Saturday, May 5, 2012

What Is NGDP Targeting?

A complete layman's guide to NGDP targeting

When I began blogging this January, I wrote a "mission statement" of sorts in my first post. I promised to keep you the reader "up-to-date on developments in the academic side of economics, such as emerging debates and trail-breaking research." I also pledged to dutifully avoid dryness and to introduce and explain jargon when it would serve a useful end. There's a tension running between those two objectives, perhaps inevitable but worth vigilance nonetheless -- so today I will take one of the most important new "big ideas" in economics, and take it out of the academic cloudscape, and put it within reach of the layman.

This is NGDP targeting. I've discussed it in the past on this blog -- see here -- but all of my writings approach it as an academic idea, and suppose to some extent the reader's familiarity with basic concepts. I believe in this idea and think that it should be put in practice. To that end, I feel that the best thing I can do for the cause is explain the idea and its merits to a broader audience.

The basic idea is that the Federal Reserve should use its power to stabilize the economy such that the nominal gross domestic product grows along a stable and predictable path at all times.

There's a lot in that single sentence, and this post will help you understand (1) the background concepts, (2) the concept of NGDP targeting itself, and (3) the merits of the policy.

The background concepts

If you understand what the Fed is, what its goals are, how it sets and conducts monetary policy, how monetary policy affects the economy, and what nominal GDP is, skip this section.

The Federal Reserve (informally, "the Fed") is the central banking system of the United States, and the institution charged with setting and conducting the nation's monetary policy. Along with fiscal policy, determined by the U.S. Congress, monetary policy uses the supply of money to help stabilize the economy through the business cycle (the economy's pattern of "boom and bust"). A full description of the Fed and monetary policy goes beyond the scope of this post, but note that ever since the Humphrey-Hawkins Act of 1978, the Fed has had two objectives, what economists call the "dual mandate": maintaining the maximum sustainable level of employment and keeping prices stable. In practice, this means that the Fed has conducted policy aiming for an annual inflation, or increase in prices, of 2 percent; simultaneously, the Fed seeks to prevent large increases of unemployment, which often comes from a recession, or from unemployment falling so low that inflation rises -- a scenario which happens when a booming economy "overheats."

The Fed conducts monetary policy by targeting a desired level for the federal funds rate, the interest rate at which major banks lend to each other overnight to they can meet the level of bank reserves they are required by law to hold. The Fed does this by purchasing or selling Treasury bills, the debt of the U.S. federal government, on the open market -- when the Fed sells T-bills, it contracts the supply of bank reserves, making them more scarce, and thus banks will charge more to lend them out to other banks. The Fed does this because the fed funds rate acts as a "base rate" upon which other interest rates are determined by markets; when the fed funds rate goes up, the supply of bank reserves goes down, raising interest rates on mortgages, car loans, and all of the other forms of debt which exist in the modern American economy. When interest rates rise, lending, investment, spending and economic activity generally slows; therefore the fed funds rate is the tool with which the Fed prevents inflation and unemployment from rising above their desired levels.

Nominal gross domestic product, or NGDP, is the sum of all spending in the United States economy, measured in the dollars as you and I use them. More technically, NGDP is the market value of all final goods and services purchased in formal exchange. Another way to think about NGDP is as the sum of all the goods and services produced, counted in their quantity (what economists call "real GDP"), and then multiplied by their prices (the "price index" or "deflator"). In this way, NGDP combines a "real" variable, or one which measures something actually being produced, and a "nominal" variable, one which considers prices and not actual production. GDP accounting, like the Fed, is also complex and not the focus of this post, but it is important to understand NGDP as a concept.

What is NGDP targeting, and how would it work?

NGDP targeting is having the Fed determine a path along which NGDP would grow, and using its monetary policy tools to effect that end.

In essence, the Fed would pick a rate of growth which reflects the sum of its desired rate of inflation and the historical rate of growth in real output; in the United States, this level is in the neighborhood of 5 percent per annum -- 2 percent inflation, 3 percent real growth. This would mean that the Fed would change targets from the fed funds rate to NGDP.

In some discussions, an NGDP target implies an initial level of NGDP, beyond which it grows at a constant rate. In such a scenario, the Fed would seek to hit that initial level using its monetary policy toolkit and then proceed from there.

The Fed could target NGDP using a variety of tools. It may, in fact, not need to go any further than the tools it has -- buying or selling Treasury bills in the open market, which affects the economy through the regular transmission mechanism of monetary policy, as discussed in the second paragraph of background.

Some economists have also advocated for the Fed to create a market on which traders could buy and sell "NGDP futures," securities which are priced by what the market participants expect the rate of NGDP growth to be. Similar futures markets exist in real life: for oil, for agricultural commodities such as corn and rice, etc. By purchases or sales of futures in this market, the Fed could set expectations for NGDP growth -- when the futures market anticipated low NGDP growth, the Fed would buy securities and raise expectations up to the target level.

In economics, expectations tend to be self-confirming; if people expect the economy to grow at a certain pace, then they will spend, hire, borrow, etc. in line with that expectation, and meaning that their expectations have now become reality. Therefore an NGDP futures market would help the Fed not only setting expectations for NGDP growth, but through this expectations channel, making NGDP actually grow at that rate.

The virtues of NGDP targeting

So why are academic economists actively discussing the idea of NGDP targeting? What are the main arguments of those who say that such a policy would be better than what we have now?

I see several main arguments for NGDP targeting. First, it would prevent most recessions, which are caused by a decline in aggregate demand. Second, it makes the credibility of the Fed aligned with, and not against, the policy which will counteract recession. Third, it would better handle those recessions which are caused by shocks to real variables, for example, if the supply of oil fell swiftly. Fourth, it would better handle positive real shocks -- that is, those which increase aggregate supply, which might occur if productivity were to rise faster than expected. Fifth, it would avoid the onset of impotence which comes from the "zero lower bound," when the Fed can push its target for the federal funds rate no lower than zero, and it is "out of ammunition" in terms of its conventional policy options. Sixth, the stabilization of aggregate demand would lead to higher long-run real growth in that risk-averse firms could make marginal investments which expand their production. Let us consider each of these arguments in greater depth.

First, NGDP targeting would prevent most recessions. Most recessions are "nominal" -- that is, they are not caused by a sudden change in "real" variables, like the resources we have available to us or our ability to use them productively. Recessions occur because of declines in aggregate demand, the sum of individual decisions to spend money, or rather in declines in expectations of future demand. What this means is that if the Fed maintains stable expectations for future demand, then nominal recessions will not happen. When nominal recessions do not happen, neither do all of the high costs which come from recessions -- unemployment, lower output, government debt, etc.

Second, it makes the credibility of the Fed aligned with, and not against, the policy which will counteract recession. Economists speak of the "credibility" of a central bank because if it is strong, then the central bank has greater control over market expectations. A low-credibility central bank must "prove" it wants something, because if markets do not believe that something will happen, they will not make it so by aligning their expectations with policy. For example, if a low-credibility central bank wanted to end high inflation, it would have to "prove" this by pushing down demand so sharply that the economy went into recession and inflation "reset" at the desired lower rate. A high-credibility central bank would be able to merely adjust expectations higher or lower, without the recession -- this is the story of Paul Volcker in the early 1980s -- and thus having low credibility requires costly behavior on the part of the central bank. For this reason, central banks are often unwilling to act in ways that may jeopardize their credibility, even if doing so would be "better" for the economy. In this way, the Fed's inflation target is poor policy, because every time the Fed wishes to support aggregate demand in the face of recession, it risks sending inflation above target and thus damaging its credibility. Instead, NGDP targeting would be better policy because if real growth fell, and the economy went into recession, the Fed's "default" policy -- i.e. the one it would take to preserve its credibility -- would be to increase inflation to support aggregate demand. By changing the "default" option, the Fed's incentive becomes to protect the economy from unnecessary costs, and not to protect its credibility at the economy's expense.

Third, an NGDP target would better handle recessions caused by real shocks. Real shocks, or changes in the resources available to us or our productivity, are not a product of monetary policy. They happen because of geopolitical uncertainty (oil), or bad growing seasons (corn or rice), or other reasons which, given that Chairman of the Federal Reserve Ben Bernanke is "not quite God," the Fed cannot prevent. But monetary policy can handle the event of a real shock in good and bad ways, in better-planned and more poorly-planned ways -- just as you, facing a surprise leg injury, could do things that would help you heal more quickly (say, icing, stretching, and going to physical therapy) or less quickly (say, running a marathon). NGDP targeting offers the best policy response to a real shock because a real shock reduces aggregate supply, causing prices to rise and real output to fall -- and since NGDP targeting considers, in essence, prices multiplied by real output, real shocks do not change NGDP is demand stays stable. In effect, an NGDP target can differentiate between real and nominal shocks -- and it responds to both optimally. In comparison, a real shock is a "threat" to the credibility of an inflation target, because prices "want" to rise, and so enforcing an inflation target requires reducing real output, and plunging the economy into recession, at a time when it is already suffering the supply shock.

Fourth, an NGDP target would better handle positive real shocks. Real shocks, however, do not always reduce the supply of resources or productivity -- sometimes, they increase either of these. For example, the introduction of computers to workplaces increased the rate of productivity growth substantially during the late 1990s. This was most certainly a good thing, but it presents a problem for some forms of monetary policy, because it reduces inflation. If the central bank targets inflation, it would be forced to ease policy to raise the rate of inflation, even when the economy was already booming in real terms. That is dangerous because, combining the explanation in Reason 3 and this, inflation targeting tends to magnify the booms and bust (and not stabilize) the economy in the face of real shocks. An NGDP target, on the other hand, would respond to an increase in productivity optimally: because an increase in aggregate supply causes output to rise and prices to fall, NGDP is unchanged if demand is unchanged, so the NGDP target limits the extent of the boom and prevents it from generating asset-price bubbles. What happens, both for positive and negative real shocks, is that the balance of real and nominal growth within the NGDP target shifts -- but policy is not forced to respond in ways which run against the ultimate goal of economic stability.

Fifth, an NGDP target would avoid the zero lower bound problem. When the central bank sets an interest rate, it can run into a problem during recessions, because it cannot set the "nominal" interest rate below zero, even if that is "what the economy needs" to return to stability. This is the "zero lower bound" problem, and it is most likely to happen during deep recessions -- we're at the zero lower bound today. That means that monetary policy is made useless when it is most needed. It would be like taking away the lifeboats at the moment when the Titanic hit the iceberg, as economist Scott Sumner once vividly put it. The NGDP target resolves the zero lower bound problem because if a real recession happened, the central bank would increase the rate of inflation, which means that it would have effectively cut the "real" interest rate (i.e. the interest rate minus inflation) without pushing the "nominal" interest rate to the zero lower bound.

Sixth, if firms are "risk-averse" -- meaning that they will systematically prefer low-risk, low-reward scenarios to high-risk, high-reward ones -- then an NGDP target will lead to a higher long-run rate of real growth. (This is my own argument for NGDP targeting.) Economists believe that firms in the United States are risk-averse, and that's not a bad thing in itself, but it does imply that nominal instability has real costs in the long run. In other words, if the economy is constantly swinging from boom to bust in NGDP, then if I run a business, I won't invest as much as I would have if the economy grew with more stability, because I am afraid that if I invest (say, I open a new storefront) and a recession comes, then my investment will lose value. Now expand this consideration of one firm to the entire economy: when NGDP growth is unstable, firms make fewer investments -- factories buy fewer machines, merchants open fewer stores, companies hire and train fewer employees, etc. -- and what this means is that, in the long run, the economy grows more slowly because productivity increases more slowly. Alternatively, stable NGDP growth, which would be achieved by targeting NGDP, could increase the rate of long-run real growth because of the tendency of firms toward risk aversion when they invest.

Further reading on NGDP targeting, and sources

Bennett McCallum, "NGDP Targeting."
Nick Rowe on the blog "Worthwhile Canadian Initiative."
Scott Sumner, "Re-Targeting the Fed."
     (See also: Scott Sumner's blog, "The Money Illusion," and his report, "The Case for NGDP Targeting")
The Economist, "Changing target."
     (See also: The Economist's "Free Exchange" blog.)
Christina Romer, "Dear Ben: It's Time for Your Volcker Moment."
David Beckworth's blog, "Macro and Other Market Musings"
Matt O'Brien, "A Rebellion at the Federal Reserve?"
More resources here.


  1. Evan: Very good post.

    I like your sixth argument. Far too many economists ignore this possible relation between short run business cycle policy and long run growth.

    Vivek Dehejia and I once argued something similar:;col1

  2. Yep, Vivek's and my argument is the same as your sixth, except we find we don't need to assume risk-averse firms. The non-linearity of the short-side rule Q=min{Qs;Qd} does exactly the same thing as non-linearity (risk-aversion) in firms' utility function, when firms's sales are sometimes demand-constrained, as they will be in a sticky-price business cycle model.

  3. @NickRowe:

    Thanks. I will read this in full at some point -- just read the first few pages -- and I'm glad to know that I'm not the only one making this argument about the long run. Risk-aversion to me is the more intuitive assumption, but the risk of loss b/c of sticky prices for firms is a powerful argument for demand stability that I will keep in mind and invoke the next time we hear the recessions-are-cleansing Schumpeter line. Now that I think about it, in fact, has the question ever been put this way: so suppose that there are two pools of individuals, those who run single-person firms, and those who are looking to start single-person firms. When people move in between these two pools, they send signals. When someone moves from the "looking to start" pool to the "running" pool -- that is, they create a firm -- they send a signal +1 for a given industry. When someone moves from the "running" pool back into the "looking to start" pool -- that is, they close a firm -- they send a signal -1 for a given industry. Those in the "looking to start" pool receive all of these signals and interpret them. When they see an increasing signal in a given industry, those who move from "looking to start" into "running" are more proportionally more likely to enter that industry. Conversely, when they see a decreasing signal for a given industry, firm creation is less likely to happen in that industry. Now here, I think, is the kicker: those signals can be interpreted in three ways, (1) individual firm failure or success not representative of the industry, (2) industry failure or success, (3) macroeconomic failure or success. Signals 1 and 2 are valuable; signal 3 is noise. The more signal 3 you have relative to the amount of signals 1 and 2, the less efficient the market-entry and market-exit mechanism becomes.

    In other words, the "creative destruction" mechanism is not enhanced by a lot of signal 3; it matters why firms are being created or destroyed if new firms are to be created efficiently.

  4. I think it ate my comment. Trying again.

    Evan: your argument about signalling is new to me, but there's a lot of literature I haven't read. It's reminiscent of Lucas '72 signal extraction (where AD shocks will also cause a suboptimally small supply response to real/relative demand shocks), except applied to investment and growth.

    Your model does face the same problem as Lucas' however. If agents have contemporaneous information on aggregate shocks (they know if there's a recession on now), they ought to be able to subtract the noise from the signal and figure out how much is due to (3).

  5. Evan,

    Very excellent post, I think this is a terrific resource.

    It might be helpful to emphasize the practical advantages of level targeting versus growth rate targeting. Only level targeting returns the economy to it's pre-recession growth trend line, whereas growth rate targeting basically gets us what we have today - an agonizingly slow recovery. I guess this difference is only relevant in those situations where the central bank permits a collapse in NGDP, but that's the situation today, so I think it's worth mentioning.

  6. For signal processing, would it help if there were two industries that the firms could go into? An asymmetry in the magnitude of the shock on the two industries would cause more confusion in terms of allocation. Considering how opaque modern economies are, it shouldn't always be clear to every market participant how the shock is distributed over the industries. As a result, firms do not allocate themselves efficiently across industries. This would also get around the rational expectations of how much a macroeconomic shock affects the industry. Yes, on aggregate, the impact is predictable. However, due to the presence of many heterogeneous industries, the impact is not.

  7. NGDPLT (4.5%) reduces the size of govt. And this is a positive thing.

    1. It removes the Fed from decision making. It gives us Friedman's PC running the show.

    2. It gives us a CAP on growth. And like all caps, it gives us a pie that must be split up.

    3. The PIE focuses the private sector's attention like a laser on increases in public expenditures that eat up the PIE.

    4. So when RGDP is running 4.5% and inflation is 0%, a increase in pay for public employees = rates have to go up, less RGDP for private sector = veto public employee wage increase.

    5. So when inflation is 4.5% and RGDP is 0%, public employees STILL do not get a cost of living wage adjustment.


    Under such a target since 1998, the real effect would be that 22M Federal, State, and Local public employees would be earning 25%+ less than they earn today.

    We'd have nearly $7T less US debt and almost $500B a year in savings.

    The real effect is that to get wage increases public employees would have to fund those increases by making productivity gains... meaning firing public employees.

  8. I think the sixth point is interesting, but I don’t quite feel satisfied.

    Let’s say that firms project the joint distribution of expected RGDP, PGDP, nominal firm revenues, and nominal firm costs to their relevant planning horizon. There’s some sort of correlation or cointegration among these variables.

    You’re arguing that firms make decisions today to earn profits in the future based on a utility function so that they trade off expected profits and the expected variance of those profits based on their risk aversion. I think that’s reasonable (as an aside, you might enjoy Tyler Cowen’s book on risk and business cycles). If the firm has expectations that either RGDP or PGDP will be volatile over their relevant horizon, then through the correlations it will make their profits more volatile and so they would pursue a more prudent course of action. I’m with you so far on this.

    However, in a world of heterogeneous expectations for the distributions of PGDP and RGDP, there is no reason a priori to expect that just because one firm believes risk has increased that they all do. Rather, if we had an options market on NGDP, then we could calculate the implied volatility of expected NGDP from market participants (like the VIX and I would expect the two to be strongly correlated). You really care about aggregate NGDP volatility, not individual NGDP volatility expectations (presuming we could also construct implied RGDP and PGDP volatility expectation indices, then there would be an empirical question as to whether firms pay more attention to one or the other).

    Regardless, your argument is implicitly that NGDP targeting would keep an NGDP implied volatility index stable. I don’t think that’s the case theoretically, but I do agree that a stable NGDP implied volatility index would be a positive thing that would boost long-run growth.
    I don’t deny that the central bank could keep the level on target, but if they only made purchases of futures contracts, then they would not be able to prevent the effect of risk aversion impacting the option prices that make up the volatility index. E.g., if some investors believe that NGDP is going to fall, then they could buy puts on NGDP ( if you’re not familiar). Since the central bank would only control NGDP expectations and not NGDP itself, it still may turn out to be profitable to make these trades. If more investors go into this trade, then it will result in a higher implied volatility of NGDP, even if the expected level is held by the central bank. If implied volatility of NGDP rises, then it will have all the negative effects that you mention.

    Alternately, the speculators may bet on the relative split between PGDP and RGDP. Higher implied volatility of RGDP or PGDP individually could lead to negative effects even if NGDP expectations are contained.