Concurrent validation design uses the current employees to give validity to a test. The test scores are then compared with the current performance of the employee. Predictive validation design uses applicants before they are hired to give validity to a test. The test scores are then compared to later performance on the job.
The nine selection methods discussed will be affected by the validation design used to different degrees.
Interviews – Given the subjective nature of interviews it is hard to determine the effect of the different validation designs. Predictive design would offer a more insightful test because the current employees will have become more homogenous and may present characteristics and skills that hey acquired on the job and did not necessarily have before.
Background Checks and References – There is inherently a low validity for background checks. Applicants choose the names to give the employers and would not readily offer the name of a person that they did not believe would represent them in the best light. Also, the reference person is most likely someone who has some type of relationship with the applicant and they would most likely not want to give any damaging information to the prospective employer. The predictive design would be more useful in the background check and reference check selection. Current employees will adapt to the culture that has been created in each department and in the company as a whole and this may trickle into their life outside of work. Thus, they may get similar types of responses from references.
Physical Ability Tests – These tests are better served by predictive validation. If a position requires and applicant to lift a box from a palate and push it on to a conveyor belt, then they will need to have the specific strength to do that. Now if you look at a current employee who has been lifting the boxes and pushing them on to conveyor belts for 3 years, they most definitely have built up the strength to complete that task. However, the important fact is what the current employees strength level was before they were hired and had the 3 years to build up their strength.
Cognitive Ability Tests – These abilities, like strength, can build over time. Employees will continue to get better and faster at tests given to test their ability in these areas. The cognitive ability tests are better validated by predictive design.
Personality Inventories – These types of evaluations are inherently hard to validate. Applicants and employees are prone to giving responses that they think the test giver wants to hear. Concurrent validation could be used, however, the current employees who perform well may have conformed to match what the company was looking for instead of coming in with those traits. Current employees would also be prone to giving answers that they believe the test giver would like to hear. Using predictive validation will be more accurate because the test values all applicants before they have had any influence from the company, other employees or their managers.
Work Samples – This type of evaluation is better validated by predictive design. For example, if you have a programmer who has been working for the company for 5 years and is a top performer, their score is bound to be higher on a work sample test than a new employee who has been with the company for 2 months. If you took a sample of the same top performer before they started with the company then their sample would have been closer to that of the employee of 2 months.
Honesty Test – These are tests that, like many others, can be faked. People are generally honest or dishonest. It is fairly black of white. There are varying degrees, but in my opinion any degree of dishonesty creates a dishonest person. Are there dishonest people currently employed by the company? Yes. Are there honest people employed by the company? Yes. Can either side fake their tests to be seen in a different light? Yes. This test is not effected by the switch from concurrent to predictive design because it will be virtually impossible to get a standardized set of results to base subsequent tests on.
Drug Tests – Drug test are another type of test that is very black or white when it comes to the results. The applicant or employee is either using drugs illegally or they are not. This is not affected by using concurrent or predictive design.
Companies that have rigorous selection methods give the impression that they must be good companies to work for because it is very hard to get into the company. The company spends a lot of time and money of the hiring process to make sure that they have the right fit for every position. This can be reinforced by having a very well planned and well executed orientation and training process.
If one relies solely on supervisory evaluations as the criterion in validation tests there will be predictors that are over and under estimated. Physical ability and cognitive ability tests are likely to be overrated. Supervisors will expect new employees to have very high scores on these tests based on the current employees. The supervisors are not taking into account the development and time that it took the current employees to achieve the current levels. Work sample may be overrated as well. There may be a higher than possible expectation from the supervisory stand point.
If a company relies solely on social network data there can also be issues with over or under estimation. A supervisor could request to be copied on all system idea emails. This type of action could lead to overestimation of their involvement in the development process. There could also be situations where two employees may sit right next to each other and might discuss ideas. Then they may bring these ideas up in a group meeting. These types of interaction would not be gauged and would lead to an underestimation of the contribution from those specific employees.
The supervisory data and the social media data are bot useful. However, the usefulness multiplies when they are used together and added to traditional predictors.