I did a recent post criticizing Bernanke’s defense of paying interest on reserves. This policy was introduced in October 2008, and even the Fed viewed the policy as contractionary in intent. Indeed Susan Woodward and Robert Hall called the Fed’s explanation a “confession of the contractionary effect of the reserve interest policy”. The term ‘confession’ is rather telling, as it implicitly pushes back against the widespread view that the Fed was doing all it could in 2008 to stimulate the economy. Today I’d like to discuss the market view of IOR.

Back in 2010, Louis Woodhill suggested that the Fed announcement of interest on reserves, as well as two subsequent increases in IOR, had a very negative effect on the stock market:

A valid way to gauge whether events are “good” or “bad” for the economy is to look at the stock market’s reaction to them. The day that Lehman Brothers collapsed, the S&P 500 went down 4.71%. Three days later (i.e., at the fourth market close after the event), the S&P 500 was down by a total of 3.61% from its pre-Lehman close.

At the time of the Fed’s IOR announcement, the S&P 500 was down by a total of 12.18% from its pre-Lehman close, 15 trading days earlier. However, the day that the Fed announced IOR, the S&P 500 fell by 3.85%, and it was down by a total of 17.22% three days later.

On October 22, 2008, the Fed announced that it would increase the interest rate that it paid on reserves. The S&P 500 fell by 6.10% that day, and it was down by a total of 11.11% three days later. On November 5, 2008, the Fed announced another increase in the IOR interest rate. The S&P 500 fell by 5.27% that day, and it was down by a total of 8.60% three days later.

I have some serious doubts about Woodhill’s way of doing event studies, particularly the four-day windows for each shock. But before getting into interpretation, let’s do some math:

1. If we take the three negative IOR shocks cited by Woodhill, and look at the four-day windows that he describes, the declines in the S&P500 are 17.22%, 11.11%, and 8.60%. That adds up to 36.93% decline. But there is a compounding factor that (I think) reduces the total declines to about 32.75% (someone tell me if my math is wrong.) The intuition is that two consecutive 10% drops at up to 19%, not 20%.

2. Of course 2008 was a catastrophic years for the stock market, as this is when the Great Recession got going. The total decline in the S&P500 from December 31, 2007 to December 31, 2008 was 38.5%.

3. Thus the overwhelming majority of the stock market decline of 2008 took place in twelve trading days, immediately following three contractionary IOR announcements, and only a small part of the decline occurred during the other roughly 240 trading days.

What can we make of all this? Let me start by criticizing Woodhill, then defend him, and then give you my own view.

Let’s start with the four-day windows. In an event study, a one-day window would be more appropriate. Why should the market reaction have taken four days? That time frame seems cherry-picked. Yes, even the single day drops were pretty large, much larger than a normal single day movement in the S&P500. But this was a very tumultuous period for the economy and the stock market, and large daily moves were pretty common in late 2008. So this is not statistically significant.

If I were to defend Woodhill I’d point to the fact that interest on reserves was a new and unfamiliar policy. It was not well understood by the markets. Indeed it wasn’t even well understood by the Fed (which is why adjustments had to be made in October 22 and November 5, to make the policy more effective.) The Fed certainly wasn’t loudly publicizing the fact that it was contractionary, you had to read the fine print. Perhaps the markets noticed the effects of IOR were contractionary. Thus over a period of several days they noticed monetary conditions tightening as banks were less anxious to move excess reserves out into the real economy, given that they were now earning more interest on excess reserves.

And yes, 2008 was a very volatile period for stocks, but a pretty big share of that volatility came in the twelve trading days cited above, when most of the total decline of 2008 occurred. So IOR may well have caused some of that volatility.

On the technical question of event studies, my views are somewhere in between, but a bit closer to those of Woodhill’s critics. I’m not comfortable with the four-day windows on stock prices. I also recognize the extreme volatility of stock prices in 2008, having seen the same thing in my study of the Great Depression.

However I still put some weight on Woodhill’s argument. In the 1930s, some of the very biggest stock price movements occurred immediately after monetary shocks. Thus the largest 2 day stock rally in American history occurred right after Hoover announced a change in gold policy that would lead to 1932’s QE policy, and the next day Congress signaled its approval. There are too many such “coincidences” to be dismissed.

And suppose you are one of the Fed people who think that IOR helped the Fed to achieve its 2008 policy objectives. The fact that almost the entire stock market crash of 2008 occurred in just a few days after these three IOR announcements certainly doesn’t give much reason to think IOR helped the economy.

For what’s it’s worth, I think there’s about a 10% chance that Woodhill is basically right, in the sense that at least half of that 32.75% stock decline was linked to tighter money, particularly IOR. But that’s still really bad news for IOR! Suppose you were told that some foreign policy move would lead to a 10% chance of a nuclear exchange with North Korea—would you be reassured that 10% is a low probability?

The Great Recession is the economic equivalent of a major foreign policy disaster. If there’s even a 10% chance of IOR having caused this disaster, that’s really bad.

I also think there’s about a 50% chance that at least half of the stock declines over the three one-day windows were due to IOR. This is partly because in later years we saw IOR (including negative IOR) clearly impacting stock prices in various countries. And even those three one-day windows add up to a bit over 15%, or slightly less with compounding. That’s still huge! How would you feel if the Dow fell 3000 points in the next three days?

So even in a world where Woodhill is only half right, or even 25% right, he’s still basically right. Interest on reserves was a huge policy mistake. And I think there’s at least an even chance that he is at least 25% right.

Screen Shot 2017-03-21 at 12.34.25 PM.png

Update: Commenter “dlr” presents some very powerful counter-evidence against this post:

Like you, I believe that monetary policy was the most proximate cause of the 2008 crash. But I think this theory about IOR is way more unlikely than you do, and I would give less than a 1% chance that this data-mined version of an “event study” accurately portrays the information available from markets.

October 6, 2008. Europe was down 5% and S&P Futures were already down 2.5% at 8:15am when the Fed announced IOR. Futures rose on the release before declining again, and were still down 2.5% after trading started, before finishing down 3.8%. Over the weekend, both Fortis and Hypo Bank had to be rescued, and banks in every market were dropping heavily before 8:15.

October 22, 2008. The Fed IOR announcement came at 10am. The S&P was *already down* by 4.1% before the announcement. By 11am it was 1.5% *higher* than its 10AM tick, before closing down 6.1%.

November 5, 2008. The Fed IOR announcement came at 10am. At the time of the announcement the S&P was down 1.5%. In the 30 minutes subsequent to the announcement, it immediately rose to almost flat on the day. It did not start dropping below its pre-announcement level until after noon.

There was never a single Fed IOR announcement that was immediately followed by sharp drop in markets. This seems like it should be extremely persuasive counter-evidence to someone like you, who favors well run event studies the focus on trading just after the announcement before infinite confounding variables enter and take the “event” out of event study.