Stanford economics professor John Taylor posted this graph, which shows the simultaneous decrease in state- and local-level borrowing and increase in federal stimulus spending:
Much of the stimulus money went to cash-strapped states who were able to temporarily lower the amount of money they needed to borrow in order to finance their over-spending. The most common argument against stimulus spending is that rational economic actors will decrease their present consumption by the amount of additional government spending in anticipation of inevitable future tax increases (to tax is to spend–deficits are future taxes!). This idea is known as Ricardian equivalence, named after 18th century economist David Ricardo. It is criticized on a number of grounds, including that it does not factor in population growth or international trade considerations, and that real people do not actually behave as purely rational economic actors. These counter-arguments are all valid and apply generally to Ricardo’s theory, which should not be interpreted as an apodictic law of the universe. However, as the graph demonstrates, stimulus spending was almost completely offset by reductions in state spending, not individual spending. Interestingly, I’m pretty sure that Paul Krugman used a similar graph that showed total government (state and local) expenditures to illustrate that stimulus had not been tried. Still, many Keynesians are going around saying that the stimulus worked, it just wasn’t big enough. This argument is only valid if additional stimulus would not have also been offset by decreases in state and local government borrowing.
This graph should not be interpreted as an indictment of Keynes’s theory of the money multiplier, but rather of our political system, which seems incapable of producing the kind of “targeted, timely and temporary” increases in government spending that are required in order for stimulus spending to be effective.
A Keynesian might respond that the stimulus prevented thousands of state and local government workers from losing their jobs, which is probably true. However, Keynesian theory does not predict a multiplier when government spending remains constant.
David Henderson adds comment here.