Dear Michelle
1. I am using a salary survey from CUPA (which stands for College and Universal Professional Association for Human Resources). The survey is filled with data from various Universities from all over the United States. When I run a report I ask it to compare "University X" (that I am working with) against 15 other Universities that they feel are similar to theirs in size, location, etc. The survey takes each person at "University X" and compares that persons salary with the average of other peoples salary in that same title at the other 15 Universities. The data then is downloaded into a spreadsheet, the spreadsheet automatically calculates the mean, median etc. for the 15 Universities and tells me how close "University X" employees are to that data.
Autumn Jane: What you have described is known as a Position Analysis. I have attached a sample for your reference.
2. 80% of the midpoint in this case is referring to the CUPA survey. Typically a compensation philosophy is that employers want to pay their employees close to 80% of the midpoint. So, CUPA averages, for example, everyone’s salary in a "Professor" position for the 15 "peer Universities". Then it takes the Professors salary at "University X" and compares it to the other "Peer University" Professors. If the Professor at "University X" is paid within 80% of the average of all 15 peer Universities, then they consider themselves to be paying their Professors "fairly".
Autumn Jane: Referring to the Position Analysis, the different market percentiles i.e. P90 / Q3 / MD / AVG / Q1 / P10 are also midpoint values of the market. If a company decides that their pay policy is Market Q3, it basically means Q3 is the midpoint and it is 100%. Taking yours as example, when you take your salary and divided by average, this is known as a Market Index Analysis. For this index, any job(s) that fall below 80% is your underpaid job(s) and any job(s) that falls above the 120% is your overpaid job(s). Any job(s) that has an index of between 115% to 120% are call “ceiling jobs” as they will max out or reach the max range in near future. The market index analysis makes your salary planning very much easier.
3. I am not sure what you mean by market chart?
Autumn Jane: Have attached a Market Chart for your reference. It is used to compare your overall Pay Practice against the Market.
4. If I am understanding you correctly are you saying that regression analysis can be used for internal analysis to find the OUTLIERS and OUTLIERS are determined by looking at the salaries that fall above or below the 80% and the 120% line?
Autumn Jane: Yes, your understanding is correct. The only concern is whether the 80% and 120% is the right range your company wants to adopt. A company can decides to go with 65% and 135%. The decisions depend very much on the company pay philosophy, market demand and supply, etc.
5. If so, is this because the salary is at least two standard deviations from the mean?
Autumn Jane: Yes, one of the reason.
6. We should use year in position not year of hire. Is that correct?
Autumn Jane: Yes, more accurate but No, because it is still not by job sizing.
7. I am hoping to do a more detailed multiple regression with ONLY the people that are determined to be OUTLIERS. Is that acceptable?
Autumn Jane: Yes.
8. a. If we have to do multiple regression on the OUTLIERS to see why they are OUTLIERS, we would have to get more data (confounding factors; job points, education etc.) on those people from “University X” and from CUPA. I don't think CUPA supplies the job points for the 15 peer Universities. Can it be done without the confounding factors on the CUPA people?
Autumn Jane: Is the CUPA survey quite similar to the attached Position Analysis? If yes, without doing multiple regression, you can already have a good analysis by looking at the position profile i.e. the age, years in company and years in position.
b. If we were able to get more data/confounding factors for the people in CUPA so that we were comparing apples to apples, can you tell me how we would do multiple regression analysis with that data? Can you send me a spreadsheet with multiple regression? Can you explain how multiple regression in this format a little more for me?
Autumn Jane: You may want to Google “multiple regression”. A lot of details and examples for you to pick up more tips.
Regards
Autumn Jane