An Academic Perspective on

Technical Analysis
Christopher J. Neely
Federal Reserve Bank of St. Louis

The Technical Analyst

European Conference 2006
February 9, 2006
1
 Disclaimer
Disclaimer: The views expressed are
my own and do not necessarily reflect
official positions of the Federal Reserve
Bank of St. Louis, or the Federal
Reserve System.

I thank Mark Hoeman of Hoeman Capital Management

for helpful discussions on the realities of technical analysis.

2
 A few more caveats
• I cannot tell you how to make money.
– I am not giving investment advice.
• Academic research on TA is only loosely
related to TA practice.
– I am very aware of this. Any conclusions in this
talk relate to academic research and not real TA.
• My knowledge of FX trading is pretty
modest but I do know a bit about economic
research.
3
 What do academics want to do?
• Determine if TA works and if so, why?
• Why is TA a puzzle?
– The Efficient Markets Hypothesis.
• Why should people earn unusual returns with
publicly available information like past prices,
volume, etc?

4
 Academic attitude on TA
• Old attitude: TA is as reliable as astrology.

• Newer attitude: Markets are complex.

– We need to think about risk and information
problems. 5
 Academic attitude on TA
• Most common attitude: If it really works,
why don’t you go make some real money?

6
 How is TA studied?
1. Test simple, extrapolative trading rules
on daily or higher frequency data
• Oldest, most frequent study.
2. Theoretical models justifying TA.
• Why should TA work?
¾ Asymmetric information, sequential, opaque
trading.

3. Empirical microstructure studies.

• Severe data requirements. 7
 Why do we need to test TA?
• We observe that bumblebees fly.
• But should they fly in theory?

• This isn’t as crazy as it sounds; we use

reality to revise theory. 8
How would academics test TA?
• What do academics need?
– Ex ante rule, good data.
• How does the rule do over time?
– Are returns fairly stable?
• Does the rule take excessive risk?
– Sharpe ratios, CAPM betas, CVaR

9
 Problems with testing TA
• How do we choose an ex ante rule?
– Textbooks, newsletters will produce rules
that are probably profitable on recent data.
• How do we encode judgment?
– We can’t encode judgment.
• Firm, high-frequency data was—until
recently—difficult to get.
10
 Problems with testing TA
• Prices must be transactable.
• Limited set of rules that exclude
judgment.
• Data snooping bias in rules.
• Publication bias in results.
• One must realistically assess risk.
– Academic models of risk are poor.

11
 Results of testing TA
• Starting in the early 1980s, researchers
found that a variety of trend-following
trading rules produced excess returns in FX
markets.
– Filter, MA, channel, ARIMA rules
– Daily or weekly data.
– Early profits were in the 7-12 % range.
• The returns seem to have disappeared in the
1990s.
12
 Results of testing TA
• Typical results from portfolios of MA rules for the
DEM, JPY, CHF and GBP vs. the USD.
– Use portfolios of mean returns to double MA rules (1,5),
(5,20) and (1,200).
– Rules trade about 10 – 30 times a year.
– Calculate rolling annual returns over 1976-1990 and
1991-2004.
– Returns are about 8.3 to 10.2 percent per annum in 1976-
1990 and -1.5 to 2.0 percent per annum in 1991-2004.
• Returns net out modest transactions costs of 5 bp per trade.
– Returns are noisy though.

13
 Results of testing TA
• Typical results from portfolios of MA rules for the
DEM, JPY, CHF and GBP vs. the USD.

14
 Results of testing TA
• Researchers find much less success for TA
in equity markets.
– Cannot trade on index data.
• Can use futures index data.
– Individual stocks must have a lot of liquidity.
– Transactions costs are much higher in
equities.
– What is the right equity benchmark to beat?
• It is hard to beat the buy-and-hold.

15
 Theoretical models of TA
• Asymmetric information, sequential trading
and opacity are important components.
– Individuals have asymmetric information and
price changes reveal that.
• Two simple examples:
– Would you rather eat dinner in a crowded
restaurant or one with nobody in it?
– Information cascades can create false inference.

16
 What explains the success of TA?
• Academics really don’t measure risk well.
– Risk is inherently difficult to measure because it
involves extreme, unusual events.
– Sometimes everything looks good UNTIL
disaster strikes.

17
 What explains the success of TA?
• There are limits to risk-arbitrage trading.
– Liquidity problems (e.g., LTCM.).
– Principal agent problems: Short horizons
• Central bank intervention
– Profits of trend-following daily FX rules mostly
disappear when we take out the days of Fed
intervention.
– BUT intervention does not generate trading
profits but it does react to strong trends.
18
 An Aside on FX Intervention
• All major central banks have made
substantial excess returns on their USD FX
intervention.
• Central banks “buy low and sell high” with
an investment horizon of years.

19
 An Aside on FX Intervention
DEM/USD and
“fundamental value”

Fed buys Fed sells

low high

Profits in Cumulative Fed

billions profits are
of USD strongly positive

20
 An Aside on FX Intervention
• It isn’t just the Fed that has made profits from
its USD intervention.
• I could put up similar graphs for Germany,
Japan, Switzerland, and Australia.
– “The Case for Foreign Exchange Intervention:
The Government as an Active Reserve Manager”
– WP 2004-031B, Revised July 2005

– http://research.stlouisfed.org/econ/cneely/
21
 Recent Work on TA
• TA research has stalled rather badly.
• Extrapolative rules probably no longer work
in FX markets.
– There is a lot of noise in the data.
• Why don’t the rules work anymore?
– Data snooping or market adaptation?
• There has been some good empirical
microstructure work.
– Are round numbers special? (Carol Osler, Brandeis)
22
 What can I learn from you?
• How would you like to make me less
ignorant?
• What are the recent trends in TA?
• What sorts of rules have become popular?
• Any suggestions for how to test TA?
– High-frequency data sources, good rules.
• Why isn’t there (more) long-term arbitrage
to fundamentals?
23
 What can I learn from you?
• If you give me your email address for a
brief questionnaire, I promise no Viagra
ads or requests for assistance from my
brother, the Nigerian banker.

• I will send you a short (10 minute)

questionnaire asking your opinions on
technical trading.

24
Thanks for your attention.
The End

25