Wednesday, March 27, 2013

Time to Update the Priors: Not All Policies Matter for All Outcomes

. Wednesday, March 27, 2013

In response to my previous post Mark Thoma and I had a bit of a roundabout on Twitter. His first point is that he wasn't addressing our article in his blog post. I mentioned in my post that that was obviously the case, even though he quoted Farrell's summary of it before launching his discussion on unrelated topics. But through the Twitter discussion several interesting topics were raised. Thoma asked me two main questions:

1. Is it your argument that peripheral crises can never lead to systemic crises?

2. Is your answer to #1 completely independent of crisis response policies?

My response to #1 is that the likelihood is exceedingly low. I base that on the fact that not a single peripheral crisis* from 1970-present (using the IMF's database) has turned into a systemic crisis. Obviously crisis management policies have varied quite a lot, but the global (i.e. systemic) outcome has not.

That leads me to believe that policy considerations are not very important from the perspective of preventing systemic contagion from peripheral crises. In peripheral crises, policy matters a lot locally -- i.e. for the peripheral country experiencing the crisis and maybe one or two others connected to it -- but not systemically. In systemic crises, policy in the core matters a lot globally; policy in the periphery matters less**.

Thoma disagreed. Obviously that's fine, but I'd like to explain why I think he's wrong. Since I've heard others make it in plenty of other contexts, I thought I'd make the point here. These Tweets of his seems to sum up his view, which I believe is a common view:

We went around like that for awhile. So here's my position: if policy mattered for global outcomes things would likely be much more worse right now than they are. Why? Because the Cyprus deal is basically the worst possible policy from the perspective of preventing contagion. Cyprus has been hacked off from the global financial system. Capital controls are in force. Its entire banking system has been closed for two weeks, its second-largest bank will be closed permanently, and its largest bank will be forcibly restructured. Foreign depositors, bondholders, and shareholders are being gutted.

If you were trying to start a systemic crisis, this is the way you would do it: repudiate a huge chunk of claims, and close the capital account. And yet markets around the world are up today. Even in Russia.

Or take the examples of Iceland and Ireland. Iceland repudiated the debt of its banks, imposed capital controls, and told international investors to take a hike. Once again, this is a recipe for contagion yet systemic crisis did not result. Ireland did the opposite: it guaranteed the debt of its banks, did not institute capital controls, and paid off international investors. Systemic crisis also did not result. The opposite local policy response produced the same global outcome. Only the local outcome varied.

Contrast those cases (and all the other eurozone cases, and Argentina, and E Asia, and etc.) with the US in the Fall of 2008. A couple days of dithering -- of the sort that the eurozone has made its speciality -- lead to an immediate and profound downturn in global markets, including the largest single-day evaporation of wealth in absolute terms in history. The US tried to kick the can down the road, but couldn't because it is the core node; the EU has been able to repeatedly kick the can down the road because those crises are in the periphery.

I conclude from this that policy always matters locally, but it only matters systemically when the crisis is in a core node. No matter what the policy response to peripheral crises is, systemic contagion is exceedingly unlikely. I am prepared to hear arguments counter to this, but they must go beyond assertion or "I'd rather not chance it". Evidence is preferable, but I'd even countenance an evidence-free logical argument with a clear causal mechanism. These are every bit as rare as the claim that systemic outcomes depend upon local policies is common.

*Note that we're making it hard on ourselves by calling essentially every country but the US and UK "peripheral". This is not a common view (or was not prior to the crisis). Many thought that Iceland, Ireland, Cyprus, and other tax havens were significant global financial centers (or rapidly becoming so).  We're not talking about Somalia here... we're talking about OECD countries. Yes, we realize this is a bold claim. We think we have evidence and argument to support it.

**Remember that Spain was winning awards for prudential regulation a couple of years before the crisis. Ireland was held up as a model too. In the end that didn't help them.


Latinamericanist said...

Let me apologize in advance, I haven't read your paper yet (it looks great, though, & I will) so this may have been covered.
I don't have a strong view of this either way and am perfectly willing to be convinced. The strongest view of contagion would be the "Commanding Heights" view of the ripple effects of the Asian crisis:
Asian Crisis --> Russian and LA crisis --> LTCM crisis which by all accounts was very close to being systemic.
With which step in this do you disagree?

Kindred Winecoff said...

Confession time for me too: I've never seen Commanding Heights! I know, I know. So I don't know the particular argument made there, but I know general arguments made about that event.

Anyway, we do talk about Asia in the paper. We think it constitutes the strongest case against our position, so we take it seriously. First, we show that the claim that Asia had a broad systemic (as opposed to local) effect does not appear in equity market indices or economic growth rates. (Or aggregate trade statistics, but that got cut from the final version of the paper.)

Okay, but maybe we just got lucky. LTCM *could* have been worse, right? I don't think so, but even if it was it wouldn't contradict our argument. So I'd disagree with *every* step in that chain.

First, the Asia crisis was an exchange rate crisis due to balance of payments issues, not a banking crisis, and it did not transform into a (global) banking crisis. The drop in demand from the Asia crisis may have had an impact on Russia, but that was also an exchange rate (not banking) crisis. In other words, there was no *financial* contagion, in the sense that firms collapse and are unable to make payments to their counterparties, thus rendering the counterparties insolvent as well.

So it comes down to LTCM, which was a) not a bank; b) lost $4bn total (ie less than JP Morgan lost on a single trade -- the "London Whale" -- last year); c) was easily absorbed by other American financial firms, which is the *opposite* of contagion.

So we think that this episode shows the strength of our account: the US financial system was able to absorb these losses quite easily, and thus prevent a further spread of crisis. Remember that the late-1990s was the Golden Age of US growth in both the real economy and the financial sector.

What was the risk of a systemic crisis resulting from these episodes? We cite some literature which estimated probabilities, and found that it was very low (less than 1%). This is borne out by the fact that there was no general run on US banks, nor an uptick in bank closures, in the years during and after the Asia crisis.

But if it had led to a broader crisis, it would not have been the result of financial contagion from Asia or Russia. There was no financial contagion from those places. LTCM did not go down because it was tied to financial counterparties in Russia or Asia; it went down because it made bad bets and lost. It is thus more analogous to Bear Stearns than anything else: if financial contagion had started, it would not have originated in Russia/Asia, but in the US.

Latinamericanist said...

I think that's a good answer. I'd say perhaps the weakest part is the claim that LTCM was easily absorbed. In the conventional telling - be it "when genius failed" or "commanding heights" - the whole thing was a close shave and disaster was only averted by the good connections and crisis management of Rubin and his team.
I'm inclined to believe that those accounts tend to overstate events in favor of the narrative (not so much unlike the election narrative story in your DeLong post) but that is the narrative you're working against.

Kindred Winecoff said...

I don't know about the truth of the narrative. It doesn't seem so hard to get 10 bankers in a room and tell them that if they don't pony up several hundred million each to buy LTCM then they'll have to pony up several hundred million each (or more) to take losses on LTCM. The terms of the bargain were probably a bit contentious -- Bear Stearns refused to participate -- but in the end they all made money on the deal, and all of them could easily afford it.

Time to Update the Priors: Not All Policies Matter for All Outcomes




Add to Technorati Favorites