One of the tenets of market monetarism is that acivist fiscal policy is a waste of time, since monetary policy can do all the stabilisation that’s needed. On Twitter last night Joe Wiesenthal at Bloomberg wondered what MaMos would think of a strict balanced budget rule. Would that leave monetary policy able to achieve nirvana?
I thought: surely MaMos would accept that monetary policy works with long and variable lags, the phenomenon, emphasised by Milton Friedman, that this blog is named after? And that therefore we’d get macroeconomic volatility from imperfect control?
Believing in Friedman’s dictum would not undermine MaMoism, since if both fiscal and monetary policy worked with the same long and variable lags, and in the same way, ie if they were perfect substitutes as an instrument, then there would be no need for both.
However, Noah Smith pointed me to this post by Scott Sumner, the leading MaMoist, which contains the phrase ‘long and variable lags is a myth’. The argument in this post seems to me to be full of holes. So, while I wait for an anti-MaMo Matlab program to finish, I will explain why.
First off, it seems to make the argument that mainstream macroeconomists wrongly identify the interest rate (or the money supply) as the sole monetary policy instrument. And therefore – I think – it follows that all the VAR evidence showing how shocks to monetary instruments take time to have their full effect on variables central banks might care about is misguided.
Well, no. Certainly, if one maintains the argument that expected future monetary policy is potent [which, provided you believe in forward-looking expectations is reasonable], and one is prepared to contemplate that there are shocks to expected future policy, VAR based studies that identify only shocks to the contemporaneous instrument are not enumerating all the volatility in the data injected by monetary policy. [And in fact work by Gurkaynak et al, and others, seeks to identify shocks to both current and expected future interest rates in just this way]. However, it does not follow from this that the VAR incorrectly measures the effects of shocks to the instrument. Moreover, as the work that does identify expectations shocks shows, those also take time to have their full effects. In other words, they work with long lags too!
Second, as a purely analytical matter, noting that you can manipulate expected future rates to affect the economy doesn’t preclude that those manipulations, or changes in today’s instruments, take time to have their full effects. In fact, so far as the empirics tell us, both changes in today’s interest rate and changes in expected rates take time to have their full effect [on variables like inflation, GDP, etc].
I’ll attempt an analogy.
If I am playing chicken timidly, I might be able to influence the trajectory of the other car by calling the driver’s mobile, and explaining that I will turn out of the way, perhaps persuading the driver to turn her car back towards mine. However, the other driver might take time to process what I say, and it will still take time for my turn of the steering-wheel to have its full effect on the path of my car.
I’m not sure how well that analogy went. But at any rate, the phenomenon is true of macroeconomic models of policy and the associated empirics.
And the corollary is that because the economy will be buffeted by shocks, and it takes time to respond and counter them, no policymaker would be able to generate stable outcomes for goal variables.
Backing up, I don’t accept, of course, the premise of this line of thinking at all, that monetary and fiscal policy are perfect substitutes, and therefore that the former can be dispensed with. But it’s worth bearing in mind that the logic gone through here works for fiscal policy too, both in the models, and in those attempts to identify shocks to fiscal policy today and in the future.
Sumner cites Woodford and Krugman as commenting on the potency of expectations, and uses this in support of his thesis that changing expectations changing things refutes the long and variable lags thesis. But I am quite sure neither of them believe any such thing. Estimated versions of Woodford’s model (for example, the original Rotemberg-Woodford model) behave just like my account above. And Krugman is a firm believer in sticky prices, talking interchangeably between IS/LM and New Keynesian models. Which behave just as I’ve explained above.
The only model I know where monetary policy has its entire effect instantaneously is the flexible price rational expectations monetary model. And in this case there is no point in monetary stabilisation policy at all. Money has no short-run effects on output. Optimal policy in this model is to set rates at zero permanently, obeying the Friedman Rule. If there are real frictions in this model, like financial frictions, there will still be a role for fiscal stabilisation, however.
I’m sure these mix-ups would get ironed out if MaMos stopped blogging and chucking words about, and got down to building and simulating quantitative models. Talking of which….
[Update: Scott Sumner replies on The Money Illusion here, including some priceless phrases about how his research found that there were in fact, no lags at all between monetary policy changes and their effects, and some other collectibles about there not being a NGDP futures market.]
longandvariable 
Reblogged this on DAMIJAN blog.
It seems like any expectations focused theory will have trouble demonstrating effects due to monetary behavior. A given interpretation might result from a variety of data points and Fed actions, and the same holds for a given Fed action in the context of a given set of data points. Despite the potential for circularity, there do seem to be events where a focused central bank quickly developed credibility, i.e. enacted policy without long and variable lags. FDR’s repricing of the gold standard qualifies. Sweden’s ability to reprice their exchange rates may be another. The EU’s 2011 tightening may count. Of course, interpretations of interpretations are prone to biases, but don’t these events get your investigative antlers up?
*Switzerland’s
I think you are misreading Scott’s post. But I will let him respond.
Personally, I think that if there is a change today in expectations about monetary policy (say) 10 years from today, that financial asset prices will respond today, real output and employment will take months or a year or so to change much, and the stickier prices and wages a bit longer still. But a lot of things will change long before the 10 years is up. So it looks like a lead.
Which (I think) is what you are saying.
(It’s normally “MMs” by the way.)
Aww, I like MaMos, let me have it.
[Thanks for engaging.]
Agree with all that.
One can use the semantic ‘lead’ to mean that, but still the ‘lags’ that shifts in expectations or current policy work at, in models that we have that fit the data reasonably well, mean that the economy necessarily experiences highly imperfect control.
“And the corollary is that because the economy will be buffeted by shocks, and it takes time to respond and counter them, no policymaker would be able to generate stable outcomes for goal variables.”
That’s where I think I would disagree a bit. Because some policy rules will have much stronger automatic stabiliser properties than others (e.g. price level path targeting vs inflation targeting). Commitment to a rule where the policymaker will respond in future to current shocks (even if there is, for example, a lag in the data) can help reduce the immediate effect of those shocks, even if the policymaker does not observe the shock until it’s past.
But *perfect* stability may not be attainable, of course.
(Though I doubt I’m saying anything you didn’t already know.)
“I’m sure these mix-ups would get ironed out if MaMos stopped blogging and chucking words about, and got down to building and simulating quantitative models.”
Wouldnt they just build models that agree with their school of thought, like all model builders?
You might be right that some have ulterior motives in building their models. But the beauty of this way of proceeding is that it doesn’t matter. Once it’s built we can all inspect how well it does at matching the data and compare it to other competing models.
Sumner engages in metaphysics, see his blog post on this topic raised by Yates. Expectations fairy is the key. But this is untestable.
I don’t think it is completely untestable. We can measure expectations and detect their rationality, or lack of it. Usually, such measures fail pretty badly. There are ways of rescuing RE despite these apparent failures, since all tests involve auxiliary assumptions too, but the totality of the evidence makes these rescues highly far-fetched in my view.
I think your key point is that economists have looked at indicators of expected future monetary policy and estimated that shocks to expected future monetary policy work with lags.
I think this is more convincing that the fact that standard models imply longish lags.
I will now attempt a partial defence of Scott Sumner. Standard new Keynesian DSGE models are designed to give moderately long lags. The delay (based on VAR estimates) is a stylized fact which the models try to reproduce.
Here I quote from Smets-Wouters 2007 AER VOL. 97 NO. 3 pp 586 etc
“Modeling capital adjustment costs as a function of
the change in investment rather than its level introduces
additional dynamics in the investment
equation, which is useful in capturing the humpshaped
response of investment to various shocks.”
I suppose that I should make and simulate a model before typing this, but I dare guess that, without habit formation in consumption and costs of adjusting investment, effects on output and employment would have a start large and decline impulse response function not a hump shaped impulse response function. In any case, I am sure it is possible to design a model with sticky prices and instantaneous effects monetary policy or changes in expected future monetary policy.
I guess a shorter version of this is “what dannyb2b said”. It is possible to invent a model with any properties. To answer questions about the real world, it is necessary to assess models and see if they fit the data and give good out of sample forecasts. Of course, this is what mainstream new Keynesian macroeconomists attempt to do — the theory involved a good bit of work, but the assessment (so far) required a huge amount more.
Of course “fiscal policy” has a role in MM policy. During recessions, (periods of inadequate demand when shocks have resulted in actual NGDP being below target) it is likely that short and long term interest rates will have reduced. This will make the NPV of some governmental activities which have present costs and future benefits positive that were not positive at higher interest rates. It is also likely that a gap may open between market prices of some inputs into these activities and their opportunity costs which will further increase NPVs. If governments respond rationally to these changes in NPVs, increasing expenditures during recessions. this will work together with monetary policy to restore NGDP to its target level. “Fiscal policy” consists in continuing to apply proper cost benefit analysis to government activities and avoid “austerity,” acting as if governments needed respond to recessions like credit constrained consumers.
Fiscal policy isn’t at issue here. The balanced budget rule was just what got us into speculating whether, with that, monetary policy could or could not stabilise the economy according to MMs.
From what I’ve seen, Sumner claims that any expansion in fiscal policy will be automatically ‘offset’ by more contractionary monetary policy. This contradicts your comment I think, thutcheson.
Even if “offset” (no immediate effect on GDP) it would still be optimal to undertake, as a properly done cost benefit analysis would mean that the fiscal deficit resulting from investments that passed the NPV test would be more valuable than the other expenditures that are “crowded out.” But I assume that the period of “recession” corresponds to a time when NGDP is still below target so the monetary authority would not take additional measures to “offset” the expenditure. Of course if NGDP reacts with “short and invariable lags” to purchases of NGDP futures, the period of “recession” would not last long. However, it would be good to get Sumner’s view on that.