Saturday, October 29, 2011

Open Market Operations and Non-Neutralities of Money

Matt Rognlie and I are having a conversation in the comment thread of this previous post, which I'm sure most of you have lost track of. Here's a summary of the basic issues: One of my complaints with New Keynesian economics is that it skirts around most of what is interesting for me about monetary policy and how it works. In mainstream monetary models, e.g. standard representative agent models with cash-in-advance constraints, non-neutralities of money are restricted to the effects of unanticipated money and inflation. Monetary policy matters due to distortions in intertemporal prices, for example the anticipation of higher money growth and higher inflation acts as a tax on labor supply and reduces output. Further, the nominal interest rate increases due to a Fisher effect. Mike Woodford looked at those effects and thought that they did not matter much in practice, or that they had the wrong signs, and he wrote down models where he could dispense with those types of intertemporal distortions entirely. In basic New Keynesian models we do not worry about the details of monetary exchange, it is assumed that the central bank can choose the short-term nominal interest rate at will, and monetary policy has real effects because of relative price distortions due to sticky prices and wages.

My contention is that one cannot analyze monetary policy without modeling the role of central bank liabilities and other assets in exchange, and the role of the central bank as a financial intermediary. This need not involve substituting for New Keynesian-type effects. One can easily take the approach of being explicit about exchange, the central bank balance sheet, and central bank intermediation activity, and include the sticky prices and wages if one really can't live without them.

In this paper, one of the results I get is a particular non-neutrality of money. Prices are flexible, so it's certainly not a New Keynesian effect, and it's different from what you get in mainstream monetary models. There are essentially two classes of assets - currency and various other assets (government bonds, loans) which may be fundamentally illiquid but are made liquid (though not as liquid as currency) by financial intermediaries. A standard open market purchase (think of this as normal times) will ultimately increase the stock of currency in nominal terms, with no change in the real stock of currency, but the real stock of other assets declines, those assets become more scarce, the real interest rate falls, and lending increases. Essentially, this is an illiquidity effect.

In reply to Matt's last set of comments:

1. [Here he's discussing the effect of the open market operation]
But ultimately I have severe doubts that this channel makes much of a quantitative difference. When the Fed adjusts policy through open market operations, over the short to medium term it's making purchases in the tens of billions of dollars; maybe $100 billion at the very most. Meanwhile, the MZM money stock is $10 trillion, and that's an underestimate of the true size of the universe of liquid assets. Fiscal shocks happen all the time that adjust the quantity of liquid government debt by much more than Fed operations normally do; if you're positing that this an important channel for the effects of Fed policy, it follows that the Fed is at most a minor sideshow next to the Treasury. That doesn't ring empirically true to me.

First, in my model, there is not an increase in the yield spread "between government debt and other liquid securities." To keep things simple, I put assets into two classes. In the second class there is everything that is not currency, and I assumed that all that stuff (government interest-bearing debt and loans) could be intermediated in the same way. In a more elaborate model, one might imagine assets with different degrees of liquidity, but liquidity will be priced according to how assets are intermediated. For example, we might think of a house as highly illiquid, but a mortgage-backed security (MBS) can be highly-liquid, and the MBS is essentially backed by the houses that act as collateral for the mortgage debt that gets chopped up and put into the MBS.

Second, Matt has hit on something interesting in the latter part of the above paragraph, relating to fiscal policy. The Treasury could indeed be more important than the Fed, as it can bring about changes in the total quantity of consolidated government debt outstanding; the Fed can only change the composition. In fact, under current circumstances, the changes in the composition of outstanding debt the Fed can accomplish are irrelevant. Matt seems to think that these things don't "ring empirically true." I say run with the idea.

Matt goes on to discuss how New Keynesian effects work, and how he thinks they are empirically more relevant than what I'm after. You can read the details in the comment thread in the previous post. Two comments:

1. In terms of current events, my model might tell you that our current problem is that liquid assets (the second class of assets - the intermediated non-currency assets) are too scarce, and the real interest rate is too low. New Keynesians tell us the real rate is too high. If we take the New Keynesian line, we have to take a stand on what the "natural" real rate of interest is. That would be the real rate if wages and prices were perfectly flexible. To determine what that rate is we have to determine what the shock was that was driving the recession (and the financial crisis) presumably. I'm not sure what the New Keynesians have in mind there. Also, at first glance, real rates (based on current inflation, current short nominal rates, TIPS yields) look pretty low to me.

2. Some of Matt's arguments are in terms of back-of-the-envelope reasoning about the quantities of government debt and currency relative to other liquid assets. But we know that, through the shadow banking sector, a small quantity of assets, used as collateral, can support a very large quantity of credit activity. This is part of what Gary Gorton has written about. One might not think that things going haywire with a relatively small quantity of mortgage debt could cause such a big problem, but it did. Similarly, small changes in the quantity of interest-bearing government debt outstanding, through the process of rehypothecation, can give potentially very large effects in asset markets. Collateral and rehypothecation are not in my model, but if one were to take it to the data, that might be part of what one would want to include.


  1. I don't know if this is the best place to add this, but I was looking at your August 19th post on liquidity traps, and you write, "The nominal interest rate is essentially a measure of the scarcity of money as a medium of exchange." And you talk about this a lot, scarcity of liquid assets, scarcity of liquid assets, we need more liquid assets,...

    But the nominal interest rate, like any price, is determined by two things, not just one, not just supply, but also demand. It's determined not just by "the scarcity of money as a medium of exchange." But also the demand for holding money. And, in addition, the supply of money is affected by its velocity, like the supply (in the market) of a natural resource is affected by how fast it gets recycled.

    So, you then write, ""Neither can we correct it through "quantitative easing." We cannot ease anything through swaps of reserves for long-maturity debt, as that cannot make reserves relatively less scarce under the current circumstances."

    But if QE lowers long-term interest rates, and as a result increases the velocity of money, then isn't that essentially equivalent to increasing the supply of outstanding money, just like an increase in the recycling rate of steel would be essentially equal to an increase in the supply of steel?

    "How can government action mitigate the liquidity scarcity?" Increase the velocity (recycling) of the current amount of liquid assets?

  2. Not sure what your question is.

    "But if QE lowers long-term interest rates..."

    But it does not.

  3. Applying Serlin's "logic" to another situation:

    "If my aunt had balls, she'd be my uncle."

  4. Steve, thanks for the thoughtful comments. I'll probably post a response on my blog at some point in the near future, but I have a few other posts I've been putting off and might try to finish beforehand (not to mention a lot of other work!). In the meantime...

    Monetary policy matters due to distortions in intertemporal prices, for example the anticipation of higher money growth and higher inflation acts as a tax on labor supply and reduces output.

    Evidently we have very different views about what makes monetary policy interesting!

    I've never thought that this was a very plausible channel, at least in the form it's presented in classical models. If we try to interpret the world following these models, higher inflation acts as a tax on labor supply mainly because it makes paper currency (the dominant form of non interest-paying base money) more costly, which reduces the efficiency of consumption and therefore makes labor less valuable. But I just don't think that paper currency is that important. (This was the key theme in my New Keynesian vs. New Monetarist effects post.) The cost of inflation is a trivial consideration for virtually all cash transactions, and most (licit) transactions could easily be conducted using another payment mechanism if inflation really was such a serious cost. Most estimates of the cost of departing from the Friedman rule are small, and even those are vastly exaggerated because they are usually calibrated by assuming that all paper currency is actually used for transactional purposes, when in practice the vast majority is surely held for other reasons. (If we accounted for this reality, even papers like Lagos-Wright that add a wrinkle to the traditional welfare analysis and arrive at higher estimates would have to be adjusted massively downward.)

    I do, however, think that there are significant welfare costs from the interaction between inflation and a poorly designed tax system, in which the rules for capital gains, depreciation, amortization, inventories, etc. fail to adjust for inflation. In practice this is a much more important channel than the transaction-frictions channels of traditional models. is assumed that the central bank can choose the short-term nominal interest rate at will

    This assumption seems to be true in practice, and it can be justified in any model where prices are even the slightest bit sticky (adjusting every millisecond?) and substitution between base money and other forms of liquidity is responsive to the yield spread. In fact, the central bank has remarkable power over the real interest rate as well, since inflation expectations don't adjust one-for-one with the short-term nominal rate, and in fact generally go the other direction: all else equal (including long-term nominal expectations), a temporarily lower nominal interest rate will lead to higher inflation.

    I'm more worried about models that are at odds with this empirical reality than models that leave the mechanics of monetary policy implicit. Although your model does provide a way for conventional monetary policy to affect the real interest rate---by changing the stock of Treasury securities via open market purchases---this implies that the Treasury has vastly more control over the real interest rate than the Fed, even at high frequencies. As I said before, this doesn't ring true empirically, and it is also an extremely fragile result. For instance, if the Fed keeps excess reserves in the system and sets monetary policy via IOR, your proposed channel for the Fed to affect real interest rates disappears, since the Fed is no longer implementing monetary policy via asset swaps. Will the Fed really lose its power to set real interest rates in this environment? My guess is that the Fed will have just as much power as before. (And that's what you get from New Keynesian models.)

  5. ...and monetary policy has real effects because of relative price distortions due to sticky prices and wages

    Virtually everyone would agree with this statement, including New Keynesians. Even if the statement is technically correct, however, I think that the intuition is problematic. It's true that in a world without sticky prices and wages, New Keynesian effects are impossible; but that's because nominal quantities are irrelevant in this world, and any attempt by the Fed to set real interest rates at a non-equilibrium level will immediately result in infinities. But if we consider the much more realistic example of the perfectly flexible limit---in which prices are still slightly sticky---we get a result that's very different in flavor. In this world, if the Fed sets the "wrong" policy rate for a few months, then greater price flexibility only amplifies its mistake: if the rate's too high, then deflation will increase the gap between the real rate and its natural level.

    In particular, the conventional quadratic loss function in New Keynesian models has both a y^2 term for the output gap and a pi^2 term for the inflation rate. The pi^2 term reflects relative price distortions across different varieties of goods, while the y^2 term reflects some kind of wedge between MUL and MPL*MUC, which in the simplest model corresponds to unusually large or small markups (these models adopting the convenient fiction that the distortion from markups in the steady state is eradicated through an appropriate subsidy).

    In the New Keynesian framework, such unnatural markups are not possible when prices are perfectly flexible, but when prices are even slightly sticky, the output gap (and corresponding welfare cost) that results from a particular Fed "mistake" is actually increasing in the level of price flexibility. In fact, even the welfare cost from inflation in such an episode increases with price flexibility. Conditional on a particular output gap, more flexible prices mean an increase in inflation and a decrease in the welfare cost of inflation that more-or-less cancel each other out, but an increase in the output gap means even more inflation and therefore even greater welfare costs.

    Whether or not you think the thought experiment of a "policy rule mistake" by the Fed is realistic, I think that it's a useful rejoinder to the nearly universal (but incorrect) intuition that the size of the New Keynesian effect is decreasing in the level of price flexibility. The notion that this effect depends on "relative price distortions from sticky prices and wages" is true in one sense and false in another.

    First, in my model, there is not an increase in the yield spread "between government debt and other liquid securities."

    Sure. I was thinking about hypothetical assets with zero liquidity value that provide a stream of payouts over time. These assets do not exist in your model, but if they did they would be priced according to the rate of time preference beta. In the New Keynesian model, when they exist (e.g. if we are trying to think about durable goods) these assets are priced according to a liquidity-premium-free real interest rate that depends on time preference, the expected rate of change of consumption, and the elasticity of intertemporal substitution.

    I agree completely with your suggestion that even illiquid assets may have some liquidity value following intermediation, and that they will be priced accordingly. Modeling the degree to which various kinds of real assets can be transformed into liquidity, and how this changes over time (possibly sharply), is very important for understanding the effects of finance on the real economy.

  6. "Evidently we have very different views about what makes monetary policy interesting!"

    No, I was describing what happens in the most basic, conventional monetary model. I agree that this is not very interesting, and Cooley and Hansen (1989) says those effects are quantitatively insignificant.

    "This assumption seems to be true in practice, and it can be justified in any model where prices are even the slightest bit sticky (adjusting every millisecond?)"

    No, this has nothing to do with the sticky prices. I still say it's important to map out how what the central bank does (asset swaps) translates into asset price movements. You can do that and still have your sticky prices.

    "For instance, if the Fed keeps excess reserves in the system and sets monetary policy via IOR, your proposed channel for the Fed to affect real interest rates disappears, since the Fed is no longer implementing monetary policy via asset swaps."

    Not correct. That's in the paper. The effect works in exactly the same way. In that context, the open market swaps are irrelevant, but changing the IOR matters, and induces the same movements in the asset quantities. Now the central bank is literally setting the short rate, and the quantities move, rather than setting the quantities and having the short rate move in response.

    "Virtually everyone would agree with this statement, including New Keynesians."

    I don't.

  7. Steve,

    A question and a comment:

    1. I'm with you on a number of critical comments regarding the NK paradigm. However, a quick question. Bernanke seems to think that monetary policy works through portfolio rebalancing. At least that is what he is talking about here:

    Isn't this just a limited participation/segmented markets story? I don't see your model specifically addressing this issue. Perhaps this is an empirical question, but I would be interested to get your thoughts -- especially since you have used limited participation and segmented markets in the past.

    2. I think that expectations matter for policy. For example, see Berentsen and Waller:

    In their model, monetary policy only has real effects if the central bank has committed to a price level target. One of my problems with the current Fed policy is that they aren't committing to ANY stated objective. They have no goal. As a result, saying you are going to buy $X worth of government bonds is irrelevant regardless of the mechanism by which monetary policy has real effects because expectations matter.

  8. Josh,

    When Bernanke and other people discuss QE, the seem to have in mind one of two things:

    (i) "portfolio balance," which is essentially Tobin (1969):

    In Tobin's model, you assume asset demand functions where the arguments are rates of return and wealth. In that model, if you move relative asset supplies, the rates of return have to change. Hopefully people understand why that is the wrong way to think about asset markets and the effects of central bank actions.

    (ii) Segmented markets/preferred habitat theories of the term structure: The idea here is that there is something limiting arbitrage across assets of different maturity. The theory here is essentially non-existent. In this paper, for example,

    the authors put the assets in the utility function. Not much progress on the preferred-habitat model as you can see.

    In the course I taught for advanced PhD students last year, I tried to get them interested in the idea that you might make a connection between the segmented markets literature (which was about distribution effects from monetary policy) and QE. However, my current thinking is that you need to model intermediation in a serious way - that's what it's all about - and when I think about that, I can't see why QE matters under current conditions. I don't have a formal model of that yet, as of course you need to capture the term structure.

  9. Professor Williamson,

    Very interesting argument. However, I'm a bit confused, so I would appreciate it if you could enlighten me (and allow me if it is a silly question...):

    You said in the above comment of October 31, 2011 7:04 AM, "...changing the IOR matters...Now the central bank is literally setting the short rate." So, it seems that the IOR minus inflation rate equals the real interest rate now.

    And, here you said, "if you allow the Fed to tax reserves, there's no liquidity trap. Problem solved." So, you seem to be advocating the negative IOR, which means the lower real interest rate and/or the higher inflation. This looks exactly like New-Keynesian solution.

    On the other hand, you also advocate the higher real interest rate and/or the lower inflation as a solution. How should these two solutions be reconciled?