15.

481 Financial Market Dynamics

and Human Behavior
Andrew W. Lo, MIT
Unit 4: Neuroscience and Decision Making
Lecture: Applica9ons to Finance
Practical Applications
 15.481
Trader Psychophysiology

© 2020 by Andrew W. Lo
Lecture 4 Slide 3
All Rights Reserved
 15.481
Why Focus On Traders?
§ Traders highly trained and motivated to be “rational”
§ Trading decisions are narrow in scope and time
§ In vivo vs. in vitro concerns among economists
§ Practical significance—rogue traders:
Trader Year Company Loss

Nick Leeson 1995 Barings £ 827,000,000

Yasuo Hamanaka 1996 Sumitomo $ 2,600,000,000
John Rusnak 2002 Allied Irish Banks $ 691,000,000
Jiulin Chen 2005 China Aviation Oil $ 550,000,000
Jerome Kerviel 2006-2008 Societe Generale € 4,900,000,000
Boris Picano-Nacci 2008 Caisse d'Epargne € 751,000,000
Kweku Adoboli 2011 UBS $ 2,300,000,000

© 2020 by Andrew W. Lo
Lecture 4 Slide 4
All Rights Reserved
 15.481
Focus on Emotional Mechanisms
§ Understanding the mechanisms behind the economic behavior
of individual market participants:
– What role do emotions play in the decision process?
– How risk-taking behavior depends on the emotional state?
§ How well are the behavioral mechanisms adapted to
succeeding in the modern financial marketplace:
– How emotions affect trading profit and loss?
– Is there any difference in physiology for good vs. bad performance?
§ Evolutionary and neurophysiological basis for emotion in
financial decisions.
© 2020 by Andrew W. Lo
Lecture 4 Slide 5
All Rights Reserved
 15.481
Experimental Design
§ Measure psychometric and real-time physiological data
§ Collect real-time market and trading data
§ Clean and process physiology data
§ Identify “financial events” in markets and trades
§ Construct “physiological signature” around each of the
financial events
§ Compare event “physiological signatures” for specific
type of financial events to baseline physiology
§ Stratify by event category, securities, experience, traits
© 2020 by Andrew W. Lo
Lecture 4 Slide 6
All Rights Reserved
 15.481
Physiological Data
§ Galvanic Skin Response (SCR) on the palm
§ Blood Volume Pulse (BVP) taken through photoplethysmography on the
thumb (BVP)
§ Temperature (TMP1) on the index finger
§ Temperature (TMP2) on the forehead midway between eyebrow and
hairline above the left pupil
§ Electrocardiogram (ECG)
§ Electroencephalogram (EEG)
§ Electromyogram (EMG1) on underside of wrist on flexor digitorum, and
(EMG2) on trapezius muscle group on the back of the shoulder
§ NEW: testosterone and cortisol levels via saliva samples

© 2020 by Andrew W. Lo
Lecture 4 Slide 7
All Rights Reserved
 15.481
Psychometric and Trading Data
§ Personality traits: IPIP-NEO survey
§ Zung depression (SDS) and anxiety (SAS) surveys
§ Emotional state: UWIST MACL survey
§ Internality, Powerful Others, and Chance (IPC) survey
§ Demographic data
§ Profit-and-loss (P&L) data

© 2020 by Andrew W. Lo
Lecture 4 Slide 8
All Rights Reserved
 15.481
Equipment

© 2020 by Andrew W. Lo
Lecture 4 Slide 9
All Rights Reserved
 15.481
In Vivo Measurement
Blood Volume Pulse (BVP), forehead
Electroencyphalography
(EEG), right forehead Electroencyphalography
(EEG), right forehead

Electromyography (EMG)
Electromyography Upper back
(EMG), inside arm

Biofeedback system:
NeXus-10 – Mark II
(Stens Corporation)

Skin Temperature

Skin Conductance (GSR)

Physiological sampling rate: 2048 samples/sec

Market data sampling rate: 1 sample/.001 sec to 10 sec
© 2020 by Andrew W. Lo
Lecture 4 Slide 10
All Rights Reserved
 15.481
In Vivo Measurement

© 2020 by Andrew W. Lo
Lecture 4 Slide 11
All Rights Reserved
 15.481
In Vivo Measurement

© 2020 by Andrew W. Lo
Lecture 4 Slide 12
All Rights Reserved
 15.481
In Vivo Measurement

© 2020 by Andrew W. Lo
Lecture 4 Slide 13
All Rights Reserved
 15.481
Raw Physiological Data
20 EMG: face
10
0
20 EMG: arm
0

8 SCR
7.5
92
91.8 Temperature

35
25
15 BVP
40 Respiration
35
0 20 40 60 80 100 120

© 2020 by Andrew W. Lo
Lecture 4 Slide 14
All Rights Reserved
 15.481
Artifact Elimination
§ Speech and hand movement (due to typing) is eliminated with
the help of the EMG signal recorded from muscles on the face
and arm

SCR

arm
EMG

© 2020 by Andrew W. Lo
Lecture 4 Slide 15
All Rights Reserved
 15.481
SCR Signal Analysis
Meaningful Skin Features Describing
Conductance Responses Signal Properties

Local Baseline
Skin Conductance Responses Response
Magnitude
Response Onset

© 2020 by Andrew W. Lo
Lecture 4 Slide 16
All Rights Reserved
 15.481
Market Data
1.042

1.0415

Euro
1.041

1.0405

1.04

Australian dollar
0.645

0.6445 bid price

0.644 ask price

0.6435
0 5 10 15 20 25 30 35 40
Time, min
price = (ask + bid)/2
spread = (ask - bid)
simple net return = (price(t) - price(t-1))/price(t-1) · 100%

© 2020 by Andrew W. Lo
Lecture 4 Slide 17
All Rights Reserved
 15.481
Market Event Detection
Price chart for Dollar / Yen
106.95

106.9 Low volatility

High volatility
106.85
5-min moving
average
Yen 106.8

106.75 Trend-reversal

106.7

106.65
Deviation
106.6
0 5 10 15 20 25 30
Time, min

© 2020 by Andrew W. Lo
Lecture 4 Slide 18
All Rights Reserved
 15.481
Data Analysis Approach
§ During significant events, are there changes in traders’
physiological variables?
§ Conversely, during significant physiological events, is
there unusual market activity?
§ Stratify participants by category, e.g., years of
experience, risk aversion coefficient, education, etc.
§ Apply machine-learning techniques

© 2020 by Andrew W. Lo
Lecture 4 Slide 19
All Rights Reserved
 15.481
Results
All Traders
Market Events

DEV, DEV, DEV, TRV, TRV, VOL,

Physiology Features Price Spread Return Price Spread Maximum

Number of SCR Responses 2.872 0.604 2.088 1.664 0.636 0.579

Average SCR Amplitude -0.695 0.887 0.186 -2.478 1.136 0.880
Average Heart Rate 0.905 -0.588 -0.596 -0.617 1.032 -0.270
Average BVP Amplitude Relative
0.128 -0.855 0.688 1.408 -0.523 3.619
to Local Baseline
Average BVP Amplitude Relative
1.417 -2.726 2.324 1.417 -2.211 2.886
to Local Baseline
Number of Temperature Jumps 0.988 2.820 0.650 1.777 4.868 -2.482
Average Respiration Rate 0.729 -1.173 0.671 1.155 -1.628 -0.064
Average Respiration Amplitude -1.646 -0.250 0.136 0.034 -1.487 -3.499

denotes
p-value < 0.05

© 2020 by Andrew W. Lo
Lecture 4 Slide 20
All Rights Reserved
 15.481
Results
High Experience Low or Moderate Experience
Market Events Market Events

VOL, Maximum

VOL, Maximum
DEV, Spread

DEV, Spread
TRV, Spread

TRV, Spread
DEV, Return

DEV, Return
DEV, Price

DEV, Price
TRV, Price

TRV, Price
Physiology Features

Number of SCR Responses 0.561 1.228 -0.221 -1.261 0.184 2.081 2.696 0.123 2.752 2.170 -0.665 -0.688

Average SCR Amplitude 0.030 0.344 0.744 -1.340 1.149 0.904 0.756 1.578 -1.265 -1.947 -0.012 -0.615

Average Heart Rate 0.449 1.886 -1.797 -1.635 1.836 -2.018 1.310 0.051 -0.382 -1.106 -0.311 1.283
Average BVP Amplitude
Relative to Local Baseline
-0.061 -0.747 0.334 0.015 0.164 2.623 -0.204 -1.775 0.875 1.796 -1.289 2.708

Average BVP Amplitude

Relative to Local Baseline
-2.529 -1.367 1.590 1.110 -1.165 2.301 1.917 -2.664 1.884 2.082 -1.759 2.758

Number of Temperature
Jumps
-1.010 -1.435 NA NA NA 1.644 0.837 2.902 1.145 2.445 2.633 -3.213

Average Respiration Rate -1.556 1.109 0.121 0.866 -0.102 0.093 0.079 -0.552 0.323 -0.631 -2.170 -1.528
Average Respiration
Amplitude
1.087 0.265 -1.777 -0.657 0.121 -0.607 -2.749 -1.319 2.142 0.055 -2.832 -4.533

© 2020 by Andrew W. Lo
Lecture 4 Slide 21
All Rights Reserved
 15.481
Results

© 2020 by Andrew W. Lo
Lecture 4 Slide 22
All Rights Reserved
 15.481
Results
Most Successful Traders’ Characteristics
§ SCRs confined to trading intervals
§ Low internality scores
§ Ability to “move on” after losses
Males traded more often and more aggressively;
females had better risk-adjusted returns
© 2020 by Andrew W. Lo
Lecture 4 Slide 23
All Rights Reserved
 15.481
Loss Aversion Revisited
Risk-Averse When Facing Gains Trader PerspecKve
§ A: $ 240,000 “Nervous Nellie”
§ B: $1,000,000 with 25% probability
$ 0 with 75% Probability
Risk-Seeking When Facing Losses
§ C: – $ 750,000
§ D: – $1,000,000 with 75% probability “Rogue Trader”
$ 0 with 25% Probability
© 2020 by Andrew W. Lo
Lecture 4 Slide 24
All Rights Reserved
 15.481
Loss Aversion Revisited

“Cut your losses, ride your gains”

© 2020 by Andrew W. Lo
Lecture 4 Slide 25
All Rights Reserved
 15.481
New Technology
Empatica E4
Wristband

§ Skin conductance
§ Blood volume pulse and heart
rate
§ Temperature
§ Mo7on (built-in accelerometer)

Volunteers Needed!
© 2020 by Andrew W. Lo
Lecture 4 Slide 26
All Rights Reserved
 15.481
Narrow Implications
If Pain Is Due To Permanent Shift
§ More pain to come; better get out now
If Pain Is A Temporary Event
§ Revisit the original reasons for investing
§ If unchanged, “no pain, no gain”; tough it out
§ Otherwise, more pain later!
§ Make sure the pain is manageable! August 1998
We Should Learn The Right Lesson From Pain
© 2020 by Andrew W. Lo
Lecture 4 Slide 27
All Rights Reserved
 15.481
Narrow Implications
$450

$400
Apple
$350

$300

$250

$200
Time To Get Out?
$150

$100

$50

$0
20050228 20060228 20070301 20080229 20090302 20100302 20110301

© 2020 by Andrew W. Lo
Lecture 4 Slide 28
All Rights Reserved
 15.481
Narrow Implications
$100

$90
Enron
$80
Time To Get In?
$70

$60

$50

$40

$30

$20

$10

$0
19980102 19990104 20000103 20010102 20020108

© 2020 by Andrew W. Lo
Lecture 4 Slide 29
All Rights Reserved
 15.481
Narrow Implications
Systematic Framework To Avoid An Amygdala Hijack
§ What is the main motivation for the trade/strategy?
§ Why should it work?
§ Why shouldn’t it work?
§ How can you tell?
§ How long should it last?
§ What is your “loss threshold”?
§ When do you get back in?
§ How did you do?

© 2020 by Andrew W. Lo
Lecture 4 Slide 30
All Rights Reserved
 15.481
Narrow Implications
Example: Equity Statistical Arbitrage
§ What is the main motivation for the trade/strategy? lead/lag
§ Why should it work? over-reaction, delayed price-discovery
§ Why shouldn’t it work? decimalization, low vol, DE Shaw
§ How can you tell? daily P&L patterns
§ How long should it last? from a few days to a few hours
§ What is your “loss threshold”? 3–5% MDD
§ When do you get back in? higher volatility/volume
§ How did you do? performance attribution (size, sectors, etc.)

© 2020 by Andrew W. Lo
Lecture 4 Slide 31
All Rights Reserved
 15.481
Broader Implications
§ Fear and greed do drive Wall Street (and Main Street)
§ That may not be a bad thing
§ Possible explanaTon for “intuiTon”
§ IrraTonal behavior Û emoTons gone wild!
§ Modeling raTonality may involve “mulTple minds”
§ Implies new version of Efficient Markets Hypothesis; one in
which behavior adapts to changing environments
§ But how does adaptaTon work?
Þ Evolu:on © 2020 by Andrew W. Lo
Lecture 4 Slide 32
All Rights Reserved
 15.481
Next Time

© 2020 by Andrew W. Lo
Lecture 4 Slide 33
All Rights Reserved