This is easily the most dishonest thing I have read all day. However, Donald Luskin does raise one valid point: It's not accurate to talk about the transition costs for privatizing social security as a "cost" per se. It's true the government would have to spend anywhere from $7 trillion to $8 trillion over a number of years in order to meet its current Social Security obligations while also allowing workers to divert some of their payroll taxes into private accounts. In the long run, though, assuming that privatization saves money, the government will recoup its losses. From an accounting perspective, there's no difference whether you spend those trillions now or later.

Of course, that's from an accounting perspective. In the real world, piling up massive short- and medium-term deficits could have very real effects on the economy. What those effects are -- higher interest rates, a declining dollar -- I don't know. But I don't think now is really the best time to find out, with deficits already mounting year by year.

So why not instead simply index Social Security payouts to inflation? Currently, initial benefits are calculated based on a worker's average salary, which is then indexed to the average growth in wages over time. But payroll taxes are also indexed to the growth in wages, so it's hard for tax receipts ever to overtake benefit payouts. But if we just went back to the old 1972 rules, when benefits were indexed to the rise in the cost of living, then tax receipts would exceed payouts (since under presidents not named Bush, wages tend to outpace inflation). Voila: fixed system. Note that the only reason the 1972 rules were ever revised was because stagflation in the late 1970s was forcing the system to write ludicrously high benefit checks; but now that the Fed keeps inflation under control, this shouldn't be an issue.