11 BB
11 BB
DEVELOPING AND USING WEIGHTED APPLICATION BLANKS: AN EXPERIENTIAL EXERCISE Donald B. Dawson, Northern Illinois University Albert S. King, Northern Illinois University Ralph F. Catalanello, Northern Illinois University ABSTRACT Development and use of effective weighted application blanks begins with an analysis of procedures required in the developmental process. An experiential exercise is presented which gives a hands-on experience in developing and using weighted application blanks. The exercise was developed to aid conceptual and operational understanding of procedures in weighted application blank development. The hypothetical data used were developed solely for the purposes of the exercise. Upon completion, the individual should have a keen insight into the development and use of weighted application blanks. INTRODUCTION Learning Objectives 1) To understand the logic and systematic procedures for developing weighted application blanks (WABS). 2) To give the individual a hands-on experience in developing WABS. 3) To develop an awareness of the possible uses, implications, and ramifications of using WABS. Materials Needed: 1) Pencil and calculator. attempts to identify what personal factors reliably differentiate groups of potential desirable and undesirable employees. By identifying personal factors of job applicants, a rapid screening of the applicants is possible. The results, along with other information, can be used to enhance the selection and placement of individuals in organizations.2 There are numerous studies that statistically support the notion that WABs are successful in differentiating potential desirable and undesirable employees [2; 6]. One study found that even in relatively low-level jobs with modest hiring and training costs, a large savings was obtained by using a WAB [4]. The potential to an organization for proper use of a WAB is not however, without limits. One problem is that WABS tend to lose predictive efficiency over time. Another obstacle is that practitioners and researchers simply do not know how to design and develop WABS. An experiential exercise which presents a step-by-step method for development may foster continued emphasis on their use and promote renewed interest in successful selection and placement techniques. The procedure for developing a WAB is straightforward:3 1) 2) 3) 4) 5) 6) 7) Choose an appropriate criterion. Identify criterion groups. Select application blank items to be analyzed. Specify item response categories. Determine item weights. Analyze independent samples for cross- validation. Set WAB scores for selection.
2) Read the introductory material for each section as you come to that section. Then, follow specific instructions given for each section. NOTE: Each section builds on previous ones, hence it is important to understand each section before moving on. OVERVIEW Selection and placement of individuals is one of the most difficult problems facing an organization. Improper (i.e., discriminatory) selection and placement can have far-reaching repercussions. Selection and placement techniques have been developed and refined that give management a foundation for basing selection decisions. One of these selection techniques is the Weighted Application Blank (WAB) [3]1. A WAB ". . . provides one systematic method for determining which personal factors are important in specific occupations and how to use them in selection, [3, p. 5]. The WAB Much of the material presented has been adapted from Development and Use of Weighted Application Blanks, by George W. England, Rev. Ed., Industrial Relations Center, University of Minnesota, 1971.
1 2
The results obtained through the introduction of a WAB should not be the sole basis for making selection decisions. See, Pace, Larry A. and Lyle F. Schoenfeldt, Legal Concerns in the Use of Weighted Applications, Personnel Psychology, Vol. 30 (1977), pp. 159-16 6.
3
This is the basic procedural outline as indicated in England (1971), however, some steps have been modified for this exercise.
120
Predictable
Relevant
Uncontaminated and If a criterion is biased, results obtained bias-free will be distorted, and have little meaning. While it is extremely difficult to obtain a completely unbiased criterion, it is something to strive for in any study. To be completely accurate, an unbiased criterion is Necessary. If, in analyzing a chosen criterion, it is found that it may not be as good as desired, it is wise to search for one that could be possibly better. IDENTIFYING CRITERION GROUPS Choosing an appropriate criterion and identifying criterion groups are the most important steps in developing a WAB. If an inappropriate criterion is chosen and criterion groups are not identified properly, the whole intent of the study is not being met. An appropriate criterion coupled with properly identified groups leads to meaningful results. Once an appropriate criterion is chosen, two groups of current employees are formed: a High-Criterion Group (Group I)-representing desirable employees, and a Low-Criterion Group (Group II)--representing undesirable employees. In order to make best use of the data, both groups should be as large as possible, with a minimum of 75 employees per group. These groups are subdivided into a Weighting Group and a Holdout Group. Each Weighting Group should include 2/3 of the larger group, with the remainder going into Holdout Groups. When all groups are formed, major questions need to be answered: 1) What is the distinction between short and long tenure?, and 2) How far back, chronologically, should we go to get data?
Representative
Discussion No matter what criterion is ultimately chosen, it has to be reliable. The reliability of a criterion is necessary for pre diction purposes. In this particular instance, job tenure appears, at least for the present time, to be a reliable criterion. Necessary Any criterion chosen should be a representative measure of the problem. In this instance, management has defined the problem as being one of tenure. Therefore, the criterion of job tenure appears to represent the problem.
120
2)
Equal Interval Classes--divides the responses into equal intervals within a range of responses (e.g., for Age: 21-26, 27-32, etc.). Maximum Weight Classes--trial and error approach to maximize differences of a particular response item. This method is least desirable because it may capitalize on chance differences within a particular item.
3)
Some items can easily be divided into response categories by a natural division. For example, marital status can easily be divided into the response categories of single, married, divorced (including legally separated), and widowed. For purposes of the exercise, all items are divided into response categories using Equal Frequency classes, or by a natural division. Instructions: 1) From data in Table 14 total responses in each category of "Age" for both High and Low Weighting Groups (use the worksheets provided). 2) Total both Columns 2 and 3. 3) Convert numbers in Response Categories into percentages; i.e., divide each Response Category in Column 2 by the total of Column 2. Place this number in Column 4. Do the same for Columns 3 and 5. 4) Subtract Column 5 from Column 4 for each Response Category. 5) Do this for remaining application blank items. DETERMINING ITEM WEIGHTS To determine whether desirable workers have different response patterns from undesirable workers, it is necessary to determine Net Weights and Assigned Weights. This is a straightforward process-using the charts in Tables 2 and 3. Instructions: To determine Net Weights, use Charts A & B in Table 2. Take the difference in percentages (Column 6), and locate that same percentage in Chart A or Chart B. Place the corresponding Net Weight in Column 7. If percentages are negative, use the same process, only treat the differences in percentages and Net Weights as negatives also. To determine Assigned Weights, use Table 3. Locate the appropriate Net Weight in the table, and place the corresponding Assigned Weight in Column 8.
121
Age:
Mp - Mq t Mp - Mq t Mp - Mq t
x pq =
Age:
x pq =
Age:
x pq =
r The correlation coefficient pbi depends directly upon the differences r between means Mp and Mq. If pbi is significant at a specified level of confidence, a difference between desirable and undesirable employees exists for that particular item. Therefore, cross-validation supports use of that item in the final WAB. Instructions: From Table 6, determine whether r is significant at p .05 level of r confidence. Then compare your results with those that follow. Note: If pbi is calculated as a negative, treat it as a positive number. Item Age Average Tenure Distance of Residence Number of Previously Employed Years .5000 yes r pbi Significant at p .05 yes yes yes
Do the items clearly differentiate between desirable and undesirable employees? What confidence can you hold in the results? SETTING WAB SCORES FOR SELECTION The practicality of a WAB is that selection decisions can be enhanced by determining which applicants appear to be more desirable employees. It is difficult in practice, however, to set rigid cutting scores for selection. Rigid cutting scores constantly need to be validated [7, Section 5, Paragraph B], and fluctuations in supply and demand of potential applicants might influence the number of potentially desirable employees. One alternative is to develop a range of scores in which more desirable applicants appear at the top of the range, and the more undesirable appear at the bottom. Those appearing more desirable can be evaluated further to assess other qualifications. EXTENSIONS AND APPLICATIONS A WAB can be used as a screening device to rapidly identify potentially desirable applicants. It may be used with other tests and measures as a predictor of possible success on the job. The WAB provides a quick and inexpensive technique to process many applicants and determine which appear to be more desirable. Those applicants scoring high can be evaluated on other qualifications to assess their potential to the company.
where: Mp = mean of values in the favored (desirable) category. Mq = mean of values in the remaining (undesirable) category. t p = standard deviation of the total sample. = proportion of cases in the favored (desirable) category, (i.e., number of cases in desirable category/total number of cases. = proportion of cases in the remaining (undesirable) category.
For a good discussion of the Point-Biserial Coefficient of Correlation, see J. P. Guilford, Fundamental Statistics in Psychology and Education, (New York: McGrawHill, 1978), 6th ed.
122
[3]
[4]
[5]
[6]
[7]
123