Early Days of the Treasury Tax Model on Capitol Hill
By Pete Davis
In early March 1974, I walked out of the Treasury Department with a magnetic tape on my arm containing the Treasury Individual Income Tax Model and headed a few blocks over to the basement of the Commerce Department, behind the aquarium, where a Univac 1108 sat ready to run Congress’ first computer model.
Watergate caused this. In late 1973, IRS Commissioner Don Alexander came under pressure from White House Chief of Staff Bob Haldeman to conduct special audits of nearly 400 taxpayers on the so-called “White House Enemies List.” Don knew his refusal might not stop Haldeman, so he called Larry Woodworth, Chief of Staff of the Joint Committee on Internal Revenue Taxation. Larry called House Ways and Means Chair Wilbur Mills (D-AR), who called President Nixon to dissuade him from auditing his enemies. Larry also became concerned that he was completely dependent upon Treasury for estimates from the Individual Income Tax Model. He wanted his own people to run that model, and that’s why the Joint Committee on Taxation hired me.
When I joined the staff, Larry and the top attorneys on the staff were deeply engaged in an audit of President Nixon. To everyone’s dismay, Nixon had inflated the value of many deductions, and his tax attorney had backdated the date of his contribution of congressional papers to Whittier College by a few weeks to cover up his negligence in missing the effective date of the 1969 Act, which limited the deduction for contributions of congressional papers. We knew the President was in serious trouble, and nothing leaked. With some trepidation, we expected our official report to Congress would be a bestseller that summer, but it was overshadowed by the Supreme Court ruling on the White House tapes and by the Watergate hearings leading to impeachment. I got the tax model up and running as John Doar and Hillary Clinton worked under armed guard a few floors below me in the Old Congressional Hotel, at the corner of New Jersey Ave and C St. S.E., now a parking lot.
Back then, the Ways and Means Committee was in the habit of waiting several days or a week to get the results from a Treasury computer model run. Joint Tax had its own estimators, all four of them, but they literally operated with eyeshades and huge electric calculators that would take a minute or so of chunking to do one long division. I never ceased to be amazed at how accurately they could estimate proposed income tax law changes by hand from Statistics of Income data from those big volumes IRS would publish each year. The problem came when several proposals were combined, the interactions could be tricky, and that’s where the model excelled.
The model had its own problems however. Treasury’s extrapolation of the latest IRS SOI data sample two or three years out to current levels produced spurious results in many data items, such as the child care credit, that weren’t targeted for lack of sufficient degrees of freedom. We didn’t have names, addresses, or Social Security numbers, but it became painfully obvious if one tremendously wealthy taxpayer had paid himself a dividend or if another used her enormous municipal bond interest income to make large charitable contributions. I quickly learned to add a special routine to my model runs to print outlier returns. I also found that seemingly simple model runs failed because of quirks in the software or in the data. Just because one of the hundred and ten data items had a title, like “investment interest expense,” didn’t mean that it contained useful data, and just because the software contained a term, TOTEXEMP, didn’t mean it contained the total number of exemptions. I had always been a fan of quick ballpark estimates and idiot checks to bound the results of physics labs, and those skills saved me from many an egregious mistake. More importantly, Treasury’s Roy Wyscarver never failed to patiently take my calls, even a few at 4 a.m., and to help me figure out why my model run wasn’t producing valid output.
The Tax Reduction Act of 1975 (back in the days when we had simple titles for bills) stood out in the history of the model because it proved to the Ways and Means and Finance Committees that the “computer” could inform their decision making on a real-time basis. We had gone home for the holidays in 1974 as President Ford’s “Whip Inflation Now” program and Fed rate hikes attempted to cool an overheated economy, but we returned in January to find that the economy had fallen off a cliff. A stimulative tax cut had to be enacted as soon as possible. Within weeks, the Administration and Congress had agreed upon the outlines of an immediate tax rebate and exemption credits for individuals and an investment tax credit increase for business. The markup of the rebate started with the Ways and Means Committee asking for a few computer runs rebating 5% or 10% of 1974 income tax up to a maximum of $500 or $1,000. Those were rejected because they either cost too much or failed to confer enough money on those in the lower income brackets. That night, I had the bright idea to compute a matrix of revenue estimates with percentages of rebate in the columns and maximum amounts down the rows. The Committee was so intrigued with the table in the next morning’s markup that members started engaging in the estimating process by speculating about applying an income phaseout to maximize the benefit to lower income taxpayers. As we spiraled in toward a solution, it became obvious that some more runs would be required to establish the phaseout range. I ran back to the Hotel, ran them over lunch, and ran back to 1100 Longworth with an armload of printouts at 2 p.m. Mr. Ullman and the rest of the members were shocked and pleased to find their questions answered over lunch, and we quickly settled upon the rebate that was enacted in early May. I believe the 1975 Act still holds the record for the fastest tax cut from inception to enactment.
The only problem with the ensuing popularity of the tax model was that members came to expect all of their estimating inquiries to be answered over lunch, even when they couldn’t be estimated by the model. I remember one evening when Commerce’s 1108 crashed, and I couldn’t produce any of the runs I had promised Larry for that day’s markup. I whispered in Larry’s ear at the markup table on-camera in front of the full Ways and Means Committee that I couldn’t produce any runs until the computer came back up. Larry thought for a second and turned to the Committee and stated, “The computer is sick!” All day, staff came up to me as said, “You don’t look sick,” and I’d try to explain that the computer was having a bad day, not me. The New York Times actually wrote a 6-inch sidebar about the sick computer.
The Earned Income Tax Credit, also enacted in the 1975 Act, posed a challenge. The Model could estimate the nonrefundable portion of it with no problem, but what about estimating the refundable portion of the credit. The Statistics of Income sample included some returns, which after extrapolation and law changes, would reside on the Model as nontaxable, but they certainly didn’t include the nonfilers with earned income that might qualify for the credit. How many people would “come out of the woodwork” to claim the refundable credit? Fortunately, HHS had some Trim Model runs that gave us some confidence in imputing those nontaxable returns to the Model. A hand estimate based upon other HHS data was done as a backup. When those came out nearly the same, we went ahead with our $8.125 billion one-year EITC estimate fearing that we could be off by billions. It’s not all that often that you can directly determine the accuracy of a revenue estimate after the fact, but credits are an exception. Two years later, we were pleased to find the Statistics of Income found our estimate to be quite accurate.
As the years went by in the late 1970s, a very good working relationship developed between the tax staffs and the model. Lawyers relished the opportunity to simulate tax shelters for the Tax Reform Act of 1976. Economists started attending legislative drafting sessions. Members of Congress became more sensitive to the economic effects of their proposals, and a sheaf full of model estimates often fostered compromise and consensus and exploration of policy possibilities.
Pete Davis is President of Davis Capital Investment Ideas. He has served as an economist with the Joint Committee on Taxation (1974-1981), the Senate Budget Committee (1981-1983), and Senator Robert C. Byrd (1992).