David R. Henderson  

DeLong on Horwitz on Bastiat

Spotted at the Johnson Center... Anti-EMH and Antitrust...

Warning: I've got to leave for a seminar in a few minutes and so I'll be brief. I won't explain everything Bastiat said, everything Steve Horwitz said, or everything Brad DeLong said. I just want to home in on a mistaken way DeLong used to judge Horwitz's argument. I also won't bother spending time talking about Brad DeLong's tone and I would ask commenters not to bother either. The easiest thing in the world--and, frankly, the most boring--is name-calling.

A few days ago, Steven Horwitz wrote the following:

Even with those flaws, the Administration's accounting is still one-sided. What it doesn't consider are the jobs lost due to the very policies that are "saving" jobs. Government can only spend what it takes from the private sector one way or another, either through taxation, borrowing, or the redistribution effects of inflation. For every dollar that government spends, there is one less dollar being spent somewhere else in the economy. The jobs that weren't created because the private sector lacked access to capital due to increases in government borrowing should be offset against whatever jobs the stimulus supposedly is creating.

The problem, of course, is that what was never created cannot be seen and therefore cannot be counted. The French economist Frederic Bastiat once defined economics as the art of seeing the unseen. It may be true that we can "see" it by recognizing the unseen effects of policies, but if you can't count what you can't see, you'll always lose out in the numbers game. The result is that estimates of the net employment effects of government programs will always be biased in favor of the program's effectiveness. The inability to count what we can't see should give us long and serious pause when reading about the jobs "created or saved" by the stimulus package.

In response, Brad Delong wrote (I'm reproducing what I regard as his key paragraph):

In both these cases, it is true that we cannot see the jobs that haven't been created. But we can see the price changes that caused those jobs not to be created--just as we cannot see the neutrino, but we can see the Cherenkov radition emitted as a neutrino passes and triggers the creation of an electron moving faster than light. In the first case, we detect crowding out through the fall in bond prices and the rise in bond interest rates accompanying the non-creation of private sector jobs. In the second case, we detect crowding out through the rise in wages and the wage inflation accompanying the non-creation of private sector jobs.

Focus on the last sentence. It's wrong. And, ironically, given the way DeLong dumps on Bastiat, it's wrong because, in a sense, DeLong forgets to look at what is unseen. If wages are not falling, then that well could be due to extension of unemployment benefits and some of the additional spending in the stimulus package. DeLong has arbitrarily chosen zero real-wage increase as his baseline. But in a readjustment, what Arnold Kling calls a recalculation, there's a case to be made for some real wages to fall. At those lower real wages, some of the currently unemployed would be employed. Those jobs that aren't created, therefore, are a cost of the stimulus package. No one, including Steve Horwitz, claimed that there was a one for one. So the jobs not created by the private sector are indeed a cost of the stimulus package.

And notice how carefully Steve Horwitz put it:

The result is that estimates of the net employment effects of government programs will always be biased in favor of the program's effectiveness.

He didn't say that there are no new net jobs created, just that we need to net out the job loss. Would DeLong really disagree with that? Hard to believe he would.

Comments and Sharing

CATEGORIES: Fiscal Policy

COMMENTS (21 to date)
Travis writes:

Your commentary is astute, but was it worth your time to show Brad DeLong's pretensions at economics are just that?

Dnaiel Kuehn writes:

DeLong doesn't dump on Bastiat - he just doesn't want to see him misapplied.

My huge concern is that I was always under the impression that all of these estimates net out gross job loss ANYWAY. Horowitz simply asserts that they don't net out job loss without offering one iota of evidence that that is the case.

The estimate of jobs "saved or created" is ultimately derived from empirical estimates of fiscal multipliers. And empirical estimates of fiscal multipliers ALWAYS net out "unseen" losses. That's the whole point of the estimation exercise!!! That's how you can potentionally get a multiplier of less than one.

That oversight by Horowitz is far more egregious and meaningful than any sort of tone problem that DeLong is guilty of. It's the basis of Horowitz's whole argument after all. You'd think he'd put forth some evidence that the estimates are gross employment estimates and not net employment estimates.

Daniel Kuehn writes:

If I'm mistaken somebody please correct me. This is absolutely central, and I honestly want to know the correct answer. Multiplier estimates absolutely subtract out GDP losses: they are net effects. And I was always under the impression that these job estimates were derived from the multiplier estimates - and therefore, that they too are NET CHANGE estimates.

Ryan Vann writes:

I think the real issue is that the proxies used are simply unreliable. Delong never convincingly explains why wage rates and treasure rates would capture crowding out effects. Why use such a figure when you could simply use something like government job vs private job losses, or look at stats in areas where stimulus is heavily focused? My guess is they wouldn't support his viewpoint.

Horowitz just provides rhetoric, and not much in the way of technical analysis.

Ryan Vann writes:


I'm not sure I completely understand what you are asking, but I've always know multiplier (every activity technically has a multiplier) calculations to be done by taking the change in income over the change in something else, in this case fiscal spending. That is part of the reason using multipliers to approximate crowding out, to me, appears a messy endeavor on an aggregate basis, and I really don't think much of the calculations.

I'm not sure what exactly a Romer or a Bernstein does in the event of a negative growth rate (whether they use absolute changes in income or what).

Daniel Kuehn writes:

Ryan -
I'm not sure what you mean. How else would you capture crowding out - a net effect - if not by looking at change in income in response to change in spending?

Honestly, measuring gross effects would be a lot harder than measuring net effects, wouldn't it? You have identification issues measuring net effects, of course - that's your big obstacle. But assuming you have a strategy for proper identification finding one net effect is a hell of a lot easier than finding the positive gross effect and the negative gross effect and combining the two.

Yet for some (as yet) unexplained reason, Horowitz seems to think that what they do is look at the gross positive effect and ignore the gross negative effect.

Disaggregation can always be enlightening, there's no question of that. That doesn't make aggregation meaningless - particularly in an economy-wide recession. If this were just a housing bust and limited to that I might think differently. WSJ recently reported that only 20% of job losses were in construction.

Ryan Vann writes:

Crowding out isn't a net effect its a composite that can have net implications.

Daniel Kuehn writes:

Ryan -
Well of course. But I think we've all - Bastiat, DeLong, Horowitz, Henderson, me, and you - established that that what we care about is the net effect and that that net effect includes the negative gross effects and the positive gross effects.

We no more care about the non-net implications of crowding out in isolation than we care about the non-net implications of the jobs created by shovel-ready stimulus work.

Steven Horwitz writes:

First of all, folks, the name is Horwitz. With one "o."

Second of all, the original post was a 500 word blog post for PBS's Nightly Business Report. It was designed for a lay-audience, not for other economists. I had 500 words to make the point, there simply was not space for the "on the one hand" type subtleties that go with a serious argument made to other economists.

There have been a lot of good points raised here and elsewhere about my argument, but please keep in mind who I was writing for and the space limitations I faced.

It was intended to be a rhetorical piece, not a journal article nor even something I might post at Coordination Problem. People are asking for things that article could not possibly have done.

Ryan Vann writes:

Net income was negative for the last 2 years; we wouldn't conclude that stimulus was negative on net because of that. There is a gaping chasm between saying there is a no crowding out as evidence by a net increase in income and saying there is a net increase in income. We do care about non-net effects, if we are at all concerned with market composition or time.

spencer writes:

I suggest you actually look at the data before you make these type of analysis.

Your point is that real wages are falling.

But in 2009 real average hourly earnings actually rose 3.5%.

But we would never expect you to be confused by the facts, now would we.

David R. Henderson writes:

I did not say that real wages are falling. I took it that DeLong had the numbers right. Apparently he didn't. My point is that many real wages should fall.

Loof writes:

In regard as to whether creating jobs as a goal of economic policy can be good for government. Steve Horwitz’s topical You Can’t Count What You Can’t See derived from Bastiat’s definition of economics as the art of seeing the unseen provides a negative answer. L’ll provide a positive view.

Horwitz uses a Friedman story:
“In observing hundreds of Chinese workers clearing land for a new building using shovels, Friedman asked his hosts "Why are they using shovels? Why not use heavy equipment like an earth-mover?" The Chinese official said "If we did that, we'd lose all of those jobs!" Supposedly Friedman said "Oh, you're trying to create jobs! I thought you were trying to build a building. If you want to create jobs, why not take away their shovels and give them spoons?"”

Friedman’s argument is off base. He goes to the silly extreme: use “spoons”. The aim of the officials wasn’t to clear land with pure economic efficiency or solely create jobs; it was about clearing land AND creating jobs. The Chinese, practically wise, found the middle way – and would politely ignore Friedman’s smartass comment.

For the same taxpayer cost it’d be easy for government to use an earthmover for a week; not employ hundreds for a month. The key economic unseen difference: the employed people typically would take money earned, buy in local markets and the money would circulate through many invisible hands: the butcher, the baker, etc. and so create employment for more people to make money and business profit – and, not just provide an opportunity for a nonlocal earthmover owner to maximize profit; and, to be fair, perhaps more efficiently clear the land with a minimum of local employment. Relative to communities, it’s unseen trickle down economics vs. unseen ripple around economics.

Typically “earthmovers” legally grease, “lobby”, the right hand of politicians who get government economists to do a cost/benefit analysis relative to efficient causation for a trickle down economy. In Asia, it’s not uncommon to also grease the left hand to eliminate the competition from other movers. By legal hook and as a downright crook Greedy Archimedes’ unseen hand gains leverage to move the world when given a place to stand – and use his earthmover.

Niklas Blanchard writes:


I just want to point you to this, because I use the story quite often...but I don't think it's really attributable to Friedman:

"This is the point of the apocryphal story of an engineer who, while visiting China, came across a large crew of men building a dam with picks and shovels. When the engineer pointed out to the supervisor that the job could be completed in a few days, rather than many months, if the men were given motorized earthmoving equipment, the supervisor said that such equipment would destroy many jobs. “Oh,” the engineer responded, “I thought you were interested in building a dam. If it’s more jobs you want, why don’t you have your men use spoons instead of shovels.”"

Creating Jobs vs Creating Wealth

The thought problem is that the supervisor in our story is actively destroying value. Ostensibly you could use the argument that the "ouput gap" that exists at a (this) particular moment would negate this...but it's still destroying value if the marginal cost of employing the people is greater than the marginal cost of capital investment in another "earth mover"...which it almost surely would be.

[Note: Incorrect html coding corrected. --Econlib Ed.]

Steve Horwitz writes:

In response to Daniel's question, which he also emailed me, I'll reprint my email reply here:

All the data on "jobs created or saved" is not, as far as I can tell, making any attempt to net out jobs destroyed or not created.

See here for example for what appears to be a fairly neutral analysis: http://politicalmath.wordpress.com/2009/11/05/dirty-stimulus-jobs-data-exaggerates-stimulus-impact/ In particular: "This led to a great deal of mockery over the “save or create” turn of phrase, but the administration set out to actually measure the number of jobs that were saved or created by having recipients of the stimulus funds fill out a form in which they indicate how many jobs that particular chunk of the stimulus created (that form can be found here)."

This is what I was referring to in my original blog post about "jobs created or saved." There is absolutely no evidence out there that the Administration has attempted to do any "netting." After all, the whole point of my NBR post was that they *can't*. But in answer to your question, they *aren't*.

Andrei Bolkonski writes:

Help me understand something, please. The discussion of crowding-out and netting seems a bit beside the point to me.

If the stimulus increased real rate of inflation (compared to a scenario where the stimulus hadn't occurred--which may be simply no inflation as opposed to the deflation we would've experienced), doesn't this, by itself, disadvantage the private sector compared to the public sector (whose increase in cash offset inflationary effects) when it comes to hiring? If so, you couldn't capture this loss by looking at bond rates as DeLong suggests.

In other words, DeLong seems to be expecting the private sector to act, with regard to hiring, the same as if the stimulus hadn't occurred. But in a real sense, the private sector has less capital, and so that's a problematic assumption.

Bob Murphy writes:

Daniel Kuehn,

You've been asking a good question all along, but I think Steve's answer is right: When it comes to the administration's announcements of how many jobs they've saved or created, they are naively counting them all up, and then assuming that's a net figure.

I think you're correct, that the Keynesian macro models that Romer et al. used to justify the stimulus in the first place, are true general equilibrium models that simulate everything and then give you the net figure.

But DeLong is busting Horwitz for ignoring "empirical reality" (or whatever). DeLong isn't claiming, "Steve you idiot, can't you see that the theoretical models already netted out that loss?"

No, DeLong is saying, "We can count the jobs created with our own eyes, and we can see that there haven't been any jobs destroyed because if there were, we'd see the telltale signs of rising wages and interest rates. But we don't, so no jobs destroyed."

So I'm not sure how you would classify all this in your framework. Would it satisfy you if Horwitz and Henderson said, "OK sure, they technically consider the possibility of job destruction, but then they rule it out with a totally BS argument. So in effect they ignore the unseen." ?

Bill N writes:

Neutrinos aren't the source of Cherenkov radition. The source is the highly visible electron. http://en.wikipedia.org/wiki/Cherenkov_radiation He's wrong there and could just as easily be wrong about everything else. There no laws of physics to verify his thesis, so it's all hand waving.

Ryan Vann writes:

Prof Murphy,

We haven't seen net job creation though (negative rates for a while), so the argument isn't even as respectable (albeit naive) as "there was a net increase in jobs and no increase in wages, etc, therefore no crowding out." If, on net, jobs have been decreasing, why would wages increase, with or without crowding out?

Daniel Kuehn writes:

Bob -
I don't see why you make the claim that it's a BS argument.

Obviously looking at wage rates and interest rates isn't the final word in the conversation. But it's an excellent rule of thumb.

Anyone making a claim that's as out in left field as Horwitz should be prepared to answer the question of why wage rates and interest rates are stagnant if crowding out is a serious concern. That's not the end of the discussion, but it's a reasonable point that we still haven't seen a reasonable answer to.

There's also the full employment question that DeLong raises that Horwitz doesn't even address. Bastiat's arguments clearly apply at full employment. Horwitz offers no reason for thinking they apply when we are substantially below full employment.

I'd also note that my correspondence with Horwitz over email was considerably longer than what he posted - and I cited the portions of the CBO and administration reports that noted that they are providing net employment effects, rather than gross employment effects. Indeed, the CBO report went on for about half a page discussing the issue of crowding out. The discussion was even more detailed and thorough than the one Steve provided!

I haven't heard back from him since sending him those pieces.

Brad DeLong writes:

So is your argument really that if not for the stimulus package wages would be falling--and falling wages would be inducing employers to hire lots more workers?

That hasn't happened in the U.S. economy since 1921. In the U.S. economy since 1921, falling wages and prices have deepened depressions by multiplying bankruptcies. Deflation is really bad juju.

I confess I am surprised to see somebody calling for it, even here.

Comments for this entry have been closed
Return to top