Crashing Computers

In the mid-1980s, officials at the Internal Revenue Service (IRS), prompted by the need to handle an ever-growing mountain of tax returns more efficiently, decided to overhaul the agency's aging computers. Give us enough money, they told Congress, and we'll not only bring in $40 billion in extra tax revenues a year but we'll also make it possible for taxpayers to file their returns and get their tax problems solved with little more than the flip of a switch.

Almost 10 years and some $3.4 billion later, an ambitious scanning machine prototype and upgraded telephone systems are in place in many IRS locations, but not much else has changed.

Of the more than 200 million tax forms that will flood the agency before this year's filing season is over, 95 per cent of them will be processed on keyboards connected to computers that in some locations chug along on 1940s-style vacuum tubes.

The billion-dollar computer modernization program clearly hasn't lived up to the high expectations the IRS had for it--a failure that Arthur A. Gross, the agency's head of computer services, described in testimony in January before the National Commission on Restructuring the Internal Revenue Service, a bipartisan congressional panel.

Two new systems that together cost more than $300 million have been scrapped in the past two years. An attempt to collaborate with the Commerce Department on a system that would let taxpayers file via the Internet disintegrated recently in the midst of interagency recriminations.

What's more, despite its statutory obligation to protect taxpayer data from the prying eyes of outsiders, the IRS lacks "the critical mass of intellectual capital" to fix its computer problems without help from private experts, Gross said.

How could this have happened? Where did all that money go? "It hasn't been wasted," deputy IRS commissioner Michael P. Dolan, the agency's top-ranking career employee insisted in an interview. Individual taxpayers now have the option of filing returns via touch-tone phone or from their tax preparer's or accountant's computer. Businesses can make their tax deposits electronically, he continued, and over the past three years, the IRS has been able to answer an increasing number of taxpayer calls because of wiring and equipment upgrades at its 29 service centers. "There are a lot of things currently on the ground and operating that are giving the taxpayer a real, live, in-the-flesh benefit," he said.

The IRS has spent at least $1 billion replacing aging mainframes, data storage systems and wiring for the telephone hotlines at its service centers, according to an agency spokesman. Some $400 million was spent to consolidate the agency's service centers, while another $400 million has gone toward developing new computer programs for IRS field agents involved in collection efforts, the spokesman said.

The IRS experience with computers mirrors many of the problems other federal agencies have had managing technological change.

As recently as 1994, for example, the Federal Aviation Administration (FAA) was trolling Radio Shack stores for spare parts to keep its air traffic computers running, the General Accounting Office (GAO), Congress's investigative arm, reported. Meanwhile, the FAA's plan to upgrade its computer system faced so many delays that the agency would "spend twice as much as originally planned for a system that will meet few of its original requirements," the GAO said. The National Weather Service's plans to improve its weather prediction system and the Defense Department's Corporation Information Management initiative have come in for similar criticism.

Some of the IRS's difficulties can be blamed on turf wars, bureaucratic snafus and congressional micromanagement, congressional aides and technology analysts in tax reform groups say. But former IRS employees and private technology experts familiar with the inner workings of the agency say the IRS's culture of secrecy and its tradition of stubborn self-reliance in managing massive technical projects have caused many of the problems.

Cutting Edge No More

By any measure, the tale of the agency's inability to make its modernization dreams come true isn't for the fiscally faint of heart.

And the IRS's failure is all the more striking because it was once regarded as one of the federal government's technology pioneers. In the mid-1950s, it became the first civilian agency to install keypunch computers. Those machines, the era's state-of-the-electronic-art, so revolutionized the processing of tax information that historians have labeled the '50s and '60s the IRS's Golden Days.

By the late 1970s, however, the mini-computer revolution was gaining steam; the advent of silicon chips and integrated circuitry quickly made the IRS's equipment seem clunky. In 1985, an attempt was made to replace the agency's aging computers with the sort of streamlined, high-speed technology that was by then in use at the Defense Department and the government's intelligence-gathering agencies. That plan flew apart at the seams.

Despite inadequate testing and IRS field administrators' warnings that more than a thousand software programs designed for the new service center replacement system needed to be rewritten, the agency's top brass put the system into use nationwide just before that year's filing season. The system worked so poorly that harassed IRS employees took to stuffing unprocessed returns--some of them still containing checks--"into wastebaskets and ceiling ducts," Shelley L. Davis, the IRS's official historian from 1988-95, reports in her recently published Unbridled Power: Inside the Secret Culture of the IRS.

In the wake of what Davis calls "the Filing Season From Hell," Congress began pumping money into two dozen projects that fell under the rubric of Tax Systems Redesign, all aimed at upgrading or replacing faltering equipment.

In 1989, the IRS announced that it had yet another plan for overhauling its computers, this one called Tax Systems Modernization. This blueprint involved the same 24 computer projects along with six similar ones.

Three years later the projected price tag for the decade-long modernization program had swollen from $4 billion to $8 billion, the GAO has reported. At the same time, the carefully contrived, centrally administered plan that Congress believed it was buying didn't really exist.

"In order to make a credible pitch to Congress, the IRS had to couch what it was doing in terms of a massive central effort to provide a unified, common system," said Thomas P. Giammo, a retired information technology consultant. As assistant commissioner for information systems at the U.S. Patent and Trademark Office from 1988-93, he overhauled a similarly aging computer system to manage the modern-day flow of paper. Giammo continued: "But [the IRS] could never organize itself internally to accomplish that because the group of people put in charge of the effort didn't then, and don't now, have enough clout to force people on the periphery of the organization to follow their lead."

Hello, Give Me Martinsburg

It may not win kudos for management, but the IRS is still capable of efficiently processing tax returns and mailing out refunds, agency officials maintain. Much of the "hardware"--the computer terminals, keyboards, hard drives, modems and such--the IRS uses daily are less than 10 years old and unlikely to fail altogether. The software programs, or mathematical operating instructions, that tell the machines what to do are much older, but they still work well enough to get through many more tax seasons, the officials said.

But these older software programs can make the IRS computer system play havoc on the taxpayer who hits a snag.

Ever wonder, for example, why when you call the IRS with a question about a tax-due notice you'll get one answer and then, when you call back for clarification, you often get an entirely different answer to the same question?

That's because the agency's customer service representatives are working with nine databases. While all nine databases, or collections of digitized data, may reside in one or more of the same huge mainframe computers at a central IRS location, few of the them can communicate with one another. Employees at 15 of the IRS's 29 customer service centers can't even call up all nine databases on the same computer screen.

Even if you're lucky enough to reach an IRS customer service rep who can call up information from more than one database at a time, you're not guaranteed an answer based on the most current information about your case. Only one database, called the master files, located in Martinsburg, W.Va., contains all the information for each individual taxpayer. No one but IRS employees at that center can update the master files. Because the Martinsburg center adds data to its records only once a week, on Saturday, new information can take as long as 10 days to show up on the rest of the IRS's screens, Gross told the commission.

Dolan, the IRS deputy commissioner, argues that the nine databases don't present a problem for the agency because the odds are slim that taxpayers will need information from all nine at the same time. "A particular call might take [a customer service representative] to one or two of the databases in order to get an answer, but not all nine," he said. "The image that has been drawn recently of our people running around the office to nine different terminals isn't correct."

Linking all the databases so that information can be passed back and forth between them and stored in more than one location, also wasn't a top priority in the IRS's modernization scheme, he said. "The first thing we worried about was did we even have the [telephone] answering capability, the telecommunications infrastructure, the whole suite of [communications tools] that would help that taxpayer with a question," he said. "One of the things we had not been able to do prior to a year ago was respond well [by phone] to a taxpayer who was remote from one of our service centers.

Investments in telecommunications upgrades were investments we thought we needed most. If a taxpayer can't get to you, then you're not serving them." The IRS applies a similar reasoning to the millions that it has spent in recent years on computer systems that don't work as planned, or that don't work at all.

In August 1995, the IRS paid the Commerce Department's National Technical Information Service (NTIS) $10 million to build Cyberfile, a computer system that would let taxpayers file their returns via the Internet. IRS officials believed NTIS could deliver the system in time for the 1996 tax season.

Despite warnings from the IRS acquisitions staff that the agency had failed to perform a thorough enough procurement analysis of NTIS, the IRS top management wrote NTIS an additional $7.1 million check for the project in December 1995, according to the GAO. Meanwhile, major U.S. software companies such as Intuit Inc., the Menlo Park (Calif.)-based maker of Turbo Tax, built a dormant connection to the IRS Internet site into their tax programs; with special instructions from the IRS, the connection could be activated in minutes. Five months later the IRS ditched the Cyberfile project.

Cyberfile failed because "the project was hastily initiated. Development and acquisition were undisciplined, and Cyberfile was poorly managed and overseen," the GAO reported. "Only after NTIS informed IRS in April 1996 that the $17.1 million had been obligated and that the system still was not finished, did IRS stop to reconsider the project."

The Document Processing System (DPS), a massive project that would allow IRS employees to scan tax returns into their computers rather than entering the data by hand, went through four contractors before the agency canceled it last year. Total cost: around $298 million, including $13 million-$14 million paid to the last contractor, Bethesda (Md.)-based Lockheed Martin Corp., to shut the system down, an IRS spokeswoman said.

The IRS decided to cancel the DPS project because it was going to require an investment of more than $1 billion more before it became truly usable, Dolan said. "Some people will say, `Well, that's an investment you walked away from.' But we learned a lot from DPS," he said. The IRS is using some of the techniques developed during the DPS project to speed up the initial processing of tax forms, he said. "For us to have spent $1 billion on something that would have consumed so much of our discretionary capital wouldn't have been nearly as prudent as calling a halt to it early," he added.

Another huge scanner project, the Service Center Recognition/Image Processing System (SCRIPS), is still being used at the IRS's processing centers in Austin, Texas; Cincinnati; Kansas City, Mo.; Memphis; and Ogden, Utah. Five years ago, the IRS predicted that the system would save $17 million in labor costs and increase employee productivity 20 per cent. Two years ago, the agency lowered its savings estimate to nearer $5 million. Last year, the IRS predicted that SCRIPS would actually cost more than it saved. In January, the GAO pegged the eventual cost of the system, which it said has "critical weaknesses," at $288 million over its 10-year life-cycle.

Another IRS Shopping List

The primary value of scanner systems such as DPS and SCRIPS lies in their ability to turn notations on paper forms into digital images, thus eliminating the drudgery and cost of typing tax return data into computer databases by hand. Some computer experts maintain that the technology doesn't exist that can outdo the human eye. "The truth is that while computers are very good at doing certain menial tasks, they are still not able to perform many tasks that humans find simple, like reading numbers on a paper tax form," Jaron Lanier, a visiting scholar at Columbia University's computer science department, wrote recently in The New York Times.

Those sentiments aren't universal, however. Many insurance, financial services and credit card companies have replaced banks of keypunch operators with one or both of two types of imaging technology. The first type of technology, which takes a picture of a document and reproduces it in exact detail on the computer screen, is practically foolproof, computer experts say.

The second type, called optical character recognition (OCR), requires the computer to scan the document and translate what it sees into the same digitized characters that a human would type in on a keyboard.

Optical character recognition isn't 100 per cent accurate. "In some cases, you need a great deal of manual intervention in order to validate that what the computer interpreted is what was intended" by the author of the document, said Gerard H. Barloco, vice president for information services at USAA, a worldwide insurance company based in San Antonio. Still, he said, "it's true that the optical character recognition technology is in many cases very good."

Other computer experts are blunter. "The hard part every company has to deal with is having to scan and index documents they don't design. The IRS designs the damn documents, and they should be able to scan them," said the chief of computer services for one of the largest investment houses in the world, who asked not to be identified.

The IRS had hoped to find a way to combine both the picture taking and optical character recognition capabilities of advanced computer systems in its aborted document processing system, Dolan said. Eliminating both paper handling and keyboarding costs "is really the solution we've been after," he added.

The most important goal, however, is to cull tax return data reliably, accurately and as cheaply as possible. "Today that means using seasonal workers who can enter the data manually," because IRS officials have become convinced, after considerable consultation with government experts, that optical character recognition technology hasn't advanced far enough to read handwritten information with the degree of reliability the agency demands, Dolan added. "The notion that everybody files a clean, two-sided form that can glibly be read by OCR is not the experience we've had," he said.

Ensuring that tax forms are taken out of envelopes and sorted correctly is also a major IRS concern, Dolan said. "You realize the kinds of overhead that are required to take tax documents from the mail and make them computer-ready. . . when you look at the fallout we've experienced when a return gets lost," he added, referring to the foul-ups that occurred in 1985.

The $64,000 question is why, if reliable imaging technology is widely used in industry, the IRS didn't call in a top consulting firm from the private sector to design its scanner systems. It's a question that has been asked about many facets of the tax systems modernization project.

Dolan said the IRS wanted to call in outside consultants to help modernize its computers as far back as the early 1980s, but they were overruled by Treasury Department and Office of Management and Budget officials in the Reagan Administration. Some Members of Congress aren't buying that excuse, however.

"Traditionally the IRS has had an insular culture," said Rep. Rob Portman, R-Ohio, who heads the National Commission on Restructuring the IRS with Sen. Robert Kerrey, D-Neb. "The second problem is the agency's inability to properly define its mission or its strategic direction. . . . Having nine different computer databases, for example, is clearly unacceptable. It's not only terribly inconvenient for the taxpayer, but it's also a waste of taxpayer money."

In early February, after the GAO ranked the IRS near the top of its list of federal programs at "high risk" for waste, fraud and abuse, Rep. Steve Horn, R-Calif., went even further, calling the IRS "a basket-case agency."

"The entire agency seems to have this mind-set that makes them hunker down, think of excuses and resist change, rather than trying to improve," Horn, chairman of the House Government Reform and Oversight Subcommittee on Government Management, Information and Technology, said twice in separate hearings.

The IRS is sworn to protect the confidentiality of tax data, but too often the agency clings to its code of secrecy to sweep less-than-useful projects under the rug, historian Davis argues. "The IRS mind-set is `Either we can, or we must, do everything inside the agency,' " she said in an interview. "The reasoning is `If we hire competent outsiders from the technology world, heavens forbid but they might learn how the IRS processes tax returns, and they might learn how we select which returns for audit. And if people knew how we choose returns for audit, they'd all start to scam us.' "

In the IRS's defense, many technology experts say it faces a tough challenge in trying to create a computer system for the 1990s and beyond that can draw on information stored by technology designed almost 40 years ago.

"Building a brand-new computer system in an environment where the old system has to keep functioning smoothly until the new system works, in an era where the technology changes from month to month, has been a major problem for lots of organizations, especially in the federal government," said Bob Gellman, who was formerly chief counsel to the House Government Operations Subcommittee on Information, Justice, Transportation and Agriculture, and is now a private consultant.

"We can't close the place down for a year while we boot up a new system," Dolan said. "Someone likened our problem to flying a 747 and trying to remake it in the air, and I think that's just about the truth."

The National Commission on Restructuring the IRS, now holding another round of hearings on the IRS's modernization efforts, expects to issue a report on its findings by June.

That's just about the time that the IRS hopes to present to Congress a pared-down list of priorities in the technology arena, Dolan said. The agency has been working for a year to redesign its computer modernization plans, and "I feel very confident that in the late spring or early summer designs of a more incremental, more focused, more disciplined approach will take shape," he said.

So will his agency ask Congress for even more money for yet another attempt to devise and install 21st-century technology? "Oh," Dolan said with a sigh, "it's way too early for me to predict that."