they almost always underestimate the direct cost of government programs

And that's not to mention the routinely-unanticipated indirect costs!

Here's a nice piece from Veronique de Rugy in Reason...

Congress says that the health care package it passed at the end of 2009 will cost roughly $900 billion over 10 years—and will somehow end up saving taxpayers money in the long run. If you think that sounds unlikely, you’re right.

With the federal government, massive cost overruns are the rule, not the exception. The $700 billion cost of the war in Iraq dwarfs the $50 billion to $60 billion that Mitch Daniels, then director of the Office of Management and Budget, predicted at the outset. In 1967 long-run forecasts estimated that Medicare would cost about $12 billion by 1990. In reality, it cost more than $98 billion that year. Today it costs $500 billion.

Nor is the problem limited to Washington. In 2002 the Journal of the American Planning Association published one of the most comprehensive studies of cost overruns, looking over the last 70 years at 258 government projects around the world with a combined value of $90 billion. The Danish economists Bent Flyvbjerg, Mette Skamris Holm, and Soren Buhl found that nine out of 10 public works projects had exceeded their initially estimated costs....Budget busting occurred throughout the seven decades studied, with the totals spent routinely ranging from 50 to 100 percent more than the original estimate.

How did the United States do? According to the Danish researchers, American cost overruns reached an average of $55 billion per year....The military, too, has a long history of cost overruns....

Strangely, lawmakers never seem to anticipate these extra costs even when the excesses take place under their noses. The Capitol Hill Visitor Center, an ambitious three-floor underground facility originally scheduled to open at the end of 2005, was delayed until 2008. The price tag leaped from an estimate of $265 million in 2000 to a final cost of $621 million....