1. PM systems evidence irrational behavior when discrimination determines employee fortunes Irrational systems are disordered Order in the system, if any, is not correspondent with administrative claims as to how the system is intended to operate How should performance ratings behave in a rational system? 1.Correlate with LOS? 2.Correlate with pay increases? 3.Correlate with retention decisions? Should anything other than LOS and job performance be a better predictor of pay increases or decision to retain an employee?

2. Southern Bank >$300 billion in Total Assets Rapid expansion through mergers and acquisitions Merged/acquired 12 institutions in Florida from 91-94 In 2014, total assets ~ $1.5 trillion and world’s largest bank according to market capitalization (~250 billion) Over 3000 affected by involuntary layoff Defendant in age discrimination case ADEA protects workers of 40 years of age and older Approximately 350 plaintiffs filed suit Focus was on entire system for which we had data Discovery was problematic College recruitment, compensation, promotion, layoffs

3. 25,721 employees in dataset data file spanning 1991-1994 PM System 1-5 point graphic rating scale Annual, top-down, single-source system Many missing appraisal observations Depositions of HRM staff asserted that PM ratings and length of service were the only criteria for reduction in force decisions Naturally we focused on the tenability of this assertion We began with establishing prima facie evidence

4. E MPLOYMENT S TATUS Total Retained Involuntary Layoff Voluntary Termination and Other Terminated - Reason unknown A GE G ROUP Under 40 years of age Count155395116768365016 31.0%19.0%33.4%16.7%100.0% 40-49 years of age Count7795475112982135 36.5%25.6%23.9%14.0%100.0% 50-59 years of age Count4763832832071349 35.3%28.4%21.0%15.3%100.0% 60+ years of age Count127228176118649 19.6%35.1%27.1%18.2%100.0% Total Count29352109264614599149 32.1%23.1%28.9%15.9%100.0% Table 4. Employment Status by Employee Age Group Age of employees in 1994 was one month older than employees in 1991 It could not have been due to a higher voluntary termination rate among older employees because their rate of voluntary and other terminations was lower than that of employees in the youngest age group (23.9%, 21%, and 27.1% versus 33.4%).

5. A GE G ROUP 1991 L AY - OFF R ATIO 1992 L AY - OFF R ATIO 1993 L AY - OFF R ATIO 1994 L AY - OFF R ATIO Under 40 years of age Count77/1630710/2263118/167146/1599 % Within Age Group.047.313.070.028 40-49 years of age Count63/842366/114582/86136/815 % Within Age Group.074*.319.095*.044* 50-59 years of age Count35/511265/74158/53425/501 % Within Age Group.068*.357*.108*.049* 60+ years of age Count17/144176/30323/15012/139 % Within Age Group.118*.580*.153*.086* Total Count192/31271517/4452281/3216119/3054 %.061.340.087.038 Table 6. Proportional Involuntary Lay-off Rates of Each Employee Age Group From 1991-1994. 11 of 12 differences in lay-off ratios are statistically significant 10 of 12 differences in lay-off ratios demonstrate adverse impact on ADEA employees Aggressive lay-offs in 1992 produced a rate among 60+ employees that was almost twice that of those under 40 years of age

6. Employment Status Employment Status Years of Service Job Performance ADEA Status Figure 17. Southern Bank Decision Model of Employment Termination Model 1 Coefficient was not statistically significant. Model 2 Coefficient is statistically significant.

7. Among those for whom PM data were available, no rational logic could explain the pattern of empirical relations observed Identifiable, legitimate factors could not explain ratings Relations with employee LOS went from positive to negative from 1990-1994 Relations of pay and performance ratings went from positive to zero among ADEA employees but not among non-ADEA employees 44.8% of employees designated as PNE (position not evaluated) were subject to involuntary termination r = -.75 between positions converted to PNE and presence of payroll record in 1994 Ongoing college recruitment programs hired several thousand new employees over the 1991-1994 period while hundreds of laid-off employees had reapplied for their old positions and were denied employment Both adverse impact and adverse treatment were viable allegations that could be substantiated by evidence

8. The series of analyses we reported in this document repeatedly support the allegation of age- related bias being manifest in a wide variety of employment decisions at XXXXXXX during the 2- year period immediately subsequent to the acquisition of large institutional competitors. The demonstration of age related bias was pervasive in most decisions involving employee retention including the designation of job as “position not evaluated” which was correlated with involuntary layoffs; the reversal in appraisal ratings relationship with age subsequent to acquisition; the higher promotion rate among employees younger than 40 years of age; the reversal in the relationship between age variables and percentage increase in hourly pay and the consistent demonstration of a negative age-pay raise relationship subsequent to acquisition; the fast-tracking of new management talent through the college recruitment program; and the often improbably high over-representation of older employees among those experiencing involuntary layoffs. This last effect was robust across most acquired institutions, job groupings, and years subsequent to acquisition. Finally, these last series of tests conducted with the largest population of XXXXXXX employees available to us for analysis suggested a pattern of force reduction that systematically reduced the proportion of employees above the age of 40. Whereas younger employees tended to engage in higher rates of voluntary termination, older employees were designated at a disproportionate rate by XXXXXXX for involuntary termination through layoffs. A definitive test of two decision process models clearly indicated that the best explanation for employee status as a “layoff” versus “retention” was membership in the more senior age group (i.e., ADEA class membership) at the time of the decision.