An Email Exchange on Efficient Markets
The following is a slightly edited version of an email conversation
between Eugene Fama (perhaps the most prominent financial economist alive
today), and myself (Ivo Welch), with John Cochrane and Jay Ritter weighing
in, too.
Some background: Eugene Fama was instrumental in pioneering the concept
of Efficient Markets. Richard Thaler, who is repeatedly mentioned in the
exchange, is a leading economist scholar who is proposing a view of finance
that argues that markets are less efficient. This is sometimes called
"behavioral finance," although "imperfectly rational finance" would be a
more apt description.
From: Eugene F. Fama
Date: 10/14/2002
To: (long list of recipients, incl. Ivo Welch, Jay Ritter, Richard Thaler, etc.)
Hi guys:
I enjoyed your review of the IPO literature in the August JF. But I think
it's spoiled by repeated unsubstantiated use of the term "internet bubble."
What exactly do you mean by a "bubble" and what is the scientific evidence
that this was one? I'm surprised that this term didn't ring a bell with the
editor as a blatant example of loose talk that shouldn't be allowed in a
scientific journal. Except on this point, you are clear when you are
stating personal opinions, and I have no quarrel with that.
Regards,
Eugene F. Fama
From: Ivo Welch
Date: 10/16/2002 10:40am
To: (same long list)
hi Gene: Thanks for your comments. I think you are correct, in that we
should have stated once that we are using the term bubble just loosely. The
justification is that the reader (public+profession) already has the
appropriate association. Even you knew what we meant (plus, it seemed to
have kept you awake! ;-) ), and it must have been easier reading "during the
Internet bubble" than a constant reiteration of "slowly beginning with the
Netscape IPO of 1995, happening mostly in the tech sector, Nasdaq increasing
from about 1000 in 1995 to 4600 in March 2000, and then descending back down
to 1200 today."
However, I am curious myself now:
- What sort of evidence would it take for you to conclude that a bubble
exist(ed)?
- (And, vica-versa, what sort of evidence would it take for you to
conclude that no bubble exist(ed)?)
PS: Mind you, I believe that there are no bubbles where money-making is
easy. We are talking about prices deviating from underlying fundamentals,
and exploitation being difficult due to the long-run nature, high
volatility, and high transaction costs. As one of your former students, may
I suggest that we adopt a different taxonomy of Market Efficiency now?
true believer: prices are always equal to fundamentals <-- Gene Fama ;-)
mild believer: prices sometimes deviate, but exploitation is very difficult. <-- Ivo Welch
unbeliever: prices are often so far off that great bets or even arbitrage
pop up all the time. <-- Richard Thaler ;-)
PPS: I am trying to poke fun at you, too, Dick.
From: Eugene F. Fama
Date: 10/16/2002 12:11pm
To: Ivo Welch
Gene's comments were interspersed. Therefore, this email was edited.
(In response to my claim of describing mostly the period.)
You might have explained that this is all you meant by a bubble.
> However, I am curious myself now:
- What sort of evidence would it take for you to conclude that a bubble exist(ed)?
- (And, vica-versa, what sort of evidence would it take for you to conclude that no bubble exist(ed)?)
A rigorous definition of a bubble and empirical tests on more than one episode.
From: Ivo Welch
Date: 10/16/2002
To: Eugene F. Fama
Hi Gene: I agree that we should have stated at least once upfront what I
mean by use of the phrase "bubble" (although I suspect that Jay R. would
have protested), and that I believe it also was a bubble, but that I have no
scientific proof that it was such. I like your definition of bubbles as
deviations from suitable long-run expected rates of returns. John's is also
interesting and related. Can I reiterate my suggestion that I think it
would be good for the profession if you indeed wrote a taxonomy (and with it
a potential research agenda) for the rest of us?
I disagree with you, however, that we need multiple episodes: I think it is
perfectly ok to believe in rare and occasional presence of bubbles, only.
Given our asset pricing malaise, Market Efficiency is so powerful on short
horizons, and so weak on long horizons. Given that it is so difficult to
pin down a definite long-run equilibrium model for stock returns, I wonder
whether there is any hope of ever deciding based on the/any evidence whether
bubbles exist or long-run Market Efficiency holds. I wonder: will this
always remain a faith-based issue?
Let me prod you again with my questions:
- What feasible study could convince you that the Internet episode was a
bubble? not a bubble?
- Adopting a Bayesian view: I cannot prove that it was a bubble. I
cannot reject Market Efficiency if it is one's strongly held prior (NULL).
But, I cannot prove Market Efficiency on this data horizon and in this
context, either. Would you put a reasonably positive probability of the
Internet episode having been a bubble? (There was some pricing that one
would consider at least fairly odd at the time, I think.)
Maybe our difference in perception is milder than I think.
(PS: I lost a ton of money shorting Internet bubble firms. I could not
afford holding on. But, I definitely did believe at the time, too, that it
was a bubble. This is not just ex-post for me.)
Regards,
/iaw
From: Eugene F. Fama
Date: 10/17/2002 10:20
To: Ivo Welch
Gene's comments were interspersed. Therefore, this email was edited.
"Internet period" is a less charged term.
> I disagree with you, however, that we need multiple episodes: I think it is
> perfectly ok to believe in rare and occasional presence of bubbles, only.
Believe is the key word. Remember the Holbrook Working story that I tell in
class about bubbles.
Gene's next email explained: Working showed plots of cumulative iid normal
random numbers to his economist colleagues at Stanford. They quickly
concluded that these were plots of specific commodity prices and identified
the "obvious" bubble episodes.
Given our asset pricing malaise, Market Efficiency is so powerful on short
horizons, and so weak on long horizons. Given that it is so difficult to
pin down a definite long-run equilibrium model for stock returns, I wonder
whether there is any hope of ever deciding based on the/any evidence whether
bubbles exist or long-run Market Efficiency holds. I wonder: will this
always remain a faith-based issue?
Perhaps
Let me prod you again with my questions:
- What feasible study could convince you that the Internet episode was a
bubble? not a bubble?
- Adopting a Bayesian view: I cannot prove that it was a bubble. I
cannot reject Market Efficiency if it is one's strongly held prior
(NULL). But, I cannot prove Market Efficiency on this data horizon
and in this context, either. Would you put a reasonably positive
probability of the Internet episode having been a bubble? (There was
some pricing that one would consider at least fairly odd at the time,
I think.)
Maybe our difference in perception is milder than I think.
How many Microsofts among Internet firms would it have taken to justify the
high prices of 1999-2000prices? I think there were reasonable beliefs at
the time that the Internet would revolutionize business and there would be
many Microsoft-like success stories based on first-mover advantages in
different industries.
At this point, there was a brief conversation between John Cochrane
(Chicago) and myself, in which John suggested (and I agreed) that it would
help to define bubble. Gene was cc'ed. In the conversation, I suggested
that Gene would be uniquely qualified to define what a bubble means, even if
he does not believe in their presence.
From: Eugene Fama
Date: 10/16/2002 12:18pm
To: Ivo Welch, John Cochrane
Here it goes. Bubbles are special cases of market inefficiency where
cumulative returns differ predictably from equilibrium expected returns for
sustained periods.
From: Ivo Welch
Date: 10/17/2002 12:47pm
To: Eugene Fama
Hi Gene: I do not want to intrude on your time much further, so feel free
not to answer.
I looked up Holbrook Working, but could not find what you must be refering
to. (answered above) I do not remember the story. Then again, I was in your
class 18 years ago. Sheesh, has it really been so long...
I think your Microsoft10 argument is not impossible, but it is
implausible. This is for 2 reasons. First, if all these Internet firms were
about to revolutionize businesses and able to defend first-mover advantages,
I wonder why the old economy firms did not drop like crazy. Second, I
cannot intuit the process by which independent new information came out that
there would be no first-mover advantage, after all. That is, why was this
justification that Lycos would become a new Microsoft so sensible in March
2000 and so insensible in September? Naturally, I need not see the entire
way that markets function and what new information is arriving, but the
process by which markets came to refute this idea of Microsoft10 dominance
in many industries needs as much explanation as the one why they came to
this idea in the first place.
Would you allow me to post and/or forward some of your comments to students?
Also, Dick Thaler wrote me an email that he would be curious. Is it ok?
Gene granted permission for this page later.
Regards,
/ivo
From: Eugene Fama
To: Ivo Welch, Jay Ritter, Richard Thaler
Date: 10/18/2002 10:41am
Loughran and Ritter (2002, Why has IPO prcing changed over time) report that
during 1999-2000 there are 803 IPOs with an average market cap of $1.46
billion (Table 1). 576 of the IPOs are tech and internet-related (Table 2).
I infer that their total market cap is about $840 billion, or about twice
Microsoft's valuation at that time. Given expectations at that time about
high tech and the business revolution to be generated by the internet, is it
unreasonable that the equivalent of two Microsofts would eventually emerge
from the tech and internet-related IPOs?
Eugene F. Fama
From: Ivo Welch
Date: 10/19/2002 12:24
To Eugene Fama, Jay Ritter, Dick Thaler
Hi Gene: The bubble view is less based on the IPO Gross Proceeds, but on the
aftermarket valuation. That is, it is not difficult to understand why Tech
IPOs were priced so highly at initial sale; it is difficult to understand
why the entire Tech sector was so high. In January 2000, I posted on my
webpage https://www.iporesources.org/internetmadness.html the market caps of
firms that I did not understand at the time: AOL=170 billion; Yahoo=$114
billion, Akamai=$30 billion, etc.
Microsoft as a benchmark against the bubble is not correct, because it also
rode the wave of Tech enthusiasm. Even Bill Gates and Steve Ballmer came
out repeatedly claiming that they were worth a lot less. (Microsoft now has
a market cap of $271 billion, despite continuous improvement in
performance.) In fact, I might even believe that the entire market was
exceedingly high at the time. It is just that the Tech IPOs were so far
above what I considered reasonable that I felt it was here that one can
reasonably claim the presence of a bubble; the same could not be said for
the market. I agree: I have not proven the presence of a bubble, but
neither have you proven the absence of a bubble *in this context*.
So, I have to repeat myself: your view is not absolutely impossible, but
appears highly improbable to me. Oddly, on the scale of Fama to Thaler, my
own priors sit at about 85% Fama and 15% Thaler. I am a bit surprised,
though, that your posterior is so unaffected, not so much by the price
pattern, but by the price/fundamental pattern. I wonder if Newton would
have been moved by Einstein-ian relativity. After all, 99% of what we see
was explained by Newton alone. [The only thing that I dislike about this
analogy is that if Thaler was Einstein, he would have claimed that he could
see light bending around him everywhere ;-) .]
Regards,
/ivo
Jay Ritter later pointed out that Gene quotes
$1.46 billion market cap on the first day of trading. I should not
have ignored this figure, becayse in my emails, my concern was with
the issue of explaining tech valuations at the peak of the market in
March 2003 (Nasdaq briefly reached 5,000), not the IPO's "first day of
trading" price.
Jay Ritter's email was not working during the initial conversation,
so he weighed in later only.
From: Gene Fama
Sent: Wednesday, November 13, 2002 4:42 PM
To: (large number of people)
Jay and Ivo:
There is an interesting result implicit in your IPO review paper in the
August JF. Any post-issue under-perfomance of IPO stocks does not
benefit
the IPO firms since under-performance is always measured relative to the
closing price on the first day of trading (or the end of the first
month). If one factors in the strongly positive first day returns, the
long-term returns on IPOs are not low, by any measure. This is
interesting because it seems to be evidence against the hypothesis that
firms go public when they can get prices that are too high.
Eugene F. Fama
Robert R. McCormick Distinguished Service Professor of Finance
From: "Jay Ritter"
To: "Eugene F. Fama"
Cc: "ivo welch"
Sent: Wednesday, November 13, 2002 6:54 PM
Subject: Re: long-term returns on IPOs
Gene--
If the degree of underpricing (as measured by first-day returns) is
constant over time, then this is just another transaction cost of going
public and thus is unrelated to the question of the timing of IPOs. So I
presume that the issue is one of whether there is a correlation of IPO
volume, underpricing, and long-term returns. Long-term returns can be
measured in both an absolute sense (raw returns) or abnormal returns
(relative to some benchmark). If it is the case that IPO volume is high
when first-day returns are high and long-term returns (measured from the
first aftermarket price) are low, then you are absolutely right-- if
long-term returns measured from the offer price are uncorrelated with
volume, then there is little evidence of timing ability.
In fact, IPO volume and first-day returns have a very low correlation (see,
for example, the numbers in Lowry-Schwert (2002 JF)). You are correct that,
measured from the offer price to five years later, the raw and abnormal
returns on IPOs are not noticeably low, once penny stocks are deleted form
the samples. But there still seems to be some negative correlation of
volume and long-term returns, although it is not an incredibly strong
negative correlation. The exact numbers depend somewhat on whether the
1999-2000 bubble period is included. These years had incredibly high
first-day returns and incredibly low long-term returns.
The volume in 1999 and 2000 was not extreme, however. 1983, 1986, 1993, and
1996 all had higher IPO volume than 1999 and 2000, as measured by the
detrended number of IPOs, if one assumes that IPO volume should increase by
2% per year to reflect the growth of the economy.
Jay R. Ritter
Cordell Professor of Finance
P.O. Box 117168
University of Florida
Gainesville FL 32611-7168
A final word: for me, the challenge for efficient market proponents is
not only to explain why Tech stocks increased so dramatically from 1995 to
2000, but also to explain why their explanation stopped working in March
2000. The decline seemed rapid and not associated with a lot of new
fundamental information. So, I am sure that it is possible to
concoct such explanations, but I have yet to hear an explanation that I also
find plausible. So, to this day, I still believe we experienced a
bubble.
Jay R. also pointed out that the most reprinted article in the
history of Fortune magazine was Warren Buffet's Fortune piece from
November 22, 1999, in which he warned about the overvaluations of Tech stocks
and Internet stocks---well before March 2000.
Ivo Welch