Understanding predicted compensation
This is an note/information message.
The examples and coefficient table in this article draw from a specific illustrative regression model that uses six pay factors: Time in Role, Number of Direct Reports, Education, Functional Family, Global Grade, and Location. Your own analysis may include different factors, but the calculation method is identical regardless of which factors the model contains.
Overview of predicted compensation
Predicted compensation is the salary the regression model estimates an employee should earn, based on their objective characteristics. It uses a gender-neutral model, excluding gender from the calculation explicitly to ensure the predicted salary proves fair and unbiased.
The calculation applies a log-linear regression: we first predict the natural logarithm of salary, then convert back to a currency amount using the exponential function. This approach captures the fact that pay differences work in percentage terms – each grade premium adds roughly the same percentage regardless of the absolute salary level.
The factors used in any given model will vary by organization. What matters is the method, which stays the same: sum the intercept and the relevant coefficients for the employee, then exponentiate.
Formula
Where to find coefficients for a specific analysis
Export the Regression results from the Compensation model page to Excel. The gender-neutral coefficients are in the second sheet, labeled "Regression results (no demogr.)." The intercept always appears in the first data row.
Step 1 – Calculating the predicted log salary
Sum the intercept and the regression coefficients for all factors that apply to the employee:
Predicted log(salary) = Intercept
+ (Time in Role × coefficient)
+ (Direct Reports × coefficient)
+ Education coefficient [for the employee's education level]
+ Functional Family coefficient [for the employee's job family]
+ Global Grade coefficient [for the employee's grade]
+ Location coefficient [for the employee's location]Continuous variables (Time in Role, Number of Direct Reports): multiply the employee's actual value by the coefficient.
Categorical variables (Education, Grade, Location, etc.): add the coefficient for the employee's specific category. If the employee is in the reference category for a variable, its contribution equals zero – the intercept already captures that value.
Step 2 – Converting to the predicted salary
Apply the exponential function to convert the log salary to a currency value:
Predicted Salary = exp(Predicted log(salary))| Tool | Formula |
|---|---|
| Excel | =EXP(log_salary) |
| Calculator | e ^ log_salary |
Coefficient reference table
All coefficients from the gender-neutral regression model are listed below. Reference categories contribute 0 to the predicted log salary and have an asterisk (*) next to their name.
| Variable | Reference | Category | Coefficient |
|---|---|---|---|
| Intercept | — | (base) | 10.6005 |
| Time in Role | continuous | per year | 0.0136 |
| Number of Direct Reports | continuous | per report | 0.0011 |
| Education | Primary* | Skilled | 0.1722 |
| Other | 0.1357 | ||
| Masters | 0.1116 | ||
| Junior College | 0.0552 | ||
| Bachelors | 0.0515 | ||
| Trade Degree | 0.0422 | ||
| Functional Family | Support* | Top Management | 0.2656 |
| Sales | 0.1599 | ||
| Finance & Accounting | 0.0885 | ||
| Operations | 0.0814 | ||
| Global Grade | Grade 1* | Grade 2 | 0.1320 |
| Grade 3 | 0.1031 | ||
| Grade 4 | 0.1255 | ||
| Grade 5 | 0.1717 | ||
| Grade 6 | 0.3103 | ||
| Grade 7 | 0.3271 | ||
| Grade 8 | 0.3523 | ||
| Grade 9 | 0.5225 | ||
| Grade 10 | 0.4029 | ||
| Grade 11 | 0.7468 | ||
| Grade 12 | 0.8111 | ||
| Grade 13 | 0.8619 | ||
| Location | Location 4* | Location 1 | 0.0906 |
| Location 2 | 0.1151 | ||
| Location 3 | 0.2023 |
Worked examples
Example 1: Grade 5 employee in Operations
Employee profile
| Attribute | Value |
|---|---|
| Global Grade | Grade 5 |
| Functional Family | Operations |
| Location | Location 2 |
| Education | Bachelors |
| Time in Role | 3 years |
| Number of Direct Reports | 0 |
Calculation
| Factor | Value | Coefficient | Contribution |
|---|---|---|---|
| Intercept | — | 10.6005 | 10.6005 |
| Time in Role | 3 years | 0.0136 / year | 3 × 0.0136 = 0.0408 |
| Number of Direct Reports | 0 | 0.0011 / report | 0 × 0.0011 = 0.0000 |
| Education: Bachelors | Bachelors | 0.0515 | 0.0515 |
| Functional Family: Operations | Operations | 0.0814 | 0.0814 |
| Global Grade: Grade 5 | Grade 5 | 0.1717 | 0.1717 |
| Location: Location 2 | Location 2 | 0.1151 | 0.1151 |
| Predicted log(salary) | 11.061056 | ||
| Predicted Salary [exp(11.0611)] | 63,644 |
Example 2: Grade 9 employee in Finance & Accounting
Employee profile
| Attribute | Value |
|---|---|
| Global Grade | Grade 9 |
| Functional Family | Finance & Accounting |
| Location | Location 3 |
| Education | Masters |
| Time in Role | 8 years |
| Number of Direct Reports | 2 |
Calculation
| Factor | Value | Coefficient | Contribution |
|---|---|---|---|
| Intercept | — | 10.6005 | 10.6005 |
| Time in Role | 8 years | 0.0136 / year | 8 × 0.0136 = 0.1089 |
| Number of Direct Reports | 2 | 0.0011 / report | 2 × 0.0011 = 0.0021 |
| Education: Masters | Masters | 0.1116 | 0.1116 |
| Functional Family: Finance & Accounting | Fin. & Acc. | 0.0885 | 0.0885 |
| Global Grade: Grade 9 | Grade 9 | 0.5225 | 0.5225 |
| Location: Location 3 | Location 3 | 0.2023 | 0.2023 |
| Predicted log(salary) | 11.636321 | ||
| Predicted Salary [exp(11.6363)] | 113,133 |
Frequently asked questions
Why is gender not included in the predicted salary?
The predicted salary always uses the gender-neutral model. This ensures it represents what an employee should be paid based on objective, job-related factors only, making it a fair benchmark for pay equity analysis.
What are reference categories?
Each categorical variable has one reference category – its contribution amounts to zero, absorbed into the intercept. All other categories are measured relative to it. For instance, a Grade 5 employee earns approximately 17.2% more than a Grade 1 employee (the reference), all else being equal.
Why do coefficients represent percentages?
Because the regression runs on the natural log of salary, each coefficient represents an approximate percentage change. A coefficient of 0.15 means approximately 15% more pay. For larger coefficients, use exp(coefficient) − 1 for the exact percentage (for example, exp(0.15) − 1 = 16.2%).