It's a pity that the great divide-by-ten controversy has diverted everyone's attention from what ought to be the main point: that the comparison Krugman drew between tax losses to the government and the salaries of the created jobs is meaningless whether he's doing the math correctly or not.
One one side of the comparison, it's bogus to treat the pay earned by the holder of a newly created job as a pure benefit-- that is, unless someone's discovered that there's no such thing as disutility of labor after all, and I'm late getting the word. No doubt there are second-order effects of job creation as well, but wage data obviously tells us nothing at all about either their magnitude or their sign.
On the other side, if Krugman wants to treat a mere transfer (or rather, a foregone transfer) as a cost, he's also got to include the offsetting benefit to taxpayers. That is, unless he's assuming that it's no benefit at all because taxpayers oughtn't to have the money-- in which case he's engaged in something very different from the exercise in uncontroversial positive economics that he wants us to believe he's engaged in.
I think that a good rule of thumb is that any time someone speaks of the "cost" of a tax cut they are unlikely to be making an economic argument.
Krugman might wish to say that he wants to discuss fiscal policy in terms of "bang for the buck." That is, he could argue that if the goal is to increase employment by 100,000 in the year 2004, a government jobs program will result in a smaller increase in government debt than a ten-year tax cut.