Recognizing High Cost Factors in the

Financing of Public Education:

The Calculation of a Regional Cost Index

November 2000

Methodology

Construction of the Index

In order to adjust for geographic variations in the cost of educational resources, the regional cost index was generated following a methodology similar to one developed by Rothstein and Smith (discussed below) for the state of Oregon. (1) This involved the use of a statewide index based on median salaries in professional occupations that require similar credentials to that of positions in the education field. In particular, these titles represented categories for which employment at the entry level typically requires a bachelor's degree. The professional occupations selected for use in this index came from a list of 94 occupational titles developed for use by the state of Oregon.

Working from the list of 94 titles used in Oregon, we removed all titles for which insufficient wage information was available in New York State, thus the list from which the RCI was calculated was made up of 77 titles. The titles used appear in Appendix A. In addition to those titles with missing data, the final list excluded teachers, other educational positions and categories that tended to be restricted to federal and state government, since the markets for teachers and for many government positions tend not to be fully competitive. Education-related titles were also excluded in order to ensure that this index be entirely a measure of labor market costs, and not be subject to the tastes or control of districts. In other words, we sought to measure genuine labor market costs, not the results of districts’ decisions to hire especially high quality teachers, or to influence the index value in later years by choosing to pay more for staff. By basing the index on the wages earned in the labor market by professionals with similar skills, we have created a measure of costs in the sector of the labor market in which districts compete for teachers and staff, in each of the regions of the State. Because personnel salaries and benefits make up the vast majority of the costs faced by school districts, we now have an index that compares the buying power of the educational dollar in the different regions of the State.

Statewide Median Wage

The first step in generating a regional cost adjustment from this list of titles was to establish a statewide median wage figure to which median wages in each labor force region could be compared for indexing purposes. This statewide median wage was calculated by taking the total number of employees in each of the 77 titles for the whole state (for example, the total number of people working in the title "pharmacist" across the state), and multiplying by the median hourly wage for that title (12640 pharmacists * $30.35). This result was then summed for all titles, and this result was divided by the total number of employees in all 77 occupations (1.23 million). This produced a weighted statewide median wage of $26.02 per hour for the professional titles making up the index.

Title Weightings

It was important to avoid the possibility that the index could be skewed due to compositional differences in the percentage distribution or mix of the individuals occupying the 77 selected titles. In other words, if professional wages in the titles selected were found to be identical in two labor force regions, but 60 percent of the employees in region A occupied the 10 lowest salaried titles (vs. a 10 percent employee representation in these lower salary titles in region B), a simple summation of wages could lead to the erroneous conclusion that professional service costs were far higher in region A than in region B. In short, such "apparent" cost differences would be due totally to differences in the title composition of the workforce, not to true wage differences in those titles.

This problem was avoided by weighting the wage for each title based on the relative importance of that title in the group of 77 titles statewide. Thus, in determining the regional differences in median wage, we assume that the "mix" of jobs in each region is the same as the "mix" in the state as a whole. These title weights were then applied to each region, therefore making the distribution or service "mix" of titles a constant across the state. As an example, if computer programmers made up 10% of the total number of employees statewide in the 77 titles, then 0.10 compositional weighting was assigned to computer programmers in every region. This title weighting procedure thus imputes to every labor force region precisely the same mix of employees across the 77 titles in every region.

Title weights were generated by dividing the statewide number of employees in a given title by the total number of employees in the 77 titles of the index. For example, the number of pharmacists was 12,640, and this was divided by 1,232,080 (the total number of workers in the state in these 77 titles). This yielded a title weight of .0103 for the pharmacist title. (Since this was performed for all the titles in the list, the sum of all title weightings equals one.)

Final Calculation of the Regional Index

Once the title weights were determined, they were incorporated into the data set for each of the nine regions. The median hourly wage for each title was multiplied by the title weight. This result was summed for all 77 titles, yielding a regional median wage. This regional median was divided by the statewide weighted median professional service wage to yield the final professional service wage index for each region. These were normed on the North Country.

When median wage data were missing for a title in a given region, two alternatives were explored for "plugging" these holes. One method involved a simple substitution of the state median wage for a given title for the missing wage information in a particular Labor Force Region. However, it was quickly recognized that this method of wage attribution was biased "upward" (toward the median) in low cost areas of the State, and biased downward in high cost areas of the State. The alternative which was finally selected was based on the creation of a similar regional cost index, using a smaller set of occupational titles (those titles in which no data was missing in any region of the State, n=47). That index was used, in conjunction with the statewide median salary information for any occupational title that was missing salary information in a specific region, to estimate the missing regional salary item.

Data

Although the list of professional occupations used to create the RCI was based on the work of Rothstein and Smith in Oregon, the actual wage data used was provided by the Bureau of Labor Statistics. The data was extracted from the 1998 Occupational Employment Statistics (OES) Survey, an establishment survey in which employers report the number of employees and wages for each title they employ. As noted by U.S. Department of Labor analysts, "Establishment surveys have little information on the demographics of their employees, but …wages and earnings tend to be more accurately reported in establishment surveys as they are based upon administrative records rather than recall by respondents…These factors make establishment data the natural choice…"(2)

The entire 1997-1998 Occupational Employment Survey data file for New York State was made available to staff through the NYS Department of Labor, thus data was provided for all 661 occupational titles for nine regions of New York State, as well as a statewide total for all titles. The OES data uses estimates of wages based on "straight-time, gross pay, exclusive of premium pay. Included are base rate, cost-of-living-allowances, guaranteed pay, hazardous-duty pay, incentive pay including commissions and production bonuses, and on-call pay. Excluded are back pay, jury duty pay, overtime pay, severance pay, shift differentials, nonproduction bonuses, and tuition reimbursements." (3)

The Bureau of Labor Statistics arrives at these estimates through the use of an annual mail survey of about one-third of establishments state- (and nation-) wide in the following sectors: agricultural services; mining; construction; manufacturing; transportation and public utilities; wholesale and retail trade; finance, insurance and real estate; services, as well as others. (4) The survey is repeated in a three-year cycle, where as the cycle continues, data from the third of establishments surveyed in current years builds on previous years’ data, in a process called wage-updating. This results in detailed and precise estimates of wage levels even in small job categories or geographic regions. In the fourth year, the survey cycle starts over.

Due to the fact that wage data is built-up over three years, the approximations of wages become increasingly accurate and are most precise in the third year. This year's index calculations are based on the most accurate data-year in the cycle, and thus inspire confidence that the results are a good representation of the variation in professional service costs around the state. The triennial nature of the data also suggests that the RCI need only be updated in those years in which the most accurate data in the cycle are available.

It should be noted that results of New York and Long Island were combined. A single median wage was calculated for this labor force area, because there is evidence that these two areas actually function as a single labor market. With professionals, especially education professions, moving to jobs across the lines between Long Island and New York, it is necessary to consider this entire region as a single area, with similar wages and costs.

Data Reliability

The index results reported here are based on the most recently available data, while those reported in last year's Report to the Regents were based on the previous year's data (thus it was based on the 1996-97 Occupational Employment Survey). Since the index was calculated in each of these two years, we are able to compare our results in order to determine how stable and reliable the index values are. It was found that even with the difference in the data accuracy mentioned above, where we expect the wage data used to calculate this year's index values to be more accurate than those used last year, the results of our (Rothstein and Smith's) procedure are quite stable (see comparison of values below).

When the results of the analysis for the two years were compared, they were found to be only about 2% different from one another. Chart A depicts the results. When one considers that the previous year’s data was a rougher approximation of the "true" level of wages in the 77 titles, this small inter-year variation seems to indicate that the method of calculating the index is quite reliable. We can thus expect that when the index is calculated on a three-year cycle, as the nature of the survey data suggests is most appropriate, the results will be good measures of the variation in costs, with a minimum of variation introduced by the method of calculation itself.

In order to probe the small differences between the two indices further, closer study was made of the data for the labor force region that experienced the largest change in its index value, the Capital District. This region was found to be further from the state mean (unnormed RCI=1) in the most recent analysis than it was last year. When salaries and number of employees for each title were compared, for example, it was found that, while there were a number of salaries and titles that experienced slightly above-average, but moderate, growth in the Capital District, a single title was found to drive a large portion of the change. This was the title Other computer scientists and related, whose median hourly wage of $17.33 (1997) increased to $22.60 (1998).

While this change may be partially due to an improvement in the OES data as the responses of more establishments are incorporated into the survey, we should note that this growth might be due to real changes in the economy. It seems likely that this job category is the one into which many Internet workers fell, and given the fact that the Internet was experiencing extremely rapid growth in this period, it is not surprising that this could influence the result. In fact, in the face of such major shifts in the underlying economy, the stability of this index is quite remarkable.

Data Validity

Since the delivery of a market basket of professional educational goods and services (assuming agreement on these things) can be presumed to vary in cost depending upon the area of the state in which they are purchased, a wage-cost index of some type had to be constructed. Unlike the Chambers cost index based upon teacher salaries (discussed below), the cost index developed here was based upon professional salaries of titles over which districts have no control, and which are further assumed - unlike teacher salaries - to be governed by open-market labor forces. In effect, such an index assumes that in communities in which the entry-level cost of such professional titles is relatively high, teacher salaries will be proportionally higher as well.

Although, as is mentioned above, the concept of a correction for the differences in the cost of operating school districts in New York State has a long history, this index is a relatively new tool. Thus, it is necessary to determine whether or not it is valid, i.e. that it measures what it claims to measure. Several tests were performed in order to determine validity.

As reported in last year’s Report to the Regents, an important preliminary test of construct validity was conducted last year in order to determine how well the index performed. As a first step, each district's pupil teacher ratio was correlated with each district's total expenditure per pupil. The correlation between the district's unadjusted expenditure per pupil and the staff ratio was found to be - .474, i.e., the larger the pupil teacher ratio, the lower the overall expenditure per pupil - precisely as one might expect. If one assumes that the true educational purchasing power of these district expenditures is more accurately measured by cost adjusted dollars based on this index rather than raw dollars per pupil – it follows that the cost adjusted expenditure measure should be more highly correlated with a staffing measure like the pupil teacher ratio than the raw unadjusted measure. When this correlational test was conducted using cost adjusted dollars per pupil, the correlation increased to -.532, consistent with expectation.

This year several other indices were compared to the RCI, in order to examine the similarity of the results. The Building Cost Index, authorized by the legislature in the 1997-98 session, is a measure which also recognizes significant differences in non-professional labor costs across the State. A comparison of the two indices shown in Chart B (using the unnormed results with New York City and Long Island treated separately) shows that although the difference in costs between Hudson Valley, New York City and Long Island and the rest of the State appears much stronger in the Building Aid results, the general pattern of the index values is quite similar.

When we investigate the year-to-year change in the two indices, we see that while the RCI changed about 2% between the first year and the second, the Building Cost Index (BCI) changed about one percent. This difference in the amount of change in the two indices may be due to the fact that the RCI uses weighted median wages to determine the index, while the BCI does not, or it could be related to the different labor markets described by the two indices. While the RCI is based on the market wages of professionals, the Building Cost Index is based on the wages of highly unionized carpenters, plumbers and electricians. The different jobs covered and the different impact of labor market forces on the two sets of workers may make the RCI more subject to change. It is important to note that since school district employees are largely professional workers, the intrastate comparison in the labor market described by the Regional Cost Index is the most appropriate one for analyzing operating costs for schools.

In another test of the validity of the RCI, the differences in per capita adjusted gross income (from NYS income tax data) across the regions did not mirror the results seen above. This is unfortunate, but we believe that it can be explained by the fact that adjusted gross income includes many sources of income that are not subject to labor market differences, i.e. Social Security and pensions, disability payments, TANF payments, dividends from stocks and mutual funds, etc. Given that these and other forms of non-wage income are associated with demographic differences across the State, we felt that the results from the adjusted gross income data were confounded with other important characteristics of New York's population, thus making a comparison of gross income across the State an imperfect match with the wage data. We are confident that should data on wage-income per taxpayer be available, instead of gross-income per return, we would find a solid correlation between the wage results representing the professional service sector that were used to calculate the RCI and the wage results which represent the entire labor market. Perhaps such data will be available in the future.

Alternative Methods of Calculating Regional Costs

There are several possible methods for developing an index to adjust for geographic variations in the cost of labor. Rothstein and Smith recommended the method used above in their work on the state of Oregon. The Rothstein and Smith method, along with the other methods which will be discussed below are all designed to do the same thing - namely, to control for specific attributes of the goods and services being purchased, so that valid inter-area cost comparisons can be made. A brief description of each method follows:

Rothstein and Smith (1997) base their index on the claim that unlike the past, when teachers were predominantly female and underpaid, salaries have increased and the teacher labor market has become a part of a broader labor pool of college educated, professional workers. Other professionals in this group include accountants, health professionals, and managers.

Rothstein et al. suggest that insight into the regional variation in teacher cost can be gained by examining the regional variation in salary of other professionals within the broader competitive labor market of similarly educated and salaried individuals. Indeed, they argue that in many communities, the overwhelming majority of teachers are employed by (and thus costs controlled by) a single agency or a very few agencies - in effect, these agencies dominate the market for the purchase of educational components. In this market situation, one purchaser can heavily influence the expenditures/costs of professional compensation; therefore, it is more appropriate to use the salaries of non-teacher professionals (which are subject to free market conditions) as the basis for price estimation. Because school districts have no control over the salary scales of other professions in their geographic area, a cost adjustment based on regional variations in these salaries is more likely to represent an accurate reflection of true variations in cost as opposed to factors under the discretion of school districts.

Rothstein and Smith acknowledge that their index is focused mainly on labor costs, and that it does not take into account variations in non-labor inputs to education. They argue that this is acceptable because the vast majority of school expenses are for compensation of staff. (They estimate the percentage at around 85% of total costs.) Since data on the myriad other costs facing districts would be difficult to obtain, and since some of the items in the other 15 % are either covered by the Building Cost Index in New York, or should be the same price across all the regions, (such as textbooks ordered from national publishers), focusing on the wages is appropriate. For those with further concerns about this approach, Rothstein and Smith suggest that the index multiplier could be applied to the portion of school aid that is directly applied to salaries. These could include 85% of operating aid, the total aid in categories that are focused on teachers and teacher improvement, and the proper proportion of other categorical aids.

One strength of this index which they do not mention, but which is nevertheless important, is the fact that an index of wages is relatively easy to explain to policymakers, district leaders, teachers, parents and other interested parties. In addition, the data is easy to obtain, the calculations are simple and as discussed above, this index yields fairly stable results that only need to be recalculated every three years. These characteristics, as well as the focus on market wages and the fact that it is not subject to district control an extremely practical means of calculating the differences in regional costs. That is why this particular method was selected.

Consumer Price Index

This method is also suggested by Rothstein and Smith. In fact, they suggest that ideally this might be used alongside the RCI method to increase the precision of the measure. However, Rothstein and Smith also acknowledge serious limitations on the use of a CPI measure to determine regional costs. The first problem is this information is collected on a different geographical basis than the wage data, making comparison difficult. Should New York attempt to collect the price data on its based on its own internal regions, the cost would be prohibitive. In addition, the costs faced by particular school districts would be difficult to measure, since such precise local data is hard to collect. Lately, the CPI itself, even the national estimates, has been criticized for failing to incorporate new shopping patterns, and to account for differences in cost due to taste. Since shopping patterns and tastes for quality vary significantly around the State, it would be difficult to avoid confounding differences in shopping patterns with differences in cost. The method of calculating the RCI that was selected successfully avoids these problems.

Chambers and Fowler (1995) suggest another type of cost index of teaching. This is based on a hedonic wage model that considers relevant conditions that may attract workers to a certain geographic area or certain positions. This cost adjustment uses a comprehensive statistical analysis (typically a series of hedonic equations) which predict the market price or wage compensation that would occur if certain personal characteristics of teachers, variations in local amenities, and job environment factors were presumed to be the same in each local geographic area. This method requires complex calculations and would be difficult to explain to decision-makers. In addition, the selection of relevant characteristics on which to measure districts and teachers could be open to significant political debate. The Rothstein method is preferred for its objectivity and simplicity.

Occupational Titles Used for Regional Cost Index

(The Professional Cost Service Index)

1. Financial Managers
2. Personnel/Labor Relations Managers
3. Purchasing Managers
4. Marketing/Public Relations Management
5. Administrative Services Managers
6. Engineer/Math/Science Managers
7. Medicine/Health Service Managers
8. Property & Estate Managers
9. Industrial Production Managers
10. Construction Managers
11. Commun, Trans, Utility Managers
12. General Managers/Top Executives
13. Other Managers & Administrators
14. Underwriters
15. Credit Analysts
16. Loan Officers & Counselors
17. Accountants & Auditors
18. Budget Analysis
19. Other Financial Specialists
20. Purchasing Agents & Buyers
21. Employment Interviewers
22. Personnel/Labor Rel Specialists
23. Cost Estimators
24. Management Analysts
25. Claims Examiners—Insurance
26. Other Management Support Workers
27. Metal, Ceramic, Matls Engineers
28. Civil & Traffic Engineers
29. Electrical/Electronic Engineers
30. Computer Engineers
31. Industrial Engineers
32. Safety Engineers
33. Mechanical Engineers
34. Other Engineers
35. Architects
36. Civil Engineering Technicians
37. Electric & Electronic Engineering
38. Industrial Engineering Technicians
39. Mechanical Engineering Technicians
40. Other Engineering Technicians
41. Chemists
42. Other Physical Scientists
43. Foresters & Conservation Scientists
44. Biological Scientists
45. Biological, Agric, Food Techs
46. Chemical Technicians
47. Other Phys & Life Science Techs
48. Computer Systems Analysts
49. Computer Programmers
50. Other Computer Scientists
51. Operations Systems Researchers
52. Economists & Market Research
53. Urban & Regional Planners
54. Psychologists
55. Other Social Scientists
56. Social Workers: Medical & Psych
57. Social Workers
58. Residential Counselors
59. Recreation Workers/Coordinators
60. Librarians
61. Curators, Archivists, Restorers
62. Occupational Therapists
63. Physical Therapists
64. Speech Pathologists & Audiologists
65. Recreational Therapists
66. Physicians Assistants
67. Pharmacists
68. Dietitians & Nutritionists
69. Med & Clinical Lab Technologists
70. Writers & Editors
71. Technical Writers
72. Public Relations Specialists
73. Artists & Related Workers
74. Designers
75. Sales Engineers
76. Insurance Adjusts, Exams, Investigators
77. Police Patrol Officers

Education Titles:

1. Education Administrators
2. Nursing Instructors
3. Teachers – Presch & Kindergarten
4. Farm & Home Management Advisors
5. Instructional Coordinators

Occupations Lacking NYS Data:

1. Agricultural Engineers
2. Agricultural & Food Scientists
3. Reporters & Correspondents

Occupations Lacking NYS Regional Data:

1. Air Pilots & Flight Engineers
2. Interior Designers
3. Financial & Statistical Analysts
4. Actuaries
5. Geologists & Oceanographers
6. Physicists & Astronomers
7. Landscape Architects
8. Chemical Engineers
9. Adjudicators & Hearing Officers

Footnotes

New York State Department of Labor. (December 22, 1999). Technical Notes for 1998 Estimates. Albany, NY: Author. Retrieved August 7, 2000 from the World Wide Web: http://www.stats.bls.gov/oes/oes_tec.htm.

New York State Office of Real Property Services. (July 29, 1999) Residential Median Sale Price Information by County. Albany, NY: Author. Retrieved July 20, 2000 from the World Wide Web: http://www.orps.state.ny.us/sales/star/med1998.htm.

Parrish, T.B., Matsumoto, C.S., and Fowler, Jr., W. (1995). Disparities in Public School District Spending, 1989-90: A Multivariate, Student Weighted Analysis, Adjusted for Differences in Geographic Cost of Living and Student Need. Washington, D.C.: U.S. Department of Education, National Center for Education Statistics (NCES Report No. 95-300).

Paternick, L., Smerdon, B.A., Fowler, W., & Monk, D. (1997), Using Cost and Need Adjustments to Improve the Measurement of School Finance Equity. In W. J. Fowler (ed.), Developments in School Finance (NCES Report 1997, pp. 151-168). Washington, DC: National Center for Education Statistics.

Reschovsky, A. & Imazeki, J. (1997) The Development of School Finance Formulas to Guarantee the Provision of Adequate Education to Low-Income Students. In Developments in School Finance, 1997. W.J. Fowler (ed.) (NCES Report 98212, pp. 121-148) Washington, DC: National Center for Education Statistics.

Rothstein, R., Guthrie, J.W. & Smith, J.R. (1998). Wyoming Education Finance Issues Report: Reconsideration of Wage Rate Cost Adjustments. Sacramento, CA: Management Analysis & Planning Associates, L.L.C.

Rothstein, R., & Smith, J.R. (1997). Adjusting Oregon Education Expenditures for Regional Cost Differences: A Feasibility Study. Sacramento, CA: Management Analysis & Planning Associates, L.L.C

State of New York. Funding for Fairness: A Report of the New York State Temporary State Commission on the Distribution of State Aid to Local School Districts. ("Salerno Commission"). Volume 1. December, 1988.

State of New York. Report and Recommendations of the New York State Special Task Force on Equity and Excellence in Education. ("Rubin Commission"). Volumes 1-3. February, 1982.

Taylor, C. (1997) Does Money Matter? An Empirical Study Introducing Resource Costs and Student Needs to Educational Production Function Analysis. In Developments in School Finance, 1997. W.J. Fowler (ed.) (NCES Report 98212, pp. 75-98) Washington, DC: National Center for Education Statistics.

U.S. Department of Labor. (1997). "Interarea Comparison of Compensation and Prices," Report on the American Workforce, p. 69-97.

U.S General Accounting Office. (1998). School Finance: State and Federal Efforts to Target Poor Students (GAO/HEHS Publication