On July 15, the IRS will launch a 45-state data-sharing initiative, but according to the GAO, there's almost no evaluation data indicating whether or not it works correctly. There's not even data on how to evaluate the data.

Data-matching's not a new concept to the IRS, but the State Reverse File Match Initiative (SRFMI) moves the data in the other direction -- that is, from the states to the federal agency. The goal of the program is, as one might expect, to figure out who's not paying her share and to get that money, thus helping to close the estimated $345 billion gross federal tax gap.

With SRFMI, the feds compare state individual, corporate, sales, and withholding tax files against federal files, looking for those who either under-reported or didn't file at all. The system's entering the third phase of its (three-part) pilot program, and 13 states are currently involved. It began in 2006.

In 2007, Senators Max Baucus (D - Mont.) and Charles Grassley (R - Iowa) -- wondering if the government was really doing it all it could to fill in that missing $345,000,000,000 -- requested that the General Accounting Office look into whether SRFMI was truly up to snuff. The GAO took up the challenge and conducted a performance audit between July and September of this year. The report was filed in early November and released to the public 30 days later.

The IRS, says GAO's audit team, is working on a plan to evaluate SRFMI data. If the GAO kept a handy pack of those little gold stars for effort like your teachers had in grade school, this is where the writers of the report would have moved their hands toward that pack... then changed their minds and kept typing, nervously eyeing what amounts to a steaming pile of raw data sitting in the middle of the floor.

According to the report (PDF available here), the IRS hasn't worked through its methodology to evaluate the the pilot program, nor does it have criteria or standards for determining the pilot program's performance. It didn't manage to document the evaluation approach itself, and the testing-sample plans didn't outline the frequency, methods or timing of data collection. And agency efforts to figure out a sampling approach are still unfinished.

That finding seems to particularly annoy the GAO auditors, who remarked that "obtaining and using SRFMI data imposes costs not only on IRS but also on the states."

What to do, what to do? Do an evaluation plan, of course, and the GAO report lays out five key features of a good one. The report records that when presented with the GAO's findings, Deputy Commissioner for Services and Enforcements Linda E. Stiff responded positively, agreeing that documenting and assessing SRFMI before further rollout was a good idea, and said that "a comprehensive evaluation plan is being developed that will incorporate the key features you recommend."