Monday, September 28, 2009

Putting Klingonomics to the Test

Arnold Kling has been promoting a macroeconomic theory he calls "Recalculation" which takes a controversial view on the efficacy of monetary policy. You can read his discussion of recalculation macro here, here, and here. Within these discussions he summarizes his view of monetary policy as follows:
In the short run, the economy is going its own way, regardless of monetary policy. Higher M leads to lower V, and vice-versa. In the long run, a significant change in the rate of money creation causes a similar change in the rate of inflation. However, the lag is long and the effect on the rate of Recalculation is small and of indeterminate sign[.]
As this and other passages from his postings show, Kling makes three controversial assertions about monetary policy in his recalculation macro theory. They are as follows:
(1) Monetary policy has no effect on expectations in the short-run.

(2) Monetary policy has no effect on nominal economic activity in the short run.

(3) Monetary policy has no effect on real economic activity in the short run.
Bill Woolsey has been all over Kling's case for making these assertions and the assumptions behind them. I will speak in a moment to Woolsey's critique but for now I want to see what the data says about the assertions (1) - (3).

Number (1) touches on an important question: can the Fed influence expectations about the future in such a way as to shape current economic behavior? Standard macro theory says yes--it is the reasoning behind the current arguments for why the Fed should be explicitly targeting some positive rate of inflation now. Kling, however, does not buy it. In order to test this question empirically, I took the monthly expected inflation series implied by the difference between the nominal 10-year Treasury yield and the 10-year TIPs yield and put it in a vector autoregression (VAR) along with the monthly GDP series from macroeconomic advisers. Nominal GDP was turned into an annualized monthly growth rate and the data used runs from 1999:1 through 2007:9. More data would have been helpful, but TIPs only start in the late 1990s. (Technical note: both series were in rates so no unit root problems, 13 lags were used to eliminate serial correlation, and corporate bond spreads were included as a control variable for the financial crisis). The two figures below show what the typical responses of expected inflation and nominal GDP to the typical sudden change or shock to expected inflation over the sample. The solid line shows the point estimate while the dashed lines show two standard deviations around the point estimate. Upon impact, the shock causes expected inflation to jump 16 basis points and occurs as the level of nominal GDP increases by 1.16 percent. In other words, a sudden positive change in expected inflation is associated with an increase in current nominal spending. Both effects persist but eventually become insignificant about 14-15 months later. (Click on figures to enlarge.)





Now presumably the change in 10-year expected inflation comes from a expected change in monetary policy, but just to be sure and to fully address (2) and (3) I have posted below some figures from another VAR I did that looks at the effect of unexpected changes or shocks to the monetary base for the period 1960:3 - 2008:2. This is a larger VAR that controls for more things. (This figure is actually an excerpt from a series of VARs I did in response to Nick Rowe's post on monetary policy and debt.) Here, the monetary base is shocked 1%. Note how all the real variables increase on impact. In other words, an unexpected positive increase in the monetary base historically has led to an increase in the short run of real economic variables. As predicted by theory, the effect of the monetary base shock eventually wears out--money becomes neutral. (Note that the price level is implicitly in these figures too: it is difference between real money and money. Here, there is a permanent effect)




So the assertions (1), (2), and (3) are empirically falsified. Of course we did not need my VARs to know this. There is already a lot of empirical evidence out there that reaches a similar conclusion. Moreover, Bryan Caplan notes numbers (1) and (2) fly in the face of everything we know from hyperinflation experiences. So why make such assertions? Bill Woolsely explains that Kling's assertions can work if prices are sticky in the short run, real income is determined by productive capacity, and real money demand is not affected by real income. As Bill Woolsely notes, this last assumptions is incredibly wrong, as many empirical studies have demonstrated over the past 50 years.

In light of the evidence I say it is time for Arnold Kling to join the ranks of the monetary disequilibrium bloggers.

Update 1:Here are some definitions to add clarity to the figures above. The real stock price series is the real S&P 500, the debt series is financial sector debt to GDP, the money supply is the monetary base, and the real money balance series is the monetary base divided by the CPI.

Update 2: Josh Hendrickson provides a nice follow up to the issues raised here while Arnold Kling assails my use of the "Dark Age Macroeconomic"-based VAR.

7 comments:

  1. David, there are some serious problems using TIPS to get expected inflation as its contaminated by other factors, see Carlson & Fuerst, Cleveland Fed Letters 2004.
    Maybe using some survey of inflation expectations or perhaps doing an ARIMA model of inflation and using residuals would get around this problem?

    ReplyDelete
  2. "In other words, an unexpected positive increase in the monetary base historically has led to an increase in the short run of real economic variables."
    Ooh...I'm not buying that bag of rotten apples David!! Last fall we had the mother of all positive base shocks and we're still waiting for the increase of real variables!!

    ReplyDelete
  3. ECB:

    Yes, I know the TIPs data is plagued with a liquidity premium and went looking for the Cleveland Fed's adjusted expected inflation series that corrects for it. However, they took it down because of the crisis. I also thought of getting survey data on expected inflation, but given time constraints I settled for the crude TIPs. So is there a place I can get survey data at a high frequency that won't cost me a fortune?

    On your second point, yes we had a huge base shock and no response from real variables. However, we also had a mostly Fed-induced negative money multiplier shock (from the interest payment on excess reserves)that offset the base shock and could have been avoided or at lest lessened. One way to interpret the 1960-2008 VAR, then, is that it shows what could have happened happened had the Fed not been paying interest on excess reserves. (Of course, given the severity of this downturn and the linear assumptions in the model, the VAR may overstate the case for the present.)

    With that said, an easy fix to the model is to include a money multiplier term. I just happen to have a paper under review right now that does just that. It questions whether monetary policy has any effect on long-term interest by looking at the effect of monetary base shocks and controls for the money multiplier. It can be found here. Any comments you have on the paper are welcomed.

    ReplyDelete
  4. I'm thinking that the Philadelphia Fed has some survey data...
    http://www.phil.frb.org/research-and-data/real-time-center/survey-of-professional-forecasters/
    Hope that helps!

    ReplyDelete
  5. I hate to just prove my ignorance, but I was puzzled by ECB's claim that there was a surge in base money over the last year and no real response.

    To me, that would suggest that if base money had remained at roughly 800 billion, then all the real variables, expecially real GDP, would have been the same. My humble opinion is that if the monetary base remained at 800 billion or so, then real GDP would have been much lower than it is today. And so, there has been a real response to the increase in base money.

    I would interpret this empirical data as showing that other things being equal, an increase in the monetary base will raise real output.

    Does the data here purport to show that there is no possible way for a financial panic to depress real output?

    I am not interested in proving the claim that if base money grows at a stable rate, then nominal income, real income, or anything else will grow at a stable rate. If that were true, it would be nice to know, of course.

    The question at hand, in my view, is whether or not any other things that might lead to lower nominal expenditure and real output can be offset by a sufficiently large increase in the monetary base, so that nominal expenditure will continue to grow on target, and hopefully, real output will suffer less disruption.

    If changes in the monetary base never impact nominal income or real output, that would seem to provide some evidence that sufficiently large increases in base money would not offset decreases in nominal expenditure and real output.

    Since this evidence suggests that changes in base money do in fact tend to have a positive impact on nominal expenditure and real output, then it remains possible that sufficiently large increases would offset other factors tending to depress those things.

    Please explain to me how I am wrong about this so that the the fact that real income fell at the same time base money rose is inconsistent with my understanding.

    ReplyDelete
  6. That is a good point Bill, you are being a better econometrican than me in interpreting my results. Yes, the results do show the effect of a monetary base shock holding all else equal(i.e. the model controls for the influences of the other variables). Therefore, they do imply as you suggest that had the monetary base been frozen since 2007 that the crisis would have been far worse.

    A good way to make your point would be to compare the present crisis to what happened in 1929-1933. Then, the monetary base grew at a mild pace while the economy was collapsing around it. Bernanke did not let that happen--and proably should have done more--and as a result the economy has fared better than it did during the early part of the Great Depression.

    Regarding your financial shock question, I only imposed restrictions on the model that allow me to pull out monetary base shocks. It is possible, however, to impose other restrictions that would allow one to pull out the effect of financial shocks. I just didn't do it.

    ReplyDelete
  7. Bill Woolsey:
    I would have thought that the reason the real GDP held up (ie only fell 2-3% and didnt fall 10% or more) was thanks to the automatic stabilizer function of fiscal policy. Around October 2008 the deficit dramatically increased
    Wonder if monetary base increase simply reflects the fiscal deficit ? (I'm guessing the folks at Kansas City would say yes!)

    ReplyDelete