In completing Sec. 129 discrimination testing, particularly the benefits test, what pool of employees to you base the testing on? Is it only employees who participate in the plan, all eligible employees, or all employees? The statute isn't clear on this and I can't find any guidance.

As you pointed out, there is no clear guidance on this issue. However, most TPAs will base the test on all eligible participants (although you can exclude those earning under $25,000 - I think that's the threshold but I don't have Code Section 129 in front of me right now). The problem with basing the test on all eligibles is that you end up with many individuals who are averaged in at $0 (i.e., those who are eligible but don't elect to participate are averaged in). This is the case even though they may not be participating because they have no eligible dependents. The exclusion of those under $25,000 allows you to exclude many who don't participate because the tax credit is more valuable.

Now, for an angle on this that I've pondered over the years. Suppose you wrote the eligibility requirements for the dependent care program to exclude anyone who has no dependents. You would need to make sure that the plan is nondiscriminatory as to eligibility, and I could make a good argument that such a provision would be nondiscriminatory. Taking that approach would eliminate the $0 amounts being averaged in for those who just don't have a need for dependent care because they have no dependents. The practical problem is gathering accurate employee census data as to who has dependents. FYI, I haven't seen anyone try this.