DDBA 8307B Week 8 Regression Write-Up Example

Why Salary Band Stopped Mattering Once Job Level Entered: A Regression Write-Up Predicting How Long a Staffing Firm Takes to Fill Job Orders, Reporting What the Model Explains and Which Predictors Survive Together

Student Name

Walden University

DDBA 8307B: Quantitative Business Data Analysis Using SPSS

Instructor Name

Month Day, Year

Why Salary Band Stopped Mattering Once Job Level Entered: A Regression Write-Up Predicting How Long a Staffing Firm Takes to Fill Job Orders, Reporting What the Model Explains and Which Predictors Survive Together

The Question and the Data

A composite staffing firm in Charlotte, North Carolina, places accounting, administrative and finance professionals with client companies. Its operations director wants to know which features of a job order predict how long it takes to fill. The data are 640 job orders filled in 2025. The outcome is days from order to accepted offer. The predictors are job level (entry, mid or senior), the salary band offered (1 to 5, from the client's posted range), the number of openings on the order and whether the client marked the order urgent. Job level enters the model as two indicator variables, mid and senior, compared with entry.

Each Predictor on Its Own

Taken one at a time, several predictors relate strongly to time to fill. Salary band correlates .41 with days to fill: higher-paying orders took longer. Senior-level orders averaged 31 days to fill, mid-level 22 and entry-level 16. Orders with more openings took longer; urgent orders were filled faster.

The Model

Together, the five predictors explain about 31 percent of the variation in days to fill. Holding the other predictors constant, senior orders took about 11 days longer than entry-level orders and mid-level orders about 5 days longer. Each additional opening on an order added a little over a day. Orders marked urgent were filled about 4 days faster, perhaps because recruiters prioritized them.

Table 1

Multiple Regression Predicting Days to Fill (n = 640)

R² = .31, adjusted R² = .30, F(5, 634) = 57.0, p < .001. Ordinary least squares in SPSS.

Why Salary Band Stopped Mattering

On its own, salary band was one of the strongest predictors. In the model, its coefficient is small and indistinguishable from zero. The reason is that salary band and job level carry largely the same information: senior orders almost always sit in the top two salary bands, and entry orders in the bottom two. The correlation between salary band and job level, coded from 1 to 3, is .78. When both are in the model, the regression asks what salary band adds once job level is known, and the answer is very little. Tabachnick and Fidell (2019) describe this as overlap among predictors: when two predictors are highly correlated, the model credits their shared contribution to whichever one explains it slightly better, and the other appears unimportant. The variance inflation factor for salary band was 2.6, which is not extreme but confirms the overlap.

The plain reading is that slower filling of higher-paid orders is mostly about the level of the job, not the pay itself. It does not mean pay is irrelevant to candidates; it means that, in this data, pay and level move together so closely that the model cannot separate them.

What Changes for Planning

For the operations director, the model suggests setting expected fill times by job level rather than by salary band, and adding about a day for each additional opening on an order. Urgent orders already move faster, so marking more orders urgent would likely just reshuffle recruiters' attention rather than speed up the firm overall. These are planning uses of the model, not claims about what would happen if clients changed their orders.

Checks

Residual plots showed no strong curvature, but residuals spread more widely for senior orders, so I re-estimated the model with robust standard errors. The conclusions did not change. Field (2018) recommends checking such plots before interpreting coefficients, and I report the robust results in an appendix.

Limits of the Model

The model explains about a third of the variation. Two thirds is related to things not in the data, such as the client's responsiveness in scheduling interviews and the local supply of candidates in each specialty. And because job orders were not assigned features at random, the coefficients describe associations among features, not what would happen if a client changed one. Cohen et al. (2003) caution that regression coefficients in observational data are conditional associations; they are useful for planning, such as setting expected fill times by job level, but not for claims that changing one feature would change the outcome.

References

Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation analysis for the behavioral sciences (3rd ed.). Lawrence Erlbaum Associates.

Field, A. (2018). Discovering statistics using IBM SPSS Statistics (5th ed.). SAGE Publications.

Tabachnick, B. G., & Fidell, L. S. (2019). Using multivariate statistics (7th ed.). Pearson.