Our policymakers need to get a firm handle on the financial risks taxpayers will be taking with the city's rail-transit proposal.

They need to assure themselves that the city's transit projections will be met within a reasonable range of error. This is particularly important for Honolulu since, on a per-capita basis, the $4.6 billion projected cost would become the most expensive rail line ever built in the U.S, even allowing for inflation and without cost overruns.

To make a sensible assessment of the financial risks of the project, policymakers need to review the experiences of other metro areas that have built rail lines.

Until recently the only official U.S. Department of Transportation comparisons of capital cost projections versus actual cost, was the 1990 Pickrell Report.

This report showed that errors for the eight projects studied averaged 44 percent. More importantly, it revealed a wide error range from the best, at 11 percent under projection, to the worst, at 83 percent over. Only one of these, the original Pittsburgh light-rail line, came in under its projected cost.

Last month, the U.S. transportation department released a new report covering projects built between 1990 and 2003 that showed that of the 21 projects covered, the best performer, San Jose, was 28 percent under forecast and the worst, Portland, was 72 percent over, for an average of 21 percent over forecast.

More important than averages is the distribution of the various error rates. For example, if the resulting costs of the 21 projects were between plus- or minus-10 percent of the forecasts it would be a reasonable indication to our policy makers of the likely accuracy of the Honolulu projections.

However, that is not the case here. These 21 projects' costs relative to forecasts errors were evenly distributed over a wide range of 72 percent over to 28 percent under forecasts. If we were to apply that range of error to the $4.6 billion projection, it would result in a spread of $3.3 billion to $7.9 billion.

The city administration will undoubtedly paint this as ridiculously improbable and wildly pessimistic.

However, each of these 21 capital cost projections was thought at the time to be reasonable by both the transit agency and the consultant who produced them. Just as our city Department of Transportation Services and its consultants also believe their current cost projections are reasonable.

In addition, the Federal Transit Administration's in-house analysts and outside consultants also examined each of these 21 capital cost projections in great detail and thought them all reasonable.

Here we have innumerable transit planners, engineers and accountants, all well educated and experienced and all believing that, as a result of their hard work, the cost projections would be, dare we say it, reasonable.

And the forecast errors have not improved. Errors of the recent rail lines built include: