.8 Million • Evidenced Based Model $53.35 to $58.35 Million Furthermore, per-pupil expenditure levels that could be used as a base cost in a funding formula range from $9,150-$9,200 (cost function) to $9,200-$9,250 (successful schools), to $10,112 (professional judgment) for an average of approximately $9,500 Finally, as a means of helping Rhode Island move these results into action, a paper on key elements for an adequate funding formula is provided. 3 Education Costs, Cost Variations, and Efforts to Determine Adequate Funding It has long been established that state education finance distribution formulas should be designed to accommodate differences in educational need by allocating different levels of financial resources across schools and districts.1 Weighted student formulas date back nearly as far, with examples of weighted pupil calculations to adjust for grade level and school size provided in textbooks dating back to at least 1951.2 At that time, primary emphasis was on the different costs of providing quality education under different geographic circumstances. Education finance scholars were evaluating the relative costs of providing curricular opportunities in high schools of varied size. Scholars and policymakers were beginning to realize that there were sets of conditions that were outside of the control of local school districts that affected the costs of operating schools. Since the Coleman report in 1966, much greater emphasis has been placed on the influence of family backgrounds, on student outcomes, and the related costs of offsetting educational deficits associated with socio-economic status of the family. Empirical research on costs, student needs and educational outcomes has been reflected for many years in the education finance literature. The goal of state finance aid distributional formulas is to provide students, regardless of their individual backgrounds or their geographic circumstance, with comparable opportunity to achieve educational opportunities. Since the emergence of the 1990s accountability movement and subsequent passage of the federal No Child Left Behind Act of 2001, the emphasis of many state school finance policies have been on outcomes and providing equitable opportunity to achieve them. This emphasis is enhanced by various types of cost adjustments. Student need driven state education finance driven formulas are rooted in the assumption that financial leverage can be applied to offset deficits that some children bring to the table by virtue of birth circumstances. Further, financial leverage can be used to create equitable conditions for learning, and ultimately more equitable student opportunities in otherwise very different environments, from the urban core to remote, sparse rural schools hours from the nearest population center. Ultimately, the education finance distribution formula must strive toward the right balance of student and societal needs. Factors affecting the “cost of education” The following illustration provides an overview of factors influencing education costs. Ideally, a need-adjusted budget allocation formula, e.g., a weighted student formula, accounts for those factors that affect costs, and are outside of control of local school districts. The preponderance within education finance research regarding costs identifies 1 Mort, P. (1924). The Measurement of Educational Need. New York: Bureau of Publications, Teachers College, Columbia Univ. 2 Mort, P. & Reusser, W. (1951). Public School Finance. New York: McGraw-Hill, p. 75. 4 two sets of factors: (a) school or district structural and location related factors; and (b) student population characteristics. Factors within school or district control include the actual student outcomes produced and the efficiency with which those outcomes are produced. Using the following illustration, one can imagine that the goal of a student need driven formula might be to identify that level of resources (cost per pupil at the center of the picture) that would be needed, given the student characteristics and school characteristics, to achieve a given level of student outcomes, if the school produced outcomes at an average (or better) level of efficiency. That is, one would not want to give a school more funding simply because they are inefficient producers. Likewise, policymakers should exercise extreme caution in allocating additional resources on the direct basis of low student outcomes. Policymakers also have to be careful not to omit major cost factors as shown on the left- hand side of the illustration that are outside of local school control. When those factors are ignored or under-compensated, it will be perceived that the school is inefficient, even when the inefficiency is outside of the control of local school officials. For example, small school size leads to inefficiency. It costs more to achieve the same outcome in a smaller school, especially when elementary school size drops below 100 pupils. That said, there may be those cases where an elementary school of fifty students needs to exist, by virtue of geographic isolation, in which case, the state necessarily absorbs the inefficiency (or closes the school and relocates all of the residents to a more populated location, an unlikely alternative). Figure 1 Factors Driving Costs of Outcomes Outside of School Control Within School Control Variable Across Schools Common Target Across Schools Poverty/ Economic Disadvantage s r o ct Student a F Language Outcomes nt Proficiency e d u St Disability Cost per Pupil n o Competitive ati Wage (Regional) Inefficiency c o L ol/ o Scale/Sparsity Cost = Spending -Inefficiency h c Remoteness S ©R.C. Wood & Associates 2006 5 Student Need Factors In education policy research in general, and in cost analysis in particular, two types of measures are used to capture differences in student population characteristics and related needs – Sociological Proxies and Individual Educational Needs. A relatively straightforward contrast can be made between marginal costs based on school or district poverty rates and marginal costs based on counts of limited English proficient students. Education cost studies, in particular cost function models, include measures of the share of children in poverty in a school district not as a measure of the individual educational programming needs of any one or a group of students in the district, but as a broad proxy measure of the socio-economic conditions in the school district, which most often relate quite strongly with educational outcome differences. Clearly, not each child identified as living in poverty or qualifying for subsidized lunch will require specific, measurable supplemental educational programs or services. Rather, it can be shown that additional financial leverage, perhaps played out in reduced class sizes or improved teacher quality, can have positive marginal effects on the outcomes of populations disproportionately from impoverished family backgrounds. By contrast, counts of children with limited English language skills relate more directly to the need for additional tutoring and language instruction involving specialized teachers in contact with specific students. Because poverty measures within education finance policies are not intended to identify individual students needs, but rather to predict the likelihood that children requiring additional learning support exist in certain schools, there is greater flexibility in how one approaches poverty measurement at the school or district level. Nonetheless, no method is best for all circumstances. Free lunch counts are based on children living in families at 130 percent the U.S. Census Bureau poverty rate, and are annually adjusted but not regionally sensitive. At this higher threshold, one would certainly expect subsidized lunch counts to be much higher than counts of children in families qualifying at the poverty level. Nationally, using subsidized lunch rates from school year 2000 and U.S Census Poverty rates, poverty rates explain about 85 percentage of variance in subsidized lunch rates, and on average, a 1 percentage difference in Census Poverty rate is associated with a 2 percent difference in subsidized lunch rate. Variance in this relationship from one state to the next depends on the numbers of families in each state that lie in the income range between the poverty level itself, and 130 percent of that poverty level. Ultimately, when selecting a proxy for vertical equity adjustment, one would like to find the proxy that predicts well educational outcome deficiencies but is not manipulable by school or district officials, and does not create perverse incentives. Clearly, funding on the basis of poor performance directly would create such incentives. The alternative is to discern which poverty or other socio-economic proxy best predicts outcome deficiencies across districts. 6 School Structural & Location Factors Beyond individual student needs, a variety of organizational, structural and geographic factors influence the cost of providing comparable services across schools and districts. Such factors include, but are not limited to: Geographic variations in the prices of educational inputs: Input prices are influenced by markets. If we take the market price for comparable teachers for example we find that it differs from one district to the next and from one state to the next. Presumably, district hiring and the uniform salary schedule they offer would tend to equalize teacher quality within the district. However, high ability teachers can be quite sensitive to local variations in working conditions and this inevitably adversely affects less environmentally desirable schools in their efforts to recruit talented teachers. Scale of school or district: Scale (size) is most often defined in terms of numbers of pupils and is most often addressed at the district level. Scale may be addressed in terms of either the scale at which productivity is maximized, at which cost is minimized, or where greatest efficiency is achieved. Scale (over sparsity or remoteness) most significantly affects annual operating costs at the school or district level because scale strongly influences the organization of staffing to deliver core services. The choice to accommodate scale inefficiencies through state policy may be contingent upon remoteness but the adjustment itself should be based on scale. Sparsity of student population: Sparsity is typically defined in terms of the number of pupils in residence per square mile. Sparsity most specifically drives costs associated with transportation, and not the core instructional budgets of schools, unless distance education alternatives are provided due to sparsity. Grade level (& Range): Some state aid distribution formulas account for differences in “costs” associated with providing educational services at different grade levels. Most such studies report higher costs at the secondary level.3 Geographic Variations in Wages Geographic variations in the prices paid by school districts for educational resources are a function of both discretionary (demand side) and cost (supply side) factors. Discretionary 3 Gronberg, T., Jansen, D., Taylor, L., Booker, K (2004) School Outcomes and Schools Costs: The Cost Function Approach. College Station, TX: Busch School of Gov, Texas A&M Univ. 7 factors are those factors within the control of local school districts. Cost factors are those factors that are outside of the control of local school districts, e.g., the availability of qualified science teachers, local market prices for utilities or for materials, supplies and equipment. The goal in establishing a geographic cost of education index is to identify specifically those cost differences outside of control of local administrators, or, for example, the different costs of a teacher given the same levels of education and experience. Historically, three basic approaches have been used to address differences in competitive wages for teachers across schools or districts or broader regions within states. The three basic approaches to adjustments include (a) cost of living adjustments, (b) comparable wage adjustments and (c) hedonic wage model adjustments. Cost of living adjustments are intended to compensate teachers and other school employees across school districts or regions within a state for differences in costs of maintaining comparable quality of living. Cost of living adjustments typically assume some basket of basic goods and services required for attaining a specific quality of living. Goods and services of a specific quality level are identified, and the price differences for purchasing those goods or services are estimated across regions in a state. The basket of goods typically includes things such as housing, food, clothing, childcare and healthcare. Without careful design and construction a problem could emerge in utilizing cost of living adjustments for adjusting school aid. It is often the case that wealthy, generally more advantaged schools or districts in and around more desirable locations will show higher costs of the basket of goods and services. Using an index based on such findings results in supporting very different rather than similar quality of life across teachers within a state. Competitive (Comparable) wage adjustments are estimated for teachers by evaluating the competitive wages of workers in other industries requiring similar education levels and professional skills as teachers. To the extent that competitive wages for similar work (at similar levels of experience, education, age, etc.) varies across regions or school districts within a state, so too, it is assumed, that competitive wages for teachers must vary. The underlying assumption is that teacher’s wages must be competitive with other local industries requiring comparable skills, or teachers might choose to work in those industries instead of education. Because local labor markets vary, competitive teacher wages must vary. Unfortunately, little is known about the mobility of teachers into other supposedly comparable or competitive professions and vice versa, and less is known about the potential role of wages in influencing mobility into and out of the teaching profession from other professions. 8 Hedonic wage adjustments focus specifically on teachers’ employment choices within the field of education and attempt most directly to provide each school district with comparable opportunity to recruit and retain teachers of similar quality. A vast body of educational research indicates that teachers’ job choices are driven primarily by location and work conditions including but not limited to student population characteristics. Neither cost of living indices nor competitive wage indices addresses work conditions of teachers. Among those work conditions that are typically considered outside of the control of local school districts are student population characteristics, crime and safety issues and to some extent facilities quality and age. A well estimated hedonic wage index should capture the negative effects of difficult work conditions on teacher choices, resulting in higher index values for the cost of recruiting a teacher of comparable quality into more difficult working conditions, assuming all else equal. Shortcomings of the hedonic approach most often relate to the availability of sufficient, detailed data to capture expected patterns of competitive wage variation in relation to teacher quality. Presently, teacher wages vary both within and across school districts primarily as a function of years of service and degree level, due to the deeply embedded single salary schedule. Instead of district level indices, comparable wage or cost of living indices might be applied to the consolidated metropolitan statistical area (CMSA), or core based statistical area (CBSA) covering a wide array of districts of varied need, but neither compensating for, nor against those needs. The downside of this approach is that districts in economically depressed regions of a state will likely be assigned lower competitive wage or cost of living indices, making it difficult to ever recruit in new, higher quality teachers from other regions of the state. In effect, the index will reinforce the depressed state of the local economy. Ideally, a well estimated hedonic wage index would capture at least some of the additional costs associated with bringing similar quality teachers into more difficult settings. Unfortunately, data issues pertaining to the measurement of teacher quality typically mute if not negate entirely this desired combat pay effect. A wage index fully accounting for teacher quality influences and how that wage index should be integrated with other cost adjustments, like additional funding for at-risk children, requires great care. The underlying premise of providing additional funding to school districts serving greater proportions of at risk children is that these children will need more contact with teachers of comparable quality if the legislature were to expect them to achieve the same outcomes as other children. That is, they need a higher quantity of teachers of similar quality. If the wage index compensates the cost of recruiting teachers of similar quality into schools with more at risk children, then the at-risk adjustment need only compensate for the costs associated with the higher quantity of teachers needed. However, where the 9 wage index does not fully capture additional costs associated with comparable quality, the at-risk adjustment must compensate for both quality and quantity. Where a metropolitan area comparable wage or cost of living index is used, with no differential for difficult work conditions across school districts within the metro area, larger weightings will be needed for at risk children in the general aid formula. Student driven weights will have to compensate for required differences in both teacher quantity and competitive wages. If a well-estimated hedonic wage index were to capture the competitive wage difference associated with disadvantaged student populations, separate weights for at risk children might be smaller because they need only compensate for teacher quantity differences. The following table summarizes the three approaches, the application, strengths, and shortcomings. First and foremost it is important to differentiate between the goals of the methods. The overall state policy goal is to attempt to implement a cost of education index, as a legitimate vertical equity and adequacy adjustment, regardless of its limitations. In this manner, the state legislature, over time, can develop, refine, and more carefully analyze the exact impact of the concept on public education in the state of Rhode Island. Figure 2 Alternative Approaches to Wage Modeling Approach Goal Data Geographic Strengths Shortcomings Unit Cost of Living Uncontrollable Basket of local Labor market Not (less) Most often costs to employees goods/ services (CBSA/ influenced by supports higher of living in CMSA) current teacher quality of living commutable compensation for teachers in distance “advantaged” districts Competitive Wage required to Wages of Labor market Not (less) Teachers don’t Wage keep a person with comparable (CBSA/ influenced by typically move to specific education/ professions CMSA) current teacher “comparable” knowledge/ skills in compensation professions teaching within a specific labor Based on Influenced by market competitive inequities across labor market local/ regional assumptions economies Hedonic Wage Wage required for Wages of School or Only approach Strongly recruiting and teachers by district to consider influenced by the retaining teacher of background localized work current single specific quality attributes & conditions salary schedule attributes conditions ©R.C. Wood & Associates 2006 10