23 Feb 2010

Can I Pass the Keynesian Turing Test?

All Posts 13 Comments

Robert Barro has a pretty interesting WSJ op ed today, in which he uses his historical analysis of government spending and tax multipliers to evaluate the Obama Administration’s stimulus package of 2009.

In a nutshell, Barro first estimates a spending multiplier of 0.4 (in the first year) and a tax multiplier of -1.1. So if the government spends $100 billion and doesn’t touch taxes (i.e. borrows and spends an extra $100 billion), then GDP goes up by $40 billion. What that means is that there is 60% crowding out. The expenditure of $100 billion directly raises GDP by $100 billion, but then other components of GDP (private consumption, investment) fall by $60 billion. Hence, the net effect on GDP (in the first year) from an additional $100 billion in deficit-financed spending is only $60 $40 billion, according to Barro’s analysis of periods of big spurts in military spending.

On the other hand, Barro finds that if the government raises $100 billion in new taxes (while holding spending constant), this lowers GDP by $110 billion.

So putting the two effects together, Barro estimates the 5-year impact of the stimulus plan. The idea is that the government “buys” some output on the front end (because the spending multiplier is above 0), but then has to forfeit output (relative to the baseline) on the back end because the higher debt requires more taxes. Barro concludes:

We can now put the elements together to form a “five-year plan” from 2009 to 2013. The path of incremental government outlays over the five years in billions of dollars is +300, +300, 0, 0, 0, which adds up to +600. The path for GDP is +120, +180, +60, minus 330, minus 330, adding up to minus 300. GDP falls overall because the famous “balanced-budget multiplier”—the response of GDP when government spending and taxes rise together—is negative. This result accords with the familiar pattern whereby countries with larger public sectors tend to grow slower over the long term.

The projected effect on other parts of GDP (consumer expenditure, private investment, net exports) is minus 180, minus 120, +60, minus 330, minus 330, which adds up to minus 900. Thus, viewed over five years, the fiscal stimulus package is a way to get an extra $600 billion of public spending at the cost of $900 billion in private expenditure. This is a bad deal.

The fiscal stimulus package of 2009 was a mistake. It follows that an additional stimulus package in 2010 would be another mistake.

OK I wanted to test whether I really understand the Keynesian mindset. So I honestly haven’t looked yet to see what DeLong and Krugman have to say about this. Here’s my guess:

My Guesses as to the Keynesian Response to Barro’s Op Ed

(1) Barro is a liar, second only to Russ Roberts in his lyinghood.

(2) Barro’s spending multiplier is way too low. He admits that he derives it from studying wartime periods, but that’s absurd. During such periods, the government enacts strict rationing measures to ensure that private consumption and investment stay suppressed, freeing up resources for the war effort.

(3) Barro’s use of the 2008 baseline is absurd. If the government had sat back and done nothing, unemployment would have continued climbing, perhaps it would have been 15% right now. That would mean lower tax revenues and more spending on social welfare programs. In essence, all the stimulus does is concentrate that unavoidable government debt increase into the beginning years, when it might obviate much of the later spending. Barro has done the equivalent of looking at a patient just diagnosed with cancer, and comparing the medical expenses of early intervention against a “baseline” of a perfectly health person’s medical costs.

(4) Barro’s spending multiplier makes no sense, both in theory and in terms of empirical evidence. In an economy with 10% unemployment and 0% interest rates, running a fiscal deficit doesn’t cause any crowding out. We would see the telltale signs if it did. So right now, when the government spends an extra $100 billion with borrowed money, that doesn’t cause a $60 billion reduction in spending elsewhere. How could it? What is the mechanism? And not only does the $100 billion raise GDP directly by that amount, with no $60 billion offset, but in fact we get further GDP gains because of the further spending by people who would otherwise have been unemployed. Barro casts aspersions on Christina Romer’s estimates of a multiplier greater than 1, but he doesn’t explain what her mistake was. He just says he can’t understand her figures and asserts his own. What pseudoscientific nonsense. Why oh why can’t we get right wing economists who know more than discredited views from the 1930s?

OK kids, go look. I promise this was off the top of my head. How did I do?

13 Responses to “Can I Pass the Keynesian Turing Test?”

  1. Beefcake the Mighty says:

    You aren't too far off:


    I have to say, isn't this a rather easy one for the Keynesians to rebut? If you accept rubbish like "idle resources" or "liquidity traps," then isn't Barro's analysis ill-conceived? Should those concepts be attacked, rather than play the Keynesians at their own game (which is what Barro is basically doing)? Who cares if the multiplier is 0.4 or 0.561 or whatever, if the entire concept is crap?

  2. BadTux says:

    Okay, here's my answer from a Keynesian mindset. Crowding out can occur only if there is a shortage of slack resources in the economy. If there are slack resources in the economy — an excess of money with a shortage of worthwhile places to invest it due to a collapse in demand, an excess of unemployed people who have a productivity of zero due to a lack of jobs due to a lack in demand — crowding out cannot occur. We currently have 0% short-term interest rates for Treasuries, demonstrating that there is an excess of money with a shortage of worthwhile places to invest it. We currently have 10% nominal unemployment and 20% combined unemployment/underemployment in our economy, people who want work (or want more work) and cannot find it. Both of these facts add up to no (zero) crowding out caused by government spending in the near term, since for crowding out to occur, government would have to be taking resources from private enterprise rather than taking from a pool of slack resources not currently used or needed by private enterprise.

    And let's not forget another factor: The printing press. The printing press adds some interesting variations to this set of equations, especially when the money supply is under severe deflationary pressure because banks are stashing money away in reserves (thereby reducing the fractional reserve multiplier) due to fears of further losses caused by a declining economy, rather than lending that money out. Right now, you can print money and it basically takes one loop through the economy, ends up back in a bank, and ends up right back on the books at the Federal Reserve again, no longer contributing to economic activity and thus no longer affecting prices in the economy (since it has a money velocity of zero once it gets stashed back in the Fed's electronic vaults again). So if we consider monetizing the debt, we can even remove the specter of future interest rate hikes causing slower growth in the future as a downside to increased government spending right now.

    In short:

    1. Barro's analysis applies only in an economy with full employment and a shortage of investment capital,
    2. We have neither full employment nor a shortage of investment capital,
    3. We have the ability to monetize a significant portion of the debt such that it does not affect future taxes, and thus
    4. Barro's analysis is not applicable to the current situation.

    And I won't even go into Barro's notion that a construction job building a superhighway for the government somehow adds less to GDP than a construction job building a commercial skyscrape, which is implicit in his notion that $1 of government spending somehow adds less than $1 of value to the economy. This is bizarre. Not only do you have a highway at the end of it, a highway which in all likelihood was built as efficiently as any that could have been built with private money if past experience is the guide, but you've also increased the value of that commercial skyscraper by a tremendous amount, since a commercial skyscraper with no highways going to it is worth, uhm, $0. Even von Mises never went so far as to ascribe a *negative* multiplier to government spending, rather, he insisted that the Free Market Fairy would make better use of those slack resources and thus have a higher multiplier (i.e. result in more GDP growth) in the end. But a *negative* multiplier? I cannot find any actual economic data, anywhere, which support such an assertion — in all of the actual real-life data we have, increased government spending resulted at least a 1.0 multiplier to GDP, even the WWII spending where the majority of the increased spending was used to build things that were literally blown up or shot out of the end of a gun had a greater than 1.0 multiplier. Maybe Barro's model says that didn't happen. But if Barro's model disagrees with reality, it is Barro's model, not reality, which is wrong.

  3. BadTux says:

    Okay, so my scoring is that you hit the Keynesian's response about 85%, a solid B score, the only thing you missed off the Keynesian scorecard was the invention of the printing press. And BTW, actual historical data showing that government spending has *at least* a 1.0 multiplier EVEN IN WARTIME. Whoa! Reality! It'd be a great idea, wot?

  4. BadTux says:

    I am baffled. You are asking me for a citation for where Mises did *not* make a claim? Uhm, how, exactly, am I supposed to do that — read the complete works of Ludwig von Mises to you and on each page, point out that he did *not* claim there was a below-unity multiplier for government spending? Probably the closest I can give you is this passage from "Interventionism: An economic analysis", where he claims:

    Public works projects are recommended as a means to provide employment. But if the necessary funds are secured by issuing government bonds or by taxation, the situation remains unchanged. The funds used for the relief projects are withdrawn from other production, the increase of employment opportunities is counteracted by a decrease of employment opportunities in other branches of the economic system.

    That is, he does *not* claim that government spending has a less than 1.0 multiplier here, he merely claims that government spending has a crowding out effect that renders $1 more of government employment resulting in $1 less of private employment. That is a completely different assertion altogether from the assertion that government spending has a less than 1.0 multiplier.

    Or is it something else that you want me to cite? Please clarify!


  5. Anonymous says:

    The Blackadder Says:

    Looks like the correct answer is (5) Pretend the article was never published.

  6. BadTux says:

    Wow, I love how you descend to name-calling, Mighty One. So mature.

    I think it is clear from the above quoted passage that Mises clearly did not believe in slack monetary resources. Yet he did, in fact, believe in slack human resources, if you read the full passage that I think he clearly admits that the majority of the unemployed during the Great Depression were involuntarily unemployed, wanted jobs, were frantic for jobs, and were not merely "choosing leisure". In that he is perhaps more in touch with reality than some of his peers such as Milton Friedman whose view of the Great Depression might be summarized as it actually being the "Great Vacation", caused by people choosing leisure over working for less.

    Thus while Mises may have never used the term "slack resources" and would have sneered at any mathematical formulation of such, to say that he had no conception of the idea specified by the term quite simply is contradicted by the man's work. As for whether Mises would have sneered at the notions of "slack resources" and "spending multipliers" as something measurable via aggregate statistics, well, yes. But that was not my point, as you very well know yet are attempting to distract attention from with your gratuitous insults (hey, you forgot to call me a cooty-head and poo-poo brain!).

    – Badtux the Snarky Economics Penguin

  7. TGGP says:

    Friedman thought the GD was caught by deflationary monetary policies. The "Great Vacation" sounds more like a real business cycle theory. Casey Mulligan's blog "Supply and Demand" has been arguing something closer to what you're saying, but he uses an odd definition of "labor supply".

  8. Beefcake the Mighty says:

    Did you happen to notice the title of the section you reference here: INTERFERENCE BY PRICE CONTROL. Von Mises clearly identifies the source of the unemployment he's talking about in the interference by govts (and their union allies) in the market's price system. Which is why he would not only sneer at "mathematical formulations" of "slack resources," but at the entire concept. Resources are not slack or idle *as such*, but idle *at some price.* It's precisely the purpose of Keynesian policies to subvert the market process of adjustments through the price system. Too bad Keynesians are too economically ignorant to grasp this simple point.

  9. BadTux says:

    TGGP, it is common for economists to believe that their own particular school of economic thought has a monopoly on all good ideas, and some other particular school of economic thought has a monopoly on all stupid ideas. Milton Freidman's money equations were brilliant and quite worthy of the Nobel that he received and I am quite happy that a student of Milton Friedman is now in charge of maintaining a stable money supply for the US. He was, however, quite the neoclassical real business cycle economist in many other ways.

    Beefy, you have now hit upon the primary Keynesian critique of classical thought, which is that the classical economists knew the price of everything, but the demand for nothing. That is, there are two parts to the supply and demand curve: the supply side, and the demand side. If the demand side collapses due to people deciding that their money is better saved than spent (rather than because prices rise), that is, if the entire demand curve shifts downward because people voluntarily choose to save more rather than spend, then employment declines regardless of what prices do.

    Keynesian thought requires accepting two things as true:

    1) One man's wage is another man's price, and vice versa. That is, if all prices and wages in a system go down (or up), it has no effect upon demand and thus no effect upon employment, since businesses are in business to make a profit, not to serve as charities hiring people they don't need in order to meet demand.

    2) Demand can go down (or up) even if neither prices nor wages have changed due to changes in people's propensity to spend — i.e., changes in how people decide to distribute their wages between savings and consumption.

    These propositions agree with actual observed reality over the past 80 years, where e.g. we've had high unemployment during periods of high inflation (see: late 70's, early 80's), and high unemployment during periods of low inflation (see: now) or during actual deflation (see: 1930's) — wage and price levels on a systemic basis seem completely irrelevant to the question of employment. What matters is demand.

    Thus the neo-Keynesian emphasis upon the importance of the demand curve and, specifically, propensity to spend, which moves the demand curve up and down. This appears to comport more closely to reality than Austrian or neoclassical thought, which appear to describe reality in a universe where unicorns are real and cotton candy grows on trees but not reality in our universe, where businesses are not charities and hire only those workers they need in order to meet demand, regardless of how cheap those workers are. If I have a sandwich shop and I need six workers to meet my demand, that's all I hire — period — regardless of how cheap the workers are. Any theory that refuses to admit the centrality of demand to the employment problem is a theory that is not worth taking seriously, at least insofar as it pretends to address the employment problem.

    – Badtux the Bemused Economics Penguin

  10. BadTux says:

    TGGP, my apologies. It was apparently Hayek who insisted that 25% German unemployment in 1932 was caused by lazy Germans choosing to take a Great Vacation rather than to work, a view also more recently subscribed to by Edward Prescott. I could swear that Friedman also basically stated the same thing, albeit not so baldly, but will readily admit that I could be mistaken.

  11. Beefcake the Mighty says:


    Fair enough, we'll have to agree to disagree here.

  12. Bob Murphy says:

    Hey Beefcake,

    I don't want to have a blog where people start swearing at each other in the comments, OK? When you do that it means I have to take time to go through and delete your comments. I don't feel like doing that so please stop swearing in the comments.