A Neo-Fisherian Challenge and Reconciliation

 Lars Svensson has a very interesting challenge to the Neo-Fisherian view. (See link for slides.) 

What happens to inflation and unemployment when the central bank (for no good reason) raises the policy rate by 175 bp?...

Sweden did, which provides  

..a natural experiment of the neo-Fisherian view: Does inflation really increase after a policy-rate increase? 

Despite roughly the same circumstances as many other countries, including the US, Sweden in 2010 raised rates 175 bp. (Top left graph). The result: Inflation fell, the exchange rate appreciated. Unemployment also rose (not shown).  

Sweden figured out it was too much too soon, (or perhaps started listening to Lars, who opposed the move)  and turned the rate cuts around. A swift sequence of rate cuts followed in 2014, bringing the target rate below zero (Top left graph below). The result: Inflation rose, the Krona depreciated, and unemployment fell. 

Lars concludes:

Monetary policy in Sweden works like clockwork and according to the textbook.

Lars points to an interesting channel that is stronger in Sweden than other countries: the household cashflow channel. Almost all mortgages in Sweden are variable rate. And Sweden, like other Nordic countries, is quite level-headed about debt (and more generally controlling incentives) in a way America is not. If you default on a mortgage, creditors seize all your other assets, and even garnish wages. None of this handing the bank the keys and driving away as Americans do. The result is that when interest rates rise, Swedes cut spending on everything else in order to make their higher mortgage payments. 

As Lars points out, this system gives a certain automatic stabilizer if interest rates move pro cyclically. But they also make out of cycle movements (which I think Lars might call "mistakes") have larger real effects.

Well, this seems pretty damning for Neo-Fisherianism, the proposition that raising interest rates raises inflation, doesn't it? 

Response: 

Not so fast. The neo-Fisherian proposition is at heart the proposition that in the long run the economy is stable at an interest rate peg. If that is true then higher interest rates must mean higher inflation, but only after all the short-run dynamics have settled down. It is entirely possible to see Sweden's experiment while at the same time inflation would eventually rise to meet a permanent interest rate rise, after transitory dynamics in the other direction fade away.  

Really, the neo-Fisherian proposition is only that a widely anticipated, slow and steady, persistent interest rate rise, accompanied by steady fiscal policy will raise inflation -- all the caveats there to tamp down common sources of transitory dynamics in the opposite direction.    

 

What do I mean by "stable?"   In every economic model, there is a steady state in which 

nominal interest rate = real interest rate + inflation. 

"Stable" means that if the central bank pegs the nominal interest rate, once the real interest rate settles back down and once other short-term dynamics settle down, inflation must eventually converge to wherever the nominal interest rate is. 

In "unstable" models of the economy, the steady state is a knife-edge. Any slight deviation snowballs and the economy spirals away. For example, the common prediction of a "spiral" at zero interest rates -- a form of peg -- follows exactly such logic. The nominal interest rate is stuck. The real rate is too high. That results in less inflation. The nominal rate is still stuck. The real rate is higher. That results in even less inflation. And off we go. 

I humorously summarized these two possibilities, and the corresponding view of the role of central banks as a seal (unstable) vs, professor Calculus (stable):

Now, the deflation spiral did not happen, despite 10 years of zero rates in the US, 12 years and counting in Europe, and nearly 30 in Japan. This is powerful evidence for long-run stability. And stability implies the the neo-Fisherian prediction that eventually inflation must rise.  (This point, and much of the discussion, is buried in "Michelson-Morley, Fisher and Occam," but the point is not originality it is to confront the Swedish experience). This is so contrary to intuition I resist it kicking and screaming as well, but at some point models and data must move priors. 

Stability is really a form of a long-run neutrality hypothesis for interest rate targets. "Neutrality" of money means that under MV=PY, if you double M, eventually P doubles. Neutrality of interest rate targets means that if you double i, eventually pi doubles.  

Lots of models are long-run stable and thus exhibit long-run neo-Fisherianism. Though dabbling in this is correlated with fiscal theory of the price level, the standard new-Keynesian model is neo-Fisherian. Indeed it is super-Fisherian. An unexpected permanent rise in the interest rate target in the standard three equation Woodford model produces an immediate equal and permanent rise in the inflation rate. 

As an explicit example, here is the response to a monetary policy shock from the most recent model I'm working on -- which includes the standard two equations of the usual new-Keynesian model: 

i = interest rate, pi = inflation, x = output, and don't worry about the rest.  That looks a lot like Sweden, doesn't it? But this is a neo-Fisherian model! If you look really really hard you can see the inflation rate eventually rise to meet the interest rate. 

Moreover, in this same model, an expected interest rate rise leads to a smooth and always positive inflation response. (This model generates the negative response by unexpectedly devaluing long-term bonds.) The next graph is the same model and also the standard NK model, faced with a fully expected and permanent interest rate rise. 

Thus, here we have an explicit model that reconciles the observation that 1) Sweden's interest rate rise and fall produced "conventional" effects but 2) the long zero bound produce no effect, despite the "conventional" prediction of a spiral. The latter is key -- nobody gets to pick and choose observations, and the "clockwork" failed miserably at the zero bound. Our model must produce both results, without too many epicycles, please. 

But this reconciliation  makes the prediction 3) that if Sweden -- or anyone -- were to try a widely preannounced, slow, steady, persistent, no-giving-up-if-the-transitory-dynamics-go-the-wrong-way-as-Sweden-did, interest-rate rise, accompanied by stable fiscal policy (included in the above calculation, and no, Turkey and Venezuela cannot lower inflation by lowering interest rates), inflation would eventually rise. 

Now, Lars points out that at least Sweden's  interest rate rise was anticipated. From  Lars' Forward Guidance, a plot of actual rates and rates forecast in the repo market: 

But, first, this does not hold on the downside. And second, is this enough anticipation? My model generates the negative effect by surprising long-term bondholders. If the debt is more than 3 years old, the rise is still anticipated. Lars alludes to an interesting new mechanism for a temporary negative effect, via households and mortgages. To be "anticipated" the rise must be baked in to households' house purchase and spending decisions when they bought the house, I think. That's a long time. 

And  my long-term debt mechanism for a temporary decline, which emphasizes the expected / unexpected distinction, is only the tip of the iceberg of frictions one might introduce to produce a temporary decline in inflation when interest rates rise. 

Reconciliation: 

In sum, once we include a multitude of plausible fractions that send inflation temporarily the other way, including the long-term bond effect and household financial frictions, a negative response to temporary interest rate rise like Sweden is consistent with a neo-Fisherian prediction that in the very long run higher interest rates produce higher inflation, and thus also consistent with the lack of a spiral at the zero bound.

But I move somewhat in Lars' direction. Just how relevant is this observation to policy? When the central bank can move interest rates, it may well want to push rates around by exploiting the temporary negative sign.  "Temporary" can be a long time.  Even if the long-run effect is positive, the central bank may move inflation up more quickly by lowering rates, pushing inflation up with the short-run negative effect, and then then quickly getting on top of inflation. Which is just what central banks classically do, and exactly what they do if the economy is unstable as well. They may never notice the positive possibility, and may never have the patience to wait for it. Thus, the neo-Fisherian possibility may be completely irrelevant in normal times. 

But when the central bank cannot lower interest rates, then the slow, preannounced, persistent, we-wont-give-up, and whatever else needed to overcome or wait out temporary forces in the other direction, may still be a useful policy for liftoff. One might indeed read the US interest rate increases -- known years in advance -- in that light. 

A note on stability: 

It is not quite true that all models have steady states in which the nominal interest rate and inflation move together one for one.  I believe Xavier Gabaix behavioral new-Keynesian model has this property (thanks to Ludwig Straub for patient explanation on this point. My comment  slides on a draft of Gabaix model summarize it compactly). If you replace x_t = E_t x_t+1 +... with x_t = delta E_t x_t+1 +... in the IS equation, delta is small and price stickiness is high, then it is possible that a permanently higher interest rate results in a permanently lower inflation rate. It does so by permanently lowering the real interest rate. The nominal rate rises 1%, the real rate declines 2%, inflation declines 1%. 

The puzzle of this result, though, as the puzzle of all models that produce a negative inflation response, is just how to get the real rate to respond so much. You can rightly scratch your head to believe that's how the world works. The standard unstable model gets around such a huge real rate response with dynamics: expected inflation is not the same as actual inflation, so the real rate does not have to move more than one for one. But things are always moving, so you must either have instability or a long run Fisherian response.  

A note on freshman physics:

The center of gravity of prof. Calculus' pendant is in the pendant, so the pendant does not go temporarily backward when he moves his hand forward. Imagine him holding an umbrella, and then the analogy works -- the bottom of the umbrella moves temporarily the wrong way, and one could exploit these dynamics to more quickly move the umbrella in the direction one wants it to go.