Scott Sumner  

Does AD influence AS?

PRINT
How the Free Market Breaks Dow... One Congressman Can Sometimes ...

TravisV directed me to a recent speech by Janet Yellen:

The Influence of Demand on Aggregate Supply The first question I would like to pose concerns the distinction between aggregate supply and aggregate demand: Are there circumstances in which changes in aggregate demand can have an appreciable, persistent effect on aggregate supply?

Prior to the Great Recession, most economists would probably have answered this question with a qualified "no." They would have broadly agreed with Robert Solow that economic output over the longer term is primarily driven by supply--the amount of output of goods and services the economy is capable of producing, given its labor and capital resources and existing technologies. Aggregate demand, in contrast, was seen as explaining shorter-term fluctuations around the mostly exogenous supply-determined longer-run trend.1 This conclusion deserves to be reconsidered in light of the failure of the level of economic activity to return to its pre-recession trend in most advanced economies. This post-crisis experience suggests that changes in aggregate demand may have an appreciable, persistent effect on aggregate supply--that is, on potential output.


It seems likely that demand shocks have at least a transitory impact on aggregate supply, if only because they impact investment. But I think Yellen is making a mistake here, which could lead monetary policy astray if we are not careful.

There are good reasons why most economists have viewed AS and AD shocks as being independent, at least as a first approximation. Deep recessions such as 1907-08, 1920-21, 1929-33, 1937-38, 1981-82, etc., are usually followed by fast recoveries. The recent 2007-09 recession is of course an exception. But should we re-evaluate the basic model based on a single exception?

Suppose the model is revised, and we now assume that deep recessions lead to a long period of sluggish growth as aggregate supply is depressed. In that case we will have not one mystery to explain (2007-09), but dozens. A more plausible explanation of the recent period is that the Great Recession just happened to coincide with a slowdown in trend growth. The unemployment rate did recover, which is exactly what you'd expect if my theory were correct. Instead, other factors not usually linked to AD seem to explain the slowdown. These include a slowdown in the growth rate of the working age population as well as a slowdown in productivity growth. Could the Great Recession have caused the sharp slowdown in productivity growth? I suppose anything is possible, but in that case why didn't previous deep recessions lead to slowdowns in productivity growth?

There's another, and in my view even more persuasive argument against Yellen's theory. The models that try to explain why AD might depress AS tend to predict lower levels of RGDP, but no permanent reduction in the growth rate of RGDP. This might reflect discouraged workers retiring or going on disability, as well as a slightly smaller capital stock. Neither factor should reduce the long run trend rate of growth in RGDP--it's one-time level reduction.

Now admittedly it's too soon to say the long run trend rate of growth has slowed. But this leads to the biggest problem with Yellen's hypothesis: long-term interest rates are extremely low, both in nominal and real terms. The overwhelmingly most likely explanation of the lower than normal long-term real interest rate is that the market expects sluggish RGDP growth to be the new normal. If that's the case, then there is really no plausible AD model to explain this slowdown. It's just barely possible that the 2008 recession depressed output by 5% today (although I doubt it) but it's completely implausible that it would depress output by 20% in the year 2030.

More likely, the US is being impacted by the same factors that have slowed growth throughout the developed world. We don't fully understand those factors, but they include a slower growth rate in the working age population, and a shift to a service-oriented economy where productivity gains are harder to achieve. Those problems cannot be fixed by "making the economy run hot" to boost aggregate supply.

Rather than trying to solve supply-side problems with monetary policy, I'd like to see central banks concentrate on refraining from causing problems by refraining from policies that generate unstable NGDP. Let other policymakers try to undo the tangle of taxes and regulations that are preventing our economy from reaching its full potential.


Comments and Sharing






COMMENTS (4 to date)
bill writes:

One quibble. I think productivity gains are harder to measure in the service sector, but not necessarily to achieve.

Scott Sumner writes:

Bill, Maybe, but how would you compare productivity gains in barbershops and steel mills, over the past 50 years? (I pick steel mills to avoid high tech examples like computers, where the gains are obvious.) Even "low tech" steel mills show massive productivity gains. I just don't see it in barbershops.

Health care is a possibility, but even there one can find lots of dispute---with some claiming our longer life expectancy is mostly due to better nutrition, sanitation, less smoking etc.

How about public school teaching? Legal work? Truck drivers?

bill writes:

Barbershops-probably no gains.
Policing-data and mapping could be making police more productive.
Architects-more productive with computers letting them start each project by copying and pasting from a prior project.
Legal work-research is faster plus copy and paste.
I chose those examples because I believe they get added to GDP just based on wages billed?


Alex S. writes:

Please feel free to tear this to pieces. It argues that AD may impact AS vis-à-vis monetary policy. Please note, all variables here are in terms of growth rates.

m + v = p + y (dynamic AD)

y* = g + ak + bn (dynamic LRAS)

Potential RGDP growth, y*, depends on the growth rates of technological knowledge, g, capital stock, k, and population, n. The a and b terms are factor share parameters greater than 0.

In the short-run, RGDP growth, y, may be influenced by deviations in actual inflation from expected inflation, E[p].

y = y* + f(p – E[p]) (dynamic SRAS)

The f term is a parameter greater than 0, replacing the y term in the dynamic AD with all the factors affecting AS:

m + v = p + g + ak + bn + f(p – E[p])

If m, p, g, n, and E[p] are relatively fixed, p = E[p], and v depends on the expected future money supply growth rate, m. E.g., if the money supply is expected to decrease in the future, then v decreases today. On the right-hand side of the equation, if v is negative, k must fall, which is a decrease in AS.

The biggest question is, why no LR adjustment in inflation, p? The market knows the Fed won’t let p fall below 0. But it knows that if p starts to rise it will result in a lower m, which causes v to fall today. Ultimately, the growth rate of the capital stock, k impinges aggregates supply, but it is the result of an aggregate demand policy [I think the problem with my argument may be somewhere here].

As evidence of this, there’s still a very large mismatch between the Fed’s model of the world and the market’s model of the world. Compare the dot plots against contemporaneous bond yield curves. The bond markets been getting it right, and the Fed’s been getting it wrong. I see no reason for this trend to change.

Put another way, suppose that the Fed’s reasons to itself, “Nominal interest rates, i, and money supply growth, m, are negatively correlated. The world is dominated by the liquidity effect. If inflation expectations are stable, increasing m will lead to lower nominal interest rates vis-à-vis lower real interest rates, r. But I’m going to ignore what m and v are doing because they don’t fit nicely into my New Keynesian DSGE model).”

The market is currently dominated by the Fisher effect, such that i and m are positively related (in other words, real interest rates are relatively constant, so that changes in m lead to changes in E[p], which lead to changes in nominal interest rates i. Therefore, the market believes monetary policy is tight, while the FOMC believes monetary policy is loose because nominal interest rates are low.
I can’t help but believe that a mismatch between market and central bank expectations would not have real effects.

At the same time, I feel like there’s got to be a major flaw in this argument. Please destroy.

Comments for this entry have been closed
Return to top