David R. Henderson  

What Caused the Panic?

Will Wilkinson on Inequality... From Poverty to Prosperity<...

In one of his posts today, Bryan asks us to consider Scott Sumner's thoughts in response to an earlier post by Bryan. (BTW, like many of the other bloggers, I'm now hooked on reading Sumner's thoughts. He has that perfect combination of thinking, knowledge, and tone that many of us strive for.) Here are my thoughts. They relate to points made by two of the commenters on Scott's post, Bill Woolsey and Greg Ransom.

First, Woolsey pointed out that Scott Sumner needed to make clear which panic he was talking about. In his comments, Woolsey writes:

I don't know what Caplan meant, but when I read the original exchange, I think you were way to quick to interpret "panic" with the stock market rather than with the financial problems that occurred before. You know, the difference between LIBOR and T-bills jumped by 300%. Crisis!!! The Great Depression is coming if we don't bail out Wall Street. No one is goign [sic] to be able to get any loans. No car loans. No loans for small business. So, Congress needs to let us borrow $900 billion to jump start the market for mortgage backed securities. That panic.

Not the rapid drop in the stock market that happened next.

That's what I remember as the panic too. And among those who panicked were Messrs. Paulsen, Bernanke, and Lazear. As I wrote on September 28:

On a Sept. 23 White House conference call, Lazear told listeners that what really led to the belief in a bailout was credit market conditions the previous Thursday, Sept. 18. Credit markets, he said, had frozen. I asked him how he could make strong conclusions about the future of the economy based on data from a day or two. His answer was that the negative returns on short-term Treasuries were scary.

As for the stock market crash, commenter Greg Ransom has a plausible explanation:

I recall one other tiny little bit of news -- about Obama taking a commanding lead in the Presidential polls. In late Sept. 2008 McCain "suspended" his campaign, then "unsuspended" it, and his poll numbers dropped like a rock, after coming neck and neck with Obama after McCain's Palin VP nomination.

Scott Sumner responds to Ransom as follows:

Greg, Good point. Unlike some of the other factors we have very accurate data on the impact of presidential elections. A win for Democrats knocks about 2% off stock values. I imagine that the odds of Obama winning rose by less than 50 percentage points in late September, meaning that the negative impact (although real) was less than 1 percentage point of stock values.

But there's a large problem with this response. Scott Sumner is generalizing from past data on Democrats rather than reasoning on the basis of this Democrat, Obama, and his plans. Maybe the market anticipated just how much he would try to expand government. One could argue that the market couldn't anticipate that, but Scott can't reasonably argue that. In that same post, he writes:

You can be sure that by the time we read that the Baltic shipping rates have plunged 90% (which I recall was in the papers last fall) the markets already knew about it.

In other words, markets are good at incorporating data very quickly.

How do the two connect? By lobbying for huge government discretion, Bush destroyed what little was left of the credibility of the small-government contingent in the mainstream Republican Party. The House Republicans and some Democrats held off the bailout on the first vote in late September and then caved on the second vote in early October. That signaled that there was even more running room for Obama.

Comments and Sharing

CATEGORIES: Macroeconomics

COMMENTS (12 to date)
frank cross writes:

Maybe Obama was worse than the mean Democrat, but was he many times worse? And of course the market would only compare him vs. the Republican and the Republicans had abandoned fiscal discipline long before the bailout

Steve Sailer writes:

One quibble on the timeline: McCain announcing he was "suspending" his campaign was very much a response to the panic in the financial markets.

Jesse writes:

Unlike some of the other factors we have very accurate data on the impact of presidential elections. A win for Democrats knocks about 2% off stock values.

Wow. Really? I feel like I'm watching a drunk looking for his keys under the lamppost. This is wrong on so many levels.

1. What data is this 2% estimate based on? The precise time frame is very important. After all, we all "know" that the stock market has actually risen more under Democratic Presidents than under Republican Presidents.

2. Stock returns are highly volatile and we only have a handful of data points, distributed over a long period of time. For example, prior to Obama only two Democrats had been elected President in the previous four decades. And of course political parties, economies, and stock markets all change over time. Given the combination of these factors, should we have any confidence in whatever point estimate we calculate for this exercise, however we do it? Obviously not.

I thought economists were trained to draw inferences logically from the data -- and to understand how challenging this task can be. That Sumner thinks "oh yes, 2%, that's the stock market reaction to a Democratic President" is a give-away that he's wandering around in the dark.

Charlie writes:

The causation ran the other way. The economy was tanking => the stock market was tanking because the economy was, these two factors gave Obama increased popularity, as he consistently won on the economy and McCain as part of the incumbent party was linked with the economic failure.

Charlie writes:

P.S. - The hypothesis is testable. There are several blips over the course of a campaign. If David is right these blips should statistically be associated with changes in the stock market. We can use the intrade prediction markets and even separate Obama from other Democrats during the primaries. My bet is that the test would fail, but I encourage someone to try to prove me wrong.

P.P.S. - Isn't this why Austrian's don't like statistical analysis though? After all, if your hypotheses are usually wrong, it's no fun to test them.

Pat writes:

Can you briefly summarize NGDP targeting? I can't get my head around Sumner's stuff. If there's a productivity shock that causes a recession, what good does a nominal GDP target do - inflation would be more volatile. I prefer an inflation target and let NGDP be whatever it is.

I could be all wet here - looking for help. Thanks

Nathan Smith writes:

It seems to me the hypothesis would have to be something like the following:

(a) Markets aren't ordinarily all that scared of Democrats. Yes, Democrats want to grow the government, and big government is bad for the economy. But opposition pressures Democrats to compromise right, Republicans to compromise left; and sometimes the ruling party compromises *past* the center. All in all, it's a wash.

(b) But having a Democrat come to power in the middle of a crisis is another matter, because a policy equilibrium is destroyed. And Obama was much more of an unknown than any other politician, having no record whatsoever and running on a completely vacuous platform of "hope" and "change." In normal times this lack of substance would mean that he probably couldn't move things very far from the ordinary political center. But at a time of crisis-- and Paulson's amazing power grab turned it from a financial crisis into a political-economy crisis-- Obama's unknown preferences could move policy a lot.

If this is the hypothesis, the tests Charlie proposes wouldn't work. The "running room for Obama" didn't exist before the September/October crisis. I'm sympathetic to the hypothesis, but even if Charlie's tests turned out to hold up statistically I would hesitate to regard that as evidence.

Jared Barton writes:

If Nathan's objection to Charlie's hypothesis is correct, then there are two ways of setting up a vector autoregression of, say, the percent change in the S&P 500 and the change in Obama's lead over McCain that should address this concern:

1) start the regression on September 24, 2008 (when McCain suspended his campaign to deal with the crisis, which we could call the start of the crisis), or

2) create an exogenous variable to the VAR for crisis (crisis=1 after 9/24/2008, 0 before), and add it to a model.

I have done both of these, as well as not worrying about your objection at all (which I think is the right answer, btw), using the nationwide polling data from realclearpolitics.com. I use the last day of the poll as the "poll's date", and average across polls that come out on the same day. Also, as many polls finish on days where no trading occurs, I use the next closest trading day. These are, in other words, fuzzy data (and that's without worrying about sample size, whether the question included leaners, and so on). I also use the S&P 500 data, obtaining from datastream.

At any rate, here are the results:

1) just doing the VAR from 9/24/2008 through Election Day, with 5 lags, indicates that the S&P 500 swings have no impact on change's in Obama's edge, but *changes in Obama's edge negatively impact the S&P 500*. This VAR, sadly, only has 11 observations, but is evidence in favor of the Mr Ransom's hypothesis.

2) doing a VAR for the entire period of data (going back to late 2006, actually, with significant gaps) and controlling for the 'crisis' period shows absolutely no impact of either variable on the other, though S&P changes are serially correlated with themselves. The crisis variable is just shy of significance at the 10 percent level for explaining the S&P 500, which is a shame, since we defined crisis, essentially, as "when the stocks started tanking!" You'd think it'd be significant, right? Negative. This VAR has 58 observations.

[It's worth mentioning that, as I'm using the actual date, coded in Stata, as the time variable, there are serious gaps in the data, so we have few observations. I have thought about treating observations in sequence with no regard to time difference and just looking at 2008, but wanted to report this first.]

3) without controlling for the 'crisis' period in any way, neither variable impacts the other at all. There are still 58 observations in the VAR.


a) we are picking the point of 'crisis' arbitrarily; that's why I don't like the crisis approaches. At all.

b) the polling data are collected over several days, but I treat them as if they happen on one day. It would be better to use inTrade data, but unless somebody can get it *for free*, I don't think it's worth $50 to settle a dispute in a blog's comments. If somebody does, though, I'll play with that data...

c) if you wish to reply to my analysis, please note that I am totally disinterested in what you *think* is the relationship among variables. Please try *going and getting some data and looking at it*. Even if you don't know how to run (and interpret) vector autoregressions, you can always start with pretty pretty graphs; I have great respect for even the simplest of graphs! [Oh yeah--if somebody has access to eViews, I'd love to see the impulse-response functions from these things; forgot how to do it in Stata.]

d) if anyone would like to play with these data, I'll send them along. I haven't saved my code to transform the data, look at graphs, and run the VARs, but I'll put them in if you like. Just post your email in the comments in a way that a Spyder can't get it but I can.

Jared Barton writes:

PS- dammit. Please disregard the above results. My lag operators are screwed-up. This is what one gets for rushing...

frank cross writes:

There is a study based on the intrade markets, by Justin Wolfers, that in some cases goes minute by minute. He did the Reagan election and others, I think. He found that Republicans are better for the stock market and Democrats were preferred by the bond market. Though the magnitude of the effect was not great.

Joe Calhoun writes:

For what its worth, I'm an investment advisor and I started getting calls from clients when Paulson and Bernanke testified to Congress. The calls ramped up after Bush's speech. Clients, who normally pay no attention to markets (that's why they hired me after all) wanted out of the market. They were scared and didn't want to hear anything that didn't involve getting the hell out of the market right now. It wasn't the credit markets or Lehman that caused the panic; it was Bernanke, Paulson and Bush.

Scott Sumner writes:

It is possible that a Democratic victory had a large effect this time around, but I think there are a number of reasons to think it wasn't a decisive factor in the stock crash.

1. McCain's economics were very erratic, and probably didn't inspire confidence on Wall Street. Didn't he vote against the Bush tax cuts?

2. The markets probably expected a Democratic victory even before the crash (according the betting odds) so the shift of odds during the October crash wasn't that great.

3. The historical data includes elections (1932, 1984) where (the markets knew that) there was a large ideological difference between the candidates.

4. The stock crash was associated with movements in other markets (such as commodities) that suggest traders were very worried about aggregate demand. Although Obama may be bad for business, I don't think anyone expected him to produce less inflation that McCain. He was certainly pro-stimulus, and McCain is a deficit hawk.

In support of your point, it can be bad for the economy to have policy adrift at an inflection point. The economy did very poorly in the long interregnums of early 1921 and early 1933, when no one was in charge.

One final point. What everyone "remembers" is what the media was talking about. Because of my unusual take on monetary policy I focused on different indicators. I was aware of the financial problems, but I was terrified in early October when I read about all the ominous reports from all over the world of collapsing commodity prices, collapsing trade and collapsing new orders for goods. That's what falling AD looks like. And we now know the collapse began in August, well before Lehman.

Comments for this entry have been closed
Return to top