Full Site Articles EconLog EconTalk Books Encyclopedia Guides

A theory of house temperatures

 George Selgin on Bernanke's Sp... There's nothing natural about ...

Bob Murphy has occasionally complained that I contradict myself on interest rates. I frequently say:

1. Interest rates are a meaningless indicator of the current stance of monetary policy.

2. Raising interest rates makes monetary policy more contractionary, on the day they were raised.

In fact there is no contradiction, but since lots of people (not just Bob) are confused on this point, I'll use a house temperature analogy. Of course this will be done half-jokingly, Bob doesn't actual hold the views about house temperatures that I will attribute to him, but hopefully the example will make it easier to see my point:

1. Suppose that on average houses are 68 degrees in the winter, and 78 degrees in the summer. People use heat in the winter, but since they are wearing winter clothing they only heat their house up to 68 degrees. People use AC in the summer, but since they are wearing summer clothing they only AC it down to 78. I think that this is a plausible assumption. House temps then represent the "stance" of house temperature policy, equivalent to the stance of monetary policy.

2. Suppose that turning up the thermostat makes houses warmer. And turning on the AC cools temps in the summer. Since turning up the thermostat raises temps, consider it analogous to an expansionary monetary policy "gesture". And vice versa for AC. Adjusting the AC or thermostat then represents raising and lowering the fed funds target.

Bob Murphy's theory of temperature would be that when people are frequently turning up the thermostat, you can expect houses to be relatively warm. And when people are frequently turning on the AC, you can expect houses to be cool. The Sumner theory is that when people are most frequently turning up the thermostats, houses will be relatively cool, even though that action makes them warmer. Bob Murphy's theory is that houses are relatively warm in the winter, because people frequently turn up their thermostats in the winter. Sumner's theory is that houses will be relatively cool in the winter, despite the fact that people turn up the thermostat more frequently in the winter, and despite the fact that turning up the thermostat does in fact make houses warmer, ceteris paribus. Bob Murphy will claim that Sumner contradicts himself on house temperatures.

I don't want to claim that pushing up the thermostat always indicates a cool house, even though it's highly correlated, because there are occasions where you push it up long enough to actually make the house warm, even in winter. But if you put a gun to my head, and asked me whether thermostat pushing is more likely to imply the house is cool or warm, I'd say cool, even though the action makes houses warmer. But most often I'd simply say that looking at how people fiddle with the thermostat controls is not a good way of judging the temperature inside a house.

Milton Friedman understood this distinction:

Low interest rates are generally a sign that money has been tight, as in Japan; high interest rates, that money has been easy.

Notice the past tense "has been tight". He knew that low rates did not necessarily imply that money was tight at that moment.

Similarly, if I was told that my neighbor had adjusted the thermostat higher 10 times in the past month, I'd assume it was winter and his house was relatively cold, even though that action made it warmer. If I was told he'd switched on the AC 10 times in the past month, I'd assume his house was relative warm, even though that action made the house colder.

Now suppose you naively looked at the correlation and assumed that turning up the thermostat actually made the houses cooler. Obviously no one in his or her right mind (except perhaps NeoFisherians) would hold that view. But let's say you did. In that case you'd end up turning up the thermostat in the middle of summer. This may be why no NeoFisherian has ever been put in charge on a central bank.

But even real world central bankers make mistakes with the thermostat. And it is partly because of exactly the fallacy that I am suggesting that Bob Murphy may suffer from. I'm up to page 410 in the Bernanke memoir, and even he seems to be confused (although when Bernanke was an academic he was fully aware of the argument that I am making here.) As a central banker, Bernanke suggested that the Fed eased monetary policy in 2008, because they frequently cut rates. In my view that's only true of January, where the rates were probably cut faster than the Wicksellian interest rate fell. But not the other 11 months, when the Wicksellian equilibrium rate fell faster. He was turning up the monetary thermostat, but the outside temperature fell so rapidly that the house was actually getting colder. He ended up with a "cool house policy stance."

Bernanke himself uses a "cold wind" analogy in his memoir. The cold wind in my example represents the falling Wicksellian equilibrium rate. Bernanke did not turn up the thermostat aggressively enough to offset the falling outside temps. Even worse (and here's where the analogy finally breaks down) the Fed actually caused the cold wind by an excessively passive policy in 2008 that unintentionally tightened policy, which lowered NGDP growth, and hence lowered the Wicksellian equilibrium rate.

Back in 2003, Bernanke claimed that interest rates were not a good indicator of the stance of monetary policy. Nor was the money supply. Instead, he insisted that you had to look at NGDP growth and inflation. He was correct. But as Fed chair he switched to the popular view that interest rates (and perhaps base expansion after 2009) were the best indicators. Thus he viewed monetary policy as being quite accommodative after 2008, when it was exactly the opposite.

To summarize, the thermostat is like the fed funds target (or the base if you are a monetarist), and NGDP growth is like the temperature inside the house. The outside temperature is like the Wicksellian equilibrium rate (if you are a Keynesian) or like velocity (if you are a monetarist.) The popular view is that the "temperature policy" of the owner is best described by the amount and type of fiddling with thermostats and AC controls, whereas the sophisticated view is that the temperature policy stance is best described by the house's inside temperature. That's the view that Bernanke used to hold, until he became a central banker. This paper by Vasco Curdia uses the sophisticated view, and shows that Bernanke is wrong about policy being accommodative after 2008.

If you are not sick of the analogy yet, we could also apply it to monetary policy failure at the zero bound. Keynesians would blame the "liquidity trap." What's the temperature analogy for the zero bound? Some would say a furnace that is out of fuel. But a better analogy is a furnace too small to fight against the bitter cold front that swept down from Canada, even going full blast.

The monetarist view is that a "fiat furnace" can never be too small. It can raise temperatures up to an almost infinite level, or at least up to 100 trillion degrees.

PS. I'm pretty sure I stole some of these metaphors from old Nick Rowe posts.

Update: Michael Byrnes has sent me a Nick Rowe post using the thermostat metaphor. He got it from Milton Friedman, but I'd guess the post I linked to is superior to whatever Friedman said about the subject.

Update#2: Market Fiscalist noticed a flaw, and suggested an improvement:

A bit nit-picky but I don't think the analogy quite works. If your thermostat is working well then just setting it to the desired setting will lead to temperature inside the house being correct without the need to ever change it. To my mind its the sensors used by the thermostat and the algorithm used to maintain the temperature that are more akin to monetary policy.

An alternative analogy might be:

- The temperature in the house is like NGDP
- The temperature outside would be like demand shocks
- The amount of power being used by the heating system is a bit like the interest rate
- When the thermostat turns on the heat its like a monetary injection
- When the thermostat turns on the a/c its like a monetary withdrawal.

So: When the outside temperature falls

A market monetarist thermostat would turn on the heat/ac in direct response to a change in inside temperature.

An "inflation targeting" thermostat would see the inside temperature drop and change the target for power used by the hearing system.

TravisV writes:

Prof. Sumner,

Do you think Market Monetarists should generally abandon explaining interest rate changes by saying "sometimes the inflation effect and income effect outweigh the liquidity effect, sometimes the liquidity effect outweighs the inflation effect and income effect"?

I've noticed it's been a long time since you've included that concept in any of your posts.....

ThomasH writes:

While it is possible to explain the interest rates are not a good measure of monetary policy, why use it only to have to explain it? What we need is a measurable target or targets (that are close enough proxies for what we really care about, presumably real income) and a commitment of the central bank to use any and all instruments that it controls to make economic results come out on target.

Conventionally short term interest rates have been the instrument used by the Fed to keep unemployment of labor an "optimum" rate and inflation on a steady path. In that world seeing interest rates much lower than in the recent path would lead on to conclude that the Fed was trying to get inflation up and or unemployment down. In an of themselves, interest rates could not tell you how far away from target the economy was. One could logically recognize that interest rates were "extraordinarily low" without concluding that monetary policy was doing too much or not enough to get the economy back on target. In terms of the thermostat analogy, the only think that matters in judging whether the thermostat should be turned up or down is whether the temperature in the house is too hot or too cold. It would be pretty crazy to look at the thermostat and conclude that the house was too warm if all the inhabitants had to wear sweaters.

Michael Byrnes writes:

I like this analogy, especially as you applied it to the zero lower bound.

Indeed, Nick Rowe has used similar analogies (these are some of my favorite posts by Nick).

B Cole writes:

Excellent blogging. I do not blame the public for believing we have an accommodative monetary policy now, since that view is repeated daily in every financial news outlet on the planet. But Ben Bernanke?

Scott Sumner writes:

Travis, I still believe that the quote is correct; perhaps I don't use it as much because I feel I've made my point. But of course there are always new readers, so that's a good point.

Thomas, Yes, and that's why I favor having the Fed switch from targeting interest rates to pegging NGDP futures prices.

Michael, Thanks, I added an update.

Ben, Yes, the media are awful, but so are many economists.

ThomasH writes:

Scott,

First, thanks for making the best of what I now think is a confused and confusing post. :)

I see the argument (which you have made, but not often) that targeting NGDP is better than what the Fed actually does (runs a loosely-tethered inflation rate target -anything between 0% and 2% is OK but anything above 2% is the apocalypse) because it is easier to defend popularly - "I'm trying to get your income back up to target" rather than "I'm trying to get the prices of things you buy back up to target."

But this argument rests on the public misunderstanding "inflation" to exclude wages and so overestimating the costs of "inflation." What I have not seen is modeling that shows that NGDP targeting is better than price level targeting (both in the target-the-expectation sense).

My sense is that either target could be better under some assumptions about a) what kind of non-monetary shocks the system is subject to, b) what kind of constraints there are on use of specific instruments of monetary policy that might act to return the economy to target outcomes once disturbed by a shock, and c) whether there were also a non-price related target like unemployment or potential output.

Don't you have any grad students looking for a thesis you could interest in this project?

Michae Byrnes writes:

@ThomasH

I'll give you several answers (basically all of which I have learned from reading market monetarist blogs).

1. What a central bank really controls is nominal spending (ie, NGDP).

If a central bank loosens policy (increases the monetary base), nominal spending will rise; if it tightens policy, nominal spending will fall.

Of course, changes in nominal spending will have some effect on the price level, but productivity growth will also affect the price level. If nominal spending rises, the extent to which this translates into a higher price level will depend on productivity (ie, because productivty growth will tend to cause prices to fall).

So a question, then, is how the central bank should respond to changes in productivity growth. Should it respond to a productivity boom by loosening and to a negative producitivy shock by tightening? I think not, because this would be pro-cyclical.

2. Inflation can't really be measured, only estimated. And it is difficult to estimate. There are a whole range of different price indices and adjustments: CPI, chained CPI, PCE, core vs headline, GDP deflator, etc. Whichever measure a central bank selects, the end result is that it ends up incorporating the errors of that measure into monetary policy decisions.

3. The reason we have "flexible inflation targeting" rather than a strict price level target is because central bankers understand, at least in theory, the dangers of running a procyclical policy. A good flexible-inflation targeter would be careful to keep inflation lower during a productivity boom and let in run higher during an adverse shock. What we have seen since 2008, though, is the opposite. The Fed hasn't pushed inflation up to (or above) 2% even though it should have.

4. The Taylor rule represents an attempt to guide flexible inflation targeting in a countercyclical direction by adjusting monetary polcy based on not only inflation but also an estimate of the output gap. So a high output gap calls for looser policy than one would expect from the inflation rate alone. Makes a certain amount of sense, but it ends up relying on yet another estimate. After all, who can say for sure what the true output gap really is? So once again, monetary policy ends up being driven to some extent by errors in the estimates.

Market Fiscalist writes:

"To summarize, the thermostat is like the fed funds target (or the base if you are a monetarist), and NGDP growth is like the temperature inside the house. The outside temperature is like the Wicksellian equilibrium rate (if you are a Keynesian) or like velocity (if you are a monetarist.)"

A bit nit-picky but I don't think the analogy quite works. If your thermostat is working well then just setting it to the desired setting will lead to temperature inside the house being correct without the need to ever change it. To my mind its the sensors used by the thermostat and the algorithm used to maintain the temperature that are more akin to monetary policy.

An alternative analogy might be:

- The temperature in the house is like NGDP
- The temperature outside would be like demand shocks
- The amount of power being used by the heating system is a bit like the interest rate
- When the thermostat turns on the heat its like a monetary injection
- When the thermostat turns on the a/c its like a monetary withdrawal.

So: When the outside temperature falls

A market monetarist thermostat would turn on the heat/ac in direct response to a change in inside temperature.

An "inflation targeting" thermostat would see the inside temperature drop and chnage the target for power used by the hearing system.

Jj writes:

That's a fantastic analogy, whoever came up with it.

Scott Sumner writes:

Thomas, I never had any grad students, as Bentley doesn't have a grad program in econ. In any case I'm retired from teaching now. There are lots of people doing work comparing NGDP and price level targeting, although I tend to forget the names. Woodford is one.

Market Fiscalist, You said:

"To my mind its the sensors used by the thermostat and the algorithm used to maintain the temperature that are more akin to monetary policy."

Yes, I realized that too, after I posted it. Perhaps I'll add an update.

Thanks JJ.

writes:

Let's say the outside temperature drops and the heater kicks in.

1. At that moment, you have one correlation, which is (heater on + cold house).

2. Some time passes, and you have another correlation, which is (heater on + warm house).

3. Some more time passes, and you have yet another correlation, which is (heater off + warm house).

One person could say, "As 1. indicates, a cold house indicates that the heater is on." Another person could make the exact opposite claim, "As 2. indicates, a warm house indicates that the heater is on." Almost no one would say something totally unintuitive like, "As per 3, the heater must be off, since the house is warm."

Sometimes market monetarism sounds a lot like #3.

Brian Donohue writes:

Bravo!

dl writes:

A better analogy is this. You're in a boat that starts to leak. You decide to start bailing out the water, but find that it is coming in faster than you can bail it out.

So, the question is "is your decision to bail out the boat reducing or increasing the water in it?"

Scott Sumner writes:

RPLong, Well it's a fact that warm houses are correlated with heaters not being on, but the correlation is not perfect. That's the MM view, and that's reality.

Matt writes:

If not high interest rates, what exactly has been making money "tight". Tightness seems like what you to explain more clearly, as it seems tautological.

writes:

If anybody is curious, I took the thermostat analogy further.

Comments for this entry have been closed