# Measuring stimulus

In this post, I disagreed with Menzie Chinn and argued that CBO estimates of the impactof the stimulus are not estimates. Charles Steele writes in a comment:

CBO’s approach *is* an analysis of what stimulus actually did; such analysis necessarily requires a counterfactual, based on an underlying model of what would have happened otherwise.

So while we might disagree with the neoKeynesian underpinnings, this is an argument about the best macro model. So… what’s your alternative model? And could you suggest an alternative approach to gauging the effects of stimulus?

I would start by saying that I’m not sure we can gauge the effects of the stimulus. It’s possible that is not measurable. Given that the range of multipliers in the profession these days is from .4 t0 1.5 (at least), it is clear that there is no consensus as to how to predict the impact of the stimulus.

That brings me to my second point that seems to be difficult to make clear. The CBO estimates are not estimates. They are forecasts based on previous estimates. They are akin to a golfer who is 150 yards from the flag and asks his caddy for advice on what club to use. The caddy knows that in the past, the golfer has averaged about 150 yards with a 7-iron, so he takes one out of the bag and hands it to the golfer. The golfer swings. He can’t see the green—it’s obscured by trees. The golfer asks the caddy to estimate how far his shot landed from the hole. If the caddy replies that he estimates without looking that the ball is surely within a few feet of the hole because the average 7 iron goes 150 yards when this golfer uses a 7-iron, you don’t call that an estimate. It’s a hope. An expectation. And it might be true. But it’s not an estimate. No caddy would say such a thing. He would wait till he could see where the ball actually ended up.

Surely where the ball goes depends on the execution of this particular swing. The wind. The humidity. How much sleep the golfer got the night before. And so on. Doesn’t the impact of ARRA depend on how it’s structured, who gets the money, the mood of the country, the expectations of increases in future taxes and so on? Yes, these things are hard to measure. So is the mood of the golfer and the angle of club as it strikes the ball. But that’s why the caddy looks and sees where the ball is. Even if the stroke appears to  be well-executed, his ex ante prediction of 150 yards can be way off. That’s why he looks.

In economics, we can’t look. We can’t say that because unemployment remains high the stimulus failed. We can’t say that because GDP grew a lot, the stimulus was a success. We understand there is other stuff going on. But if we can’t control for that stuff, then how can we know (or even estimate reliably) the effect of the stimulus?

I argue this is partcularly true if one caddy says that this golfer usually hits a 7-iron 40 yars while another, says no, no, no–it’s usually 156 yards.

I argue this is particularly true when the CBO says:

Economic output and employment in the spring and summer of 2009 were lower than CBO had projected at the beginning of the year. But in CBO’s judgment, that outcome reflects greater-than-projected weakness in the underlying economy rather than lower-than-expected effects of ARRA.

The phrase “CBO’s judgment” means “we don’t know.”

BTW, Menzie Chinn reasonably disagrees (see the his comment in the comments section as well) with me about whether there is a consensus in the profession over the efficacy of deficit-financed government spending being successful in fighting recessions. He might be right. I don’t think so. But I am happy to be find out otherwise. I’d like to see the paper or papers that established this consensus or the studies that generated the multipliers that were used by the CBO and the consensus around them. I also want to know in particular the paper that justified the multiplier of 1.56 that Jared Bernstein and Christina Romer used in their forecast of the effects of the stimulus package.