Answer: It depends on who you listen to, the Code or the IRS!
The following is drawn from Q 9:9 of my book, Who's the Employer? It deals with controlled groups, but the logic would apply equal well to overlapping controlled and affiliated service groups:
IRC §414(b) provides that for most qualified plan purposes, all employees of corporations that are part of a controlled group (whether or not they are component members of that group) are deemed to be employed by a single employer. Logically, this means that when you have overlapping groups, you end up with what amounts to one big group for qualified plan purposes. This is illustrated by the following example.
Three Corporations, Two Groups, One Employer. Corporations X and Y are members of a controlled group. Y and Z are members of a different controlled group. For income tax purposes, Y has filed an election to be treated as part of the group with X and not with Z. That election is ignored for qualified plan purposes.
All employees of X and Y are deemed to be employed by a single employer for plan purposes. All employees of Y and Z are deemed to be employed by a single employer. By the normal rules of logic, this means that all employees of X, Y, and Z are employed by a single employer.
Think of it this way. Suppose X wishes to sponsor a plan covering all its employees as defined in the Internal Revenue Code. Who are its employees? Because of IRC §414(b), its employees are everyone on its payroll and everyone who is an employee of Y. Who are the employees of Y? Because of IRC §414(b), the employees of Y include everyone on its payroll, all employees of X, and all employees of Z. Hence, if X sponsors a plan covering all its employees, it brings in the employees of Y and Z automatically.
At a recent conference, IRS representatives took a different approach, saying that overlapping groups must be tested twice to demonstrate compliance, once for each group. So, in the above example, a plan that included employees of Y would have to be tested against the XY controlled group and against the YZ controlled group. The author disagrees with this approach, based on the literal wording of the statute. There is no regulatory or other official sanction for this approach.
The facts cited in this example are the same as those in the author§s first controlled group case in the late 1970§s. The profit-sharing plan sponsor was X, who had reluctantly included the employees of Y, but had never included the employees of Z in his plan. Z, as fortune had it, had ten times as many employees as the other two corporations put together.
We had asked for a favorable determination letter for X§s plan. The IRS threatened to give us an adverse letter instead, using exactly the reasoning shown above. All three corporations, they held, were effectively a single employer. In desperation, we examined the plan language and found it automatically brought in the employees of Z if Z was treated as being in a controlled group with X. We argued that IRC §410(b) was therefore satisfied, and we simply needed to redo the allocations for the last several years to include all the eligible employees of all three. The IRS agreed, relieved in those simpler days not to have to issue an adverse letter.