Yes, there is a liquidity trap: a rejoinder to Tyler Cowen

I tend to think of Tyler Cowen as some kind of extraordinary economic sage: even though he isn’t active in mainstream research, he seems to know about every new development, and he can provide a capable survey of the literature on command. If I find myself disagreeing with him, my first instinct is to assume that I’ve gotten something wrong. (And since he’s one of the only reasons I have readers in the first place, I’m also a little biased.)

But his skeptical posts about the idea of a liquidity trap continue to baffle me. One representative example is this post, where he states:

Short-term interest rates being zero, and short-term interest rates being almost zero, are very different cases, especially for understanding nominal shocks and whether they can stimulate aggregate demand.  Unless short-term rates are literally the same as the rate on cash, asset swaps still can succeed.  And QEII isn’t be the same as simply switching the term maturity of the debt, as Krugman has suggested.  There will be nominal effects also.

In a sense, the first sentence is correct. As long as short-term rates are above zero—even if they’re slightly above zero—it is still technically possible for conventional monetary policy to do more. The problem comes when we think about the magnitude of its impact. Conventional open-market operations affect the economy by altering one key intertemporal price: the short-term real interest rate. (To be precise, they control the nominal interest rate, and since inflation doesn’t adjust one-for-one in the short term, this changes the real interest rate.) All else being equal, these operations will have an economic effect whose magnitude roughly corresponds to the change in the short-term real interest rate.

And if monetary policy has real effects through its impact on the real interest rate, you wouldn’t expect anything interesting to happen when the nominal rate is around zero. Suppose that inflation is currently 1.5%. Then a 0.05% nominal rate corresponds to a -1.45% real rate; a 0% nominal rate corresponds to -1.5%. Is there any reason to think that the change from -1.45% to -1.5% would make much more difference than the change from -1.4% to -1.45%, or from -1.35% to -1.4%? I doubt it. Yet Tyler’s claim that there is no “liquidity trap” seems to rest on the premise that as we approach the zero nominal bound, we can get a sizable economic impact from every last infinitesimal decline in interest rates—or, in mathematical terms, that the derivative of macroeconomic variables with respect to the real interest rate approaches infinity as the real interest rate hits exactly -1.5%. This just doesn’t seem plausible.

The only way that Tyler’s claim makes sense is if he has a different model for the real effects of monetary policy—maybe he thinks that the nominal interest rate itself, independent of its effect on the real interest rate, has real effects. In the lingo of macroeconomic models, this is equivalent to saying that consumption and real money balances are non-separable in the utility function. While this is surely true to some degree, the overwhelming consensus in macroeconomics is that it’s quantitatively insignificant: see Ireland (2004) for an empirical test. And even then, there’s unlikely to be any large effect from the last few basis points in the nominal rate, unless Tyler has some very strange specifications in mind.

To make this more concrete, let’s try to imagine the possible channel at work here. Nonseparability of consumption and money in the utility function usually means that base money is a complement to consumption. The idea is that you need liquidity to buy stuff, so that when the nominal interest rate is lower and it’s cheaper to hold your money in a liquid form (cash, checking accounts), you’ll buy more.

Since the Fed is paying interest on reserves, however, the nominal interest rate has virtually no effect on the costs of holding money in checking accounts with a 10% reserve requirement. This leaves us with cash. And that is where the proposed mechanism becomes really, really implausible. Are consumers going to change their spending habits in any appreciable way because it’s 0.2% cheaper per year to carry around cash? (So that if you carry around $1000 in your wallet, at the end of the year you’ll have saved $2?) I’m one of the most robotically economic-minded people I know, but this is such a miniscule effect that I’ve never even thought about it. Most purchases, particularly the ones that go down most in a recession (durable goods, residential and nonresidential investment) are not made with cash. And it is impossible to imagine that in a $14 trillion economy, a ($1 trillion)*(0.2%) = $2 billion decrease in the cost of holding cash will manifest itself as anything other than a rounding error in GDP.

So yes, conventional monetary policy can suffer from a liquidity trap. Nominal interest rates have only a minor economic impact on their own (even near 0%), and the inability to push real interest rates below a certain level is a legitimate constraint on monetary policy.

This doesn’t mean that all hope is lost: the Fed can still make a difference by shaping expectations of the future trajectory of nominal interest rates, or by making unconventional bond purchases so large that they trigger portfolio balance effects and drive down interest rates on longer-maturity assets. (You’ll hear plenty about this from me in the coming weeks.) Just don’t go around claiming that 0% isn’t a barrier—because sadly, it is.

Advertisements

28 Comments

Filed under macro

28 responses to “Yes, there is a liquidity trap: a rejoinder to Tyler Cowen

  1. What do you think about the fact that QE2 has caused bond yields to rise instead of fall — despite the fact the everyone keeps talking about QE2 as if the point is to keep rates low? It seems to me that the Fed asset purchase program has both increased inflation expectations due to the increase of the money supply and pushed China to the sidelines in the treasury market — because: why would China want to get into a bidding war against the Fed after the Fed has announced they are determined to buy treasuries?

    Yet the typical analysis I read is that the point of QE2 is to keep long term yields lower than they would be otherwise. Looking at the bond markets, that doesn’t appear to be the case — and I don’t understand why it should be.

  2. JTapp

    As Sumner and Beckworth have tirelessly pointed out for years now, interest rates are not a good measure of monetary policy. Fed would be better off targeting NGDP level, and influencing expectations of nominal income through a policy rule. Thinking only in terms of interest rates is a Keynesian error that M. Friedman “corrected” long ago.

  3. Scott Sumner

    Monetary policy doesn’t only work by affecting nominal (or real) interest rates, it also affects all sorts of other asset prices, such as nominal exchange rates, equity prices, commodity prices, and real estate prices.

    Yes, those prices can only be affected (at the zero bound) if the monetary injection is expected to be permanent. But even when not at the zero bound, monetary injections that are temporary have very little impact on asset prices or the broader economy.

    It’s wrong to think of interest rate targeting as being “conventional” monetary policy. If all you do is target interest rates, then the price level is indeterminate. Any truly conventional monetary policy also involves some sort of inflation/NGDP target, a la the Taylor Rule, at least implicitly. And of course NGDP targeting (especially level targeting) continues to work well at the zero bound. As Woodford showed, monetary policy is ALWAYS about the future expected path of policy–at the zero bound, and also not at the zero bound.

    BTW, I like your blog so far, despite my nit-picking.

    • Glad that you like the blog. And I’m certainly happy to be nit-picked (or worse!).

      Monetary policy doesn’t only work by affecting nominal (or real) interest rates, it also affects all sorts of other asset prices, such as nominal exchange rates, equity prices, commodity prices, and real estate prices.

      Almost all the asset price effects occur due to the change in real interest rates. Equity, for instance, is a claim on a corporation’s dividends; all else being equal, a corporation’s future (real) dividends will only be worth more now if the expected path of real interest rates is lower. The same is true of real estate, which is a claim on either the explicit or implicit rent value of a property, and commodity prices, which (if they are in contango) are pulled up by expected future real prices according to the real interest rate.

      There are some exceptions—if firms are leveraged with a lot of long-term nominal debt, then higher expected inflation will increase the value of the firms’ equity, even ignoring the impact on real interest rates. But in general, the “asset price channel” is just another manifestation of intertemporal substitution due to real interest rates.

      That said, in practical terms I think it can be valuable to think about asset prices as a distinct part of the “monetary transmission mechanism”, even if they are conceptually no different than reallocation of wage income across time due to a change in interest rates. This is due to my crude behavioralist intuition: I suspect that consumers are relatively quick to respond to changes in asset prices (which are set by professional traders and thus incorporate expected real interest rate changes immediately), but relatively slow to fully reoptimize their intertemporal consumption plan. This is a pretty sketchy argument on my part and deserves a lot more attention.

      I agree that nominal exchange rates are an important separate channel of monetary policy, though in theory they are also pinned down by the expected state-contingent path of the policy rate, except to the extent that there are portfolio balance effects. (Svensson’s “foolproof way” of escaping a liquidity trap always confused me. It seemed like just another way of expressing a certain policy commitment that could theoretically—though perhaps not as effectively—be expressed in a more direct manner.)

      It’s wrong to think of interest rate targeting as being “conventional” monetary policy. Any truly conventional monetary policy also involves some sort of inflation/NGDP target, a la the Taylor Rule, at least implicitly. And of course NGDP targeting (especially level targeting) continues to work well at the zero bound. As Woodford showed, monetary policy is ALWAYS about the future expected path of policy–at the zero bound, and also not at the zero bound.

      I absolutely agree that monetary policy is about the expected future path of interest rates, not just the current level. But I don’t this is the point that many critics of the “liquidity trap” are making. Tyler Cowen, for instance, has talked mainly about the fact that open market operations are still effective when the rate is “near” (but not at) zero, which is a very different point. Others seem to imply that QE is obviously meaningful, regardless of its ability to signal the expected future path of interest rates. (I, meanwhile, think that QE probably makes a difference, but only due to some very hazy and hard to understand effects on market expectations of the Fed’s policy rate.)

      Even so, there are some legitimate constraints posed by the zero lower bound. Suppose that we’ll be in a “liquidity trap” (i.e. even a zero nominal interest rate can’t equilibriate the market, and there is an infinite demand for money when the interest rate is 0%) for the next 2 years. After the liquidity trap is over, the Fed plans to return to a business-as-usual 2% a year inflation target. Then if you believe the New Keynesian Phillips Curve, it is impossible for the Fed to raise inflation above 2% in the meantime, no matter what it does. (BTW, I don’t literally “believe” in the NKPC, but I do think that something like it is true in this situation.) The only way to increase inflation now is to either (1) commit to holding interest rates at zero for longer than the discretionary optimum, even if your long-term inflation target is still 2%, or (2) increase your long-term inflation target. I’ve enjoyed your criticism of those who reflexively focus on the credibility problem (no central bank that has ever really tried to create inflationary expectations has failed), but I don’t think that it’s a complete nonissue either.

      One weakness of the liquidity trap argument is that the demand for money isn’t literally infinite at 0%, and at some point the Fed will push farther than the public’s demand for short term liquid assets can accommodate. (This was the content of Bernanke’s moderately flippant argument on page 14 of his paper on Japan.) In that case, the Fed can have some traction. But it would need to be on QE10 before that would start happening.

  4. David Beckworth

    Matt:

    The zero bound problem, the liquidity trap, and the like are only issues because “conventional” monetary policy uses the interest rate as the instrument. But monetary policy does not have to do so. It could target the base as suggested by Bennet McCallum. If it did and if it targeted the forecast such that it shapes nominal expectations appropriately this discussion would be moot.

    The zero bound was not a problem in 1933 when FDR talked up returning to the pre-crisis price level and then followed through by (1) devaluing the gold content of the dollar and (2) not sterilizing gold inflows. The only problem was the willingness to be aggressive enough. And note that FDR did not talk about shaping the expected path of interest rates. He talked about the expected path of the price level and by implication nominal spending.

    Again, let me refer you to the ever wise Nick Rowe on interest rate targeting as a social construct..

    • I think that if we look closely at the mechanism by which expectations of future nominal variables affect spending today, we’ll see that real interest rates have to be the intermediary.

      Suppose everyone is confident that at a certain point in the future (say, 10 years from now), the liquidity trap will have passed and real consumption will be pinned down almost exclusively by the supply side. Suppose further that our expectation of real consumption at that date is “C”. Holding “C” fixed, how can monetary policy affect the desire to consume today? Consumers will be substituting between any two adjacent periods according to the real interest rate, and thus their substitution between today and 10 years from now will also be governed by the real interest rate. (In reality, of course, there are credit constraints and so on that make this a lot more complicated.) If monetary policy is going to affect consumption today, it can only happen via some change in the expected path of real interest rates.

      The channel is easy enough to imagine: a target for nominal GDP implies a certain inflation rate to reach our expected level of real GDP 10 years from now, which lowers the real interest rate when we’re stuck at 0%. But why is a target for nominal GDP so special? Why can’t you just set an inflation target? Why can’t this all be done in the language of interest rates (i.e. you commit to following a Taylor rule with ha higher inflation target once you’re out of the liquidity trap)? It’s possible that a target for nominal GDP is more effective in some ways, but it doesn’t seem like there are any fundamental differences that make it better at escaping a liquidity trap specifically.

  5. David Beckworth

    Matt,

    No doubt there is an implied path for interest rates regardless of how you frame monetary policy and yes, interest rates are the all important intertemporal price. The point I was trying to make is that it seems to me the fear of running out of ammunition at the zero bound is a consequence of using an interest rate target.

    Inflation targeting is problematic for two reasons. First, it has no “memory” and there is no catch-up inflation if a target is missed one year. Thus, level targeting is better here. Second, it fails to distinguish between demand and supply shocks. For this reason, a price level target though better than an inflation target is still not as good as a NGDP level target.
    Some will say a Taylor Rule is more or less the same as a NGDP level target, but the difference is the measurement problems with Taylor Rule: which inflation measure, which output gap, and which neutral rate measure? A NGDP target is far easier to measure.

    Sorry for all the comments today.

    • “The point I was trying to make is that it seems to me the fear of running out of ammunition at the zero bound is a consequence of using an interest rate target.”

      I still don’t see why this is true. Even if you’re using another kind of target, the stance of monetary policy is still captured by the expected state-contingent path of nominal interest rate. (The advantage of using another target is that it might be a lot simpler to implement than a theoretically equivalent interest rate rule.) If the nominal interest rate can’t go below zero, that is a real constraint on monetary policy.

      It’s possible that you find other ways to implement expansionary policy—for instance, by committing to a higher-than-expected rate of inflation in the future, once the zero lower bound is no longer binding. But I don’t see any reason why stating this commitment in terms of another target like NGDP is any more effective, for the purposes of stimulating the economy now, than stating it in terms of interest rates.

      “Inflation targeting is problematic for two reasons. First, it has no “memory” and there is no catch-up inflation if a target is missed one year. Thus, level targeting is better here.”

      In my last comment, I used “inflation targeting” more as a shorthand for “some kind of rule that will reliably implement a certain rate of inflation over the medium term”, not as a specific implementation mechanism like short-term inflation or level targeting. This was definitely a bad choice of words on my part.

      I agree that you’ve stated a good reason why price level targeting is probably better than inflation targeting: it’s easier to credibly avoid significant long-term deviations from the target. In the data, however, there seems to be a lot of inflation persistence–for reasons that we can’t entirely explain–and this means that an strict price-level target might cause significant fluctuations in the stance of monetary policy, as the Fed would need to contract or expand significantly in order to bring inertia-driven inflation to the level consistent with its short-term price target. (There is the same problem for NGDP level targeting.) For this reason, the best policy is probably medium-term level targeting coupled with short-term inflation targeting.

  6. David Beckworth

    Matt:

    Okay, my last reply I promise.

    It is the path of real interest rates that matter, right? They are not bound by zero. So if monetary policy were not using an interest rate target and it successfully raised inflation forecasts (say via a price level target) such that the real rate was negative, why would nominal rate hitting zero matter? What would matter is if the monetary policy was hitting its non-interest rate target and if real rates had adjusted enough to restore full employment?

    My understanding is that inflation persistence in the data had declined in recent decades. If so, then inflation persistence can be explained by the monetary regime. The implication, then, is that with a credible price level target there might not be the inflation persistence problem.

    Let me close by again raising the other problem with inflation targeting: it treats all price changing shocks the same. A NGDP target responds only to shocks to nominal spending or aggregate demand. Here is a post I did on this point: http://macromarketmusings.blogspot.com/2011/02/inflation-targeting-gets-black-eye-but.html

  7. Scott Sumner

    Matt, Thanks for the thoughtful reply. I think our liquidity trap dispute relates to our dispute over the role on interest rates in monetary policy (during normal times), so I’ll focus on that issue.

    Start with a perfectly flexible price world. A one time permanent 10% boost in the money supply will boost all nominal magnitudes by 10%. leaving interest rates unchanged. Of course wages and prices ARE sticky, so that’s not a very interesting example. But I’ll argue it still plays a role in the transmission mechanism.

    Now go to a sticky wage/price world, and do the same 10% permanent boost in M. Now prices rise 10% in the long run, but in the short run nominal and real interest rates may fall, to equilibrate the money supply and demand until prices can adjust. So far this is in line with what you argue.

    I’d argue that in the sticky price world the expectation that prices will rise 10% on the long run will also cause all flexible prices to rise immediately. These include the price of foreign exchange, equities, commodities, and (to a lesser extent) real estate.

    In your view the change in the real interest rate is what drives all those other asset price changes. In my view changes in real interest rates are a sort of epiphenomenon–something that happens along with monetary shocks, but not an important driving force in the real economy. Instead I see:

    1. Currency depreciation directly boosting exports.
    2. Stock gains raising the Tobin q, which boosts corporate investment.
    3. Real estate prices rising relative to construction wages, which boosts construction.
    4. Higher commodity prices increase commodity production (in a closed econ model)
    5. There are also wealth effects on consumption, which I think are overrated in importance.

    Why don’t I think changes in real interest rates are as important as the other asset prices? Partly because they seem less important during the big monetary shocks, the ones easy to identify. And partly because theory suggests that in a ratex model, a powerful monetary stimulus surprise can so increase real GDP growth expectations that real interest rates actually rise on the news. If one simply assumes that low real rates mean money is easy, then one is likely to think monetary policy is fairly weak, as low real rates often persist for long periods with a weak economy.

    On January 3rd 2001 there was a big expansionary surprise by the Fed. Short term nominal rates fell a bit, due to the liquidity effect. Long term nominal rates rose sharply because of the Fisher effect. And even long term yields on TIPS rose slightly, as the increase in expected GDP growth overwhelmed the liquidity effect.

    And stocks soared about 5% in one day on the news. The stock boost was expansionary for RGDP, the slight rise in long term real rates obviously was not.

    QE2 strengthened my views on the liquidity trap. All the asset markets (except RE) responded just as I expected. Admittedly that doesn’t contradict your view that they were responding to a fall in real rates. (Unfortunately in this case the partly unrelated oil shock distorted TIPS spreads, but I’ll accept that real rates probably fell, unlike January 2001.)

    I won’t say any more, because unless you accept my view that the path of interest rates is not the right way to think about the path of monetary policy, nothing else I say will be persuasive. But that view of money is what motivates my view of the liquidity trap. There’s no zero upper bound on M, hence no trap. There is the theoretical possibility (outlined by Krugman) of an expectations trap (monetary injections viewed as temporary); as you noted I don’t see evidence for that problem holding back a determined central bank.

    Regarding David’s last comment, it reminded me of Nick Rowe’s very interesting posts on interest rate targeting as a sort of social convention. If the Fed can only “talk” interest rates, then once rates hit zero, the Fed becomes “mute.”

    • Dear Scott,

      Thanks for your very insightful reply. I agree that our debate is more fundamental and isn’t specific to discussion about a liquidity trap.

      I’d argue that in the sticky price world the expectation that prices will rise 10% on the long run will also cause all flexible prices to rise immediately. These include the price of foreign exchange, equities, commodities, and (to a lesser extent) real estate.

      Though I think there can be indirect effects (a change in monetary policy improves real growth expectations, which feeds back into asset prices), I’d argue that the direct effects here are happening almost entirely through a change in real interest rates.

      Suppose for simplicity that in period 0 there is a surprise 10% increase in the monetary base, but it doesn’t have an effect on sticky prices until period 1. Previously, everyone had expected the price level in both periods to be P (e.g. zero inflation). After the surprise, the price level in period 0 is still P, but in period 1 it will be 1.1*P. What is the direct effect on equity prices in period 0?

      Say that value of stock A’s real dividend stream from period 1 onward is V. Since we are considering only the direct effects of monetary policy, we will hold V constant in the analysis, and further assume that the company behind stock A earns nothing in period 0 (so that only the future stream matters). Before the surprise, then, the value of stock A is (P*V)/R, where R is the nominal interest rate. After the surprise, the value of stock A is (1.1*P*V)/R’, where R’ is the new nominal interest rate.

      The corresponding real interest rates are r = R and r’ = R’/1.1. Then we can write the period 0 nominal value of stock A before the surprise as (P*V)/r, and after the surprise as (P*V)/r’. In this case, the only change in stock A’s value comes through the change in real interest rates. The expectation of higher nominal prices in the future causes stock A’s period 0 nominal value to rise–but unless there is a change in real rates, the nominal rate at which stock A’s future streams are discounted will rise by an exactly offsetting amount. Everything must happen through the real rate.

      I said “almost entirely” rather than “entirely” in my second paragraph because there are a few exceptions: if a firm has nominal debt that matures in period 1 or later, then surprise inflation will erode the real value of the debt and increase the stock price in period 0. But here, unless there’s a debt overhang problem, the incentive to invest doesn’t increase unless the real interest rate changes: as long as equity gets 100% of the returns from an investment, and the real interest rate at which those returns are discounted is constant, the NPV of the project will be constant as well.

      In my view changes in real interest rates are a sort of epiphenomenon–something that happens along with monetary shocks, but not an important driving force in the real economy. Instead I see:

      1. Currency depreciation directly boosting exports.
      2. Stock gains raising the Tobin q, which boosts corporate investment.
      3. Real estate prices rising relative to construction wages, which boosts construction.
      4. Higher commodity prices increase commodity production (in a closed econ model)
      5. There are also wealth effects on consumption, which I think are overrated in importance.

      In cases 2-5, I am quite confident that real interest rates drive the direct effects of monetary policy. (Though again, there can be indirect effects–more on this later.) First of all, in this context,Tobin’s q is simply another way of stating that the cost of capital (the real interest rate) affects investment decisions. Bernanke and Gertler’s 1995 paper on the monetary transmission mechanism, for instance, referred to it as an “equivalent” formulation of the standard cost-of-capital story while citing several empirical papers that were skeptical about cost-of-capital effects:

      Further, in the relatively few cases in which evidence for a cost-of-capital effect is found, the nonneoclassical determinants of spending typically remain significant (for example, Boldin, 1994; Cummins, Hassett and Hubbard, 1994). Empirical studies that have tested the neoclassical model in its equivalent “Tobin’s q” formulation have generally been no more successful

      The same logic applies to real estate prices: they increase because a decline in real interest rates has made the future stream of rents (explicit or implicit) more valuable.

      Commodity prices increase if they are currently in contango because when the real interest rate is lower, future spot prices translate into higher prices today. (I don’t think that this necessarily encourages a positive production response, though. The mechanism through which higher present-value commodity futures prices cause current prices to rise is storage, as arbitrageurs make sure that the real price one period ahead exceeds the current price by no more than r + c, where ‘r’ is the real interest rate and ‘c’ is the storage cost. Since the cheapest form of storage is leaving stuff in the ground, you might expect production to decrease. On the other hand, if a producer knows that production contraints will be binding for the next several years anyway, this logic might not apply—so the response is ambiguous.)

      Finally, direct wealth effects also operate through the real interest rate: the only reason your assets become more valuable today is that the future income streams they represent are worth more in present value terms. If you are leveraged with nominal debt, then surprise inflation will add to your wealth through another channel, but in a closed economy the total amount of nominal debt is zero, so that there is no net wealth effect except possibly through a change in distribution.

      The January 3rd, 2001 example you mention is very interesting, but in light of the analysis above I think it’s best interpreted as evidence for “indirect” effects of monetary policy on asset prices. Here is one possible story. Ex ante, investors thought that there was a 30% chance of a severe recession and a 70% chance of business as usual. These expectations were partly self-fulfilling: precautionary liquidity hoarding, holding back on investment, and other behaviors drove down the equilibrium interest rate and raised the prospect of a recession.

      By loosening policy, the Fed signaled that it would take an aggressive approach to any recession that arose in the coming year. By breaking self-fulfilling expectations of a severe recession, this dramatically lowered the chance of a recession happening at all. Thus long-term real interest rates rose slightly: they reflected the expected path of short-term real interest rates, and in the absence of a recession this expected path was higher. But if the severe recession actually occurred and the Fed had to follow through on its promises, real interest rates would have been the mechanism through which it gained any traction.

      If you’ll pardon the analogy, Alan Greenspan was visibly brandishing a Kalashnikov in front of possible attackers. By scaring them off, he lowered the expected number of bullets fired from his Kalashnikov. But he did so by increasing the expected number of bullets fired in the event of an actual attack. Replace “bullets” with “low real interest rates” and “attack” with “recession” and you have what I think is the most plausible model of what happened.

  8. JTapp

    Scott and David, I’m hoping to convince at least one of the journalists who attend the Fed’s new press conferences to ask Bernake a question along these lines. I think Binyamin Appelbaum of the NY Times is game. I had the idea of taking up a collection via someone’s blog with the pot going to the journalist that would ask the question. If one of you is game for proposing something like that that on your blog…

  9. john.jansen

    In a former life I was a blogger and the first influential blogger to endorse me was Tyler Cowen. I guess he is a good judge of talent.

  10. Scott Sumner

    Matt, You said:

    “By loosening policy, the Fed signaled that it would take an aggressive approach to any recession that arose in the coming year. By breaking self-fulfilling expectations of a severe recession, this dramatically lowered the chance of a recession happening at all. Thus long-term real interest rates rose slightly: they reflected the expected path of short-term real interest rates, and in the absence of a recession this expected path was higher. But if the severe recession actually occurred and the Fed had to follow through on its promises, real interest rates would have been the mechanism through which it gained any traction.”

    I’d say that real interest rates would have fallen, because they fell in recessions even before there was a Fed.

    But let’s say you are right. This paragraph seems to conflict with your earlier remarks. Previously you said that a higher Tobin q is like a cut in real interest rates. I’d like to pursue the implications of that remark. It suggests that the Fed action lowered real interest rates in the stock market, but not the Treasury bond market. So far so good. In that case I’d argue that there is no zero lower bound for interest rates, because there is no zero upper bound for stock prices. The Fed can raise stock prices any time it wants through aggressive enough QE, and thus the Fed can always lower both real and nominal interest rates in the stock market (and real estate markets, etc.) There is no requirement that Treasury bond rates move at all.

    I will agree with you on one point; if you define real interest rates broadly to include real rates of return on stocks, real estate, etc, then my story does involve more of a real interest rate transmission mechanism than I suggested. But even then a strong real wage mechanism also applies, as commodity price increases and dollar depreciation lower the real wage rate (defined as nominal wages divided by the price of tradable goods and commodities.)

    I’m not sure I follow your stock example, as it seemed to contest my monetarist argument by relying on a sort of New Classical argument, where everything was neutral beginning in period one, when prices had risen 10%. And you assume no dividends in period zero, hence monetary stimulus has no effect on the stream of real dividends. That’s true, but it’s also a world where money is close to neutral. You are allowing a change in real interest rates, but no rise in real corporate profits (that I would expect to follow from monetary expansion.)

    My argument was that since money is non-neutral in the short run, monetary stimulus would raise output and thus real corporate profits for several years, and that even if real interest rates were unchanged, this would raise real stock prices. Indeed that was the point of my January 3rd 2001 example, that’s roughly what I think happened.

    I often argue that “that which has no practical implications, has no theoretical implications.” So if the question of what counts as “a change in real interest rates” is important, there must be practical implications. I’d argue those implications occur at the zero bound. Keynesians argue that the zero bound makes monetary policy ineffective, because the nominal rate can’t be cut. But if we are going to describe yields on houses and stocks at a sort of rate of return on capital, then obviously those nominal rates can be cut. There is no zero bound problem in those markets. Then the question becomes “Can monetary policy affect the prices of assets, without affecting real interest rates on Treasury securities?” And the answer is yes. So fiat money central banks are never “trapped.”

    Finally, here’s another argument for why monetary policy can work without an interest rate mechanism. Imagine a world without capital (physical or financial.) There are no interest rates. There are slightly sticky prices and very sticky nominal wages. (All firms make services.) A positive nominal shock lowers real wages, and increases output. By definition, real interest rates play no role, because the economy lacks real interest rates.

    In my view the main transmission mechanism for monetary policy is that a rise in M raises per capita NGDP relative to hourly nominal wages, and this encourages firms to utilize more hours of labor. That can occur with interest rates playing no role. A rise in M causes a rise in MV, and then hourly wages/MV falls. If wages were flexible you get the new Classical result. In the long run they are flexible, and you get only inflation.

    • But let’s say you are right. This paragraph seems to conflict with your earlier remarks. Previously you said that a higher Tobin q is like a cut in real interest rates. I’d like to pursue the implications of that remark. It suggests that the Fed action lowered real interest rates in the stock market, but not the Treasury bond market.

      Real interest rates are one channel through which Tobin’s q can change; I argued that they were the only direct way that monetary policy affects stock prices and the marginal return to capital investment. (And Tobin’s q is really just supposed to be a useful empirical proxy for the marginal return to capital investment, not a conceptually meaningful object on its own.) The purpose of my paragraph, however, was to say that there can also be indirect effects from monetary policy: by pledging to use the direct tool of low real interest rates in case of a recession, the Fed might be able to defeat self-fulfilling expectations and avert the recession entirely, thus raising the marginal return to capital investment via an indirect channel (shaping expectations to avoid a recession). To me, the important observation is that if the Fed is ever going to have this kind of indirect impact, there needs to be some channel through which it can have a direct effect on macroeconomic variables; otherwise no one would care what it did, and the indirect effects of its policy wouldn’t exist. That channel is the real interest rate.

      I’m not sure I follow your stock example, as it seemed to contest my monetarist argument by relying on a sort of New Classical argument, where everything was neutral beginning in period one, when prices had risen 10%. And you assume no dividends in period zero, hence monetary stimulus has no effect on the stream of real dividends. That’s true, but it’s also a world where money is close to neutral. You are allowing a change in real interest rates, but no rise in real corporate profits (that I would expect to follow from monetary expansion.)

      I agree that my toy model is unrealistic, and that it doesn’t capture the ways in which money is not neutral. My goal was to show why your intuition that monetary expansion should raise asset prices today implicitly assumed movement in the real interest rate; I think the example is very useful on that score.

      I think the key disagreement is about this: “You are allowing a change in real interest rates, but no rise in real corporate profits (that I would expect to follow from monetary expansion.)” Why would there be a rise in real corporate profits following monetary expansion? (Except through inflation eroding the real value of nominally denominated debts, which is an obvious channel, but one that doesn’t necessarily affect the real return on investment.)

      I agree that good monetary policy might cause an increase in future real corporate profits, and that this might be an indirect consequence of monetary policy that causes further feedback into asset prices, but for this to happen first we need some direct mechanism through which monetary policy affects real variables. My mechanism is the real interest rate.

      Then the question becomes “Can monetary policy affect the prices of assets, without affecting real interest rates on Treasury securities?” And the answer is yes. So fiat money central banks are never “trapped.”

      Sure, monetary policy can affect prices of assets without affecting real interest rates on Treasury securities—but not necessarily by much. It all depends on the magnitude of portfolio balance effects. Monetary policy can have an enormous impact on the market for base money because the market is so tiny (and the demand is so inelastic). Once it starts moving into other, much bigger markets, it has to acquire a far larger portfolio. The household sector in the US alone has $48 trillion in financial assets. And even then, there might not be a huge effect: Eggertsson and Woodford even proved an irrelevance result. In practice I don’t think that the assumptions necessary for this result quite hold, but they might not be so far from reality, and the intuition underlying the theorem is very important. It’s not clear that the Fed can make such a difference through further asset purchases, even if it expands its portfolio to $10 trillion.

      In my view the main transmission mechanism for monetary policy is that a rise in M raises per capita NGDP relative to hourly nominal wages, and this encourages firms to utilize more hours of labor. That can occur with interest rates playing no role.

      This sounds like the classic Keynesian sticky wage story. While I think that this may play some role, I’m not convinced that wage stickiness is so overwhelming that it’s the main issue facing monetary policymakers. Even so, you wouldn’t expect this transmission mechanism to deliver immediate results: monetary easing would only set in gradually, as the compounded effects of price inflation exceeding wage inflation became significant.

  11. Robert Waldmann

    I’d add a couple of things. One is that a subtle effect such as the cost of holding money affecting consumption is almost certainly tiny since there is basically no evidence that households total consumption level has anything to do with any price. In particular, econometricians have to work very hard, dig through lots of data and search over specifications in order to get non trivial correlation between the real interest rate and the rate of growth of consumption.

    The cross derivative of money and consumption in the utility function is not useful to those who wish to predict aggregate consumption, because the concept of a utility function is not useful to those who wish to predict aggregate consumption.

    The fact that most of macro-economic theory for decades has focused on intertemporal choice means that I must be saying that most of macroeconomic theory for decades is totally utterly completely useless. Yes of course that’s what I think. That’s what all people but macroeconomists think. The only odd thing is that I am a macroeconomist (obviously I can’t explain how that happened but let’s just say that some time ago I didn’t optimize correctly).

    But enough about reality. Turning to theory, money and consumption can, in theory, be substitutes. You don’t motivate the utility function at all. I will try an example, one in which money holdings are tiny.

    Consider a hyperinflation. Everyone has their wealth in interest paying assets and occasionally converts some to money then spends it as fast as they can. This means they don’t waste time comparison shopping or looking for exactly the good they want. This can cause higher spending. Here the idea is that you don’t need to hold money to consume, you need to hold money to shop.

    To add a really cheating twist, I note that some people find shopping entertaining (I’m thinking of women shopping for clothes). If they have to buy fast they get bored. So they might buy clothes and a movie ticket for the combined clothing and entertainment. This makes them spend more because the experience of shopping has become more expensive so they spend more on entertainment. Sadly for them, they have to pay the sales clerks to wait around so that they can buy things when they run into the store. I just assume that retailers don’t have lower costs because there aren’t people who are just looking.

    • I’d add a couple of things. One is that a subtle effect such as the cost of holding money affecting consumption is almost certainly tiny since there is basically no evidence that households total consumption level has anything to do with any price. In particular, econometricians have to work very hard, dig through lots of data and search over specifications in order to get non trivial correlation between the real interest rate and the rate of growth of consumption.

      Of course, I agree that household optimization does not happen in the instant, frictionless, and rational way that macroeconomists model it. Ordinary people don’t seem to be very knowledgeable about real interest rates. There are, however, some transmission mechanisms that operate more effectively: for instance, asset markets, in which prices are set by traders whose job is to instantly respond to changes in anticipated real interest rates and other relevant macroeconomic variables. I’m not saying that most of these traders are very smart, but judging by the instant and dramatic effects of monetary policy announcements, they certainly do reoptimize quickly. And if consumers are quick to respond to changes in their wealth holdings (particularly if they’re about to buy an expensive durable good, or a house), adjustment may happen faster than we think.

      As I mentioned to Scott above, asset prices aren’t conceptually a different transmission mechanism: they’re just a statement of the present discounted value of future income streams, which changes as the real interest rate changes. But in practical terms I think that they may be very important, as they readjust instantly to variables that households would otherwise never even consider (or, at best, consider with an extreme lag).

      As far as the empirical evidence on the relationship between interest rates and consumption is concerned, I think that the literature suffers from severe endogeneity problems. (Is it even possible to have a truly exogneous shock to real interest rates? Enough to identify consumers’ response? I doubt it.)

      The cross derivative of money and consumption in the utility function is not useful to those who wish to predict aggregate consumption, because the concept of a utility function is not useful to those who wish to predict aggregate consumption.

      Especially in this case, I think it’s best to view the utility function as a convenient metaphor, and mathematical statements like “the cross derivative of money and consumption is small” as a precise and parsimonious way of making a claim that would take paragraphs if expressed verbally. In fact, I spent a good portion of this post trying to discuss the underlying mechanism here in intuitive terms, to make sure that academic statements about the unimportance of complementarity in the utility function were backed up by a reasonable assessment of how consumers actually behave.

      • Robert Waldmann

        I think that this is not a purely verbal quibble. You wrote
        ” I’m not saying that most of these traders are very smart, but judging by the instant and dramatic effects of monetary policy announcements, they certainly do reoptimize quickly.”

        I agree that they do something quickly, but I would say that, if they are not moving from one optimum to another, they are not reoptimizing — that is they have to have rational expectations (not just be very very smart but the full deal) to optimize or reoptimize.

        My statement about consumption wasn’t about the speed of change. I think that models of optimizing agents are as likely as not to get the sign of changes wrong.

        My concern is about the widespread methodogical approach based on saying that people do not optimize (which requires among other things optimal forecasts that is rational expectations) but models of optimizing agents *might* be useful. Then in the next step assuming full rationality and going from there. I think the only possible justification for such an approach would be notable empirical success in predicting out of sample. I think that macroeconomists agree that we haven’t achieved that.

        How many decades more are we going to work on a project where the strongest claim we can make is that it might not be a total waste of time. ?

        I note in passing that I have written macro theory and what I wrote assumed and absolutely relied upon rational forward looking agents. Three published articles. Absolutely zero effect on my beliefs about the economy (especially not the paper published in the Reveiw of Economic Studies).

        I have come to agree with my PhD supervisor who in ’88 or so told me (and a mutual friend) something which I quote from memory

        “You know my view. I think that in the history of economic theory, the day when they came up with the idea of the utility function wasn’t an particularly good day.” — Larry Summers

        I think it is very very clear that subsequent experience hasn’t caused him to change his mind.

  12. Pingback: TheMoneyIllusion » Other activities (bumped)

  13. Scott Sumner

    I don’t think we are going to be able to reach agreement, as our vision of macro seems radically different. I see a two stage process. Monetary policy raises NGDP through a mechanism (excess cash balance effect) that has nothing to do with interest rates. I can’t even imagine someone telling the story of German hyperinflation using an interest rate mechanism. (That’s why Keynesian Joan Robinson denied monetary policy was responsible–after all interest rates weren’t low.) I tell the German hyperinflation story through the monetary excess cash balance mechanism. The hot potato effect–which is 100% consistent with micro theory. You have more of something, its value declines.

    Second of all, the story of why unexpected NGDP shocks impact RGDP in the short run (but not the long run) has nothing to do with interest rates. Sticky wages and prices are a necessary and sufficient condition for nominal shocks to have real effects. If someone wants to argue that easy money reduces real rates on short term Treasuries, and that (despite the fact that long term rates don’t fall) this somehow boosts GDP, then that’s fine. But it’s like a fifth wheel–we already have the transmission between nominal shocks and real effects from sticky wages and prices. And I see little empirical evidence that low rates boost GDP growth.

    So when you claim I need a real interest rate transmission mechanism, I don’t quite know what claim you are making. Are you saying I can’t go from monetary shocks to higher NGDP without a real interest rate mechanism? That’s obviously false, as the price level would adjust instantly (via excess cash balances) in a world without wage/price stickiness. Adding stickiness slows down the nominal adjustment, but certainly doesn’t magically eliminate it. Prices are still flexible in the long run, and the expectation of higher long run prices will immediately increase current nominal prices, if interest rates are unchanged. Interest rates play no necessary role in the rise in NGDP.

    And again, going from NGDP to RGDP only requires sticky wages and prices, interest rates need play no causal role. So I don’t understand your claim. Do we need an interest rate mechanism for NGDP? RGDP? Or both?

    • JTapp

      Scott’s response reminds me of the Bernanke speech in 2003 honoring Milton Friedman and listing Bernanke’s translation of Friedman’s 11 tenets of monetarism. Only the last one deals exclusively with thinking about interest rates, it’s almost tacked on as an afterthought. #10: “Monetary expansion works by affecting prices of all assets, not just the short-term interest rate.”
      Bernanke goes on to say:
      “The idea behind quantitative easing is that increases in the money stock will raise asset prices and stimulate the economy, even after the point that the short-term nominal interest rate has reached zero. There is some evidence that quantitative easing has beneficial effects (including evidence drawn from the Great Depression by Chris Hanes and others), but the magnitude of these effects remains an open and hotly debated question. “

    • So when you claim I need a real interest rate transmission mechanism, I don’t quite know what claim you are making. Are you saying I can’t go from monetary shocks to higher NGDP without a real interest rate mechanism? That’s obviously false, as the price level would adjust instantly (via excess cash balances) in a world without wage/price stickiness. Adding stickiness slows down the nominal adjustment, but certainly doesn’t magically eliminate it. Prices are still flexible in the long run, and the expectation of higher long run prices will immediately increase current nominal prices, if interest rates are unchanged. Interest rates play no necessary role in the rise in NGDP.

      And again, going from NGDP to RGDP only requires sticky wages and prices, interest rates need play no causal role. So I don’t understand your claim. Do we need an interest rate mechanism for NGDP? RGDP? Or both?

      This all seems a little circular. If prices are sticky, then why does a monetary shock have an immediate effect on NGDP? You say that “stickiness slows down the nominal adjustment, but certainly doesn’t magically eliminate it”; yes, but by the same token, stickiness is necessary to get real effects from monetary policy in the first place. I don’t see why these two (opposing) effects wouldn’t cancel out.

      In fact, it is easy to construct a world where (almost) everything you say is true, but monetary policy has no real impact. Suppose that RGDP is fixed exogenously (for whatever reason), and that the price level is pinned down via the standard quantity-theoretic device, modulo some stickiness. Then monetary shocks will give rise to higher NGDP at a rate determined by the level of price stickiness, but RGDP never changes. I don’t think that this is a good description of the world, but I don’t see how, exactly, your arguments rule out this kind of world.

      Prices are still flexible in the long run, and the expectation of higher long run prices will immediately increase current nominal prices, if interest rates are unchanged. Interest rates play no necessary role in the rise in NGDP.

      As I tried to illustrate with the numerical example earlier in this comment thread, an increase in the future price level (and thus NGDP) does not necessarily imply an increase in asset prices today. In fact, holding all else equal, it doesn’t cause any change in asset prices unless there is some change in real interest rates. (The proof: holding future real dividends constant, and the real rate at which they are discounted constant, the PDV of a dividend stream does not change.)

      You might say that it’s wrong to hold future real dividends constant, and I’d agree if we were trying to articulate a full model with every conceivable feedback. But here we’re just trying to reason about why monetary policy will have any real effect in the first place, and it’s circular to say that “monetary policy will lead to higher real dividends” and, ergo, it will have a real effect today through higher asset prices. (You’re assuming the existence of a real effect in order to prove it.)

      Btw, I agree that to the extent that corporations are leveraged with long-maturity nominal debt, a shock that increases inflationary expectations will lead to an increase in future real dividends and therefore an instant increase in asset prices, but (1) this seems like a much more specific and less ambitious channel than the one you’re backing, (2) unless firms are suffering from debt overhang, this doesn’t lead to any increase in their incentive to invest, and (3) there is an offsetting decline in the balance sheets of bondholders, so that there is no net increase in real wealth.

  14. Pingback: Liquidity Traps Are Very Unlikely, Even at the Zero Bound: A Rejoinder to Matt Rognlie : Invest My Money

  15. Scott Sumner

    Matt, I would certainly agree that in a world where many prices are sticky, a positive monetary shock that raises future expected NGDP will affect the real interest on at least some other assets (although not necessarily on Treasury securities, which often behave differently from other assets.) But I still don’t think the real interest rate on Treasury securities is necessarily all that important in the transmission process.

    For the sake of argument, assume prices were completely flexible, wages were sticky, and velocity was constant. A monetary shock immediately raises NGDP in proportion, and because nominal wages are sticky, real GDP also rises in the short run. Because cash balances/NGDP don’t change, there would seem to be no reason for interest rates to fall (Unless I’ve forgotten something, which is quite possible.) In this scenario wage stickiness does all the “real” work, and the excess cash balance mechanism does all the nominal work. Interest rates don’t play any causal role.

    Of course in the real world prices are sticky, and velocity does change. But to me that just means that the nominal transmission mechanism is more complex than just the monetarist excess cash balance mechanism, and it means that sticky wages are not the only factor turning nominal shocks into business cycles. Price stickiness also matters. But none of this means the real interest rate on Treasury securities is important, and I see little empirical evidence that it is important. It could well be that all the work is done by changes in riskier asset prices, exchange rates, NGDP expectations, combined with wage and price stickiness.

    Krugman’s always saying that empirical data supports his worldview. It seems to me that in both the Great Depression and the Great Recession, the interest rates on Treasury securities were a very unreliable indicator of the stance of monetary policy, and all sorts of other asset prices were much more informative. I’m a pragmatist, so that’s what drives me away from the new Keynesian transmission mechanism. It may be right at some theoretical level, but I can see a way for it to give me useful insights into the macroeconomy.

  16. Pingback: 「「流動性の罠」は存在する: Cowenへの返答」 by Matt Rognlie – 道草

  17. Pingback: 「政策短期金利がゼロ%の下限に達したとしても「流動性の罠」に陥る可能性は極めて小さい: Rognlieへの返答」 by David Beckworth – 道草

  18. Pingback: How did small losses lead to the Great Recession? | Matt Rognlie

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s