Process in Portfolio Construction: Extending TA
Thinking beyond Signal Generation
Richard Waddington
October 2015-10-15
Abstract:
Technical Analysis offers many methods for using systematic processes to
generate trading signals. In so doing, TA insulates traders from their own biases
and leads to consistent trading.
To take the next step and move beyond single-asset trade signals, the Technical
Analyst needs to adopt similarly consistent techniques for taking risk, that is sizing
single trades and then constructing portfolios of trades from his/her signals,
taking into account the goals and targets of the portfolio, the skill set of the TA
creating the signals, and the Portfolio Investors wishes.
By doing so he/she will ensure that the work that has been done in signal
generation is not lost through poor portfolio construction.
The twin benefits of this approach are:
Better Returns
&
Better Process
These combine to make your more investable.
This paper examines these techniques and sets out a framework for implementing
a robust rigorous approach to portfolio construction.
THE PROBLEM – TRADERS NEED TO CREATE A PORTFOLIO:
Traditional T.A. gets you this far:
The Technical Analyst’s Area of Expertise
Charts
T.A. Skills
List of
Positions
GOOG
AAPL
MSFT
IBM
+1
-1
+1
-1
Then you have to do construct a portfolio. The Portfolio drives P&L, not the
individual asset picks.
Inputs
List Of
Positions
GOOG
AAPL
MSFT
IBM
+1
-1
+1
-1
The Hard Part
Position
Sizing
The Goal
The
Portfolio
There are more than
6 million ways to
create this
portfolio!!!
If you are looking at the 500 biggest stocks in the US, then you have to make 500
decisions in Stage #1: Buy/Sell or do nothing for each stock.
 If you choose 4 stocks, then in Stage #2 there can be more than 6 million
ways to combine them into a portfolio.
 If you choose 8 stocks, there are over 150 million!
WHY IS IT HARD?
The best combination to unlock the perfect portfolio involves many competing
inputs:
 Healthy
Signal quality, Asset Choices,
Market Structure, Skill levels
 Unhealthy
Ego, Emotion, Mindset, Denial,
Hope, Greed, Fear, Style Drift,
Self-Perceived Status
 Important but often Ignored
Investor (Client or Trader) Risk Tolerances and Targets
Investpot exposure appetite to Sectors or Markets
 Advanced Requirements
Factor Exposure, Sub-Portfolio Groupings,
Per Asset Limits, Conviction, Liquidity, Trading Costs
The Optimal Portfolio requires these inputs to be applied
consistently every time.
In order to do this, the conscientious T.A. needs 3 things:
 A framework for defining these inputs
 A mathematical formulation of these inputs
 A way to find the BEST combination out of the 150million (for 8 assets) that
are possible
One of the hardest problems is defining what you mean by ‘BEST’
WHAT IS BEST?
Best means remaining close to what the investor/ provider of capital is looking
for. It doesn’t necessarily mean highest return, or even highest Sharpe.
Example: The funds shown below both have 19.7% annual RoR:
The LHS one has a defined process for converting their analysis into asset weights
in the portfolio, and they have stuck to that process for 15 years. That is to say:
they have a defined process for trade sizing.
The RHS one has a higher hit rate (% of trades that are ‘correct’ i.e. profitable =
65% vs 58% for LHS), and can be therefore considered to have better analysis, but
has been inconsistent in the way that they constructed their portfolio, leading to
a huge reduction in AUM from a max of 1.8bn to 120 m.
By introducing process, you make better (= more investable) portfolios.
AUM 2.2Bn USD
AUM 120m (from a max 1.8bn)
OPTIMALLY SIZING IS A PROCESS THAT MAKES YOU MORE
INVESTABLE
Investable portfolios need to take all the Investors wishes into account. That must
start with allowing the investor to define their targets, their goals and their
exposure requirements.
The T.A must step away from his/her Signal analysis and look from the p.o.v of the
Investor.
Example: If you are creating a Trend-Following strategy and want to charge fees,
then you need to be sure that you aren’t just replicating what a Trend-Following
ETF (i.e. Pacer ETF) can do for a 0.40% p.a. fee.
You must show, through looking at other factors that your ‘alpha’ is unique.
That means taking a more holistic approach to creating a portfolio.
The Trader looks down this list when making decisions
TRADERS VIEW
1
Asset Selection
6
2
Conviction
5
3
Risk Limits
4
4
Gross Exposure Limits
3
5
Sector Exposures
2
6
Factor Exposures
1
INVESTOR/CIO VIEW
The Investor or CIO looks up this list when making decisions


Portfolio Construction provides a framework that unites these very
different views.
The Trader’s decisions and convictions are traded consistently with the
requirements defined by the Investor/CIO.
EXAMPLE: 4 PORTFOLIOS WITH SAME ASSET CHOICES, SAME
LEVERAGE, BUT DIFFERENT OUTCOMES
The Same Set of Asset Choices can lead to very different Portfolios with different
P&L Outcomes for different Risk Tolerances. Note the Gross exposure is the same.
Gross
Date
198%
16-Mar
198%
30-Mar
198%
13-Apr
198%
27-Apr
198%
11-May
198%
25-May
198%
08-Jun
Asset
Original Portfolio
8536 JP
548 HK
-11%
-11%
-11%
-11%
-11%
-11%
-11%
UENV SP 8332 JP REC AU 152 HK
-20%
11%
11%
-20%
11%
11%
-20%
11%
11%
-20%
11%
11%
-20%
11%
11%
-20%
11%
11%
-20%
11%
11%
High Risk Portfolio
Asset
Gross
Date
8536 JP 548 HK UENV SP
198%
16-Mar -15.0% -10.6% 14.2%
198%
30-Mar -15.0% -12.4% 14.8%
198%
13-Apr -14.8% -12.0% 14.8%
198%
27-Apr -14.8% -11.6% 11.8%
198%
11-May -15.0% -14.0% 13.4%
198%
25-May -15.0% -15.0% 13.2%
198%
08-Jun -15.0% -15.0%
5.2%
15%
15%
15%
15%
15%
15%
15%
11%
11%
11%
11%
11%
11%
11%
8332 JP REC AU 152 HK
13.0%
12.8%
12.8%
12.8%
13.0%
12.8%
13.0%
10.4%
9.4%
13.2%
13.8%
12.0%
10.4%
15.0%
Asset Medium Risk Portfolio
Gross
Date
8536 JP 548 HK UENV SP 8332 JP REC AU
198%
16-Mar -15.0% -15.0% 15.0% 13.0% 15.0%
198%
30-Mar -15.0% -15.0% 15.0% 13.0% 15.0%
198%
13-Apr -15.0% -15.0% 14.8% 13.0% 12.4%
198%
27-Apr -15.0% -15.0% 15.0% 13.0% 15.0%
198%
11-May -15.0% -15.0% 11.6% 12.8% 15.0%
198%
25-May -15.0% -15.0%
2.6% 12.8% 14.8%
198%
08-Jun -15.0% -15.0%
2.6% 12.8% 14.8%
10.4%
12.2%
11.8%
11.4%
13.8%
14.8%
14.8%
152 HK
-11%
-11%
-11%
-11%
-11%
-11%
-11%
KEP SP
-15%
-15%
-15%
-15%
-15%
-15%
-15%
-11%
-11%
-11%
-11%
-11%
-11%
-11%
SUN SP 5 HK
KEP SP
-12.2%
-10.6%
-15.0%
-15.0%
-10.8%
-6.4%
-14.6%
-8.4%
-10.4%
-5.6%
-8.4%
-10.2%
-13.0%
-15.0%
-15.0%
-11.0%
-14.8%
-15.0%
-12.8%
-15.0%
-11.4%
SUN SP 5 HK
KEP SP
14.8% -11.7% -2.5% -12.5%
14.8% -11.7% -2.5% -12.5%
14.8% -2.6% -5.2% -15.0%
14.8% -6.6% -2.5% -12.5%
14.8% -6.6% -3.0% -14.2%
14.8% -4.2% -15.0% -12.8%
14.8% -6.4% -10.2% -15.0%
Asset Low Risk Portfolio
Gross
Date
8536 JP 548 HK UENV SP 8332 JP REC AU 152 HK
198%
16-Mar -11.8% -7.1% 11.3% 10.0%
6.6% 6.9%
198%
30-Mar -13.7% -8.0% 12.4% 12.0%
6.7% 7.8%
198%
13-Apr -14.8% -7.3% 12.1% 13.0%
8.2% 7.0%
198%
27-Apr -16.7% -8.3% 14.1% 15.0%
9.2% 8.0%
198%
11-May -17.0% -8.0% 14.9% 15.2%
8.7% 7.8%
198%
25-May -16.3% -10.8% 15.4% 14.5%
8.8% 10.6%
198%
08-Jun -15.9% -12.5% 15.9% 13.9%
9.3% 12.3%
116
SUN SP 5 HK
SUN SP 5 HK
-17.7%
-13.7%
-14.5%
-12.1%
-15.1%
-13.0%
-12.5%
KEP SP
-18.4% -7.9%
-19.4% -7.0%
-18.2% -7.3%
-18.2% -7.3%
-16.7% -8.2%
-16.5% -8.8%
-15.9% -10.0%
1398 HK
8008 HK
152 HK
2202 HK
11%
11%
11%
11%
11%
11%
11%
11%
11%
11%
11%
11%
11%
11%
2%
2%
2%
2%
2%
2%
2%
2%
2%
2%
2%
2%
2%
2%
1398 HK
8008 HK
152 HK
2202 HK
5.4%
2.6%
2.6%
6.8%
5.0%
9.4%
2.6%
1398 HK
2.5%
2.5%
2.6%
2.5%
2.6%
2.6%
2.6%
1398 HK
8.1%
6.5%
8.2%
7.3%
7.6%
7.1%
6.8%
10.4%
11.4%
4.6%
8.0%
6.0%
2.6%
3.4%
8008 HK
2.6%
2.6%
2.6%
2.6%
2.6%
2.6%
2.6%
152 HK
4.8%
4.8%
8.6%
5.8%
6.4%
4.4%
6.8%
8008 HK
2.5%
2.5%
2.6%
2.5%
2.6%
2.6%
2.6%
152 HK
10.6%
7.0%
6.5%
4.1%
5.5%
4.2%
2.8%
3.2%
3.4%
3.1%
3.2%
3.0%
2.9%
2.8%
8035 JP AMAT US
-11%
-11%
-11%
-11%
-11%
-11%
-11%
-11%
-11%
-11%
-11%
-11%
-11%
-11%
8035 JP AMAT US
4.4%
4.8%
6.6%
3.2%
4.4%
4.0%
5.2%
2202 HK
-15.0%
-15.0%
-15.0%
-15.0%
-15.0%
-15.0%
-15.0%
-15.0%
-15.0%
-15.0%
-15.0%
-15.0%
-15.0%
-15.0%
8035 JP AMAT US
6.2%
6.2%
7.0%
8.7%
8.8%
13.4%
10.6%
2202 HK
-15.0%
-15.0%
-15.0%
-15.0%
-15.0%
-15.0%
-14.8%
-15.0%
-15.0%
-15.0%
-15.0%
-15.0%
-15.0%
-14.8%
8035 JP AMAT US
3.2%
3.4%
3.1%
3.2%
3.0%
2.9%
2.8%
-17.7%
-17.4%
-17.4%
-15.8%
-14.9%
-15.0%
-15.0%
-18.4%
-19.4%
-18.2%
-18.2%
-17.2%
-16.5%
-15.9%
6857 JP LRCX US
18%
18%
18%
18%
18%
18%
18%
6857 JP LRCX US
15.0%
15.0%
14.8%
14.8%
15.0%
14.8%
15.0%
15.0%
15.0%
15.0%
15.0%
15.0%
15.0%
15.0%
18.4%
17.1%
18.2%
16.0%
15.8%
15.4%
15.3%
Medium Risk Portfolio
Low Risk Portfolio
Original Portfolio
101
96
16-Mar
23-Mar
30-Mar
06-Apr
13-Apr
20-Apr
27-Apr
04-May
11-May
18-May
25-May
01-Jun
08-Jun
15.0%
15.0%
15.0%
15.0%
15.0%
15.0%
14.8%
6857 JP LRCX US
High Risk Portfolio
106
15.0%
15.0%
15.0%
15.0%
15.0%
15.0%
15.0%
6857 JP LRCX US
Cumulative P&L March 2015- Jun 29015 for a L/S Portfolio,
showing different outcomes for Different prescribed Risk
Tolerances
111
11%
11%
11%
11%
11%
11%
11%
15-Jun
17.4%
19.4%
17.7%
18.2%
16.5%
16.3%
15.9%
IF YOU GET IT RIGHT, WHAT DOES IT LOOK LIKE?
P&L will be improved, and this is seen over time: we can also look at a snapshot
of a properly constructed portfolio vs. one that has been constructed without
using any defined method. Again these are real examples from a Fund Manager.
Portfolios Before and After Optimal Construction
 Properly constructed portfolios have higher weights for assets that are
negatively correlated to the whole portfolio, and assets that have been
assigned a higher conviction by the Trader.
 In these portfolios, conviction drives the ‘effective risk’. The contribution to
risk of an asset is driven by the conviction (the Strength of Signal).
 These Portfolios also fit any defined asset, country, sector and factor
exposure limits.
WHAT DOES THIS DO TO P&L? IT ENHANCES IT.
This graph shows the performance of an Global Equity Long/Short Hedge-Fund
Portfolio (250m USD) through the summer of 2015: Two portfolios were run in
parallel, one with the portfolio weightings optimally calculated by the ORS
system, and one without.
Note that this portfolio has constraints across market sectors, countries, and Net
and Gross exposures, all of which are preserved across the two portfolios.
Trading costs are included in the calculations for both portfolios. These are real
numbers.
 In this case the Trader’s signal resulted in good asset & conviction calls. The
portfolio that was properly sized outperformed, returning an extra 1.1%
over 3 months and having a higher Sharpe.
 If asset calls are not as successful, the effect of having lower weights for
positively correlated assets reduces the impact of the poor call on portfolio
P&L.
FURTHER EXAMPLES
The graph below shows the Cumulative P&Ls for a large (500m USD+) Macro
Hedge Fund that trades in liquid futures and FX. The nature of the fund is to take
large positions for short periods (1-3 days).
The Fund performed very well over the 3.5 years of this analysis (Red line), but as
can be seen had highly variable exposure and no consistent way of describing
their portfolio.
When the same asset selection decisions are made, at the same time with the
same execution prices, the P&L from the optimally weighted portfolio (Blue line)
is demonstrably better, both in terms of returns and Sharpe.
ORS Sized and Fund YPortfolio P&L
150.00%
500.00%
147.50%
145.00%
Fund Y Sharpe = 1.4
450.00%
ORS Sharpe=1.6
400.00%
Fund Y Gross
Exposure
350.00%
142.50%
140.00%
137.50%
135.00%
132.50%
130.00%
127.50%
125.00%
122.50%
300.00%
ORS Gross Exposure
120.00%
250.00%
117.50%
115.00%
200.00%
112.50%
110.00%
150.00%
107.50%
105.00%
102.50%
100.00%
100.00%
97.50%
50.00%
95.00%
92.50%
90.00%
01/04/2012
18/10/2012
06/05/2013
22/11/2013
10/06/2014
27/12/2014
15/07/2015
0.00%
31/01/2016
By introducing a rigorous optimal sizing process into the Funds workflow, the
Managers were able to demonstrate adherence to procedures that emphasised
the quality of their signal selection without the damaging effects of inconsistent
sizing and human biases.
This in turn made them more investable.
FINAL EXAMPLE: SIZING AS DIFFERENTIATOR
This particular Fund manager needed to differentiate themselves in a competitive
bid for assets. They showed the investor (a Government-run pension scheme) the
following graphs, demonstrating how their new Optimal Risk Sizing approach
would help them in both rising and flat/volatile markets.
The graph shows how they would have performed (Blue line) based on actual
performance (Red line), given their new Risk Sizing approach.
They won a 300m USD 4 year mandate from this pitch.
Account 1 Results With/Without Optimal Risk Sizing
3000
10000
Actual P&L
9000
Optimally Sized P&L
2500
8000
Index
2000
7000
6000
1500
5000
1000
4000
3000
500
2000
0
14-Sep-11
01-Apr-12
18-Oct-12
06-May-13
22-Nov-13
10-Jun-14
27-Dec-14
-500
15-Jul-15
1000
0
Account 2 Results With/Without Optimal Risk Sizing
2500
10000
Actual P&L
9000
2000
Optimally Sized P&L
8000
Index
1500
7000
6000
1000
5000
500
4000
0
01-Apr-12
3000
18-Oct-12
06-May-13
22-Nov-13
10-Jun-14
27-Dec-14
15-Jul-15
31-Jan-16
2000
-500
1000
-1000
0
SUMMARY: THE IMPACT OF PORTFOLIO CONSTRUCTION
When you build a framework for formulating the right inputs and then
consistently apply them to the asset selections that come from your TA signal,
you build a wall around your biases and let your quality analysis through.
You end up with a better portfolio, better returns & a better process,
all of which make you more investable.
Your asset-by-asset T.A. is
good, but when you build the
portfolio you introduce biases
that negatively impact
performance.
ORS build a wall around your
biases & lets your skill through
The Technical Analysts’ Expertise
Trade by Trade
T.A. Signals
Each Asset: Buy,
Sell, Do Nothing
List of
Positions
Portfolio Construction
You replace the
biases and
inconsistent
inputs with clear
objectives.
Portfolio
SO WHAT IS THE CALCULATION?
There are 4 levels of inputs needed to perform this calculation.
Note that implicitly you are already defining all these parameters when you
construct a portfolio, but you will be unable to be consistent about it, rather like
creating T.A. decisions by eyeballing charts rather than rigorous analysis.
1. Take inputs from the Trader that change as the asset decisions are changed
 The Positions you want to buy/hold/sell. The T.A’s skill is in making
and timing these decisions, based on Technical analysis
 Market Data
2. Take inputs regarding what the Portfolio is trying to achieve
 Portfolio Risk Tolerance and Portfolio Goals
3. Take constraints and conviction information on three levels: the Portfolio,
Sub Portfolios and Assets
 Risk and Factor Exposures
 Sub Portfolio Net and Gross Limits
 Asset constraints (min/max size) and conviction levels
4. Take at a Portfolio or Asset level a measure of the quality of the decision
making and the overall exposure limits
 This is similar to considering the historical Sharpe ratio of the Trader
It is important to consider practical limits imposed by trading costs and market
liquidity to create a Practical Optimal Portfolio, “The POP”, as well as a
Benchmark Optimal Portfolio “The BOP”. Hence you need to take into account
the volatility of each asset and the trading cost of that asset.
EXAMPLE INPUTS
1. Asset Directions and Market Data
Header
8536 JP
Position
Short
548 HK
Short
UENV SP
Long
8332 JP
REC AU
Long
Long
152 HK
Long
SUN SP
Short
5 HK
Short
2. Portfolio Risk Tolerance, showing the ‘utility’ function of the portfolio
3. Sub Constraints and Groupings for a multi currency stock portfolio
Group
Portfolio
Currency
Currency
Currency
Currency
Currency
Currency
Sector
Sector
Sector
Sector
Sector
MarketCap Factor
Momentum Factor
Identifier
All
TWD
JPY
HKD
KRW
USD
CNY
Tech
Util
Health
Materials
Finance
MarketCap
Momentum
minNet
-10.0%
-4.44%
-3.24%
0.93%
-0.64%
0.84%
-0.51%
-4.44%
-3.24%
0.93%
-0.64%
0.84%
(150)
(5)
maxNet
10.0%
-3.43%
-2.23%
1.95%
0.37%
1.86%
0.51%
-3.43%
-2.23%
1.95%
0.37%
1.86%
minGross
101.5%
22.55%
14.18%
2.11%
8.21%
1.41%
0.00%
22.55%
14.18%
2.11%
8.21%
1.41%
150
5
4. Portfolio Estimated Sharpe and Net/Gross Constraints
maxGross
120.0%
62.32%
39.17%
9.73%
22.68%
8.25%
0.67%
62.32%
39.17%
9.73%
22.68%
8.25%
450
600
20
KEP SP
Short
AND HOW DO YOU DO THIS CALCULATION?
1. Consider all the inputs, and represent them mathematically.
2. Search all possible Portfolios that comply with the input constraints above,
for the Portfolio that has the ‘Best' weighting of assets.
 ‘Best’ means the one that gives the highest ‘Utility’ of returns.
 ‘Utility’ is defined by measuring the returns against Input #1 “What I
am trying to Achieve?”
 Utility measured not as an absolute ‘Best’ but as a stable best. It is
important tio create a portfolio that is stable over time rather than just
one that is instantaneously optimal.
To perform the search, you must build a risk model to compute the utility of
returns by taking the historical data, and adjusting it for individual asset
convictions and any forward looking estimates as given by the user.
 Note for a portfolio of 10 assets, each of which can be between 5% &
15%, there are 10bn possible combinations, just looking at integer %. If
the Portfolio has to be exactly 100% invested, the number is about half
that.
Methods for performing this search of all the possible combinations of assets
include dimension reduction (PCA), Monte-Carlo and multi-generational optimiser
techniques. These are well established techniques in academic, engineering,
science and finance which go beyond the scope of this paper.
The purpose of this paper is not to describe these techniques, but rather to
describe how to create a framework in which you can use them to build your
portfolio.
There are practical considerations w.r.t the stability of solutions found by such
techniques in general, so please note that experience dealing with multi-variate
optimisers in finance is essential.
EXAMPLE OUTPUTS
The primary output is the Optimal Weightings for each asset in the Portfolio
ORS
Weighting
8536 JP Equity
548 HK Equity
UENV SP Equity
8332 JP Equity
REC AU Equity
152 HK Equity
SUN SP Equity
5 HK EquityKEP SP Equity
1398 HK Equity
-9.6%
-5.8%
9.2%
9.8%
5.4%
5.6%
-14.4%
-15.0%
-6.4%
6.6%
Subsidiary outputs can be snapshots of the portfolio correlation, historical P&ol
and other simple risk analytics to demonstrate the value of correctly sizing the
sassets int jhe portfolio.
We can look at them in a Data Visualisation tool, here it is Tableau:
The T.A. must decide what to do with the information: the simplest choice is to
use this approach to size your trades directly, but we have also seen people using
it as a tool for communicating the reasons for out-sized (or undersized) positions l
to Investors and Managers and having Pre-p&L discussions around the portfolio.
SUMMARY
The Technical Analyst can extend the power of his/her T.A. techniques to the
creation of the whole portfolio by following the same rigour and discipline in
Portfolio Construction as he/she does in signal generation.
In so doing he/she will ensure that the work that has been carried out in
generating signals is not wasted by taking inappropriate portfolio risk.
And as result the T.A. will enjoy better returns, a smoother process and better
relations with all the stakeholders in their investments.
The problems that need to be solved are mathematically difficult, and generally
require access to a reasonable amount of computing power, these days easily
available through cloud based SaaS programs.
And the Technical Analyst will have to remove from the decision making process
all the loopholes that allow him/her to game the system.
Any Technical Analyst who wants their work to be applied seriously in the context
of commercial Asset Management needs a real industrial-strength process for
converting good asset analysis into great portfolios.
This will make them investable and help them with their day-to-day process.
This starts with honesty about the objectives, clarity about the constraints and
the ability to be ego free about your work. These are all characteristics of a
successful T.A. and Portfolio Construction is a logical and easy next step for the
T.A. who has embraced these ideas.
This will lead to transparency around processes and returns that show the true
value of our analysis, thereby opening up your portfolio to further investment.
AUTHOR BIOGRAPHY
Richard Waddington
Richard has worked in mathematical finance for 20 years, starting as a junior
trader in Bankers Trust’s London FX derivatives desk, and continuing on through
stints in New York, Tokyo and Singapore as a Trader, Business Head and
technology entrepreneur. Richard’s interest has always been in putting process in
place in complex business situations.
Richard graduated from Cambridge University, having read undergraduate Physics
and Mathematics and post-graduate Manufacturing Engineering.
FURTHER READING AND BIBLIOGRAPHY
HARRY M. MARKOWITZ - AUTOBIOGRAPHY, THE NOBEL PRIZES 1990, EDITOR TORE FRÄNGSMYR, [NOBEL
FOUNDATION], STOCKHOLM, 1991
MERTON, ROBERT. "AN ANALYTIC DERIVATION OF THE EFFICIENT PORTFOLIO FRONTIER," JOURNAL OF FINANCIAL AND
QUANTITATIVE ANALYSIS7, SEPTEMBER 1972, 1851-1872.
RODIE, DE MOL, DAUBECHIES, GIANNONE AND LORIS (2009)."SPARSE AND STABLE MARKOWITZ
PORTFOLIOS". PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES 106 (30).
CHANDRA, SIDDHARTH; SHADEL, WILLIAM G. (2007). "CROSSING DISCIPLINARY BOUNDARIES: APPLYING FINANCIAL
PORTFOLIO THEORY TO MODEL THE ORGANIZATION OF THE SELF-CONCEPT". JOURNAL OF RESEARCH IN
PERSONALITY 41 (2): 346–373
ROSS & NISBETT (1977) “THE PERSON AND THE SITUATION, PERSPECTIVES OF SOCIAL PSYCHOLOGY”