Econ 138: Financial and Behavioral Economics

Structure Within Stock Market Brownian Motion:

Anomalies, Technical Analysis and Statistical Arbitrage

January 30, 2017

Reading:
M.F.M. Osborne, “Periodic Structure in the Brownian Motion in the
Stock Market,” Operations Research, 10 (3), 345–379, (1962). Skim
pp. 347–358.

Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 1/ 21
Structure Within Stock Market Brownian Motion

Agenda:
1 Brownian motion.
2 Stock prices as simple random walks.
3 The volume sequence.
4 Periodicity in time.
5 Periodicity in price-space.

Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 2/ 21
Stock Prices as Simple Random Walks
Temporal evolution of uncertainty: stochastic process

200

150
SPY (USD)

100

50

0
0 0.2 0.4 0.6 0.8 1
TIME (years)

Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 3/ 21
Stock Prices as Simple Random Walks
Temporal evolution of uncertainty: probability density

0.05
PROBABILITY

0.04
0.03
0.02
0.01 200
0 150
0 100
0.2
0.4 50 SPY (USD)
0.6
0.8 0
TIME (years) 1

Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 4/ 21
Brownian Motion

Periodic structure within Brownian motion:

1 Periodicity in time:
Evidence of periodic structure corresponding to intervals of a
day, week, month, & year.
These periods correspond to cycles of human attention span.
Periodicity is not in price sequences.
Periodicity is in price differences, especially the variance.
2 Periodicity in price-space:
Sites of preferred occupancy & reflective barriers.
(technical analysis)
Clustered (herding?) activity.

Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 5/ 21
Volume and Dispersion

Observations:
A lot of heat; not a lot of light.

When market activity is large, so is across-the-market

dispersion.

Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 6/ 21
The Volume Sequence

Observations:
Attempts to make a theory of volume.

“. . . some intervals of time have a much greater probability to

contain transactions than others of equal length., contrary to
our [assumptions].”

Non-agreement of Poisson’s theory with observations is

consistent with the observation that stocks tend to be traded
in concentrated bursts.

“. . . stocks suffer extended periods of excessive ‘interest’

(cf. volume), or lack of interest, lasting a month or more.”
(Investor sentiment).

Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 7/ 21
The Volume Sequence

Observations:
Daily volume tends to be lognormally distributed.

If volume measures interest then an increment of interest is

proportional to the interest present. (Weber-Fechner)

“People, like sheep, tend to develop more interest because it

is already there.” (Herding)

“This tendency of people is well-known to professional

manipulators, who attempt to generate real interest by
producing a semblance of interest by spurious trading.”
(Social psychology, fraud, and market regulation)

Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 8/ 21
 -L 00 - 7 2 0
Periodicity
96-,
4O IIJ
in Time -J

i-
M.F.M. b~~~~~~~~~~~~~~~~~~~~~~~
Osborne, “Periodic Structure . . . ,” (1962). Figure 9.
tZ CUi
~i .0 I

Daily Periodicity:
_ 111
T olt 2111li
ZD

H2000-
< M

a-

LU0F
JO 1000
Th trdn of th las hafh0 (33:30 P.. ha bee dobe give0
to

lOAM IL3.30 WEEK WEEK

MARCH 2 3 6 7 8 9 10 13 14
1961

Fig. 9. (a) Hourly volume rate of trading vs. calendar time, NYSE.
The trading of the last half-hour (3-3:30 P.M.) has been doubled to give
the hourly rate. (b) The across-the-market dispersion (doubled) of 'open'
“[T]he effect(filled
to 'close' of the ‘attention’
circles) and 'close'shows up in
to 'open' thecircles)
(open across-the-market
changes in logeP
for a sample of 12 NYSE common stocks, taken for the same time interval
dispersion, which fluctuates periodically in the diurnal cycle of
as (a). The sample was taken under the letter N-Nafi to Nat. Fuel Gas;
man’s attention span”.
data from Wall Street Journal. The range of uncertainty (light line)
corresponds to the largest range from interval, 2nd to 3rd, or 10th to 11th
Lecture 4 member of theSubstructure:
– Brownian Motion sample. R. J. Hawkins Econ 138: Financial and Behavioral Economics 9/ 21
 AVOLB <o 4 34 34 29 27

Periodicity
AVOLB>o in Time
37 I I 2I 24 20 Data from I1/4/60 to I2/1i/60
M.F.M. Osborne, “Periodic
AVOLB<o 5 Structure
32 26 . . . ,” 23(1962).
25 Table II.

(b)
Weekly Periodicity:
(Contingency day-over-day
table test of the changes
above data for the in ofodd-lot
dependence daily odd buying
lot buying volume
'increasing' or 'decreasing' on 'day of the week.')

Mon. Tues. Wed. Thurs. Fri. Total

AVOLB>o I24 38 53 68 59 342

AVOLB<o I2 I04 96 79 8o 371

Total I36 I42 I49 I47 I39 7I3

x2= I38; P(X2>138)K<<O-OI; n=4 degrees of freedom

is the opening block, usually only a small fraction of the total day's trading
volume.Day-over-day
Figure 9bchanges
confirms,inonodd-lot
a daily buying
basis, whatgreatest
we had onalready
Monday.in-
ferred from the monthly data of Figs. 3 and 4. Hence, we can conclude
fromOdd-lot
the data traders:
that the across-the-market dispersion of stock price changes
has a rather werewell-defined diurnal
typically retail period,
traders, notwhen prices are
professional considered as a
traders.
function of calendar, or total elapsed time, rather than a function of simple
get information on the weekend & trade on Monday.
trading time interval, for which 1 day=5/1f hours. These results have
plausible and
the quite anticipates theobvious
notion interpretation
of noise traders. that volume represents in-
terest or attention to stocks, and that prices tend to move under the impact
of4 –this
Lecture interest.
Brownian Since theR. sign
Motion Substructure: (up or down)
J. Hawkins Econ 138: of the motion
Financial and Behavioral
is Economics 10/ 21
Periodicity in Time
M.F.M. Osborne, “Periodic Structure . . . ,” (1962). Table II.

Mon. Tues. Wed. Thurs. Fri. Total

124 38 53 68 59 342
∆VOLB > 0 65.2 68.1 71.5 70.5 66.7
52.9 13.3 4.8 0.1 0.9
12 104 96 79 80 371
∆VOLB < 0 70.8 73.9 77.5 76.5 72.3
48.8 12.3 4.4 0.1 0.8
Total 136 142 149 147 139 713

Totalrow × Totalcolumn
Expected frequency =
Totaloverall
2 (Observed − Expected )2
χ component =
Expected
XN
χ2 = χ2 component n = 138.4

n=1
Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 11/ 21
 AV<o 23 I9 23 30 29

Periodicity in Time
M.F.M. Osborne,
AVO “Periodic Structure
249 28 . . . ,” 26(1962).
28 Table
DataIV.
from
29 1/4/60 to 12/2/60

(b)
Weekly Periodicity:
(Contingency table test ofday-over-day changes
the above data for in round-lot
the dependence of volumebuying
'increasing' or
'decreasing' on 'days of the week.')

Mon. Tues. Wed. Thurs. Fri. Total

AV>o 64 83 85 67 57 356
AV<o 72 59 64 8o 83 358

Total I36 I42 I49 I47 I40 7I4

X2=I3.5; P(X2>I3.5)<o.oI; f=4 degrees of freedom

it over a long week-end, make up their minds on Monday and Tuesday,

and Day-over-day
act the next day.changes in round-lot buying greatest midweek.
In addition to the daily data on prices and volumes, there is also pub-
Round-lot
lished traders:
on each date t, four sequences of numbers that can be examined as
indicators of the state of the market. These are Na( t), Nd(t), Nu(t), and
were typically professional traders, not retail traders.
NI(t), which are respectively the number of prices which advanced, the
number which declined, the
get information onnumber
weekdays & trade (all
unchanged later.
from previous day or
the last day traded), and the sum of these three, the number of issues
traded. Theanticipates the notion of rational traders.
last one, of course, fluctuates much less percentagewise, than
any of the other three. The running sum B(T) =ZE=To[Na(t) -Nd(t)
Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 12/ 21
 these four sequences are, or when they are most pronounced, it would
Periodicity in Time
appear reasonable to suppose that if there is some relation between prices
and earnings, the dates on which the information
M.F.M. Osborne, “Periodic Structure . . . ,” (1962). III. earnings becomes
Table about
available would be related to the price changes at about the same time

TABLE VII
Annual Periodicity:
(Contingency table test for advance and decline of market average in a month, for dif-
ferent months of the year. Data from i886-ig60 from reference 5.)

J F M A M J J A S 0 N D Total

N(market advance) 38 26 26 38 40 40 44 47 30 36 36 37 438

N(market decline) 36 48 48 36 34 34 30 27 44 38 38 37 450

Total 74 74 74 74 74 74 74 74 74 74 74 74 888
x2= 24.6; P(X2> 24.6) =o.oii; n=ii degrees of freedom

Now information on earnings is not distributed uniformly throughout the

but is distributed
year,Looking across months in bursts see: three months apart, with a sus-
we about
tained burst of information after the end of the year, corresponding to
a small that
those companies report
January annually (see Fig. 10).
effect.
Not knowing whether the information contained in these reports is
good or bad,the orsummer rally.to which the information has already been dis-
the extent
counted, it would seem reasonable to suppose that concentrated earnings
a decline in winter.
information would tend to increase, relatively, the across-the-market
dispersion of price changes at the time when information becomes avail-
Lectureable. Granville
4 – Brownian 66 assesses
Motion Substructure: R. J. the effect Econ
Hawkins of 138:
theFinancial and Behavioral of
concentration earnings13/ 21
Economics
Periodicity in Price-Space

An early scientific view of technical analysis:

In financial economics: support & resistance.

In economics: the Taussig penumbra:

“It may be quite clear that potatoes will be higher
than last year. But there will be a penumbra of
uncertainty. . . . A good observer has said that the
successful speculator is not necessarily a man of
wide statistical information or of much experience in
the trade. But he must be a shrewd judge of human
nature.”
– F.W. Taussig, QJE, 35(3) 394–411 (1921)

Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 14/ 21
hths. It should also be noted by comparing Figs. 11(a) and (b) t
Periodicity
effect in Price-Space
is more pronounced with a sample from the entire market t
M.F.M. Osborne, “Periodic Structure . . . ,” (1962). Figure 11.
is a sample restricted to a weekly volume greater than 50 round l
Congestion is seen at certain price levels:
(a) (b)
ALL VOLUMES WEEKLY VOLUME
>50 RD LOTS

200 _

W W
zz J

O lo 00 2- W 102
20 M
0 3 7
8 88
3 7 8 8 8
Fig. 11. Distribution ) of closing prices in eighths on NYSE, January 15,
1960, common and preferred. (a) Sample taken under A-K, all volumes.
Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 15/ 21
Periodicity in Price-Space: Gold, Support & Resistance.
Example courtesy of Grace Ryan, Bandwidth Capital

Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 16/ 21
Periodicity in Price-Space: S&P 500, Support &
Resistance.
Example courtesy of Grace Ryan, Bandwidth Capital

Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 17/ 21
Periodicity in Price-Space

Support and Resistance: A Physical Analog

Prices diffuse on a 1-D lattice.

Preferred occupancy sites are integers.

Less-preferred sites are half-integers.

Preferred sites act as “scattering centers.”

Increasing the ”temperature” (volume or dispersion)

diminishes this effect.

The preference of human decision-makers for certain

numbers over others is not a peculiarity of the stock
market.

Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 18/ 21
igh or low, is in fact a maximum or minimum. This censored-
Periodicity in Price-Space
on of the entire sample is quite appreciable for intervals of a day,
M.F.M.
month. ForOsborne,
a year's “Periodic Structure
span of data, . . . ,” to
it amounts (1962).
only aFigure 13.
few per
e censored-out fraction for a year, as an example, does not corre-
he published list of new highs and lows for the year (exactly an
f a year only as published at the beginning of the year). To
Histogramfraction
r the censored-out of percentage price
for the year, changes:
a stock must close high

65 ' '0%
t
33 8-10% Empirically, 64 stocks
- 52 6-8%
4-563 9 4-6 % had no change.
171 2-4 %
-148 -2 %
Theoretically, only
64 NO CHANGE
DOWN
25 ± 5 should have
157 <2% not moved.
142 2-4 %
-463 90 4-6 %
44 6-8% “No change” is too
16 y8- 10%
17 >~~~~~~i
10?h
highly populated: the
“ruler effect.”
13. Histogram of percentage price changes from "The Exchange"
90 NYSE listed common stocks for the month ending July 15, 1958.

the day, and at a price equal to or greater (less) than the pub-
h (low) for the preceding year.
Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 19/ 21
Structure Within Stock Market Brownian Motion

Summary:

The picture Brownian motion does not imply that there can
be no underlying rational structure.

There is underlying structure due to the nature of market

participants!

There are natural phenomena whose data present problems for

analysis quite similar to those offered by the market.

However, the most probable value of the expected change in

loge price from a random choice common stock at a random
time is zero.

. . . under specific conditions and times, one can find a sample

of stocks for which E (∆ ln P) is slightly different from zero.

Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 20/ 21
Structure Within Stock Market Brownian Motion

Follow-on work:
The accurate record of stock market ticker prices displays
striking properties of dependence. We find for example
that after a decline of 1/8 of a point between
transactions, an advance on the next transaction is three
times as likely as a decline. Further examinations disclose
that after two price changes in the same direction, the
odds in favor of a continuation in that direction are
almost twice as great as after two changes in opposite
directions.”
– Niederhoffer and Osborne (1965)a
a
V. Niederhoffer and M.F.M. Osborne, “Market Making and Reversal on the
Stock Exchange,” Journal of the American Statistical Association, 61 (316)
897–916 (1966).

Lecture 4 – Brownian Motion Substructure: R. J. Hawkins Econ 138: Financial and Behavioral Economics 21/ 21