An Economist’s Perspective on Technical Analysis WILLIAM POOLE SENIOR FELLOW, CATO INSTITUTE\ AND DISTINGUISHED SCHOLAR IN RESIDENCE, THE UNIVERSITY OF DELAWARE GLOBAL INTERDEPENDENCE CENTER FEDERAL RESERVE BANK OF PHILADELPHIA 9 OCTOBER 2015
Disclaimer 2
I am a monetary economist, not
an investment professional. I hold no professional license. Listen at your own risk!
What Is Technical Analysis? 3
Attempt to extract predictive
value from history of stock prices Examples: Dow
Theory Head and shoulders pattern Death cross Support/resistance levels Seasonal patterns
I will argue that: 4
Tech anal attempts to predict future
prices in speculative markets from their own history.
The effort is unsupported by
evidence and does not work.
Economic theory indicates that tech
anal should not be expected to work.
And that: 5
Any investor success “using” tech
analysis is due to other aspects of investment strategy. Tech anal has the potential to mislead investors, especially nonprofessional ones. It distracts investors from key aspects of portfolio management. Will discuss several key aspects from my perspective.
Data to Illustrate Issues 6
S&P 500 Index, end of month values,
mid 1986 through September 2015. Important not to work with monthly
average data. Will present several different
ways of looking at these data.
S&P 500 % Change per Month 15 10 5 0 -5 -10 -15
Find the Pattern!
-20 -25
ug A 29
86 -19 3
u A 1
9 g-1
90
3
u A 1
9 g-1
94
98
u
1-A
3
9 g- 1
u
0-A
3
7
0 g- 2
02
06
u
1 -A
3
0 g- 2
10
u
1 -A
3
0 g-2
2
9-A
u
0 g-2
14
Level: S&P 500-B Ratio Scale
Find the Pattern! 1000
8 -19
l
-Ju 1 3
90 9 -1
6
p
2
e 8-S
3
o 0-N
99 1 v
9
4 2
3
99 1 an
9-J
3
1-M
8
0 -20
ar
3
a M 1
y
0 -20
11 0 l-2
7 -Ju 9 2
5
3
ep S 0
1 -20
Level: S&P 500-A Ratio scale
1000
8 -19
6
l
3
u 1-J
9 9 1 -
p
2
e 8-S
3
99 9 -1
4
0
9 9 1 -
v
o 0-N
2
an J 9
9
r
a M 1
3
03 0 -2 31
00 2 y Ma
11 15 0 0 -2 l-2 p u e -J -S 29 30
7
S&P 500-B—PHONY!! Ratio scale Constructed by taking actual Index % changes and then randomizing order. 1000
3
u 1-J
l
8 -19
90
6
-S 28
e
9 p-1
-N 30
o
9 v-1
99
94 a
-J 29
9 n-1
-M 31
10
a
03 0 2 r31
00 2 y Ma
7
11
-J 29
20 l u
15
-S 30
e
0 p-2
Guess! 11
From B or A?
From B or A?
Gold: one of chartists’ favorites
12
13
14
Fat tails Issue for Investors: S&P 500 15
Fat tails – work of Benoît Mandelbrot, Eugene
Fama and others in 1960s and later. Over my 29-year sample period, month-end data, mid 1986 to present, 351 months:
Mean= 0.589 Normal Actual Std dev = 4.464 Dist Mean - 1 SD = - 3.9% 55.5 ……………. 43 Mean - 2 SD = - 8.3% 8.0 ……………. 15 Mean - 3 SD = -12.8% 0.35 ……………. 3
Note the large negative changes. With normal
dist, would have expected 1 in 90 years.
One of Many Examples on Internet 16
Fat tails. Sample period unknown.
Tail observations appear Insignificant, but are not.
Look carefully Look carefully
Fat-Tail Events Unforecastable
30-Oct-1987 -24.54 31-Oct-2008 -18.56 31-Aug-1998 -15.76 30-Sep-2002 -11.66 27-Feb-2009 -11.65 31-Aug-1990 -9.91 28-Feb-2001 -9.68 30-Sep-2008 -9.52 30-Jun-2008 -8.99 30-Jan-2009 -8.95 30-Sep-1986 -8.93 30-Nov-1987 -8.92 31-May-2010 -8.55 28-Sep-2001 -8.53 30-Nov-2000 -8.35 31-Jul-2002 -8.23
Average is -11.3
17
31-Mar-2009 8.20 31-Oct-2002 8.29 30-Sep-2010 8.39 31-Jul-1989 8.47 31-May-1990 8.80 30-Apr-2009 8.98 31-Mar-2000 9.23 31-Oct-2011 10.23 31-Dec-1991 10.58 30-Jan-1987 12.38
Average is 9.35
These are per month, not at annual rate
Many Examples for Individual Stocks 18
Recent VW diesel engine scandal: stock
down by 1/3. Texaco bankruptcy 1987
Shortly before, had been a AAA credit risk At the time, largest Chapter 11 in U.S. history A very odd and interesting case
Penn-Central bankruptcy 1971 Lehman, 2008 Etc., Etc., Etc., …
No tech anal can possibly predict these
cases.
Continuity 19
Many investment models assume
continuity. Examples: Portfolio
insurance model. Black-Scholes option-pricing model. In fact, market prices often move
in discontinuous jumps, up or down.
Concluding Comments 20
If tech analysis schemes are so
powerful, why do their inventors not maintain secrecy and trade with these models? Where is the statistical evidence to support these models? Why wouldn’t a successful model be programmed into a computer, which would then destroy its validity?
Warren Buffett 21
A famous Buffett quote: “I realized that
technical analysis didn't work when I turned the chart upside down and didn't get a different answer.” Buffett is the most successful U.S.
investor of the past 50 years. Pay attention to his advice.
Discussion 22
Questions? Disputes? Facts neglected?