Let’s say you’re about to purchase an employment related test. It could be for hiring entry level personnel, help make a promotional decision or for providing feedback to people on their skills. There are about 50,000 of these available and several that purport to measure the same things (leadership, EQ, etc.), so how do you choose the right one?
Professional standards require that test publishers provide important information about tests so that buyers can make informed choices. Of course, not everyone selling tests, especially on the internet, bothers with such niceties. At a minimum, here’s what you should find out about a test before you buy it (note that some of these may require you consult with someone with good knowledge of statistics):
- What is the validity of the test? Validity means the relationship between scores on the test and doing the job. For practical purposes, there are two ways to show the validity of a test:a) Is there a statistical relationship between test scores and job performance? Remember correlations from college statistics? What you are looking for here is a statistically significant (large) correlation between how someone does on the test and how well they do on the job. Look for correlations of about .30 and higher. This provides evidence that the test does what it’s supposed to do: measure how well someone can do a job. This type of data is critical when you are looking at personality or aptitude tests.
b) Do the items on the test represent the job? If you were looking at a test for an electrician, you would want to see items about Ohm’s law, different phased motors, etc. Or, if you were looking for a manager assessment, you’d want it to involve coaching, providing directions, etc. This is more than just the test looking right. They key here is that the content of the test should match the content of the job you want to use it for. Test publishers should be able to show that the test items were written based on input from job subject matter experts.
- What is the distribution of scores on the test (or the average score and standard deviation) for people in the same job as those you are going to give the test to? This is how you can tell if a person’s score is good, bad, or average. Saying that someone scored a 37 on a test is meaningless without this information.
- Is there a difference in scores between different demographic groups on the test? Test publishers should be able to let you know if group (gender, race, age) differences exist on the test. This doesn’t make a test good or bad, but it lets you understand whether particular passing points might lead to adverse impact.
- You’ll want to get some administrative details as well. What’s the time limit? Does it have online and paper-and-pencil versions? If online, how is feedback given?
If the test publisher doesn’t have this information available, it means they have not done the proper R&D work on their test and you shouldn’t buy it (unless you want to be the R&D site). Remember, once you administer a test to a candidate or employee, it’s your responsibility and you need to live with the outcomes. Be sure you have full knowledge of the tests that you are using.
For more information on pre-employment testing, test validation, skills assessment, and talent management, please contact Warren at 310 670-4175 or warren@allaboutperformance.biz.