I don't know how he knows that, but if he knows that than he's a lot smarter than me (and pretty much every other economist out there). I had a thought early on in this crisis that isn't entirely uplifting but true nonetheless. I thought "well this is going to give me something to think and write about for the rest of my career". Apparently, that's off the table now, because as Steve Horwitz succinctly puts it:
"Let's look at the record of the last three years:
TARP = Failed.
QE1 = Failed.
QE2 = Failed.
Stimulus = Failed.
Non-existent budget cuts/debt ceiling increase = Failed, at least in S&P's eyes.
Each of these has involved more government activism and each has failed."
I think it's worth reminding people why it's so hard to know these things, despite Steve's own confidence and his disappointing assertion that those of us who are more circumspect about his claims are "religious fundamentalists", hypocrites, practitioners of "state idolatry", dogmatic and insane (in the Einsteinian, rather than the clinical sense, of course).
Fiscal and monetary policy are endogenous phenomena in human society, which basically means that causality runs both ways with these phenomena. We all think that fiscal and monetary policy affects the macroeconomy, but we also think the macroeconomic conditions and expectations about macroeconomic conditions affect fiscal and monetary policy. We know with a reasonable degree of certainty that the causal relationship running from the macroeconomy to policy is negative (as GDP goes down, policy becomes more expansionary, either through standard policy rules or automatic processes like decreasing tax revenue and automatically increasing social insurance outlays). The other causal relationship is the one where we disagree, and that we'd like to identify.
First, let's consider the case where fiscal and monetary policy has no net impact on the macroeconomy at all. If we simply regress GDP on some policy variable, what sort of outcome are we going to get in this case? We're going to observe a negative relationship between the two variables despite the fact that we know (by assumption) that policy has no effect. In other words, the endogeneity of fiscal and monetary policy implies that uncorrected empirical estimates of the impact will be biased downward.
You can observe this problem in the following graphic. Let's assume for a minute that stimulus actually has a positive impact on GDP, as the large majority of economists have come to agree on in the case of monetary policy and many have come to agree on in the case of fiscal policy. I have two thick black lines - the bottom one shows what the economy would have done in the absence of stimulus, the top one shows what the economy does with stimulus. We never observe either of these lines. What we observe is staggered point estimates of what actually happens:
So the only data we actually have is the red line. If you casually comment on the red line to understand what stimulus does, of course it looks like it doesn't work at all. This is what an unfortunate number of stimulus critics do. What you need to have for a valid empirical assessment of the stimulus is the two thick black lines. Since we can't observe GDP perfectly and instanteously, we could settle for the thin blue line and the thin red line - occasional observations of actual and counter-factual GDP. Unfortunately, we don't even have the thin blue line. We only have the thin red line.
If policy were not endogeneous this would not be that much of a problem. You'd still need to observe a lot of cases (you'd need a large sample), but you could pretty much compare situations with policy to situations without policy. The reason is that when these processes are not endogenous, the expected value of the counterfactual during periods of fiscal and monetary policy is the same as the observed data without periods of fiscal and monetary policy. The problem is, that's not the case here.
The solution (sort of)
The solution to getting an unbiased estimate of the impact of stimulus is to identify variation in policy that is exogenous to what's going on in the macroeconomy and then look at the association between that portion of the variation in policy and the behavior of the macroeconomy. (This is essentially the instrumental variable method that I talked about in this recent post - although macroeconomists are less likely to talk in terms of "instruments"). This is very hard to do. There are two very well known attempts at this, one by Christina and David Romer (on tax policy), and one by Robert Barro and Charles Redlick (on defense spending). I think both are admirable attempts at solving a very tough problem, but both have very large problems with them.
First and foremost, they are each looking at a large swath of the twentieth century, combining estimates for periods where we would typically expect multipliers to be high with where we would expect them to be low. That doesn't give a very useful estimate. As Jonathan Parker recently noted in an NBER working paper: "We do not have a good measure of the effects of fiscal policy in a recession because the methods that we use to estimate the effects of fiscal policy — both those using the observed outcomes following different policies in aggregate data and those studying counterfactuals in fitted model economies -- almost entirely ignore the state of the economy and estimate 'the' government multiplier, which is presumably a weighted average of the one we care about — the multiplier in a recession — and one we care less about — the multiplier in an expansion. Notable exceptions to this general claim suggest this difference is potentially large."
So although Romer and Romer, and Barro and Redlick presumably do a good job dealing with endogeneity, they can only solve it by neglecting the fact that the size of the multiplier itself is not constant over time. This is not easy stuff. What Steve Horwitz claims to know is really stumping a lot of people. One way to get a better handle of what's going on is to compare economies that are in different situations at different times. Ilzetzki, Mendoza, and Vegh do that here (long version here), and Chinn puts his spin on their findings (and a few similar studies) here.
An oft-proposed solution by skeptics
One thing that fiscal stimulus skeptics like to raise is the forecast of unemployment produced by Romer and Bernstein in January 2009. They note that unemployment has been worse than the forecast suggested would be the case without stimulus, so stimulus has hurt the economy. Steve Horwitz has made this sort of argument.
I have never understood the appeal of this argument at all. To make this argument you have to believe a couple things:
1. Forecasts of the future behavior of a complex system can be made with accuracy.
2. Forecasts by political appointees are reliable sources of counter-factuals for empirical analysis.
Both of these claims are so absurd I find it genuinely shocking that anyone that teaches economics would even make this argument - but they do. When Romer and Bernstein made their projections I was worried that Republican politicians would jump on it later and mistakenly use it in this way, but I didn't expect economists would.
My question for Steve
Should be obvious by now: How exactly do you know what you claim to know? Can you let me in on the secret? I'm sure I'd blow my professors away this fall if I could nail down a fiscal multiplier estimate so conclusively. And I don't want to be an ideologue. I don't want to be a dogmatist. I don't want to be insane. I've never practiced idolatry toward the state - never - and I never want to. If I appear to be an ideologue or a dogmatist to you, it's because I'm ignorant, not because I'm an ideologue. Help me get past my ignorance because if there is conclusive evidence I'm wrong that you know, I want to know that too! I don't want to be unwittingly wrong!