Where's the monopsony?

President Obama, Paul Krugman and Robert Reich have all been pushing for an increase in the minimum wage.  I want to agree with them, and Krugman is certainly correct that the preponderance of empirical evidence shows that the minimum wage's impact on total employment is negligible.

But the question is, why?  Krugman's statement that human beings are not Manhattan apartments is true, and allows him to support the minimum wage while being appropriately skeptical of rent control, but it doesn't give a satisfactory answer as to why putting a floor on the price of labor would not create excess supply of labor.

There is in economic theory a set of circumstances, however, under which an increase in the minimum wage might raise employment.  If an employer has a market largely to itself--if it has monopsony power--then it will both pay its workers less than their productivity warrants and not hire enough workers to be at the most efficient level of employment.  Raising the minimum wage would then both increase pay and induce more workers into the labor market, hence increasing employment.  If government could nail the minimum wage to the marginal revenue product of the least productive  workers, the minimum wage could produce a first-best outcome--one where pay and employment levels were efficient.

For the argument to work, the demand for labor needn't be perfectly monopsonistic, but rather less than perfectly competitive.  The fact that wages and labor productivity seem to have less and less to do with each other is evidence that the demand for labor is not competitive, but it would be nice to have further, detailed evidence of the industrial organization of labor demand.