Sunday, March 1, 2015

The Problem With Data: Faculty Salary Reporting and the Management of Perception

 (Pix (c) Larry Catá Backer 2015)

Universities, and faculty organizations like the American Association of University Professors and the Society of American Law Teachers (SALT), have published variations of faculty salary surveys for some time now.  The AAUP publishes its Annual Report on the Status of the Profession (the 2013-2014 Report may be accessed HERE).  SALT publishes its salary survey (the 2012-2013 Report may be accessed HERE). Universities usually make theirs available in some form their their faculty senates (for Penn State, see SENATE COMMITTEE ON FACULTY BENEFITS 2013-2014 Report on Faculty Salaries (Informational); tables may be accessed HERE). .

These are meant to serve a useful purpose--as an important contribution to informational transparency.  This transparency, in turn is meant to paint a picture of the state of faculty earning that can be used, as an authoritative data set, to further  positions and negotiating strategies,  of university administrators, faculty, legislators and the like.  It is also a valuable mechanism for managing public opinion about the state of the university and the privilege (or lack thereof) of a key university stakeholder.

All of this is well and good, and fair game, in the context of the politics of university administration, public policy development, and the operation of wage labor markets for university faculty labor talent.  Yet, data is a relational as well as an objective measure.  For policymakers, and especially for engagement, the choice of relational elements--the way data is packaged and the choice of data types to place in relationship to each other--will have a profound impact on the way on which the data is read and understood. More importantly, if done with some calculation, the careful presentation of relationships among data (including some excluding others) can be used to manage conclusions as well. This no doubt is usually inadvertent, but perhaps not always so. 

This semiotic insight is both powerful and so deeply embedded in our culture that it tends to go unnoticed.  This post considers how the way in which these relational markers work affect the presentation and utility of faculty salary surveys.    It also suggests the ways in which they might be managed and exploited.

That the production of data tells us little, unless it is related to something else in the way I am suggesting can be illustrated with a perhaps crude example: the speed of a car traveling 60 miles an hour provides important information.  Yet that datum, by itself tells us little other than that the car is traveling 60 miles an hour.  It is only when we gather additional information, additional facts (data) that this "fact" (the datum) becomes valuable. But it becomes valuable in quite distinct ways depending on what sort of data is related to the original one (car traveling at 60 miles an hour). And depending in what is related to the original "fact" quite distinct inferences are possible. Consider how our understanding of the original fact is tempered by  different sets of additional data: (1) the car was traveling in a school zone; (2) a number of other cars were clocked at 40 miles an hour at the same time; (3) all other vehicles clocked were trucks; (4) the car was a police vehicle chasing a suspected criminal; (5) all other cars were clocked at noon, this car was clocked at midnight; (6) all cars were clocked on surface roads but this car was clocked on the highway, etc. Facts, then, tell us little in the absence of meaning acquired in relation to other facts.  

1.  Salary Reports that are limited to one administrative unit provide limited information that may not be useful in many contexts.  It is important of course to know what the range of salaries are within a unit.  Thus it is useful to know the internal wage differentials within a school of liberal arts or a college of law.  But that data has little relational value.  It may be useful in determining policy with respect to the rationale for differences within the unit, but it helps little with respect to assessing the salaries assigned.  Unit salary information becomes more valuable when additional data is included relating to other comparable institutions.  This information contextualizes unit data within wage labor markets in that field, but perhaps little else.  To that extent, the data is useful in disciplining  either institutions  at the lower end of the wage labor market or in countering faculty demands at the higher end of that market.  With the addition of more facts, the survey becomes more useful in different ways--(1) salary break outs (pay, benefits, discretionary additional funds, etc., (2) longitudinal information (trends over time); (3) distinctions among classes of faculty (tenure, tenure track, fixed term faculty) and among different classes of each category (full, associate, assistant professor, lecturer, etc.). These have traditionally served universities well in managing salary patterns within administrative units, especially when data is available from other institutions,  But beyond its internal utility, the data is of limited use, especially within the sometimes wide context within which such data might be employed.  

2.  Salary Reports that do not disaggregate data with respect to disparately impacted groups tends to bury potential differences in salary patterns.  Many salary reports have been doing a better job of disaggregating data by gender.  But sometimes it might be useful to disaggregate among other important categories.  This is especially true where a university has in place muscular programs of diversity.  The usual reasons for failures of disaggregation tend to go toward issues of privacy--that there are so few "data points" that disaggregation will reveal personal information.  Yet the very deficiency is an important datum in itself.  The "fact" of an absence of data may be as important a piece of information as other data.  Indeed, in the context of diversity efforts, that information may be most telling.   

3.  Salary Reports that fail to broaden salary information beyond a specific class of employees tends to be less useful then salary reports within the wider context of university pay systems.   Salary reports tend to be class based.  If the function is merely to discipline wage labor markets within specific sub-markets (e.g., the market for law professors, etc.)  such precision might be valuable, if the appropriate context for data analysis is established.  But for university wide salary surveys, such salary class divisions might be less valuable--especially where such data is used in part to establish policy for matters such as legislative funding, decisions about allocation of employment dollars among different employee classes int he university, and setting tuition levels. Some surveys, like that of the AAUP, have begun to include comparative data on administrative and staff salaries.  These are still in a formative stage of development.  But it bears emphasizing, that faculty salaries, standing alone, may not paint an accurate picture  of wage labor markers within the university or more importantly, of wage labor structures within the university.  It does not suggest either the relative pay differentials among classes of employees, nor does it help understand changes in rates of higher among them.  Those data are crucial for assessing university performance, and especially the administrative and policy decisions of its senior administrators.  This is not to suggest a negative, but merely to offer that, like other large institutions with responsibility for substantial sums, such data ought to be appropriately generated to help both policymakers and boards of trustees better assess the performance of their senior administrators.  It is also useful for university stakeholders in better performing their roles in monitoring and shared governance.

4.  The larger the size of the salary category, the less helpful it is in local context.  Especially in large institutions, university wide aggregations are useful for making macro policy determinations.  Yet at the level of the unit, or in multi-locational institutions, at non-main campus locations, the information may be not merely less useful but also detrimental.  For example, consider a large multi-campus university with locations in both dense urban areas and quite sparsely populated rural areas.  Where unfiltered data is presented on employment of fixed term faculty, it might appear either that all employees of a certain employment class ought to be paid the same wage, or that the "better paid" fixed term faculty in urban locations are "making more" when in fact cost of living differentials may effectively drive down the compensatory value of the differential. The presentation of the data can go far in suggesting that these differentials are a matter of concern, or that they are not.  

5.  Salary Reports that tend to exclude benefits might also distort  wage trends.  It has been customary to exclude benefits and other "fringe" benefits form salary surveys.  Yet that exclusion sometimes tends to distort an accurate reading of the salary data.  Consider, for example, a situation where, as is now more common, a university grants an across the board 3% salary increase, but at the same time effectively increases faculty contributions to health care programs, and reduces the coverage of the now more expensive benefits programs.  The overall effect might equal or exceed the percentage of the faculty salary increase,  Yet that net effect of income reduction through benefits programs changes over salary increases is not noted.   To that extent, salary reports, in the absence of full transparency especially by benefits administrators, tends to paint  false picture.  That is lamentable in the context of increasing efforts to use transparency as a fundamental tool of shared governance. 

6. Salary Reports tend to paint only a partial picture of working conditions.  Wage labor markets are generally understood as dynamic in the sense that wage levels are tied to some extent to productivity.  It has become something of a controversial issue in the last few years about the rate to which senior administrators, or the institution, are capturing greater percentages of the productivity gains int he form of increased revenues or whether the allocation of productivity gains are adequately shared with those who produce them. This has become an important issue (see here and  here). Salary reports do not consider the issue--and that tends to distort inferences derived thereby, suggesting increases in salary without a commensurate "payment" in terms of value added.  Indeed, the reverse may be true--salary increases my be net negative as a result of increases in benefits reductions, and may go even more net negative when the salary increases (to the extent their are some) are not considered in relation to productivity gains by the university--and the allocation of those productivity gains among faculty and other portions of university operations. Not that universities ought not to profit from gains in faculty productivity--quite the reverse, they should.  However, where universities begin to re-allocate productivity from faculty, and not share those productivity gains in the way they had been in the past, that is a set of data that is quite important to consider in understanding the "picture" the salary report paints.

7.  Salary Reports may be most important for their assumptions and premises.  While it is conventional practice to start with the data as the "meat" of a salary report, and then draw inferences from that data, that approach is sometimes less helpful.  Much of the data presented tends to be derived from sources and reflect assumptions that might significantly color the data, and constrain the value of inference making.  It is always useful to start with the premises underlying the organization of the report, and perhaps more important, to consider the caveats that are usually provided either in the introduction, or more likely in the footnotes that may be provided to such reports.  e can, by the skillful deployment of premises manage data to suggest any number of possible inferences.  It is important to understand that the data may (usually inadvertently to be sure) also manage through its choices of inclusions and omissions, through its determinations of categories, and the like,   to make it possible to suggest some conclusions but not others. It is always useful to remember that facts tell us little; relations among facts tell us more; and determination of what constitute facts and what does not, what to include what to exclude, how to label and how to ignore, may have significant effects on the "meaning" of "facts" which are invited to consider. What Salary Reports do tell us is that the issue of productivity and compensation is both locational and contextual, that it suggests both important micro policy and critical macro policy touching on university function within the broader economic and political sphere.  Transparency, honesty and reasoned discussion might be the most one can hope for in the context of reviewing these additions to the conversation about the university and its operation.   

No comments:

Post a Comment