The IRS Computer System Is the Oldest in the Government
(Bloomberg View) -- Taxpayers who rushed to complete Form 1040 by Tuesday’s deadline can take some comfort from the fact that they’re exceedingly unlikely to get a follow-up visit from the Internal Revenue Service. Over the past 50 years, audit rates have fallen pretty steadily. Today, the average taxpayer has one chance in 200 of getting audited.
The downward trend has led to concerns that the IRS needs more funding to do its job correctly and competently. That’s true, but before Congress throws money at hiring more auditors, it should take aim at the agency’s antiquated computer system. It has helped drive audit rates to all-time lows, but it desperately needs an upgrade.
After the creation of the first permanent income tax in 1913, IRS agents would scrutinize returns by hand, laboriously poring over the numbers, checking the math, and flagging returns that looked suspect. This was insanely time-consuming. It also led to a very high audit rate. One taxpayer in 10 was subjected to a “field examination” from IRS personnel, according to one estimate from 1926.
This was possible because so few taxpayers actually filed Form 1040; most didn’t because their income fell well below the threshold. As government spending increased from the late 1930s onward, more and more Americans found themselves paying income tax. That trend only intensified in the postwar years.
And therein lay a problem: How could IRS agents possibly check so many tax returns, much less run audits on this scale? As audit rates plummeted in the late 1940s, concerns over lost revenue fueled a search for solutions. Conveniently, a means of fixing the problem appeared at that precise moment in history: the computer.
The first mainframe computers may look like dinosaurs now, but they offered a way of reviewing tax returns on a mass scale. In the late 1950s, the IRS began using computers to correlate and compare the information submitted by taxpayers on their 1040 forms with the income figures supplied by employers. In 1959, the Washington Post, capturing the mood of the moment, warned of “brain machines” that would soon audit taxpayer returns; the Wall Street Journal called them “robot revenuers.”
By the early 1960s, the IRS had amassed a staggering amount of computing power in the service of compiling, collating, and auditing returns. The stars of this brave new world consisted of a number of IBM mainframe machines, many of which resided in a nondescript brick building in Martinsburg, West Virginia. This was the heart -- or brain -- of the new order. And it triggered serious warnings that the jig was up.
“The Martinsburg Monster is going to get us all,” Charles Seib wrote in Harper’s Weekly. “The imaginative taxpayer, who in the past has had at least as good a chance as a devotee of Russian roulette, now must fact the certainty that all the chambers are loaded and the trigger set.”
Under the new system, the data contained in paper tax returns -- along with all the other forms issued by employers, brokerages, and banks -- would be coded onto paper punch cards by an army of clerical workers. The data would then be “read” by the computer and stored on thousands of magnetic tapes.
The Martinsburg facility’s L-shaped array of mainframe computers and tape-reel machines became something of a destination for those looking for a glimpse of the future. “So many visitors come to gawk,” explained the New York Times in 1964, “that a glass inclosure was built for them to sit in.”
Throughout the 1960s, the IRS labored to build what it called the “Individual Master File” of all taxpayers, each identified by his or her Social Security number. This database would contain all the information relevant to each taxpayer. “By 1966,” predicted Seib in Harper’s, “every tax return in the nation will be under the Monster’s cold, electronic eye.”
This proved both a blessing and a curse for the IRS. Between 1963 and 1967, the Times reported, the number of taxpayers who reported any interest income rose 45 percent; the total amount of interest and dividend income reported to the IRS rose by $2.8 billion in the same period. “The hot eyeball of the computer was the goad to virtue,” reported the paper.
At the same time, the sheer number of possible leads on underpayment of taxes threatened to overwhelm the very human staff responsible for overseeing audits. The IRS solved the problem in two ways.
First, it began using the computers to issue automated letters demanding that taxpayers correct errors. In effect, it relied on form letters to achieve what field audits had formerly accomplished. At the same time, it began using historical data to build algorithms designed to sniff out suspect returns. Such returns might not contain any errors, but because of certain red flags -- an unusual deduction, a deviation from some norm -- the computer could flag the return. The higher the “Dif,” or “discriminate function,” score, the more likely something fishy was taking place. Audit rates drifted downward, a trend that continues to this day.
Though the IRS has periodically upgraded its computing system, today’s system is still running the same code, which was written nearly 60 years ago. Most of it is in the Assembly programming language, which the IRS itself has described as “antiquated” and “inflexible.” Worse, the number of programmers who can understand and maintain the code behind the Individual Master File (IMF) dwindles with every passing year. According to the Government Accountability Office, the IMF and its business counterpart (the BMF) are the oldest computing systems used by the federal government. (The runner-up in this dubious contest is the software used to coordinate the nation’s nuclear weapons.)
Plans to replace the IMF with a twenty-first-century equivalent known as CADE (Customer Account Data Engine) have faltered. The transition is now well behind schedule. As a consequence, the likelihood of a catastrophic computer failure during tax season increases with every passing year. That may not pose quite the same danger as an errant missile, but the prospect of lost refund checks, unnecessary audits, and other errors suggests that the time has come to bring the IRS into the 21st century.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
Stephen Mihm, an associate professor of history at the University of Georgia, is a contributor to Bloomberg View.
For more columns from Bloomberg View, visit http://www.bloomberg.com/view.
©2018 Bloomberg L.P.