Predicting Job Performance

After employees are hired, a key activity that an HR analytics wants to conduct is to explore what factors predict post-hire employee performance the best. One reason to do this is to test how valid the interview processes are; a second reason is to identify the factors that will predict who is likely to turn out to be a high performer. 

The dataset we are going to use is the combination of HR information collected throughout the interview stages, specifically the scores and ratings of the assessment tests. The company is a financial firm which hires a cohort of graduates each year. The variables are broken down into the following categories:

  • HR and resume-related information
    • Grad ID
    • Gender
    • Education
    • BAMEyn – Black, Asian, or minority ethnic
    • Work experience: Have work experience before or not
    • Job function – HR, finance, marketing, sales, risk management, legal, or operation.
    • Year 1 performance rating (Response variable)
  • Assessment tests
    • Personality test scores
    • Competency test scores
    • Aptitude test scores
  • Onboarding
    • Induction day – Attend induction day
    • Induction week – Attend induction week
    • Onboard buddy

Because we will use regression analysis to predict the performance rating, for the categorical variable “Job Function”, we will create a set of dummy variables to represent each function.

Call:
lm(formula = Year1performanceRating ~ 
                 Gender + EducationHighest +     
                 BAMEYN + WorkExperience + ACPersonalityO + ACPersonalityC +     
                 ACPersonalityE + ACPersonalityA + ACPersonalityN + ACRatingINTCOMPA +     
                 ACRatingINTCOMPB + ACRatingINTCOMPC + ACRatingINTCOMPD +     
                 ACRatingINTCOMPE + ACRatingAPTnumerical + ACRatingAPTverbal +     
                 InductionDay + InductionWeek + OnBoardingBuddy + FinanceDummyV +     
                 MarketingDummyV + SalesDummyV + RiskDummyV + LegalDummyV +     
                 OperationsDummyV)
Residuals:     
Min        1Q       Median       3Q      Max 
-2.05797  -0.30754  0.00088  0.41912   1.62688 
Coefficients: 
                      Estimate Std. Error t value Pr(>|t|)    
(Intercept)          -4.8225077  1.2015570  -4.014 7.61e-05 ***
Gender               -0.8578881  0.1306267  -6.567 2.34e-10 ***
EducationHighest      0.1384466  0.0825704   1.677 0.094669 .  
BAMEYN               -0.0193234  0.1039189  -0.186 0.852616    
WorkExperience        0.0606132  0.0892045   0.679 0.497368    
ACPersonalityO       -0.0034899  0.0037530  -0.930 0.353190    
ACPersonalityC        0.0079011  0.0026536   2.978 0.003149 ** 
ACPersonalityE        0.0097542  0.0035029   2.785 0.005710 ** 
ACPersonalityA        0.0009825  0.0030945   0.317 0.751104    
ACPersonalityN       -0.0009245  0.0022080  -0.419 0.675739    
ACRatingINTCOMPA      0.3638584  0.1381901   2.633 0.008913 ** 
ACRatingINTCOMPB      0.1306136  0.1197009   1.091 0.276099    
ACRatingINTCOMPC      0.0903146  0.1402235   0.644 0.520032    
ACRatingINTCOMPD      0.3592834  0.1125143   3.193 0.001561 ** 
ACRatingINTCOMPE      0.3456275  0.1654996   2.088 0.037629 *  
ACRatingAPTnumerical  0.0099152  0.0070480   1.407 0.160550    
ACRatingAPTverbal     0.0077312  0.0100409   0.770 0.441940    
InductionDay          0.3567033  0.1412323   2.526 0.012078 *  
InductionWeek         0.0234061  0.1286484   0.182 0.855757    
OnBoardingBuddy       0.2318439  0.2521455   0.919 0.358601    
FinanceDummyV         0.1540589  0.2537409   0.607 0.544223    
MarketingDummyV       0.6236678  0.2194201   2.842 0.004794 ** 
SalesDummyV          -0.0636855  0.2959641  -0.215 0.829777    
RiskDummyV            0.9228754  0.2446009   3.773 0.000195 ***
LegalDummyV           1.1510105  0.2548264   4.517 9.12e-06 ***
OperationsDummyV      0.4018696  0.2333447   1.722 0.086090 .  
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.6991 on 292 degrees of freedom  
(42 observations deleted due to missingness)
Multiple R-squared:  0.6639, Adjusted R-squared:  0.6352 
F-statistic: 23.08 on 25 and 292 DF,  p-value: < 2.2e-16

The analysis provides evidence that the personality and the competency tests are a valid selection technique and are significant predictors of performance. The result also reveals that investing in the induction day has benefit too.

One thing to note is that those whose job function in marketing, risk management, and legal tend to account for a fair bit of variance in the performance appraisal ratings. This would probably need some investigation in order to understand if there is systematic differences in performance scores given in different departments. Also, we notice that the aptitude test doesn’t predict performance. If we assume that it should predict performance, it may be the case in which verbal and numerical reasoning might have an influence on jobs in a varied fashion across different functions.

Also, it seems that female graduates do not perform as well as male graduates. Before drawing any conclusion, we would need to consider whether there are functional variations in how appraisal ratings are applied in practice. Also, we may explore if departments vary in female and male ratio.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s