Fall 1995 Report of the Senate Committee on the Economic Status of the Faculty

January 25, 1996


This report examines the economic status of the faculty of the University of Pennsylvania. We begin with a review of how salary decisions are made. We then compare Penn salaries and overall compensation with those of similar institutions, and with an absolute standard (the Con sumer Price Index). We then examine evidence for inequalities of salaries within schools or broad disciplinary areas and what proportion of the faculty face cuts in pay measured in real dollars. The final two sections draw explicit conclusions from the evidence presented and make a series of policy recommendations. All of the analyses reported below exclude the Medical School whose salary structure and incentives are quite differ ent than those of the rest of the University.

How Salary Decisions Are Made

Salary decisions are made by deans and department chairs. The Provost sets an overall target percentage and a range of acceptable raises which may be assigned without his review. For the academic year 1995-96, the overall percentage raise was to be 3.5% and there was another 1.2% available for adjustments at the deans' discretion. The issue of merit is addressed by having a range of permissible raises; for 1995-96, the range was from 2% to 7%. The issue of fairness is addressed by having the Provost review any raises below 2% or above 7%.

At present, there is no formal policy for salary raises or review over an extended period, analogous to the current year-by-year policy. Basic protection of long-term salary interests is based on market demand (faculty who are sought after by competitive programs); no such protection seems to exist for faculty achieving only an internal standard of excellence.

How Well Are We Doing Compared to Other Institutions and Other External Criteria?

We make three comparisons in this section: Penn salaries versus those at comparable institutions based on two sources of data; Penn's overall compensation compared to that of other comparable institutions; Penn salaries versus the cost of living index. Data comparing Penn with other in stitutions are only available through the 1994-95 academic year.

Penn Mean Salaries and the Mean Salaries at Like Institutions

Using data made available by the AAUP, we compared Penn's salaries to those of eleven other private research universities which include business schools [Carnegie-Mellon, Columbia, Cornell's endowed colleges, Duke, Harvard, MIT, Northwestern, Rochester, Stanford, Chicago and Yale] The AAUP has developed a weighted index which adjusts the salary structures of each of the other universities to Penn's mix of Assis tant, Associate and Full Professors. In 1994-95 Penn ranked fourth on this index (behind Harvard and Stanford, and just behind MIT). This was a small advance in Penn's ranking from ten years before (1984-1985) in this grouping when we were ranked sixth (using the 1994-95 faculty distri bution). However, Penn's relative dollar gain on the index over the ten year period was larger than nine of the other eleven institutions. Our sala ries were reported to be 77% higher in 1994-95 than they were in 1984-85, while the median gain over that period was 69%. In 1984-85 our sala ries were 90% of those at Harvard, the highest ranked school; in 1994-95 our salaries were 94% of Harvard's salaries. This pattern was roughly the same at each academic rank.

Penn's weighted AAUP salary index was also substantially higher (24% higher) than was the salary index at public research universities: the index average across the main campuses of nine major public research universities (Purdue, Berkeley, UCLA, Illinois, Michigan, Minnesota, North Carolina, Texas and Wisconsin) was $63,550 compared to Penn's index score of $78,512.

This index for the entire University facilitates comparisons, but there is some risk that the disciplinary mix of faculty across campuses may af fect the index's interpretation. (For example, if Penn had a much larger proportion of its faculty in its business school than did other institutions, and business faculty earned higher salaries than did other faculty, institutional salary comparisons would make Penn look better than it should if the issue was relative earnings by faculty within a discipline.) We do not have publicly available data that adjusts the comparison index for the number of faculty by discipline. However, private data supplied by the University does make comparisons within four broad disciplinary classifi cations [engineering, humanities and social sciences, sciences, management] among an elite group of public and private research universities.

These data show there are roughly comparable rates of gain in salaries over ten years across three of the areas (science, humanities and social science, and engineering) with Wharton doing particularly well. Also, Wharton has made major gains compared to other business schools while faculty in the sciences, humanities and social sciences have done a little better than average, and engineering has just improved at about the aver age in that discipline. These results are, overall, less optimistic than those from the AAUP report, although still satisfactory. The AAUP results may be overly influenced by the large number of Wharton faculty here, compared to the number of business faculty elsewhere.

What about total compensation?
How well does Penn do comparing salary plus benefits compared to other institutions?

A good proportion of faculty compensation comes in the form of benefits. Using the AAUP index, in 1994-1995, overall compensation at Penn was $100,162, 27.6% more than the average salary of $78,512. Penn's ratio of total compensation to salary of 127.6% was the highest among the comparable schools, between 3.4 and 6.8% higher than others of the top five [Harvard 124.2, MIT 124.1, Chicago 123.9, Stanford 120.8].

Using the overall compensation figures rather than salaries alone improved Penn's ranking from fourth to third in the set of twelve institutions, and it also sharply improved its absolute dollar position. While Penn's mean salary was 94% of top ranked Harvard's, Penn's total compensation was nearly 97% of Harvard's.

We admit to some wariness about reporting these numbers since (a) Penn's reported ratio is not consistent with overall benefits rate charged to salaries in 1994-95 (over 30%) and (b) we are quite unsure as to whether all reporting universities include the same items in their overall compen sation rate.

How Well Are We Doing Compared to the Cost Of Living?

Penn's mean salaries over the past ten years have more than kept pace with the cost of living index (CPI) published for Philadelphia by the Bu reau of Labor Standards. Figure 1 uses 1985-1986 as a baseline for the mean salary of continuing faculty while using the CPI data as of September 1985 for the cost of living baseline. Over the ten year period the accumulated annual raises for full professors total 79% (annual compound rate of 6.0%), while the CPI has increased only 46% (compound annual rate of 3.9%). Comparisons for continuing assistant and associate professors show a similar pattern, with assistant professors doing somewhat better than associate or full professors

A comparison of the AAUP salary index in 1994-95 versus the index in 1984-85 (based on the mean salaries paid to all faculty in those years, rather than just continuing faculty) shows the same results. There was a compound rate of increase of nearly 6.25%, compared to the CPI com pound rate of increase of 3.9%.

There are two additional approaches to CPI comparisons, which may be worth some attention. First we may look at the median raise rather than the mean raise, on the grounds that what the faculty member in the middle of the distribution earned is a better measure of the average than is the mean which would be biased upwards by the high salaries of a few faculty. In 1994-1995, for example, the mean raise for full professors was 5.5%; however the median raise was only 4.4%. We do not have medians for previous years so we cannot make a long-term comparison, but we assume that the pattern of higher mean than median raises is consistent over time. If we assume that the median increase was consistently 1% less than the mean raise, then the comparison of the faculty raises to CPI index would be less favorable (a compound rate of 5.0% for continuing full professors), although still substantially better than the CPI of 3.9%.

A final approach is to shift focus from annual aggregate raises across the faculty, to the performance of individual faculty over a multi-year pe riod. Here we examine the median raise of full professors who remained at Penn for the five- year period between 1990-91 and the current aca demic year. The median annual compound raise is a little more than 4.3%, compared to an annual CPI rate of increase of 3.0% over the five-year period ending in September, 1995.

What Is the Nature of The Distribution of Salary Raises Among the Faculty?

It is important that the average faculty member progresses well, compared to faculty elsewhere and compared to the Philadelphia area CPI. However very few faculty earn the average salary within their discipline. We are also concerned with the distribution of salaries among faculty. We examine the distribution of salaries against two criteriathe equality of the distribution, and the proportion of the faculty whose raises match the increase in CPI.

Unequal salaries have many causes (rank, longevity, market demands across fields and subfields, productivity), some of which will be seen as legitimate. Thus the fact of inequality does not define the presence of inequity. Nonetheless, the discussion of whether inequity is present needs to begin with the facts of inequality. Our first measure of inequality compares the ratio of total salaries which are earned by the top one-fifth of the faculty and the bottom one fifth of the faculty. To decrease the effects of disciplinary differences all comparisons are made within a school, or for SAS, broad groupings of natural science, social science and humanities. We have been given these data with the understanding that we may not disclose specific within area numbers. We begin this analysis considering only full professors, since the numbers of assistant and associate profes sors within schools and disciplinary groupings are often too few to estimate these statistics with any stability.

Across 10 schools/areas where there are at least 9 full professors, the bottom one-fifth of the faculty earn between 14% and 17% of total sala ries, while the top fifth earns between 25% and 29% of total salaries. The weighted average ratio of the top to bottom 20% is 1.74, that is: the top one-fifth of full professors within an area earn 74% more than the bottom one-fifth of full professors. In eight out of ten groupings, including all of the larger areas, the ratio is between 1.64 and 1.89. At this time we do not have data for these ratios over time, and thus cannot say whether the level of inequality in the 1995-96 salary distribution is greater or less than it has been in previous years.

Another way of looking at the same issue is to consider the distribution of raises over a multi-year period. In this analysis the raises given by each of the schools/areas to continuing full professors over the 1990-91 to 1995-96 period are compared. The median increases vary among schools (from 18-30% over the five year period), which may reflect differences in the wealth of schools. Also the range of raises granted faculty vary a great deal between schools. For example, the interdecile range (the difference between the 10th and 90th percentile in raise granted over the five year period) is as little as 14%, and as high as 31%. The interquartile range (the difference between the raises granted to the faculty members at the 75th and 25th percentile) varies from 6% to 13%. There is no particular association between the median raise and the range of raises granted (the correlation between median raise and interdecile range is .12). Thus, one interpretation of these differences in ranges suggests sharp differ ences between schools in their criteria for allocating salaries, rather than any differences in available funds. Some allow much larger variation in raises granted than do others.

A second criterion for examining the salary distributions incorporates the idea of a safety net. We assume that only under rare conditions would the University choose to reduce the real salary of a faculty member, by giving a raise smaller than the increase in the cost of living. This criterion should be met in any given year, but would be particularly relevant over a longer period. We compare the CPI with the raises given to continuing assistant, associate and full professors in 1995-96, and with raises given to full professors over the five-year period between the 1990-91 and 1995-96 academic years. We only present data for which there were at least 15 people in the rank in the school/area.

All of the schools/areas have made sure that at least 75% of their faculties at each rank have received CPI-level raises in 1995-1996. However, not all of the schools/areas have made sure that at least 90% of their faculty have received CPI equivalent raises. In 1995-96, when the CPI for the previous year was 2.6%, 3 out of 6 schools reported that at least 10% of their continuing full professors received less than the CPI increase. Two of 5 reporting schools said that 10% of their associate professors, and 0 of 4 reporting schools said that 10% of their assistant professors received raises below the CPI.

All of the schools/areas have assured the top 75% of their continuing full professors of raises of at least 16% over the 1990-1995 period, the in crease in the CPI for the period. However six of 10 reporting schools/areas (including all those with at least 6 full professors continuing for the en tire period) have allowed the bottom 10% of their faculty to fall below that criterion.

Some Conclusions

We have done our best to make sense of the data that were made available to us or that we were able to gather independently. However we rec ognize that mistakes may still have been made in the process of analysis. Nonetheless, we believe that the following are likely to be true:

  1. Penn's mean salaries are competitive with those of other like institutions, with some substantial relative gains in the past ten years. This is particularly true if overall compensation is considered. Average salaries have also increased well above the rate of increase in the CPI. Penn's strong performance reflects particularly large gains in some schools, but at least competitive gains in all schools and areas.
  2. There is substantial inequality of salary and percentage salary raises within each school and area, and the degree of inequality varies a fair amount among them. The inequalities within schools/areas are seen whether one does an analysis of the distribution of total salaries in a single year or over time analysis of percentage raises given to continuing full professors.
  3. There is a percentage of the faculty who fall below the rate of CPI increase, and thus who can be said to be taking a cut in pay in real terms. Out of 15 comparison groups defined by rank and school or area, five included 10% of their faculty with under CPI raises in 1995-1996. (About 15% of all faculty of the University received raises less than 3%.) There are conditions of lack of productivity, (or of University-wide budget problems in some earlier years) which might justify raises less than the CPI. Some schools or areas may believe that more than 10% of their fac ulty are performing so poorly as to be deserving of cuts in salary, and can justify these results. However this common pattern of more than 10% of the faculty in an area/school receiving a real cut in pay in a year when the total salary pool available permitted raises of 4.7% (the sum of the regu lar funds of 3.5% + 1.2% in special opportunity funds) well above the CPI of 2.6% has to raise concerns.
  4. The central administration has clear policies about the ranges of acceptable percentage pay raises, and about the average pay raises to be granted across the University. In the past these have been somewhat ignored by individual schools which have given average raises often more and sometimes less than the recommended amount. The Provost has indicated that he intends to enforce the University wide limits in the future more strictly than they have been enforced in the past.
  5. The central administration has no explicit policies about long-term raises, neither to assure that there is some floor below which individual raises will not fall, nor to consider what levels of inequality of raises will be acceptable.
  6. The central administration restricts publication of information about salaries for reasons of individual privacy and to control the level of in dividual complaints about salary. At the same time the central administration provides no information to the schools which would allow them to take into account long-term individual salary histories, inequalities and inequities.

Policy Recommendations

  1. The Provost, in consultation with the faculty, should establish a formal policy for salary raises over a multi-year period, analogous to the current one-year policy. At the very least it should establish a floor for raises over a five-year period with exceptions granted only with explicit permission of the Provost. It might also define a median raise and an expected range of acceptable raises over the five-year period. (The establish ment of such a policy would also require that the central administration provide detailed long term information about salaries to deans and depart ment chairs, so that salary decisions might be made in light of such information.)
  2. There should be no publication of individual salaries. However distributional information about salaries within broad disciplinary areas should be made available to all faculty. This might include the median and the interquartile range (25th percentile and 75th percentile) of both salaries and percentage increase in salaries. This information might be aggregated by rank and by school/area, with no data reported for groupings that included less than 20 faculty which might reveal individual information. We recognize that this may increase tension between administrators and faculty; but the current circumstance is structurally out of balance, in which individual faculty have little information, except for University -wide averages, on which to base any claim of inequity. They are entirely subject to the decisions of administrators with minimal information to use as the basis for negotiation.
  3. Annually the Provost establishes a floor for raises acceptable without explicit review by the Provost. In years when the available funds per mit average raises 1% larger than the CPI, that floor should be the CPI for Philadelphia for the previous 12-month period at the time raises are cal culated. This will ensure that no faculty member receives a cut in real pay without agreement by the Provost that such a cut in pay is justified.
  4. In any year in which the entire salary pool is not 1% higher than the CPI increase, then individual schools should implement salary policies which do not exacerbate existing inequalities of salary among faculty who performed at a satisfactory level. One such policy would provide nearly uniform percentage raises, so that raises for some do not impose the burden of a cut in real pay on others. Another such policy would pro vide higher percentage raises for less well paid faculty, for example providing uniform dollar raises among all faculty, thus decreasing inequality. This policy would continue to allow lower raises to those who did not perform at a satisfactory level.
  5. Through several measures we have now begun to establish the degree of inequality within schools/areas. The inequality is substantial. We are unable to determine whether or not this inequality may reflect some inequity, or is a legitimate product of market and other forces which lead to such differentiation. It is our intention to investigate this issue in the coming year. We will focus particularly on the tension between salary policy based on merit and salary distributions over time that are fair. Our view is that salary differentials among individuals should reflect real and important differences in productivity and genuine differences in the market. There should not be a situation in which the cumulative effects of year-to-year decisions, over the long run, result in unacceptably large salary differentials among individuals who differ only marginally in produc tivity (based on the quality of individual scholarship, teaching and citizenship). A great University will always have a few scholars of exceptional renown, but its greatness depends on the overall excellence and well-being of its entire faculty.

We will consider whether, in order to avoid such problems of poor distribution, either within departments or schools, we believe that long-term salary policies need to be enacted, and a pool of money needs to be supplied (perhaps at the Provost level) to correct for long-term disparities at the individual, department or school level.

The current report will be submitted to the Provost; we will report on the administration's response to it in the Spring.

Senate Committee on Economic Status of the Faculty

Robert C. Hornik (communication), Chair
Roger M. Allen (Asian & Middle Eastern St)
Elizabeth E. Bailey (public policy & mgmt)
Charles E. Dwyer (education)
Laura L. Hayman (nursing)
Jerry S. Rosenbloom (insurance)

ex officio:

Faculty Senate Chair
William L. Kissick (med)

Faculty Senate Past Chair
David K. Hildebrand (statistics)

Faculty Senate Chair-elect
Peter J. Kuriloff (education)


February 6, 1996

Volume 42 Number 19

Return to Index for this Issue.

Return to Almanac Homepage.