Use of the Dexterity Test as a Preemployment Exam

We have collected data from Western Electric on some of their production workers.   In this assignment, you must evaluate the dexterity test as a preemployment exam. The data include productivity in units/hour, education level, gender, their tenure with the firm, whether they quit in their first 6 months with Western Electric, how they were hired, and their score on the dexterity exam that was given prior to their employment.  You will focus only on gender, productivity and preemployment exam scores.

We have data on the gender of the employees.  Suppose that we decided to use a cutoff score on the dexterity test to determine employability, as well as an education level requirement.  We want to consider two scenarios.  First, we will use only the dexterity score as an employment screen.  Would any of the possible cutoff scores lead to a problem of adverse impact, as defined by the EEOC?  (This is the 4/5th rule, which you can find in the BNA sources in the library).  Please provide a chart that indicates your result.  You can make this chart by finding a passing rate for each exam score cutoff.  You can indicate violations of the 4/5th rule by taking 80% of the men's passing rate and comparing it to the women's passing rate.  To compute this for EVERY possible cut-off score, you need to make a formula that relates to the results from the pivot table that you will make.  (1) Move over several columns from the pivot table and use the highest score as the row. Suppose that you move to column H.  In this cell (H14), make it equal to the cell for men (for example =B14); move over one column(I14) and make this equal to the cell for women in the row (for example =C14).  Now move up to the 2nd highest row value.  How many pass the test if this value is used?  It is the number in the previous row-cell (eg H14) plus the number in the men's next highest cell (eg B13).  This formula will then give you the number passing the exam of men for the current score.  If this formula is copied to all cells in this column above the last row, you will have the number passing at each score.  Now move over past the women's column and make a new formula that makes the percentage passing at each level.  It will be the number in column H for men divided by the total number of men taking the exam (the total of men in the pivot table).  If you put a $ in front of the row (eg B$15), then you can copy this formula for all scores for both the men's column and the women's column.  This will give you the information that is necessary to determine any violations of the 4/5th rule for every score. 

Can we show that the dexterity exam is sufficiently correlated with productivity that it can be used even if there is adverse impact?  To answer this question, we need find the correlation of the dexterity test and output for men and women separately.  Is it sufficiently correlated with output that you would feel confident in defending its use even if there is adverse impact?

These test results are based on the output and tests for current employees.  Can you guess the likely cutoffs that were used to choose employees in previous years?  Suppose that we increased the cutoffs a little.  Can you determine the increase in the number of good employees who would be rejected incorrectly (type II error)?  Can you determine the increase in efficiency in rejecting poor employees (reduction in type I error) that would result?  What do you have to assume about the definition of  "good" and "poor" employees to do this analysis?

The Western Electric data are available on an EXCEL spreadsheet.   You should print out the variable information that is located in the SPSS file information file.

The purpose of this assignment is to introduce you to validating a preemployment exam.