“My beef with Scott Sumner” OR “Dude, you’ve already won the debate!”

Pleasantries: brilliant blog. Anyone remotely interested in monetary policy, a.k.a. “the only remotely possible technocratic method to manage an economy”, should read it.

Ok. Prof. Sumner is absolutely right that the Fed made a mistake in fall of 2008. This is illustrated by the sharp decline in inflation expectations as illustrated in this graph (stolen from here):

But Sumner often slips from this valid critique into a claim that the Fed continues to make a mistake (e.g. here: “Throughout much of 2009 I kept challenging liberals to push harder for monetary stimulus.”). This is not a valid argument. For example, the Fed made a mistake in the late 1920’s trying to prick what they thought was a stock bubble. Their actions helped precipitate the Great Depression. Flippantly: this doesn’t mean policy is still too tight today. More relevantly: its not the reason why policy was too tight in 1930, 1931, 1932 and early 1933. The mistakes they made in those years where independent of the original mistake.

The sharp up-tick in inflation expectations through 2009, as seen in the graph above, suggests the Fed is no longer making a mistake; policy is no longer too tight.

What to do about high unemployment (illustrated below)?
First, unemployment has — hopefully! — topped out. This is more evidence that policy is, if not appropriately loose, at least loosening. Second, we don’t know the mechanisms linking macro policies to employment. The Fed doesn’t (and can’t!) determine the unemployment rate in any given month even if completely ignored inflation. Its hubris to suggest that it can. Third, even if money policy’s effects on inflation expectations is immediate, in a world of very slow moving wages, there’s long lags between policy changes and employment changes. The current policy could be producing improvements in the unemployment rate maybe today, maybe in three months, maybe in nine months… we don’t know.

All this said, I support Sumner’s goal of moving the Fed towards history dependent policy (aka “level targeting”). This policy, if even implementable, would make mistakes like the one the Fed made in fall 2008 less severe. I’m just not sure why Prof. Sumner insists on arguing policy is still too tight today, nearly 16 months later, when he doesn’t need to in order to make a case for this policy change.

12 thoughts on ““My beef with Scott Sumner” OR “Dude, you’ve already won the debate!””

  1. Re: the fed making a mistake, it assumes that Fed, given sufficient monetary looseness could have supplied sufficient money supply to satisfy the increased money demand. Excess reserves already started increasing in September 2008, and only spiked further after the Fed loosened. (By September 1st, excess reserves had gone from a normal level under 2 billion to 60 billion. Lehman only failed later that month, so the housing crisis was already becoming a financial crisis). Given that the Fed’s expansion only increased excess reserves, what could the Fed have done? Basically, what is the counterfactual, what policy should the Fed have proposed, and why wouldn’t it have gone into more excess resserves, given that excess reserves have already increased by 30 times in a single month?


    2) Regarding 1929, the counterfactual to the Fed not popping the bubble is to let it expand further which would only make things worse. The bubble’s collapse played a bigger role than the interest rates at the time, which were still relatively low before the crash. The highest real discount rates got in 1929 was 6.1628, they were as high as 7.3520 in 1927. (See global financial database discount rates). The world has seen bubbles in the current crisis, the last recession, the Japanese depression, and in 1929 all being major contributors to massive contractions in output in each country. The Fed isn’t the main cause, bubbles that lead to major crises do.

    If you want to defend a EMH explanation for high stock prices in 1929, you have to explain why it didn’t reach the same (nominal!) level until late 1954. Some kind of new information needs to arise in October 1929 that means rationally expected fundamentals or discount rates change so drastically that asset prices only recover a quarter of a century later (!). Or you accept the much more logical choice that there was a bubble, that stock valuations deviated from any kind of reasonable valuation based on discounted fundamentals. Given the 1929 bubble, a crash was inevitable.

    I also support level-targeting, but I don’t think that we’re close to a policy loose enough to approximate level targeting. To counter the accumulated periods below the price level target (less than 2% inflation say), we need to have higher inflation to compensate to reach the price-level target. This is the main point of Eggertson and Woodford, and I believe Sumner’s point as well.

  2. Yeah, “the Fed made a mistake” is never made very precise. Here’s some possiblities:
    1. They control inflation expectations. Normatively, we should have constant expectations. Expectations dipped, therefore they must have made a mistake.


    2. Optimal policy is level targeting. Under level targeting, out of equilibrium, you can have wild swings in inflation, but play play the game… With credible policy, we know the Fed will do error correcting. Since its credible, the public knows this, so they won’t make dramatic changes in prices. Knowing this, the Fed doesn’t have make drastic error corrections. In equilibrium, inflation is stable. The Fed’s error is not level targeting.


    3. Optimal policy is a forward looking Taylor rule where the lower bound is dealt with by nonstandard policy instruments. The Fed made bad predictions about future inflation or output gaps. Implementation error.


    4. Policy is a Taylor rule, but the Fed’s mistake is to give up on monetary policy at the lower bound.

    Yglesia and Klein seem to be arguing the 4th point and in his more cogent moments Sumner argues something like 2. For some reason, Sumner seems to think that stoking the flames of criticism 4 will get the Fed closer to implementing 2. I don’t get it.

  3. “This is the main point of Eggertson and Woodford, and I believe Sumner’s point as well.”

    The problem I see with this is perhaps subtle and academic. A commenter at Sumner’s said I was being pedantic, or something. Anyway, the problem with this is that Eggertson and Woodford argue for a policy stance or maybe its better called a policy regime. Essentially, they say the IF the Fed and the public are playing the game described in my point #2 above, THEN level targeting is optimal. My argument all along is that the Fed and the public are not playing that game. The public understand the monetary policy via Taylor rules and interest rates and maybe, MAYBE!, an inflation target. Trying to switch games in mid-stream wouldn’t work.

    If the Fed could sit the public down to a mug of beer and explain their rationale for changing to level targeting and elicit head nodding and hand shaking, then we’d be ok. But they can’t. My reading of history is that the public comes to understand the game its playing with the Fed by observing the Fed over a longish period of time. (Have you seen Cogley’s paper with Sargent on learning from the Great Depression… I think their results suggest the GD *still* has a substantial impact on expectations or at least it did until very recently.)

    Put another way, I don’t think Eggertson and Woodford’s analysis applies to the period of time that the Fed and the public are transitioning to the new game. The examples from history of monetary regime change that pop to mind don’t suggest that doing so is a pretty thing in the short run.

  4. While I’ve lectured on the causes of the Great Depression to undergrads, I’d defer to you on this. I will say that in the original post, my point was that the mistakes made earlier by the Fed don’t necessarily have anything to do with mistakes made later.

  5. http://www.themoneyillusion.com/?p=3977#comment-13486
    “Ambrosini probably won’t read this, but I’ll respond to his post anyway. Money is still too tight for several reasons:

    1. Two year inflation expectations are still way too low.
    2. It’s NGDP, level targeting, that matters anyway, and we need lots of catch-up in a world where many wages (like public employees) are inflexible.
    3. Tight might today creates a severe recession, which creates a massive budget problem, which RAISES long term inflation expectation.

    I realize your argument is in some ways more persuasive to good economists than my messy argument. I’d rather defend your position. And I agree that the “sell by date” of my argument is approaching, but I still don’t think we are there yet. When things are this depressed you can get rapid recovery without a breakout of inflation. I lived through it in 1983-84. We can do it again. If I’m singing the same song in 2011, put me out to pasture.”

  6. First of all, calling one indicator (TIPS minus Treasuries) _the_ measure of inflation expectations seems premature to me. There might be other reasons for this spread (e.g. CPI-related issues, liquidity), and there are other indicators that have been used in the past, e.g. gold. Yeah, gold bugs might be nutty, but their expectations still count just like anyone elses. Second, as I have been harping on, we don’t truly understand monetary policy without understanding the role of the banks. I find it kind of crazy that monetary thought used to be all about M0, M1, M2… Friedman & Schwarz’ (and Bernanke’s) mechanism for monetary contraction during the GD was about bank failures… yet now even a historically minded guy like Sumner basically thinks the Fed is the sole power that determines the money supply.

  7. push- I see things a little better now about your disagreement with Sumner.

    My two cents- what does a Taylor rule mean at the lower bound? If the Taylor Rule is signaling negative interest rates, then that seems to suggest QE, but how much? Is there a mapping of monetary base to implicit negative interest rates, or do you just keep loosening until inflation increases enough to signal tightening? If it’s the latter, how rapidly do you loosen. I think these questions should matter in all cases 1-4, regardless of whether you want a Taylor-rule, level-targeting, or almost any policy rule.

    Re: the GD, I agree that the Fed’s actions before 1929 had nothing to do with subsequent mistakes post-1929.

  8. Thanks TGGP. If Sumner reads my comments he should know that I’m refreshing his comments every 5 minutes except for bathroom breaks.

    ssendam, I agree about TIPS spreads not being a perfect measure of expectations, but I think other measures show basically the same message. And yes, banks matter, but only instrumentally. Its like saying that Barry Bonds’ swing doesn’t matter, only his bat. I think you can talk about the mechanics of his swing without talking about the physics of his bat. Similarly, you can talk about money without talking about banks (or any mechanism for that matter). I know you’ll find that too simplistic, but it would be more useful if you could tell us why having a good story for the banks is necessary for understanding money. By that I mean that its easy to imagine a world without banks were monetary policy works exactly as it does now (e.g. imagine the Fed just has everyones money balance in a computer… actually are we too far away from this?).

    sraffa, good question about the Taylor rule at the lower bound. You can imagine some heuristic like you suggest, but I don’t know of any functions that translate negative interests implied by the Taylor rule to amounts/qualities of QE. Dissertation idea!

  9. Sure. What is the story we tell undergrads about what happens when banks make loans? The textbooks say that it gets redeposited and re-loaned out.. until the total effect is that base money gets multiplied by the inverse of the reserve ratio. In effect, banks “create” money by making loans. Now this never made sense to me the first several times I read it, but this was the mainstream of economic thought until a few decades ago. And so we have money multipliers, reserve ratios as instruments of monetary policy (China just did this: http://online.wsj.com/article/SB126331431653926353.html ), etc.

    Now this suggests that if banks somehow figured out how to make more loans, then maybe that would effectively create “more” money. Well we’ve had financial innovation, people took on second mortgages and HELOCs, the whole so-called shadow banking system… At the micro level, what happens to households when HELOCs become available? They get a bunch of cash right now, which gets repaid over time.. that sounds similar to the effect of a monetary injection.

    Finally, look at what the economic historians say (e.g. Kindelberger, or Alan Taylor’s recent paper). They talk about credit expansions and contractions and how they are related to booms and busts… well credit doesn’t actually exist in our standard monetary models, much less expansion or contraction. Everyone but academic economists take it for granted that credit matters, that financial innovation (and therefore the banks) can expand (or contract) credit. Sumner, I think, would simply say the financial crisis was a shock to velocity, but where does velocity come from? Does financial engineering increase velocity? I don’t think our understanding is complete until we know how this works.

  10. ssendam, thanks for your reply. I think credit is an important area of study. I think the mechanisms underlying monetary economics, like banking, are important areas of study. I just don’t think understanding them is necessary for understanding monetary policy.

    Suppose the Fed used a forward looking Taylor rule for policy (ignore the lower bound). If there’s a credit crunch happening, then this should increase the expected output gap and/or increasing the expected inflation gap. They’ll decrease interest rates as the Taylor rule directs, as they would do for any event that changed the output and inflation gaps in these directions.

    I suppose you could argue that we need to understand the underlying mechanisms (like banking and credit) so the Fed can do a better job at measuring expectations and forecasting. In other words, the Fed’s failure was like my #3 above and they need to avoid making such bad mistakes at forecasting.

    I’m skeptical that they can become better forecasters and maybe this is what’s wrong with forward looking Taylor rules. Perhaps, then, we need to find optimal policies that are easier to implement (like level targeting?). But in any case, I don’t think understanding the underlying mechanisms will make policy any better. That said, we might want to understand these mechanisms because its intellectually satisfying to do so, though.

  11. It is precisely because of the lower bound that study of credit is important. Gaps in our knowledge have caused huge timing and size errors in the implementation of QE.

  12. I suspect you have to reinvent macro for every item that is managed technocratically via government monopoly in an economy.

    Since money is effectively so in the US, we need a macro monetary policy theory. If housing were managed so, we would need a “macro theory of housing” and so on.

Comments are closed.