Halden Zimmermann | Attendance Improvement Program

7
EXP 420 Project Assessment of the Attendance Improvement Program at the Ford Motor Company

Transcript of Halden Zimmermann | Attendance Improvement Program

Page 1: Halden Zimmermann | Attendance Improvement Program

!!

EXP 420 Project !

Assessment of the Attendance Improvement Program at the Ford Motor

Company

Page 2: Halden Zimmermann | Attendance Improvement Program

I. EXECUTIVE SUMMARY ! A study was performed to review and analyze the increased absenteeism at the Buffalo Stamping Plant (BSP) in Woodlawn, New York. A recent Detroit Headquarter communique revealed an increased absenteeism rate among the hourly employees at BSP. The BSP Senior Management subsequently decided to explore the underlying causes of the observed increase in absences and to generate appropriate recommendations.

The Ford Motor Company instituted a union-approved Attendance Improvement Program (AIP) in 1984 to improve employee attendance. This program specifically targeted employees that have less than 10 years of service (See the AIP Guide in Appendix G).

In order to determine what the underlying causes may be for the increased absenteeism rate, historical absenteeism data beginning July 2000 through October 2001, was reviewed for all of the employees at the BSP plant. The hypothesis of this study is that employees with more than 10 years of service contribute to the increased controllable absences:

Ho: Controllable Employee Absences of µ <=10 Years of Services - µ >10 Years of Service >= 0 Ha: Controllable Employee Absences of µ <=10 Years of Services - µ >10 Years of Service < 0

Historically, Ford management believed employees with less than 10 years of service are more likely to utilize the four occurrences (one occurrence may be up to 9 consecutive days or 72 hours of absence) that are allowed each rolling twelve month period before an employee enters the AIP. This likelihood was based on the fact that employees with less than 10 years of service received less vacation days than employees with more than 10 years of service. These facts supported the enrollment of employees with less than 10 years of service in the AIP remedial program to correct their absenteeism record. Employees with 10 or more years experience are not contractually obligated to participate in AIP.

The results of the analyses revealed that the absenteeism rate of employees with greater than 10 years of service were not statistically significant when compared to employees who have less than 10 years of service. However, upon inspection of the data using time-series, there does appear to be the potential for a Type II error, i.e. failing to reject the null or base-line hypothesis when it is false (see Appendix E and F). There may be a need to inspect the data further to determine whether employees are likely to “abuse” the allowed occurrences available to them, as well as the potential that plant foremen are miscoding absences.

Given these observations, we recommend that select employees with more than 10 years of service be admitted to the AIP (with union management approval). These employees are those who are abusing their contractual rights such that they are chronically absent (more than 1000 hours of absenteeism). Additionally, the plant must retrain foremen regarding the absence codes recording and evaluate the use of a "reward" program for all employees that demonstrate consistent attendance and do not abuse the amount of occurrences that are allotted to them. !!!!!!!

! 2

Page 3: Halden Zimmermann | Attendance Improvement Program

!!!!II. INTRODUCTION !

The goal of this project is to examine and evaluate the expected controllable absenteeism between employees with 10 years of service or less and employees with more than 10 years of service at the Ford Motor Company Buffalo Stamping Plant in Woodlawn, NY. A recent report on absenteeism released from the Ford Motor Company’s headquarters in Detroit, Michigan, shows that BSP has experienced an increased rate of absenteeism. BSP management’s strategy is to reduce controllable absenteeism and ultimately, the costs of absenteeism that contribute to the unfavorable cost variances experienced by the plant. !III. OBJECTIVE

The objective of the analysis is to review the available absenteeism data of all hourly employees at the BSP. This data will be collected and compiled from two major systems: the Timekeeping & Work Order System (TWOS) and the Plant On-Line Reporting System (POLR). These systems collect the hourly attendance of each employee in the Ford Motor Company plants by tracking the hours worked and/or the hours absent. Each hour (or partial hour) of absence is assigned a code by the plant foremen.

Ford’s absence codes can be segregated into two categories (see Appendix A for a complete list of codes and their definitions by category):

Controllable – absences which may be controlled by plant management, and Contractual – absences dictated by the United Auto Workers Union labor contract

Different statistical methods as well as charts and graphs will be utilized to analyze the available data to reveal trends and relationships. In this analysis we will test the hypotheses that:

Ho: Controllable Employee Absences of µ <=10 Years of Services - µ >10 Years of Service >= 0 Ha: Controllable Employee Absences of µ <=10 Years of Services - µ >10 Years of Service < 0

This hypothesis is established to statistically test whether the expected amount of controllable hours of absence by employees with more than 10 years of service have more controllable absence vs. those employees with less than 10 years of service.

We assume that this data will provide valuable information that will be used by BSP's management.

!

! 3

Page 4: Halden Zimmermann | Attendance Improvement Program

IV. DATA DESCRIPTION

This study included both controllable and contractual absenteeism data collected from July 2000 to October 2001. The coded data was compiled from Ford’s database of employees during this sixteen-month period. The absenteeism data is compiled into two data sets: amount of absence by employee by absence code, and amount of absence by absence code by month.

Within the employee data set, additional variables were formed to segregate the employee sample into analyzable subsets. These variables were created as "dummy" variables with the intent of homogenizing the employees within these subsets, and to determine what relationship, if any, these variables have to the BSP absenteeism.

!V. DATA ANALYSIS PROCESS & METHODS

To achieve the objective of this project, the employee and monthly data were inspected and "cleaned" such that the Overtime Refusal data was deleted. This omission will have no affect on the outcome of this study as there were no observations during the sixteen-month period. !

There were 5 phases to the analysis process: !Phase I:

Test the validity of the observed sample data. This was done by randomly selecting individuals within the employee sample and validating the absenteeism hours with Human Relations and their foremen. The employee sample appears to be consistent with the absenteeism recorded for each employee. !Phase II:

Calculate an employee's age on 10-31-2001, years of service on 10-31-2001, and age on their hire date (rounded up to the nearest integer). A Y2K error was detected for employees whose date of birth occurred prior to 1940. These errors were corrected before completing these calculations. !Phase III: ! Creat dummy variables to segregate the following employees: ▪ Employee Years of Service: 10 years or less, more than 10 years ▪ Employee Years of Service: less than 5 years, between 5 and 10 years, between 10 and

20 years and more than 20 years ▪ Employee Age: less than 20, between 20 and 30, between 30 and 40, between 40 and

50, and more than 50 ▪ Employee Gender: Male or Female ▪ Employee Race: White, Black, Asian, Hispanic or North American Indian ▪ Employee Skill-Level: Unskilled, Skilled-Direct, Skilled-Indirect, Retired ▪ Employee Status: Active or Inactive (Retired/Terminated)

! 4

Page 5: Halden Zimmermann | Attendance Improvement Program

Phase IV: !Use MS-Excel and StatPro to calculate summary statistics and correlations among

all of the variables (see Appendix B). Used StatPro to create scatterplot charts of select paired variables and their correlations, time-series charts of the absence codes, and side-by-side box plots of Controllable and Contractual Absences with their respective component absences. !Phase V: !

Constructed and performed hypothesis tests of specific correlations and differences of sample means to determine the statistical significance of these estimators. Upon completion of each test a decision re the null hypothesis is provided at an appropriate confidence level. !VI. DATA ANALYSIS RESULTS !Phase I: The sample data was compiled and arranged so the employee subsets could be analyzed statistically. The data was then arranged in a table format so that different subsets could be analyzed against the Controllable and Contractual absenteeism codes. The sample data is available on a diskette labeled EXP420 Project Data within a WinZip file titled "Total Sample Data & Subsets" provided with this project report. !Phase II: Box plots of Controllable and Contractual Absence and their respective component absence codes were constructed to determine the influence that particular absence codes may have on the sample statistics (see Appendix C). The Controllable Absence box plot reveals a significant amount of outliers associated with Medical Absences. Additionally, the Contractual Absence box plot revealed a significant amount of outliers associated with Temporary Lay-off Absences (TLO). These outliers in both the Controllable and Contractual Absences indicate that there may be unique individuals that should be removed from the statistical sample due to their circumstances or there is a high degree of miscoding occurring that is distorting the outcome of the statistical analysis. !Phase III:

Correlation of Sample Variables !One particularly important correlation of Controllable Absence with the dummy

variable "10 Yr Split" was statistically analyzed with an initial 95% confidence level (see Appendix D) using the following hypothesis test: !

Ho: Population correlation ρ = 0, where ρ is the sample correlation Ha: Population correlation ρ ≠ 0

! 5

Page 6: Halden Zimmermann | Attendance Improvement Program

!!The outcome of this statistical analysis was that there was a significant statistical

correlation between these variables at a 99% confidence level. !Hypothesis Testing !Based on the result of the correlation analysis, the primary hypothesis test was

constructed of the difference of the means for the two groups of employees with controllable absences: !

Ho: Controllable Employee Absences of µ <=10 Years of Services - µ >10 Years of Service >= 0 Ha: Controllable Employee Absences of µ <=10 Years of Services - µ >10 Years of Service < 0

The outcome of the hypothesis was to fail to reject the null hypothesis at a 95% confidence level. !

To solidify our results, we performed an additional two-tailed hypothesis test comparing the difference of the means of the same two groups in Phase IV using the following hypothesis: !

Ho: Controllable Employee Absences of µ <=10 Years of Services - µ >10 Years of Service = 0 Ha: Controllable Employee Absences of µ <=10 Years of Services - µ >10 Years of Service ≠ 0

The outcome of the hypothesis was to fail to reject the null hypothesis at a 95% confidence level. Based on these statistical analyses, the data would suggest that employees with 10 years of service or less have a greater affect on the total amount of Controllable absences and should continue to be subject to the AIP program. Additionally, the employees with more than 10 years of service should not be subject to the AIP program. !Phase IV: ! Upon failing to reject the null hypothesis above, there is greater probability of committing a Type II error, i.e. failing to reject the null when it is in fact false. As a result, further analysis may be warranted over all of the variables. !!VII. DISCUSSION !

Examination of the data leads us to conclude that controllable absence associated with higher seniority individuals (greater than 10 years) are not relevant. Based on the data, people with higher seniority do not have more controllable absence issues than lower seniority individuals (less than 10 years). !

The probability of having a Type II error warrants additional analyses of the remaining variables and their relationship with Controllable Absence.

! 6

Page 7: Halden Zimmermann | Attendance Improvement Program

!VIII. RECOMMENDATIONS ! Based on our analysis and findings it is recognized that there exists a problem with absenteeism related to the following categories. !

These are the XXXXX number of recommendations that are proposed:

1. Institute the AIP Program across the plant. 2. Categorize employees as to “low,” “medium” or “high” risk for absenteeism based on the

data that is available. Calculate the Expected Absence of each employee and place them into each of the XXXX risk categories

3. Provide incentives to those employees that do not abuse the absences. 4. Given that the absenteeism has hurt the profit maximizing business strategy of the

company this study recommends ……. 5. Use a set of rules for hiring, discipline, and

! 7