Monday Matters
22 August 2005
In my report to the faculty at the Fall Semester 2005 Faculty Meeting I briefly described the methods we used to calculate faculty salary increases for this year. I stated that I would go into more detail about these in today's edition of Monday Matters . In this column I will review the philosophy underlying raise decisions, the process used to make these decisions, and a few statistics about these raises. I will start with a caveat: no system is perfect, but most or all of the administrators in this College believe that it is important to have a reward system based upon a rational philosophy, even if a perfect system can't be found. As a faculty member I would rather have explicit information about how raises are calculated even if I question the way that these raises were calculated. That is the reason for the information that I provided in the College Faculty Meeting and the reason for this edition of this column.
Last year the University administration provided us a pool of raise money and asked us to use this to address salary inequities. The largest raises last year were provided to individuals who were below the College median salary for their rank as well as to individuals who have been employed by the University for many years but whose raises have not kept pace with the gradual increase in the median salary. This increase in the median salary has been due to natural inflationary factors as well as to increases in starting salary offers that were necessitated in order to recruit talented faculty members. After reflection by the Dean and informal discussion with his advisors, followed by a more formal discussion in Administrative Council, this year we decided to implement the first year of a multiyear plan designed to provide the largest salaries for the faculty members who are deemed the most productive in the college.
The criteria used to rate productivity and the subsequent productivity ratings assigned to faculty members are likely to be the two most controversial issues in the whole raise process. This year we used a system that would give equal weight to the ratings provided by annual performance review committees and a separate set of ratings provided by the department chairs. The department chair ratings were to take into account criteria that might not normally be addressed by the APR committee, including willingness to be involved in College and department initiatives, demonstrated willingness to work for the goals of the college without personal remuneration, exhibited leadership, assumption of extra responsibilities when needed, participation in procurement of extramural funding and other entrepreneurial activities, and other criteria deemed appropriate by the department chair. The development of the ratings themselves, the most important aspect of the entire raise process, occurred at the department level. In cases where the department chair sent me two or more separate APR ratings without aggregating these into a single rating, I aggregated by calculating an average rating in which each of the individual ratings were weighted equally. An interesting side note: In some instances APR committees provided ratings that had very little variability. In these cases the ratings contributed very little to the final productivity rating but rather the department chairs' ratings became the most important factor in the decision-making process. This was not by administrative edict but rather a by-product of APR ratings providing little variability.
For my part, the first step of the process was to convert every salary to a nine-month salary so that we could make fair comparisons. I then created a regression model that predicted salary from rank, a variable indicating whether or not the individual was on a tenure track, and the academic department. After constructing this model I could easily calculate regression residuals. For those of you who have been removed from regression analysis for many years, I'll remind you that a residual is simply the portion of the score that cannot be explained by the model. Residual analysis is one of the most important analytic tools when conducting a regression analysis for it allows you to study variability after accounting for factors that you already expected to contribute to the total score variability. In this case, it allowed me to study salary variability that could not be accounted for by differences in rank, track, or department. Given the productivity model that we adopted, we want the variability in salary residuals to be associated with the variability in productivity ratings.
The next step I took was to rank the individuals within each department based on the ratings provided to me by the department. I then divided everyone into one of five approximately equal categories so that the individuals with the highest rankings were in the top category, the individuals with the next highest rankings were in the category second from top, and so on. Thus, at the end of this process every individual in the College had a productivity rating from one to five. As I stated during my report to the faculty, these ratings should not be construed as being associated with a rubric ranging from not productive to highly productive. The fact is, most if not all members of the faculty are productive. The best we can say about these productivity ratings is that they represent relative productivity within our College based on a certain set of criteria and are designed to be variable.
Interestingly, when I correlated the residuals for the past year's nine-month salary with the productivity ratings, using a simple estimate of linear relationship, the correlation was zero. That is, after taking into account rank, track, and department, salaries were unrelated to productivity ratings. The philosophy of the raise model that we adopted suggests that we should enhance this correlation among salary residuals and productivity ratings.
This year the Provost's instruction to the deans was to use the money being provided by central administration for salary adjustments so that these raises rewarded productivity and varied from 0% to 6%. We obtained permission to add college money to the salary adjustment money and to alter salary increases so that they ranged from 2 to 15%. Thus, we were able to provide a raise for everyone and simultaneously increase the upper end of our ranges. This provided more flexibility for making adjustments. We received just over 3.5% of our salary base from the University and we added an additional $150,000 of College funds for the raises.
We used the two sources of funds to create two components to each raise: an adjustment due to University money and a potential adjustment due to College money. The raises paid for from the University funding were made using the scale below, where 5 is the top productivity category.
|
Category |
Raise |
|---|---|
|
1 |
2.0% |
|
2 |
2.75% |
|
3 |
3.5% |
|
4 |
4.25% |
|
5 |
5% |
The College money was used to adjust for inequities highlighted by comparing residuals to productivity ratings. The extra college money was used to provide an additional raise amount for those faculty members who rated in the top two productivity categories. The amount of the adjustment was based on both the productivity category and size of the residual. Those faculty members who had high productivity ratings and also had the highest negative residuals were the ones who received the highest raise. Regardless of residual, everyone earning one of the top two productivity ratings received some of the College money in their raise. For example, individuals who had high positive residuals (i.e. already near the top of the pay scale) but were in the top productivity category received an extra $2,000 in their raise. Individuals who had high positive residuals and were in the second highest productivity category received an extra $1,000 in their raise. These amounts were increased for smaller residuals so that when we addressed the faculty members with the largest negative residuals, those in the top productivity category received an extra $5,000 in their raise and those in the second highest productivity category received an extra $4,000.
As a result of using this system for determining salary raises, the total raises ranged from 2% to 15%. The five number summary for the distribution of faculty raises in the College is:
|
Minimum |
2% |
|
1st Quartile |
3.5% |
|
Median |
4% |
|
3rd Quartile |
10% |
|
Maximum |
15% |
After calculating raises, I once again calculated residuals by regressing the nine-month salary on rank, track, and department. I then computed the linear correlation of these residuals to the productivity ratings. This time, the correlation was just under .4. Thus, in the first year of a multiyear project we were able to go from a zero correlation among residuals and productivity ratings to a positive correlation, albeit still a rather weak one. One reasonable goal for future years would be to increase this correlation even further, but that is subject to discussion when we set our priorities for addressing the salary distribution in the next years of our salary adjustment efforts.
Another important goal of this process is to raise the entire salary distribution in the College. In my view, raising this distribution more appropriately rewards individuals for the type and amount of work we do in this College and at the same time establishes the College as a more inviting place to work. This progressive spirit can be a strong recruitment mechanism so that we can to continue to attract high quality faculty members as we did this year.
As always, feel free to drop by or write me a line if you have any questions or comments about this column. I just ask that you keep in mind that productivity ratings were established at the department level, so I have little to offer about individual ratings. That said, I know that each department chair would be happy to meet with a member of the faculty to discuss ratings, productivity, or goals for the future. It is, in fact, your efforts that have made it possible for the College of Education to provide long overdue performance raises that begin to reward the hard work of our faculty.
Until next time,
Mike