Calling the Keynesians’ Bluff
Friday, February 1, 2013
It’s time to enter the inner sanctum of Keynesian theory and see what (little) it’s worth.
As a rule, the latest economic figures are not that important in themselves, but they shed fresh light on long-run trends. It may or may not be true, as Dean Baker suggests, that the 0.1 percent decline in GDP in the fourth quarter of 2012 (according to preliminary Bureau of Economic Analysis estimates) was caused by a decline in government, especially military, spending. But the real question is why the economy is so feeble that it couldn’t take a few defense cuts in stride. The big picture, which the latest data make a little clearer, is the ongoing poor performance of the Obama economy and the Keynesian paradigm that has informed the Obama administration’s economic strategy.
An influential group of Keynesian economists remains undaunted in its advocacy of fiscal stimulus. Paul Krugman, in particular — far from thinking he has anything to apologize for in his full-throated support for stimulus back in 2009 — regularly crows vindication and blames his opponents for not surrendering. Krugman and other Keynesians would say that they continue to be confident because the data support them, but they can only say that because their theoretical prejudices enable them to look at the data in some very odd ways. They wax triumphant over low inflation and low interest rates, which, for theoretical reasons inscrutable to non-economists and debatable by economists, they regard as proof that “aggregate demand” is the problem and stimulus would work, while dismissing as irrelevant the fact that deficit spending at a rate of well over $1 trillion per year since 2009 has left us with anemic growth and 12 million people unemployed, and many more abandoning the labor market in discouragement, while median incomes fell by upwards of 7 percent during Obama’s first term in office.
The political effect of Keynesianism has been to provide a warrant for fiscal irresponsibility. This isn’t necessarily a left/right issue: Keynesianism encourages conservatives to prioritize tax cuts over spending cuts, and liberals to focus on increasing spending more than raising taxes. Doubtless politicians have always been tempted to run deficits, but Richard Wagner and James Buchanan, in Democracy in Deficit, show that prior to the Keynesian Revolution, politicians largely restrained themselves, borrowing during wars and balancing budgets or better in peacetime.
There have been many attempts to estimate the government spending 'multiplier,' and the results are all over the place.
Keynes once said that “practical men, who believe themselves to be quite exempt from any intellectual influences, are usually the slaves of some defunct economist.” In this case, it’s true. Today, we’re living in a Keynesian brave new world. Savings, which Keynes disliked, have been way down for a generation in the United States. That’s one reason median incomes have stagnated, while high-saving Asia has enjoyed surging growth. With scant domestic savings, U.S. investment has to be financed mostly from abroad. Governments run ever-growing welfare states on borrowed money. Even in the 1990s, when the United States was supposedly running surpluses, the unfunded liabilities of Social Security and Medicare were mounting. Now those are still mounting, and we’re running huge primary deficits too. But Keynesian economists keep telling us not to worry, and saying that it’s more important to promote economic recovery than to control the deficit.
It’s time to call the Keynesians’ bluff, to enter the inner sanctum of Keynesian theory, and see what (little) it’s worth. I saw the basic Keynesian models first when I was an undergraduate, and they made me uneasy. Fifteen years or so later, after grad school and some policy jobs, and having studied the models a few more times and taught them to undergrads myself, I’m still an unbeliever. Here’s why.
The Keynesian Cross
The models I’ll present here are taken not from Keynes’s General Theory of Employment, Interest, and Money but from undergraduate textbooks. These textbooks took the models from economists contemporary with or slightly after Keynes, who were trying to make the ideas in Keynes’s confusing but highly fashionable book clear and accessible to themselves, one another, and new students of economics. Keynes’s book was fashionable because it raised a heterodox standard of revolt against orthodox laissez-faire economic doctrines, which were wrongly believed to have caused the Great Depression. (The real cause was inept monetary policy by the Federal Reserve, plus rampant intervention by the Hoover and Roosevelt administrations, trade protectionism, and, later on, high taxes, union activism, and political risk due to Roosevelt’s populism and experimentation.)
The economics profession’s great mistake was to treat Keynes as a guru rather than a gadfly. Keynes’s skepticism about the ability of the price mechanism to bring the macroeconomy continuously into equilibrium, taken in small doses, is a useful corrective to the complacency to which theorists of economic equilibrium are prone. It’s good, now and then, to stress the special resistance of workers to nominal wage cuts, or the casino-like vagaries of financial markets. Most subtly, there is a deep tension, which Keynes glimpsed, between the desideratum that prices be flexible enough to equilibrate supply and demand in particular markets, and the desideratum that prices be stable enough to ease the burden of economic planning for individuals. But a “general theory of employment, interest, and money” is exactly what Keynes’s book of that name wasn’t — or at least it wasn’t a good one. Attempts to express that theory mathematically and graphically have to falsify it in order to make sense of it, and in the process they lose some of the gadfly-style heterodox insights that Keynes sprinkled throughout the book. One school of macroeconomics, called the “post-Keynesian” school, would reject the models shown below because they’re not Keynesian enough. But they have a firm foothold in the economics curriculum, and they’re the basis for prominent Keynesian economists’ analyses.
Our first target will be the Keynesian cross, shown in Figure 1. This chart first appeared in Paul Samuelson’s 1939 paper “A Synthesis of the Principle of Acceleration and the Multiplier” (p. 790), and was later popularized by Samuelson’s 1948 textbook, Economics, which dominated the market for a generation. Economists find that when the logic of an argument becomes moderately complex, a graphical representation is often indispensable for making it clear, so bear with me and see what you can make of the graph below. Then read my explanation. Then take another look at the graph.
Figure 1: The Keynesian Cross
The vertical axis in Figure 1 represents expenditures; the horizontal axis represents income or GDP. E(Y) means “expenditures as a function of income,” and each point on the line represents what people would spend (E) if they had a certain income (Y). The slope of E(Y) implies that when people make extra money, they spend more, but their spending does not rise proportionally with their income. ΔG means “change in government spending,” i.e., fiscal stimulus, and E(Y)+ΔG is the expenditure function E(Y) plus ΔG.
The 45° line contains points where E=Y. E=Y must hold, it is claimed. That is, expenditure must equal income, because one person’s spending is another person’s income. But the economy, without stimulus, must also be on the E(Y) curve, since spending is a function of income. Therefore, income is determined by the intersection of E(Y) and the 45° line — in other words, Ye. With stimulus, income is determined by the intersection of the 45° line with E(Y)+ΔG — in other words, Ys.
According to the Keynesian cross model, then, government spending can raise income. Not only that, it can raise income by a lot more than the increase in government spending itself. In Figure 1, ΔY=Ys-Ye > ΔG.
The same point can be made using a story. (You may have to take my word for it that the chart and the story are saying the same thing, though kudos if you do get it.) Suppose a stimulus bureaucrat goes out and spends $10 million building a bridge to nowhere. The $10 million is paid out to contractors, workers, and materials suppliers, who receive it as income. They save, say, 20 percent, and spend the other $8 million buying houses, cars, food, and so on from other people who receive $6.4 million in income, save 20 percent and spend $5.12 million, and so on.
If everyone saves (on average) 20 percent of any extra income they earn, we can make a precise prediction — using a nifty mathematical technique called resolving a geometric series — of how much government spending will grow the economy. This is called the “multiplier” effect of government spending. Income, by this theory, will rise by exactly 1/MPS times the increase in government spending, where MPS is the marginal propensity to save. In this case, 1/(20%) = 5. Each extra dollar of government spending yields four more dollars of private sector spending as well. The Keynesian cross captures the intuition at the back of most proposals to stimulate the economy through government spending.
Well, it isn’t true. There have been many attempts to estimate the government spending “multiplier,” and the results are all over the place — with all due respect to Valerie Ramey, this is a case of “ask a stupid question, get a stupid answer,” because much depends on what kind of government spending — but none of them come close to confirming the optimistic predictions of the Keynesian cross model. And we really should have known that before taking the model to the data. A little thought exposes the flaws in the model’s logic.
At the deepest level, the model’s relationship to time is hopelessly confused. In the chart, expenditure is a function of income. But expenditure when and income when? The model seems to assume, and only works on the assumption, that expenditure now is a function of income now. But that can’t be right, because expenditure and income are both flows, which can only be measured over time. And — this is the key — causation runs forward in time. My spending this year cannot be determined by my income this year, because I won’t know for sure until midnight, December 31 what my income will have been, and by that time my spending decisions are water under the bridge.
Maybe spending depends on income with a lag, but what lag? A week? Three months? Ten years? If you start picking arbitrary lags out of the air and trying them on for plausibility, you may get the following flash of insight: spending doesn’t really depend on income at all, it depends on wealth and expected future income. And that’s the beginning of a whole different model.
If you start picking arbitrary lags out of the air and trying them on for plausibility, you may get the following flash of insight: spending doesn’t really depend on income at all.
Think of it in a different way. A construction worker gets $20,000 working on a bridge to nowhere. How much of it does he save, and how much does he spend? That’s a meaningless question, if we don’t add over what time period he is to spend or save it. In the next ten minutes, he probably won’t spend any of it. He keeps — saves — 100 percent. In the very long run, he and his family and heirs presumably spend 100 percent. What else would they do with it? In the shorter run, say, a few weeks, he might spend next to nothing, or he might spend much more than 100 percent, if, say, the $20,000 is just enough to supply the down payment on a $100,000 home. (In this case, the change in his spending is 500 percent of the change in income that induced it.)
Keynes gave the “marginal propensity to consume” an important role in his theory. The marginal propensity to consume is supposed to be the portion of an extra dollar of income that is spent as opposed to saved. But again, over what time period? There is no non-arbitrary way to choose the time period. Moreover, as soon as time is brought into the story, it is clear that the multiple rounds of spending which a dollar of fiscal stimulus is supposed to generate might take a long time to play out. Meanwhile, the savings that supposedly disappear in the Keynesian cross model in reality go into bank vaults and then (usually) go back out as loans, so that money keeps circulating too.
Keynes borrowed the idea of equilibrium from microeconomics and especially from the Econ 101 model (then still rather new: Keynes studied under Alfred Marshall, the first to popularize it in his 1890 textbook Principles of Economics) of how supply and demand determine price and sales volume in markets for particular goods. The concept of equilibrium is atemporal: it leaves time out of account, implicitly “jumping over” the dynamic process by which prices and quantities might gradually converge to equilibrium. That may be all right if you’re thinking about the long run, when the dynamics have played out. Keynes wanted to understand the short run, and to set to one side the supply-and-demand equilibrium by assuming “sticky prices,” thus allowing for inefficient unemployment of resources. But since multiple rounds of spending would take years to play out, it makes little sense to combine them with sticky prices or to ignore savings being lent out again. In short, the Keynesian cross model does not deserve to be taken seriously.
The IS-LM Model
Our next target is the IS-LM model, which first appeared in John Hicks’s 1937 paper “Mr. Keynes and the ‘Classics’; A Suggested Interpretation.” It was the most famous thing Hicks ever wrote and was the main reason he got the Nobel Prize in 1972, but Hicks later downplayed it and doubted its usefulness. It is supposed to explain how interest rates and GDP are determined by fiscal and monetary policy. It starts with two stories about how interest rates are determined.
First, the interest rate must balance savings and investment. Savings have to equal investment, because when people save, they are in effect trying to transfer current resources into the future via the financial system, and they can only succeed, collectively, in doing so if there is corresponding activity that transfers real value into the future via investments, e.g., building houses or factories, or training workers. For a given level of income, people will want to save more when interest rates are higher. At the same time, higher interest rates reduce investment, since only investments that are expected to earn a (risk-adjusted) return higher than the interest rate are worth pursuing, and the higher the interest rate, the fewer such investments there are. So, as interest rates rise, boosting saving and reducing investment, there will be some point where savings and investment are brought into equality with one another. Investment doesn’t depend in any obvious way on income, but saving does — more income leads to more savings — so the higher income is, the lower the interest rate needed to balance savings and investment. We can represent this relationship as a downward-sloping curve in the interest rate–income space — or the IS curve in Figure 2.
Second, the interest rate must balance demand for money, or “liquidity,” with demand for bonds. The advantage of holding bonds is that they pay interest. The advantage of holding money is that it is liquid, easy to use in transactions. As income rises, people want to spend more, so they want more money for transactions. Assume the money supply is fixed. The higher are people’s incomes, the higher the interest rate that is needed to induce people to hold only the available money supply — so this relationship can be represented as an upward-sloping line in the interest rate–income space, the LM curve in Figure 2.
But wait a minute. These were supposed to be stories about how interest rates are determined. How can both stories be true at the same time? Are interest rates determined in bond markets, and do they depend on the money supply? Or are they determined by real savings and investment behavior, leaving no role for the money supply? In fact, we can reconcile these by making GDP a variable of the system, leaving us with two equations in two variables. The whole model is shown in Figure 2.
Figure 2: The IS-LM Model
In Figure 2, the IS and LM curves jointly determine both the interest rate and GDP. For example, given LM, if IS1 applies, interest rates will be r1, and income, Y1. If IS2 applies, interest rates will be r2, and income, Y2.
So what is IS-LM good for? Mainly, it’s used to show how the government can manage the macroeconomy through fiscal and monetary policy. Of course, that only works if the government is able to move the IS and LM curves around. Monetary policy moves the LM curve around. A bigger money supply means interest rates have to be lower, or income higher to induce people to hold it. That shifts LM to the right. Fiscal policy, so the story goes, moves the IS curve around. If the government borrows more, either to spend or to cut taxes, that pushes the IS curve up, since for a given level of income it takes a higher interest rate to raise savings and/or lower investment to the point where savings suffice to finance investment and government borrowing. That said, if “Ricardian equivalence” holds — that is, if people see that the government is running deficits and save more for the future to pay the higher future taxes that will be needed to pay for the deficits — then fiscal policy can’t move the IS curve at all.
Paul Krugman offers the following defense of the usefulness of IS-LM:
In early 2009, when the WSJ, the Austrians, and the other usual suspects were screaming about soaring rates and runaway inflation, those who understood IS-LM were predicting that interest rates would stay low and that even a tripling of the monetary base would not be inflationary. Events since then have, as I see it, been a huge vindication for the IS-LM types — despite some headline inflation driven by commodity prices — and a huge failure for the soaring-rates-and-inflation crowd.
Yes, IS-LM simplifies things a lot, and can’t be taken as the final word. But it has done what good economic models are supposed to do: make sense of what we see, and make highly useful predictions about what would happen in unusual circumstances. Economists who understand IS-LM have done vastly better in tracking our current crisis than people who don’t.
So IS-LM helped Krugman to forecast that the monetary stimulus and large budget deficits of 2009 wouldn’t lead to soaring interest rates and runaway inflation. OK, that’s useful. But wait a minute. Take another look at Figure 2. In the IS-LM framework, if fiscal policy shifts the IS curve to the right, then it will raise GDP — in that sense, “stimulus works” — but it will also raise interest rates. Just the opposite of what Krugman predicted. Meanwhile, Ricardian equivalence would predict just what Krugman says he predicted using IS-LM (so I wonder whether Krugman’s successful predictions actually reflect a closet belief in Ricardian equivalence). According to Ricardian equivalence, deficit spending makes people save to pay future taxes, so interest rates stay put. What’s going on?
Well, it turns out that Krugman is using a modified version of IS-LM, shown in Figure 3.
As Krugman explains:
Why is the LM curve flat at zero? Because if the interest rate fell below zero, people would just hold cash instead of bonds. At the margin, then, money is just being held as a store of value, and changes in the money supply have no effect. This is, of course, the liquidity trap.
And IS-LM makes some predictions about what happens in the liquidity trap. Budget deficits shift IS to the right; in the liquidity trap that has no effect on the interest rate.
Well, all right then. But notice that Krugman isn’t exactly applying IS-LM. He’s importing into IS-LM an insight he got from just thinking about how bond markets work. The flat portion of the modified LM curve, moreover, is hard to interpret. The LM curve is supposed to represent combinations of r and Y that are consistent with a fixed money supply and equilibrium in the bond market. But if money is draining out of circulation into people’s mattresses, that’s not exactly an equilibrium.
Finally, one could believe the chart in Figure 3 and still expect the deficits of 2009-2012 to lead to soaring interest rates. It all depends on how far the IS curve moves. But IS-LM can’t really tell us that. It can’t tell us where the IS and LM curves are. And we can’t find out by looking at data either. Like ordinary supply and demand curves, IS and LM curves are not directly observable. But supply and demand curves are sometimes indirectly observable, because “instrumental variables” — things we can be pretty sure affect only one side of the market directly — sometimes come along and conduct natural experiments for us. For IS and LM, that doesn’t happen, because too many inter-related causes are tangled up together.
Figure 3: The IS-LM Model (Modified)
And did Krugman say something about inflation? Did he say that IS-LM helped him predict low inflation? I think he did. “Those who understood IS-LM were predicting” that monetary policy “would not be inflationary.” But the funny thing is that IS-LM says nothing about inflation. The price level isn’t anywhere in this model. This is still a sticky-price, or stuck-price, model, which rules out inflation and deflation by assumption. So how did Krugman really make the successful predictions he attributes to IS-LM? As Tyler Cowen suggested, “Perhaps Krugman’s successful predictions have come from a broader framework and not from Old Keynesianism per se?”
There are many other problems with IS-LM. First, it purports to show how “the” interest rate is determined, but there are a lot of interest rates in the economy, and they do not necessarily move together. In particular, long- and short-term interest rates can behave quite differently from one another. If, as seems plausible, savings and investment depend mainly on long-term interest rates, while money markets are driven mainly by short-term interest rates, it might not make sense to draw IS and LM in the same space. Second, is the interest rate on the vertical axis real or nominal? That is, should we take into account that current policy affects expected inflation? Third, IS-LM, like the Keynesian cross, has a strange relationship to time. It’s a short-run model — that’s why it can ignore the price level and allow GDP to be below its potential — but does it make sense to keep using the same IS-LM story for four years straight, or have we moved into the medium run or the long run by now? No one knows.
Finally, for all its vagueness, the one thing that IS-LM must predict if it’s of any use at all is that stimulus works. According to IS-LM, trillion-dollar deficits from 2009 to 2012 must have made the economy better. Better than what? Than it would have done without them. How would it have done without them? Well, of course, no one knows, and in that sense, “stimulus worked” is not a falsifiable claim. But mere extrapolation from long-term trends would suggest that moderate unemployment and growth is the usual state of the U.S. economy, and that it normally bounces back from recessions, so if we’d done nothing we’d probably be moving toward the long-run trend. Instead, we’ve seen anemic growth and high unemployment. The economists who now, ex post, claim that they know the economy would have been even worse without stimulus are making a strong and prima facie implausible claim. Since they did not predict, ex ante, that things would get nearly as bad as they did, why should we believe their confident, counterintuitive counterfactuals now? A likelier guess is that the U.S. economy would have stayed closer to the trend line without the government’s unusually large policy response, and if it is doing worse than usual now, that shows that stimulus doesn’t work, which reflects poorly on theories like IS-LM, which predict that it would.
The AS-AD Model
The last Keynesian model we’ll look at is called AS-AD, or “aggregate supply–aggregate demand.” Amitava Dutt traces the history of the AD-AS model here: it seems to have appeared first in a 1947 textbook and to have made its debut in the academic journals in O.H. Brownlee’s 1950 paper “The Theory of Employment and Stabilization Policy.” It is important chiefly because whenever economists talk about “aggregate demand,” typically this is the model they have in the back of their minds.
Even if your familiarity with economics is dim, you’ve probably heard of demand and supply. Typically, the demand and supply concepts are applied to one particular good, say, shoes. Demand curves slope down because it takes a lower price to induce buyers to purchase more shoes. Supply curves slope up because it takes a higher price to induce sellers to provide more shoes. Equilibrium occurs when the price is just at the right level to make the quantity demanded equal to the quantity supplied.
Well, the AS-AD model is superficially about the same, as shown in Figure 4.
Figure 4: The AS-AD Model
Instead of supply, demand, price, and quantity, here we have aggregate supply (AS), aggregate demand (AD1 and AD2), the price level (P), and GDP or income (Y). AD and AS are the demand and supply curves for everything — all goods and services, measured in inflation-adjusted dollars. In this way, the demand-and-supply chart that is the staple of microeconomics is converted into a macroeconomic model.
I have drawn two AD curves, AD1 and AD2, in order to illustrate the putative consequences of fiscal stimulus, which is often advertised as a way to boost the economy by increasing “aggregate demand.” Unlike IS-LM, the AS-AD model is not silent about prices. Maybe when Krugman was predicting low inflation, he was really using AS-AD.
Except, in the AS-AD model, if the government really can shift AD to the right (debatable), then it will boost the economy, raising GDP from Y1 to Y2, but it will also raise the price level. So stimulus works, but contra Krugman, only at the cost of inflation. OK, you can redraw the AS curve to avoid that prediction. You can redraw the curve in a lot of ways. That’s the trouble with these models. They’re easy to manipulate because they’re not well-grounded, either in empirical data or in rigorous theory that starts from individual maximizing behavior. To lapse into jargon for a moment, they have no “microfoundations.”
It is fairly easy to see why the demand curve for shoes slopes downward. A demand curve is defined with the help of a ceteris paribus, or “all else held equal,” condition. If the price of shoes rises and everything else stays the same, particularly consumers’ incomes, consumers will buy fewer shoes, using the pairs they have longer and keeping fewer in their wardrobes, saving their limited cash for other items. But if we’re talking about the price for everything, the ceteris paribus condition doesn’t make sense. We can’t say that prices rise and consumers’ incomes stay the same, because everything includes both what one buys and what one sells, so if the general price level rises, one’s income rises too. In that case, why should one buy less? And why should AD slope down?
Because — so the story goes — a higher price level reduces real money balances, that is, the real purchasing power of the money you have in the bank, to below your desired level. You start spending less and saving more, to bring real money balances back up to the level you like. Well, that’s plausible. But when people are in a lot of debt — and in case you haven’t noticed, a lot of people are right now — the opposite story applies. Prices rise, your income rises, and your real debts are reduced, so you consume more. The AS curve is derived in the same way, namely that lower real money balances make you work harder, and in the same way, debt turns the story on its head.
Even if we put the debt issue to one side by assuming your real money balances are positive, your reaction to a rise in prices depends on expectations. If you interpret a rise in prices as temporary or one-off, you should cut your consumption, either to save for when prices fall or to boost your real money balances. But if you interpret a rise in prices as an increase in inflation, you should probably do the opposite and spend more of your money now, while it’s more valuable, rather than letting its value deteriorate further. In this case, AD will slope upward.
So there isn’t really any good reason to think that AD slopes down and AS slopes up, and with that, the AS-AD model basically vanishes in a puff of smoke. Actually, is it really sensible to talk about “aggregate demand” at all?
When microeconomists talk about demand, they’re not just talking about how much people buy, they’re talking about a curve, which is really a continuous set of hypothetical statements of the form “if the price were P, quantity demanded would be Q.” How do they know this curve exists? Or in other words, how do they know this set of hypothetical statements is (roughly) true? Well, first of all, by introspection and deduction. Being human, we know that humans have many wants and rationally try to satisfy them as well as we can using scarce means, and from this we can deduce that people will buy less, ceteris paribus, as a good rises in price. Second, as mentioned above, we look for instrumental variables to conduct natural experiments for us.
No corresponding justification is available for the AD curve (or AS). We don’t know by reason that AD must slope down. As we saw, it depends on whether cash balances are positive or negative (and on their distribution), as well as on how a price rise affects expectations of future price trends. Nor do instrumental variables conduct natural experiments for us, because the macroeconomic time series is too short and the web of plausible causal links is too thick and tangled. If you like, we know aggregate quantity demanded: we know how much people actually buy. But we can’t claim to know that aggregate quantity demanded is a function of the price level, or if it is, what is its functional form.
The Confidence Fairy vs. Keynesianism
Probably no model as loose and vague as those presented above could be published in an academic journal today. Standards have risen. Publishable articles need the “microfoundations” I mentioned above — they need to start with the individual and build up to macroeconomic phenomena by clear deductive links — and/or they need to be testable in the data. Unfortunately, the Keynesian cross, IS-LM, and AS-AD models linger on in the undergraduate textbooks, though sometimes presented with skeptical commentary or pushed to the appendices of chapters. It’s hard to say why. One reason is that the better models which have come along (a) are often too technical for undergraduates to understand, and/or (b) don’t add up to a complete system the way the old Keynesian models sort of do, albeit superficially and rather spuriously. Perhaps there’s a certain lingering veneration for Keynes’s great name. And textbooks must be spokesmen for the field, so a certain inertia results, as everyone tries to summarize what everyone else thinks.
Don’t get me wrong: I think most economists who use these models to give policy advice more or less believe them. They wouldn’t write down such slippery and untestable models themselves. They know better. They’ve imbibed the higher standards. But the old Keynesian models get, so to speak, “grandfathered in” to the body of respectable economic theory. They’ve been part of the curriculum for decades, and still are, especially at the undergraduate level, so despite various disavowals by Nobel Prize–winning leaders of the field, top economists go on assuming they must be at least partly right. Their vagueness and lack of empirical traction protects them from positive refutation. How much credence should we give to this professional semi-consensus among economists? Remember: doctors bled patients for centuries in obeisance to the ancient theory of the four humors, a treatment we now know is worse than useless. Trust the experts only as far as they can convince you with data and reasoning.
The truth is that most of the real knowledge that macroeconomists have achieved in the past few decades has come at Keynesianism’s expense, by taking into account that people take the future into account. Krugman has coined a term which may serve as a concise mnemonic for these developments — he likes to write about the “confidence fairy.” Unfortunately, Krugman uses the term to mock the notion that the weakness of today’s economy has something to do with expectations for the future. But if we let Marx name “capitalism,” we may as well take Krugman’s snappy label for rational-expectations macroeconomics and run with it. Keynesian economics is myopic economics. It is economics for a world where people get their take-home pay and immediately spend most of it. It is economics for a world where people watch the government spending more money and don’t think about how the government is borrowing to pay for it and will have to raise taxes sometime to pay down its debts (or at least reduce the deficit). It is economics for a world where sellers go on mechanically charging the same prices, rather than cutting prices to get surplus goods out the door. Confidence-fairy economics recognizes that real people are not myopic. They think ahead, and that makes the Keynesian models break down.
The savings that supposedly disappear in the Keynesian cross model in reality go into bank vaults and then go back out as loans, so that money keeps circulating too.
Robert Skidelsky’s book Keynes: The Return of the Master, one of several books that try to spin the 2008 crisis as a vindication of Keynes, pulls a fast one by distinguishing the period 1951–1973, when Keynes was supposed to have been the “coach” for economic policymakers in the West, from the period after 1973, when the free-market Washington Consensus prevailed. He then cites various statistics about economic performance that favor the earlier period over the latter. But in the United States at least, the 1970s were the heyday of Keynesianism’s influence — in 1971, Richard Nixon, a Republican, said “We are all Keynesians now”— while the Eisenhower administration in the 1950s was not Keynesian, being committed to balanced budgets in spite of recessions which it saw as a cure for lingering wartime inflation. In academia, Keynesianism was dominant in the 1950s and waning in the late 1970s, but policymakers took some time to get the message. The Keynesian 1960s featured big fiscal and monetary stimulus that led into the stagflation of the 1970s. In Britain, Keynes’s native country, his legacy was decades of chronic economic weakness which saw Britain steadily lose its economic leadership, until Margaret Thatcher, a free marketer, halted the slide.
By the 1980s, the rout of Keynesianism was complete enough that even policies that could have been defended in Keynesian terms were not. Ronald Reagan and George W. Bush, like John F. Kennedy, were big-spending tax-cutters, but Reagan and Bush argued for their policies mainly in terms of their supply-side impacts, rather than defending them as Keynesian-style aggregate demand management. Nor did many macroeconomists offer Keynesian defenses of Reagan and Bush. This might have reflected the partisanship of Keynesian economists (i.e., they were Democrats who didn’t want to defend Republican presidents) but mostly I think it was in good faith and reflected the intellectual defeat of Keynesianism by the monetarism of Milton Friedman (among others) and the rational expectations of Robert Lucas and Thomas Sargent (among many others). Even the “New Keynesian” school that emerged in the 1980s, which provided new justifications for the Keynesian doctrine of sticky prices, was not so much about vindicating fiscal stimulus, or the old Keynesian models like IS-LM and AS-AD that lingered in undergraduate textbooks, but rather, sought to check an anti-Keynesian reaction that had gone too far, to the point that it would actually (for example) have discredited Milton Friedman’s explanation of the Great Depression as a consequence of contractionary monetary policy. (Friedman was Keynesian enough to think sticky prices were a factor in the macroeconomy.) But fiscal stimulus was decidedly out of favor before 2008, and for good reasons.
Figure 5: The Great Investment Bust
In troubled times, people sometimes turn to an old-time religion for solace. Since 2008, there has been something of an Old Keynesian revival, in which Krugman is the most influential figure. These advocates know they can’t compete on the plane of rigorous high theory and have turned in ad hoc fashion to a variety of common sense or behavioral explanations, or sometimes crude appeals to history like Skidelsky’s mentioned above, to make the case for stimulus policies such as those that have left the United States with (to repeat) 12 million unemployed and declining median incomes. What Marx once said of the two Napoleons who ruled France — “history repeats itself, the first time as tragedy, the second time farce” — summarizes the likely career of this second incarnation of Keynesianism.
The Long Run
Citizens and politicians should learn their lesson — that fiscal stimulus doesn’t work — and focus on dealing with long-run problems. Not only will fixing those problems benefit us in the long run, it’s probably the best thing we can do for the economy in the short run too. As shown in Figure 5, gross private domestic investment has recovered somewhat since the nadir of the Great Recession, but it's still far below its 2006 peak; since 2009 it has been lower in real terms than at any time between 1999 and 2008 (pending final 2012 data). Investment is crucial to our future: it’s the source of capital that can make workers more productive and raise living standards. But more investment would also create more demand for workers right now. Why aren’t firms investing? Low interest rates should encourage investment. Even more, record-high corporate profits should encourage investment. Superficially, it’s ironic that corporate profits are doing so well under a relatively anti-business presidential administration, but this is actually just what we should expect. Investment is low, therefore capitalists face little competition and can make more money. But what’s good for some corporate incumbents is bad for the rest of us. There’s some force — but not much — in the Keynesian argument that investment is low because demand is weak and more capacity isn’t needed. Investment is oriented toward the future, sometimes the rather far-off future, and presumably there would be plenty of long-term projects worth financing with today’s cheap money if firms felt confident that the economy will get back to normal and taxes and regulations won’t expropriate them ex post. The great investment slowdown is a complex whodunit, but the guiltiest-looking suspect is political risk.
To encourage investment, the government needs to find a way to credibly recommit to letting investors keep their money if they succeed. Mainly, that involves appeasing the confidence fairy by making it clear how the country will pay its bills. Generally speaking, markets allocate resources better than governments, so the more government spending is held down, the better the economy does. The 1990s, a rare oasis of fiscal discipline and (as a result) a booming economy, are the example to emulate. But for a given trajectory of spending, it’s better to know how it will be paid for, so that private sector actors can plan around it, than for the economy to be burdened with the uncertainty that deficits create. All this is fairly easy to see, once you take off the Keynesian spectacles.
Nathan Smith is a professor of economics and finance at Fresno Pacific University. He is the author of Principles of a Free Society; Complexity, Competition, and Growth; other articles; and blogs at Open Borders: The Case.
FURTHER READING: Smith also writes “A Face for the Faceless” and “What If Justice Demands Open Borders?” Tino Sanandaji explains “Why Keynesianism Works Better in Theory Than in Practice.” Arnold Kling discusses “16 Tons of Keynesian Economics.” James Pethokoukis contributes “New Evidence that Obama’s Keynesianism Has Had Little Positive Impact.” Nick Schulz says there is “No Such Thing As Aggregate Demand.”
Image by Dianna Ingram / Bergman Group