| Experts debate Y2K cost | Year 2000 resources |

How much will it cost to fix the year 2000 problem? Probably less than we've been hearing, wrote Paul A. Strassmann in his Computerworld column on May 18. Strassmann used first-quarter 1998 filings with the Securities and Exchange Commission to estimate that companies are budgeting only $50 billion to fix year 2000 troubles - a surprisingly low 8.3% of their annual information technology budgets. That column triggered an E-mail debate among other millennium experts, Leon A. Kappelman, Capers Jones and fellow columnist Ed Yourdon, who say the cost of fixing the problem will be much higher than Strassmann's projections. What follows is a virtual roundtable discussion on the subject:

And these project leaders tell us that their enterprises are spending about 38% of one year's IT operating budget dealing with their Y2K problems - and that that number only marginally includes embedded systems and desktop/PC problems.


Strassmann: Before you place your trust in project managers, you need to be convinced that they have no conflict of interest in reporting higher year 2000 costs.


Kappelman: Just like management may have a bias for reporting a lower one. The truth is somewhere in between, perhaps? Seems that both samples pose validity challenges.


Strassmann: Arguing that "the truth is somewhere in between" does not recognize that the biases do not have equal validity. When I weigh the biases of project managers reporting to SIM and management reporting to the SEC, all I can say is that management may be held accountable for what they say. It becomes a matter of record and subject to litigation and liability suits. Project managers filling out SIM questionnaires cannot and would not ever be held accountable for what they say. Furthermore, the project managers claims remain anonymous, while SEC filings are on the public record and signed by a CFO.


Kappelman: Agreed. Permit me to clarify: I didn't imply they were equal, else I would have said "in the middle" rather than "in between." Still, the decision of what one enterprise includes as a Y2K budget may be very different from the choice of another.

In light of no standardized definitions for things such as "Y2K compliance" or "Y2K expense" and what I think is management's bias toward underreporting this figure, I suspect the SEC figures are on the low end of the scale.


Capers Jones (chairman, Software Productivity Research): I've been studying estimating accuracy for more than 20 years. For large software project estimates, these are the approximate results from among our clients: Out of every 100 estimates, about 70 will be extremely optimistic, 25 will be within plus or minus 10%, and 5 will be conservative by more than 10%.

Since optimism and underestimates are the norm for software, and since year 2000 estimates are more complicated than normal software estimates, I suspect the SEC reports may be optimistic. What often happens is that the project manager initially creates an accurate or conservative estimate. When it is presented to senior management or clients, the conservative estimate is rejected and arbitrarily replaced by a more aggressive estimate and schedule.

Ultimately we have to get down to numbers. Is it $300 billion-plus as alleged by Gartner Group, or is it closer to $50 billion for U.S. corporate accounts?

My estimate for overall U.S. costs prior to the Y2K event is for $177 billion since I also included hardware upgrades, database repairs and other costs outside of software. I estimate that another $497 billion was for post-2000 litigation, damages and recovery. These are not in the SEC data, I suspect.


Ed Yourdon (chairman, Cutter Consortium): Strassmann's information is interesting, but among other things it assumes that companies are reporting accurate data to the SEC. I've heard from a fairly reliable source that at least one very large New York bank is spending five times as much on Y2K as it's reporting, in order to disguise the magnitude of its problem and to avoid scaring people. Also, the language in the SEC disclosures was often so ambiguous that it was difficult, if not impossible, to tell what companies were really doing.


Strassmann: If we had lots of cheaters, the regression analysis would have shown it. To my surprise, this was not so - the relationship between the numbers reported to the SEC and my approximation of the IT budget was a relatively tight fit. To get such a result would require 85 SEC filers to uniformly share an identical underreporting bias. The probability of that happening is hard to conceive.

As I noted in my Computerworld column, while the SEC numbers have greater credibility than anything published so far, I do not feel we should depend entirely on the SEC data until we see a larger sample from companies that have passed through the testing phases of Y2K remediation.

As noted by Machiavelli: Always assume incompetence before looking for conspiracy.