Cross Tested allocation

[pmacduff]

This plan did fine last year with cross-testing with the goal to get the most $ to DAD of the contribution . For the current PYE, they lost a NHCE who was younger and increased salaries on the younger sons (who are owners). I gave everyone 5% except Mom & Dad (who are 63 and 66 respectively) and gave them 15%. All of my tests fail miserably no matter how I run (accrual vs. allocation), etc. (I use Relius)

Here are the demographics:

HCES:

SON - AGE 33 COMP $150400 - GREATER THAN 5% OWNER

SON - AGE 39 COMP $168600 - GREATER THAN 5% OWNER

SON'S WIFE - AGE 47 COMP $13,000 - OWNER BY ATTRIBUTION

MOM - AGE 63 - COMP $9600 - GREATER THAN 5% OWNER

DAD - AGE 66 - COMP $147000 - GREATER THAN 5% OWNER

EMPLOYEE - AGE 36 - COMP $91828 - HCE COMP > LIMIT IN PRIOR YEAR - NO OWNERSHIP

NHCES:

EMPLOYEE AGE 31 COMP $48443

EMPLOYEE AGE 55 COMP $51571

EMPLOYEE AGE 59 COMP $56606

I'm thinking maybe I'm not using Relius correctly to test because it seems like if I give everyone 5% except Mom & Dad, I should pass easily. What am I missing? Any suggestions appreciated!

[Mike Preston]

Use restructuring. Slam dunk.

Plan 1: Mom, Dad and 31 year old NHCE.

Plan 2: Everybody else

Each has ratio percentage in excess of 70% (actually, each is 100%), so 410(b) not an issue.

Test plan 2 on the basis of contributions (everybody is getting 5%). Test plan 1 on the basis of cross-testing. 5% to 31 year old is equivalent to 68% for 63 year old, so I'd say that 15% is pretty safe.

[Archimage]

If I am not mistaken, Relius does not have restructuring capabilities so you will have to play with Relius to get it to work or do it in a spreadsheet.

[Tom Poje]

just trick Relius.

restructuring = divisions.

then when you test rate groups, select division you want, testing method (accrual or allocation)

and then enter the number of HCEs, NHCEs (non benefiting non excludables) from the other division.

I would print the avg ben % test (though not used since the one division would be treated as 0) - however it does give you the count of HCEs and NHCEs just to make sure you have entered the proper numbers.

[Mike Preston]

Or in this case you can just write:

Plan 1: 2 HCE's out of 6 = 33%; 1NHCE out of 3 = 33%; pass 410(b) with 100% ratio

Plan 2: 4 HCE's out of 6 = 66%; 2NHCE out of 3 = 66%; pass 410(b) with 100% ratio

Plan 1: Highest HCE (age 63) 22 years older than NHCE, so percentage allowed to HCE is 1.085 raised to the 22nd power which is 13.6067, so HCE allocation can't exceed 5% * 13.6067 = 68.03%

Plan 2: Everybody gets 5%.

Print the above out and attach to anything the IRS asks for.

Note that in this case, the 63 year old can receive an allocation of 68% of $9600, or $6,528 before imputing permitted disparity.

[pmacduff]

Thanks all for your replies...

Mike - your suggestion is the easiest and I will use for this case but Tom I'll also try your suggestion for Relius as well, just to see if I work it out and for future reference and use.

Thanks again!

[wsp]

just trick Relius.

restructuring = divisions.

then when you test rate groups, select division you want, testing method (accrual or allocation)

and then enter the number of HCEs, NHCEs (non benefiting non excludables) from the other division.

I would print the avg ben % test (though not used since the one division would be treated as 0) - however it does give you the count of HCEs and NHCEs just to make sure you have entered the proper numbers.

Anyone know if ASC does this?

[Mike Preston]

All systems do it, in one manner or another. Worst case scenario is that you have to have a separate case for each testing group. As indicated with Relius, you can input additional body counts. With ASC you can tell it to aggregate cases. But the only system I know that handles it somewhat automatically is DATAIR. The reason they handle it automatically is that they have the ability to run all the tests against each and every division. Assuming each test fits on one page (it usually doesn't) that means you have 6 tests time two divisions or 12 pages of output. In reality it is more like 36 pages of output. But if you scour those 36 pages and find at least one that says PASS on it in the first 18 and one that says PASS on it in the second 18, all is well.

I like my little writeup better.