If you wanna run hot you gotta run steady

Canadian economist Richard Lipsey would always insist that the Phillips Curve is a curve, and not a straight line. Maybe because wages and prices are stickier down than up. Or maybe because unemployment can't go below 0%. Or both.

Here's why that matters for macroeconomic policy:

Suppose the central bank wants 2% inflation on average, over the longer term. One way to get that result would be to flip a coin, every year, or every decade, or whatever. If it's heads, the central bank "runs the economy hot", to get 3% inflation. And if it's tails, the central bank "runs the economy cold", to get 1% inflation. If the Phillips Curve were a straight line, you would get exactly the same average unemployment rate, over the longer term, as you would if the central bank kept inflation steady at 2%. But if the Phillips Curve is a curve, and curves the way we normally draw it, you would get higher unemployment on average with that random monetary policy than you would if you kept inflation steady at 2%. And the sharper the curvature of the Phillips Curve, the worse unemployment will be, on average, with a random monetary policy.

That's why macroeconomic demand stabilisation matters. Recessions are bad, and booms may be good, but the bad and good don't cancel out. Relative to where you would be if you stabilised demand, recessions are bigger than booms.

If you add "hysteresis" to this simple story, then macroeconomic stabilisation matters even more. Because a recession has costs that persist even after the recession is over, and those long term costs of the larger recession are bigger than the long term benefits of the smaller boom.

Now maybe 2% inflation is not the right inflation rate to target. Maybe it should be 3%. But that's a separate question, about the Long Run Phillips Curve and whether it's vertical or sloped, and the point I'm making in this post about the Short Run curvature still stands. It's better to target 3% every year than to flip a coin every few years between 2% and 4%.

And maybe inflation is the wrong thing to target. Maybe it's better to target a level-path of NGDP. But again that's a separate question. Or maybe only a partly separate question. Because if supply shocks cause the Phillips Curve to shift up and down, keeping NGDP on target would stabilise unemployment better than an inflation target. And the point of this post is that you need to stabilise unemployment if you want to keep unemployment low on average over the longer term.

And maybe it's not really about Phillips Curves at all. Maybe it's about Milton Friedman's "Plucking Model": there are no booms, only recessions. (What looks like a boom is just the absence of a recession, because recessions can be big or small but booms are all the same, because big recoveries follow big recessions, and small recoveries follow small recessions, but the size of the decline doesn't match the size of the "boom" it follows). Which is like an extreme version of my curved Phillips Phillips Curve story above, where booms are smaller than recessions. And Friedman's Plucking Model always reminds me of Frances' story about her old car: most of the time it ran OK, but if something unexpected happened it was always something bad. The only good news was absence of bad news. Which is how things are with designed systems: they sometimes break down and run worse than the Toyota designer intended, but rarely "break up" and run better than the designer intended. And market economies are a bit like that, though their "designer" is the invisible hand. There can be bad monetary shocks, but the only good monetary shock is no monetary shock.

You can read this post as a critique of "running the economy hot". What do you mean by that, and how can you do it sustainably over the longer term? Or you can read it as a comment on J.W. Mason's recent good post (which I partly agree with and partly disagree with, as always). Or you can just read it on its own.

Comments

[ssumner]

Great post. Nothing to add.

[louis]

Good to see you back, Nick!
Your post is insightful as always, and the framework makes sense.
Still, I'd like to try harder to 'steel man' the case for running hot, which its proponents often don't make explicit.
I see a few variants of the argument:
1) a view that growth in the economy's potential is not exogenous to macro policy, and that periods of time where AD is strong can catalyze increases in TFP. I have no idea what evidence there would be for this, but this would be the case for small overshoots vs a steady path for rate of inflation as you advocate.
2) a distributional view that workers have more bargaining power over real wages when unemployment is tight to the lower constraint, and perhaps a dynamic view that times such as these draw more of the long-term unemployed into the workforce with some lasting benefit
3) a view that in recoveries, central banks like the Fed have been generally premature to 'take away the punch bowl' and undershoot their target, and since CBs dont have perfect knowledge of economic conditions or perfect control over the price level, restraint on tightening is on balance the most prudent course, with risk of a temporary overshoot on inflation seen as less concerning than risk of keeping employment lower than it potentially could reach.

In some ways, Greenspan's restraint from tightening policy in the mid-1990s, tolerating the possibility of being 'late' to tamp down acceleration in inflation in order to feel out the 'true' NAIRU is the paradigm for this approach.

[Nick Rowe]

Thanks Scott!

Thanks louis!
Let me try to respond:

1. If we had *very* strong "hysteresis", so that a positive shock, driving the unemployment rate down temporarily by 1%, would let us keep the unemployment rate *permanently* lower by 1%, we would have multiple Long Run Phillips Curves. That makes policy a bit trickier then just stabilisation, because we want an initial shock to get us on the best LRPC, then stabilisation thereafter to keep us there. But if the hysteresis is just that TFP growth is higher when unemployment is lower, I don't think that affects the policy conclusions in this post.

2. There's a very old debate on whether real wages are procyclical or countercyclical. AFAICT, it depends on whether wages or prices are stickier, plus labour force compositional effects, plus lots of measurement problems. But I can't get really excited about the answer, from a policy perspective, because recessions are bad regardless of what happens to real wages. And the important distributional issue is the unemployment itself, which tends to hit the poorer hardest, and less the real wages of those who remain employed.

3. Well, if the CB undershoots 2% on average, it becomes a de facto 1.5% (or whatever) inflation target. Which the economy (and expectations) eventually adjusts to. And whether that's better or worse than a 2% or 3% target is really another question about whether the LRPC is vertical or not. But a Price-level path target would have an advantage in holding CB's accountable for fixing their past mistakes in targeting inflation, so that one-sided errors can't keep on accumulating.

[Scott Sumner]

Wages tend to be countercyclical when the economy is impacted by demand shocks and procyclical when impacted by supply shocks. That supports the claim that wages are especially sticky:

https://econpapers.repec.org/article/ucpjpolec/v_3a97_3ay_3a1989_3ai_3a3_3ap_3a706-20.htm

Shorter version: Never reason from a price change.