You Use Statistics to Assess Pay Equity
How do you measure pay equity? Continuing in this series about our Gartner’s EPIC (Equitable Performance-Based Integrity Compensation) program this, is the next question. Well, dust off your memories of statistics because we are going to dig in here! And I must confess something right away. My undergraduate degree is in statistics, and I taught classes for years to master’s students. Thus, my confession is that this is FUN to me! Really fun.
Before I get started. No stress here. If you have bad memories from your class, I think statistics is taught the wrong way. It is taught as a math class. But do you know what the math is? If you can add, subtract, multiply, divide, square, square root, and take the absolute value of – that is as complicated as the math gets. Statistics should be taught as a language and logic class. But I digress …
The Best Method Is Regression
The answer to how do you measure pay equity is … regression. Specifically, multi-variate regression. But what does that mean? When you use regression, you are asking yourself what predicts something. The most common example used to explain regression is to try and predict the sale price of a home. That is what you want to predict. You put in variables (hence the word multi-variate) that you think predict a home price – square footage of the house, lot size, # of bedrooms, location, age of home, etc. You then put in this information about a set of homes that have already sold, and the method will tell you which of those variables were in fact significantly related to accurately predicting the sale price.
When you use this technique to assess pay equity, you again put in all the variables that you believe should predict pay equity (see What is Relevant Information When Assessing Pay Equity? on what those variables are) and you also add gender. What you don’t want is for gender to significantly predict pay differences. If it does, well, that is not a good thing.
Is Regression the Only Technique You Use?
Nope. Because regression comes with some limitations. You must have a comparison group of at least 30 employees, where at least five need to be women. Otherwise the mathematics does not work. Our recommendation (see Fundamental Fix #1 to Assessing Pay Equity – Flip the Unit of Analysis) where we suggest that the analysis should be done across a company, and not country by country, is one way to increase the size of the comparison group. Think about your own company. You might have two developers in EMEA and another 28 in the US. Combining them together you have a large enough group size to use regression. What do you do though if you only have 15 developers of which 3 are women? How do you assess pay equity in this situation? That is the focus of my next blog … so stay tuned. And check out my other blogs as well as those from my colleague Debra Logan.