This study is also, in part, a response to an article that appears in Volume 3, Number 10 of the Clarion, a publication of the Pope Center for Higher Education Policy, (see http://www.popecenter.org/clarion/1999/Jul_Aug/cover.html). In this article, Jon Sanders concludes that UNC-CH "pays professors the fifth-highest average salary among public universities of its kind...." The Sanders study "adjusted university salaries according to the costs of living (COL) in each institution’s location," using data that were "provided by the ACCRA Cost of Living Index for the Fourth Quarter of 1998." Our study uncovers serious flaws with the findings of the Pope Center study and provides alternative, more appropriate methods for comparing salary and total compensation levels at UNC-CH with those at other universities.

        The major findings of this study are:

        First, the Sanders study errs by using ACCRA data only for fourth quarter of 1998 when separate cost-of-living data for Chapel Hill are not available. While ACCRA provides cost-of-living data for 450 different communities in the United States, it does not cover all localities in all quarters. As a remedy, we use data for the 17-quarter period from 1995:I to 1999:I to produce a more accurate measure of the cost of living in Chapel Hill and other university communities that we consider.

        Second, based on our cost-of-living-adjusted estimates, UNC-CH ranks 14^th in adjusted salary and 14^th in adjusted total compensation among the 17 universities in a peer-university comparison group. UNC-CH ranks 55^th based on cost-of-living-adjusted salary and 58^th based on cost-of-living-adjusted total compensation among 85 Research I Universities.

        Third, regional cost-of-living indices tend to overstate cost-of-living levels in areas where people find it attractive to live. A comparison of professors’ salaries to salaries of other professionals can remedy this problem. Using data from the U.S. Bureau of Labor Statistics, we constructed wage indices for university towns that show how well professors are compensated relative to other professionals in the same location. Based on this approach, UNC-CH ranked 13^th in salary and 15^th in total compensation for full professors among the 17 peer institutions.
 

                                                                    INTRODUCTION

      This study is undertaken at the request of the Chancellor’s Committee on Faculty Salaries and Benefits at the University of North Carolina at Chapel Hill (UNC-CH). The Committee is charged with an examination of the gap between faculty salary and total compensation levels at UNC-CH and at a set of 16 peer institutions. The charge requires a recommendation to the Board of Trustees on what it would take to make UNC-CH faculty salary and total compensation levels competitive with the peer institutions.

        This study is also a response to an article that appears in Volume 3, Number 10 of the Clarion, a publication of the Pope Center for Higher Education Policy, (see http://www.popecenter.org/clarion/1999/Jul_Aug4/cover.html). In this article, Jon Sanders concludes that UNC-CH "pays professors the fifth-highest average salary among public universities of its kind...." The Sanders study "adjusted university salaries according to the costs of living (COL) in each institution’s location," using data that were "provided by the ACCRA Cost of Living Index for the Fourth Quarter of 1998." Our study uncovers serious flaws with the Pope Center study and conclusions and provides alternative, more appropriate methods for comparing salary and total compensation levels at UNC-CH with those at other universities.

        There are three major problems with using ACCRA data to adjust salary and compensation levels. First, ACCRA data are not available for all quarters for all locations. In particular, they are not available for the Chapel Hill community in the quarter chosen by Sanders for his study. Second, data are completely missing for some university communities such as Charlottesville, Virginia. Third, high cost-of-living locations are often locations that offer highly desirable amenities. Failing to take location amenities into account in a cost-of-living index may lead to overestimation of the cost of living in some locations.

        To remedy these problems, we present two separate methods for adjusting faculty salaries. The first method uses up to 17 quarters of ACCRA data per location to develop a more accurate cost-of-living index. The second method uses data from the U.S. Bureau of Labor Statistics (http://www.bls.gov) on salaries of comparable professionals to develop a salary index. We apply each of these methods to compare salary and total compensation levels at U.S. universities.

        Most tables in this study report separate rankings for full professors, associate professors, and assistant professors. Our discussion of the results will focus on full professor salaries. Many universities, especially those of high rank, treat assistant and associate professors differently than UNC-CH treats them. Top ranked schools typically pay assistant professors poorly and provide them little chance of attaining tenure. Some top ranked universities no longer grant tenure to associate professors. At UNC-CH, associate professors are generally tenured. Thus it is only at the level of full professor that valid across-institution comparisons can be made.

        Our study is organized as follows. In the next section, we present our method for using ACCRA data to make cost-of-living adjustments to faculty salary and total compensation levels. We also present rankings of universities on the basis of both cost-of-living-adjusted salary and total compensation. In the following section, we present our method for computing wage-adjusted salary and total compensation rankings. The concluding section examines the robustness of our rankings across methods.

       In Volume 3, Number 10 of the Clarion, Jon Sanders concludes that UNC-CH "pays professors the fifth-highest average salary among public universities of its kind...." The Sanders study "adjusted university salaries according to the costs of living in each institution’s location..." using data "provided by the ACCRA Cost of Living Index for the Fourth Quarter of 1998."

        In stark contrast, our study shows that Sanders substantially underestimates the cost of living in Chapel Hill with the result that he substantially overestimates the rank of UNC-CH cost-of-living-adjusted salaries among our peer institutions. We conclude that UNC-CH ranks 14^th among 17 peer institutions when the ranking is based either on cost-of-living-adjusted salaries or cost-of-living-adjusted total compensation. UNC-CH ranks either 55^th or 58^th among 85 Research I Universities depending on whether the ranking is based on cost-of-living-adjusted salaries or cost-of-living-adjusted total compensation.

Salary and Compensation Data

        The salary and compensation data used in this study come from the annual survey of the American Association of University Professors (AAUP) and cover full-time instructional faculty at all Category I (doctoral-level) universities reported in the March/April 1999 issue of Academe. The data exclude faculty in pre-clinical and clinical medicine.

        The AAUP sample includes a total of 191 Category I universities, but does not contain data for Boston, Duke, Howard, and Utah State Universities and the University of Tennessee at Knoxville. The data set also divides Cornell University into its "Endowed" and "Statutory" colleges. We use salary and compensation data from the Association of American Universities (AAU) to add Duke and Utah State to our data set. Our full sample comprises 193 institutions, including the two separate colleges at Cornell.

        We also make comparisons of faculty compensation at UNC-CH to faculty compensation within two sub-samples of schools. The first sub-sample is a set of 16 Universities that UNC-CH calls its "peer institutions." These are Chicago, Duke, Emory, Florida, Illinois-Urbana, Michigan, Ohio State, Pennsylvania, Stanford, Texas at Austin, California-Berkeley, UCLA, Vanderbilt, Virginia, Washington at Seattle, and Wisconsin-Madison.

        The second sub-sample is the set of 85 Research I Universities. The Carnegie Foundation for the Advancement of Teaching designates some doctoral universities to be Research I Universities based on the number of graduate programs offered, the number of doctoral degrees granted, and the amount of federal support garnered by the institutions. Our second sub-sample adds three schools to the list used in the Pope Center study: Duke, Utah State, and Hawaii at Manoa.

        Salaries are reported on a nine-month basis for the 1998-99 academic year. Total compensation includes salaries plus major fringe benefits. Major fringe benefits are those "…where the institution (or state) makes a definite payment of a specified amount on behalf of and for the benefit of the individual faculty member. They include social security, retirement contribution, medical, dental, and group-life insurance, disability income protection, unemployment and workers’ compensation, tuition for faculty dependents, and other in-kind benefits (such as subsidized housing, cafeteria plans, and moving expenses)." (Academe, p.35)

        Because fringe benefits represent a large fraction of total faculty compensation and because that fraction varies substantially across institutions, we also rank universities on the basis of total compensation. Doing so tends to lower the ranking of UNC-CH because it pays benefits equal to 19 percent of salaries, a substantially lower fraction than the 24 percent paid, on average, by AAU members. In fact, UNC-CH is in the lowest one-tenth of all Research I Universities in terms of the fraction of total compensation paid as benefits. This is striking, since it is much less expensive for the institution to provide many fringe benefits instead of salary compensation due to their favorable tax treatment.
 
ACCRA Cost-of-Living Data

        Our cost-of-living data come from the ACCRA cost-of living survey (ACCRA Cost of Living Index, Volumes 28-32). ACCRA cost-of-living indices are based on consumption patterns of managerial and professional households, which are assumed to be the same in every region of the country. Such households spend 28 percent of their consumption budget on housing, 16 percent on groceries, 8 percent on utilities, 10 percent on transportation, 5 percent on health care, and 33 percent on services and other consumption goods. ACCRA indices place greater weight on the costs of homeownership than do indices calculated for urban clerical workers. The survey does not adjust for regional differences in taxes. For more information about ACCRA indices, see http://www.accra.org/costofliving/.

        To provide a more accurate estimate of the cost of living in Chapel Hill, we use different observations from the same ACCRA data than Jon Sanders used in the Pope Center study. Sanders chose to use ACCRA data only for the fourth quarter of 1998 when a separate cost-of-living index for Chapel Hill is not available, even though the data for Chapel Hill were available for each of the preceding two quarters. Because of this shortcoming, he measured the cost of living in Chapel Hill with the ACCRA index for the cities of Raleigh and Durham. Sanders thus assumed that the cost of living in Chapel Hill is the same as that in Raleigh-Durham.

        Separate cost of living data for Chapel Hill are available for 18 quarters between 1988 and the first quarter of 1999 including the second and third quarters of 1998, the third quarter of 1996 and the third quarter of 1997. These ACCRA data indicate that living costs in Chapel Hill are substantially higher than in Raleigh-Durham. For example, the index for Chapel Hill was 17.5 percent higher than that for Raleigh-Durham in the second and third quarters of 1998. The following figure plots the ACCRA indices for Chapel Hill and Raleigh-Durham. It shows a large and growing gap between the two cost-of-living indices. Our approach based on a historic average understates the current high cost of living in Chapel Hill.

Where UNC-CH Faculty Live

        Sanders advocates the use of a triangle wide cost index for adjusting UNC-CH salaries. Perhaps this is because he believes that the faculty at UNC-CH are widely dispersed thought the Triangle area. We examined this assumption and found it invalid. According to zip-code information provided by the UNC-CH Office of Institutional Research and the North Carolina Office of State Planning, 71 percent of UNC-CH faculty members lived in the Chapel-Hill-Carrboro community in 1998. Only 12 percent lived in Durham and 3 percent in Raleigh. Most others lived in parts of Chatham and Orange County that are close to Chapel Hill. If one wishes to measure the local purchasing power of UNC-CH salaries, then these statistics make it clear that one should use cost-of-living data for Chapel Hill to adjust UNC-CH salaries.

        Our procedure for creating cost-of-living indexes for UNC-CH and other universities in our sample entails two steps. First, we match each university to a city or town for which ACCRA data exist during the 17-quarter period from 1995:I to 1999:I. By using the 17-quarter period, we expand number of communities for which there is ACCRA data to 450 and reduce the necessity of matching university communities with distant ACCRA communities. When the ACCRA data did not permit an exact match, we used cost-of-living data for a nearby community similar to the university community, as in the Sanders study. In a few cases, we matched university communities to an average of neighboring communities. The university-community matches on which we base our findings are reported in Table 1.

        Link to Table 1:  University-ACCRA Community Matches

        The second step of our procedure is to create a cost-of-living index for each university community. For each university community, our index is the average of available ACCRA cost-of-living numbers available between 1995:I and 1999:I. For example, the cost-of-living index for Chapel Hill is the average of the index numbers that ACCRA reports for 1996:III, 1997:III, 1998:II, and 1998:III. Our cost-of-living index for Chapel Hill is likely to underestimate the true cost of living in Chapel Hill. This is because the relative cost of living in Chapel Hill has been growing during the last 17 quarters (see Figure 1) and the averaging over time diminishes the influence of the most recent data.

Cost-of-Living-Adjusted Rankings

    Table 2 reports Cost-of-Living-Adjusted Salary and Total Compensation for UNC-CH and the other 16 universities in the peer group. In this group of 17 institutions, UNC-CH ranks 14^th in adjusted salary and 14^th in adjusted total compensation. As Table 2 shows, properly accounting for cost-of-living differences places UNC-CH very near the bottom of the list.
 

        Rankings do not reveal the true extent of the problem. They do not show the amount that salary and total compensation at UNC-CH fall below salary and total compensation at other peer institutions. Table 3 presents an index of cost-of-living-adjusted salary and total compensation for each peer university. It shows that adjusted salary and total compensation at UNC-CH are 37 and 44 percent below the levels at Duke University, the first-ranked among the peer institutions. It also shows that adjusted salary and total compensation at UNC-CH are 11 and 15 percent below the average levels for the members of the peer group.

        Link to Table 3: Cost-of-Living-Adjusted Salary and Total Compensation for Peer Institutions

        We also compare cost-of-living-adjusted salary and total compensation at UNC-CH to levels at other Research I Universities. Table 4 shows that UNC-CH ranks 55^th among the 85 Research I universities based on cost-of-living-adjusted salary. Table 5 shows that UNC-CH ranks 58^th among the 85 Research I Universities based on cost-of-living-adjusted total compensation. It is interesting to note that Texas A&M University, Virginia Polytechnic Institute and State University, the University of Georgia at Athens, Virginia Commonwealth University, and the University of Alabama at Birmingham all rank above UNC-CH on both lists.

        Link to Table 4: Rankings of Research I Universities Based on Cost-of-Living-Adjusted Salary

        Link to Table 5: Rankings of Research I Universities Based on Cost-of-Living Adjusted  Total Compensation

        Finally, we compare UNC-CH to other Doctoral Institutions. Table 6 reports cost-of-living-adjusted salary and total compensation for our full sample of 193 Doctoral Institutions. Entries in the table are thousands of cost-of-living-adjusted dollars. This table shows, for example, that cost-of-living-adjusted salaries and total compensation for full professors at UNC-CH are 18.5 percent and 21 percent below the values for full professors at the University of Virginia. Indeed, cost-of-living-adjusted salaries and total compensation are lower at UNC-CH than at UNC-Greensboro.

        Link to Table 6: Cost-of-Living Adjusted Salaries and Total Compensation at 193 Doctoral Institutions

Discussion

        We also constructed a cost-of-living index for UNC-CH faculty salaries based on a weighed average of the ACCRA indices for Chapel Hill and Raleigh-Durham. We assigned a weight to Chapel Hill of 71 percent based on the fraction of faculty who lived in the Chapel Hill-Carrboro community and 29 percent to the Raleigh-Durham index. We did not adjust any other university communities, such as Palo Alto (Stanford University), where there may be large differences in living costs for small variations in residential locations. This special adjustment changes the UNC-CH compensation ranking among peer institutions by only one position, and it increases the salary ranking from 14^th to 12^th among the same institutions. These are trivial changes.

        When choosing between job offers from different universities, a professor is unlikely to focus solely on the dollar amounts of the salary offers. The professor is likely to consider the local purchasing power of each offer in its respective community as well as the amenities offered by each community. Economic theory warns us, however, that it may be difficult to construct a measure that accurately adjusts total compensation for both differences in living costs and the value of local amenities.  In the following appendix, we address some of these issues.

        Link to Appendix 1:  Methodological Problems with Measuring "Competitiveness"

        The chief concern of the Chancellor’s Committee on Faculty Salaries and Benefits is to assess the competitiveness of UNC-CH salaries and total compensation. While adjusting for across-community differences in the cost of living may introduce as well as eliminate biases, it should be completely clear from the above results that a careful adjustment of UNC-CH salaries and total compensation does not improve the relative ranking of UNC-CH.
 
 

        The prices of geographic specific resources, such as land, tend to measure the value to workers and consumers of living in a particular area. Competition for land and housing by workers desiring to live in particular areas determines the prices of land in each area. Workers tend to locate in the area that offers the best "package" of wages, prices, and local amenities. They choose an expensive-housing area if they value the amenities available in that area.

        Adjusting wages for the total cost of housing would produce a wage index that failed to take account of the amenity values of locations. It would ignore, for example, that many people live in expensive San Francisco because they find the area attractive. It would be necessary to compensate many San Francisco residents to entice them to live in lower-cost (and less desirable) Oakland or Richmond. Adjusting wages for the total cost of housing would ignore the fact that housing is expensive in San Francisco because many people find it a more desirable place to live.

        Wages in the private labor market will tend to reflect amenity values. Employers located in geographic areas that provide few amenities must pay workers more to attract them to these areas. Employers located in areas that provide excellent amenities can pay workers less. Due to competition, private sector salaries will reflect the implicit value workers place on particular areas. Across-area differences in private sector wages will reflect both cost-of-living differences and location-amenity differences.

        To compare salaries across universities while allowing for differences in the cost-of-living and the amenity-value of locations, we measure faculty salaries in an area relative to salaries of other professional workers in the same area. This is a free- market approach for comparing compensations across geographic areas. In essence, we are assuming that the market for private workers has, through Adam Smith’s invisible hand, already solved the problem of comparing salaries across areas.

        To clarify this approach, consider two simple hypothetical examples. Suppose the only jobs in Chapel Hill and Charlottesville are teaching (at UNC-CH or UVA) and computer programming. Suppose computer programmers in both Chapel Hill and Charlottesville are paid $50,000 per year. Since computer programmers are relatively mobile, one would expect that computer programmers are roughly indifferent between living in Chapel Hill and Charlottesville. Otherwise, one would expect to see many computer programmers applying for jobs in the preferred location. Further suppose that professors at UNC-CH and UVA are paid $75,000 and $85,000 respectively. On an absolute scale, UVA professors are thus better paid.

        On a relative scale, UVA professors are also better paid than UNC-CH professors. UNC-CH is paying 1.5 computer-programmer salaries for a professor (1.5 = $75,000/$50,000). UVA is paying 1.7 computer-programmer salaries for a professor (1.7= $85,000/$50,000).

        Consider a second hypothetical example in which faculty at UNC-CH and the UVA are each paid $75,000 but programmers are paid $50,000 in Chapel Hill and $40,000 in Charlottesville. In this example, professors at each university are equally well off on an absolute dollar scale. However, professors at the UVA are better off on a relative scale. Professors at UVA are paid 1.875 ($75,000/$40,000) what computer programmers are paid. Professors at UNC-CH are paid 1.5 ($75,000/$50,000) what computer programmers are paid. In relative terms, UNC-CH professors are paid only 80 percent (1.5/1.875) of what UVA professors are paid. A 25 percent salary increase for UNC-CH professors, from $75,000 to $93,750 would be needed to equate relative professor salaries.

        The actual method that we use is somewhat more complicated than these two simple examples. We use wage data from fifteen professional occupations and regression methods to construct a weighted wage index that can be used to compare to faculty salaries across geographic regions. The metropolitan areas from which we construct a wage index for each school are listed in Table 7. Appendix 2 provides complete details of our approach. Do note that this approach uses a single index for the Triangle area, so the UNC-CH rankings are relatively less dependent on the relatively high cost-of-living in Chapel Hill compared to the cities of Raleigh and Durham. Unlike the cost-of-living adjustment used by Sanders, the wage index for the Triangle does reflect, in part, living costs in Chapel Hill.

        Link to Table 7:  Metropolitan Areas for Wage Index

         Link to Appendix 2: Methods for Wage Index Construction

        Table 8 provides wage-index comparisons for salaries and total compensation across the 17 schools in the UNC-CH Peer Group. The results are very consistent with the cost-of-living adjusted rankings presented in Table 2. UNC-CH full professors rank 13 out of 17 on the basis of salary and 15 out of 17 on the basis of total compensation. UVA is ranked first for both salary and total compensation while Duke is ranked third for salary and second for total compensation.
 

        Table 9 reports a wage index that is constructed using UNC-CH full professor salaries as the base for the index. Using UNC-CH as the base makes it easy to see what it would take to move UNC-CH up in the rankings. We see that UVA salaries are about 24 percent higher and total compensation is about 28 percent higher. University of Michigan is about 7 percent higher in salaries and 9 percent higher in total compensation. To attract faculty like those at Duke, we would need to increase total compensation by nearly 25 percent.

        Link to Table 9:  Wage Adjusted Salary and Total Compensation Indexed to UNC-CH for Peer Institutions

        It is important to recognize that these salary comparisons using professional wages are not artifacts of the particular way we combined occupations to construct the local wage indices. To demonstrate this, Table 10 contains information on the rankings of UNC-CH full professor wages and total compensation using each of the fifteen occupations. Each row displays rankings for single occupation based upon comparisons using annual salaries, instead of the composite index used above. We exclude the University of Hawaii from this part of the analysis.

        The consistency of the rankings to the use of alternative comparison occupations is quite remarkable. Only for two comparison occupations does the UNC-CH full professor compensation ranking fall in the top 30 among the 85 Research I institutions. For seven out of the fifteen comparison occupation rankings, there are at least 50 schools providing higher total compensation to full professors than UNC-CH provides. Among the smaller group of seventeen peer institutions, the UNC-CH total compensation falls among the top ten only two times. For those two times UNC-CH is still in the lowest half, with rankings at 9^th and 10^th. In eight out of the fifteen comparisons, total compensation for UNC-CH full professors is one of the four lowest among the 17 peer schools. Given these comparisons and rankings, one would need to use strained and convoluted arguments to conclude that UNC-CH full professors are among the better paid in the U.S. The two simple facts emerging from this analysis are that compensation to full professors at UNC-CH is in the lower half of all Research I universities and in the lower third of the compensation rankings among its peer schools.
 

        The relative wage index does remedy some of the shortcomings of the cost-of-living index, but it also has some drawbacks. Consider the wages of computer programmers, for example. Why would companies be willing to pay more to programmers living in San Francisco than in Atlanta? The simple answer is that we would not expect them to pay more to those in San Francisco unless the San Francisco programmers were more productive than those in Atlanta. Thus programmers living in San Francisco must be more productive. If one uses programmer salaries in San Francisco and Atlanta to define relative wages, one would probably have too high of a reference-wage index for San Francisco. The possibility of this type of error is clearly high for occupations like programming that have easily transferred products (e.g., computer programs). It may be much less of an issue for other occupations where a worker’s product is more localized. The relative wage approach could lead to an understatement of the true wages of professors in expensive areas. Because workers choose where to live, it is difficult to control for productivity differences in workers across areas. It is unclear how such problems affect relative salary rankings.

        While there is no perfect way to adjust salary or total compensation for the cost of living, it is important to realize that all of the comparisons we have done lead to the same conclusion. By any and all of the measures we consider, UNC-CH salary and total-compensation levels are low compared to those at peer institutions.

        Given that there is no absolute, correct approach for adjusting salaries across areas, one needs to recognize that there is no firm answer to the question of how UNC-CH salaries compare across areas. But the overall picture that emerges is that UNC-CH salaries are low compared to those at peer institutions. The only situation under which UNC-CH salaries look relatively "high" is when one adopts Sanders’ incomplete approach. By absolute salary numbers, by complete cost-of-living adjusted salaries, and by comparing university salaries to professional wages within geographic areas, UNC-CH salaries seem low.

        We close our analysis with a head-to-head comparison of considerable interest. The following table reports salary and total compensation for professors at Duke University and UNC-CH. There are no cost-of-living adjustments used for this comparison. The table shows that full professors at Duke are paid an 18 percent higher salary and 24 percent more in total compensation than their counterparts at UNC-CH. Since professors at Duke and UNC-CH live in the same community, none of the problems associated with cost-of-living adjustments to compensation levels affect this head-to-head comparison. For professors deciding on whether to move to UNC-CH or Duke, these numbers clearly indicate that Duke will most often win any head to head competition for faculty. This is not a far-fetched comparison, as oftentimes Duke and UNC-CH are trying to hire the same professor.
 
 

	Average Salary  Thousands of dollars				Average Compensation  Thousands of dollars
Rank	Full	Associate	Assistant		Full	Associate	Assistant
Duke University   UNC-CH	104.8   88.7	68.7 65.2	57.9 51.2		129.9 104.7	86.3 78.0	70.1 61.4
Duke as a Percent of   UNC-CH	118%	105%	113%		124%	110%	114%