Nondiscrimination Testing Revisited

Yesterday, I read an article written by David Godofsky of Alston & Bird (a large law firm headquartered in Atlanta) on nondiscrimination testing in 401(k) plans. If you are interested in this topic, I would commend you to read it. It is very well written. It's almost interesting. And, like very few pieces that I have seen on this topic that has been written by an attorney, David gets it right, even the math.

You see, David has an unfair advantage. I'm not sure when he became an attorney, but I do know that he has been an actuary for a really long time. So, unlike most attorneys that I know, he gets the numbers. And, unlike many actuaries, he gets the law.

So much for praise of an attorney. While he gets the technical stuff right and keeps it interesting, I am going to explain to you where David goes wrong. He compares testing to a peanut butter and jelly sandwich, and puts that sandwich in a positive light. I know this will be anathema to many, but peanut butter is the single worst food ever created. It smells bad, it tastes bad, and it has a bad texture. That, by the way, is either opinion or fact, and the early voting returns say it is fact, although I am afraid that opinion may come on strong. In any event, to add to the problem, he doesn't even serve anything to drink with his abomination.

Retirement plan nondiscrimination is clearly more elegant than that which applies only to 401(k) and similar plans. If Mr. Godofsky is serving PB&J with his testing, the testing that I refer to is more suited to foie gras served with sauternes.

You get the picture? His is packed in a metal lunchbox with Fred and Wilma staring at you. Mine is presented by a trained server and is only for the most discriminating among us. And, oh yeah, mine costs more than his.

Before I move forward, though, I need to warn you that just like foie gras and sauternes, neither most actuaries nor most attorneys understand nondiscrimination testing. I taught this lovely topic for more than ten years, so I saw the misconceptions.

Now, we get technical and explain to you why this process should not be left to the sous chef and the apprentice sommelier, but rather to the masters. While I am not going to go into the really gory details, this is probably about to become the single most technical topic (although I prefer to think of it as elegant) that has ever appeared in this blog.

It all started with Internal Revenue Code Section 401(a)(4) which spells out one of the requirements that a trust that forms part of a retirement plan be qualified (exempt from tax): "[I]f the contributions or benefit provided under the plan do not discriminate in favor of highly compensated employees (within the meaning of section 414(q)). For purposes of this paragraph, there shall be excluded from consideration employees described in section 410(b)(3)(A) and (C)."

That sounds pretty simple doesn't it? You wonder what nondiscriminatory means in this context, don't you?

I'm so glad you asked. You see, without explaining everything, there are two ways that a plan can be nondiscriminatory in the amount of benefits or contributions that are provided:

1. Have a safe harbor design
2. Prove it

Let's suppose that we choose to prove it. In the simple case, we will consider each employee in the controlled group. And, then for each one, we will calculate both that employee's normal and most valuable rate of accrual. Normal generally means life annuity at normal retirement age. Most valuable generally means qualified joint and survivor at the age at which it has the greatest actuarial value. This is not a simple determination. 

Then we take the employees and we put every one of them on a grid defined by his or her normal and most valuable accrual rates (by the way, if you think this is too easy so far, there are multiple measurement periods that can be used to do these calculations, and we can choose to impute permitted disparity or not). Let's assume that in our grid, the highest rates are in the upper left or northwest corner. We get to take people with similar accrual rates and group them. Then, we define a rate group for every highly compensated employee (or group of highly compensated employees). A rate group consists of every employee who sits neither south nor east in our grid of the rate group in question. Once we have done that, we determine a ration percentage for that rate group defined loosely as the ratio of 1 to 2 divided by the ratio of 3 to 4 where 1, 2, 3, and 4 are as defined here:

1. the number of nonhighly compensated employees (NHCE) in the rate group
2. the number of nonexcludable nonhighly compensated employees in the controlled group
3. the number of highly compensated employees (HCE) in the rate group
4. the number of nonexcludable highly compensated employees in the controlled group

So, now we have a percentage for a rate group. We need to get a percentage for every rate group. And, then we need to take the minimum of all those percentages. And, after that, we need to compare that minimum to the threshold percentage for the coverage test that we used and if our minimum is above that threshold percentage, then we pass.

Hold on, what's all this about coverage percentages?

Well, you see, a plan also needs to satisfy the minimum coverage requirements of Internal Revenue Code Section 410(b). And, this section has two tests (you only need to pass one). But, the tests have different threshold percentages.

Let's look at them. Under the ratio percentage test, you again take the ratio of 1 to 2 and divide it by the ratio of 3 to 4 like we did above, except that now in the numerators, we have numbers benefiting under the plan. To pass the ratio percentage test, your threshold is generally 70%, unless you are using snapshot data which is a subject for a comparison of restaurants rather than foods (snapshot data looks like fast food while complete data looks like fine dining). But, suppose you fail the ratio percentage test, or when you are doing your nondiscrimination testing, your rate groups don't clear the 70% hurdle. Then, you will want to use the average benefit test.

It's a nice little test. It has two parts, the average benefit percentage test (ABPT) and the nondiscriminatory classification test (NDCT). And, the NDCT has two parts. To run the ABPT, you combine generally all plans of the employer and determine an accrual rate (or contribution rate) much like we did above for each employee in the controlled group. Then, you take the average over all the NHCEs and divide it by the average over all the HCEs and take the average for the NHCEs and divide it by the average for the HCEs. If it's at least 70%, you pass this part.

On to the NDCT. One part of it is that the group of people covered by the plan has to be a nondiscriminatory classification. It's hard to know exactly what this means, but two examples may help. A plan that covers all the employees working in Grand Fenwick probably has a nondiscriminatory classification. A plan that covers all the employees with purple hair and orange eyes may fail to be nondiscriminatory.

The other part is a version of the ratio percentage test, but with a lower passing threshold. What is that threshold? I'm so glad you asked because you have to figure it out. Just follow these simple steps:

1. Determine the percentage of employees in your controlled group who are NHCEs. Call this the NHCE Concentration Percentage (NHCECP). 
2. Round down to the next lower integral percentage unless you are already at an integral percentage.
3. Subtract that answer from 100.
4. Multiply that answer by 0.75.
5. Add that answer to 20, but don't let your new answer go above 50.
6. This is your safe harbor percentage.
7. Subract 10, but don't let this answer go below 20.
8. This is your unsafe harbor percentage.
9. Halfway in between the two is your midpoint.
10. If your ratio percentage is at least equal to the safe harbor percentage, you pass.
11. If your ratio percentage is less than your unsafe harbor percentage, you fail.
12. If you are between the two harbors, you are in the Sea of Facts and Circumstances and only the IRS can tell you if you pass or fail.
13. The midpoint is generally the threshold for your rate groups.

That sure was fun, wasn't it?

Well, sometimes after you have done all that, you still don't pass. And, that's when you really need to rely on tricks of the trade.

Don't do it yourself. This is dangerous stuff. Talk to the master.

.