II Jornadas de Contabilidade e Finanas do ISVOUGA Papel da Contabilidade e Finanas nos Desafios da Atualidade

FINANCIAL RATIOS APPLIED TO PORTFOLIO MANAGEMENT:

ELECTRE III METHODOLOGY IN BUY AND HOLD STRATEGY

Antonieta Lima, Vasco Soares Novembro 2012

Portfolio managment key issues

Define which investment(s) to undertake, or in what asset(s) to invest, is often not clear given a set of alternatives; Investment decision may be influenced by all kinds of

constraints, generating a multicriteria decision problem;

Expected returns on various asset classes are key inputs in portfolio decisions;

Determine the best method to estimate expected returns, in a

more rigorous and assertive way, is a difficult task.

Objetive of the paper

Transform portfolio managment into a multicriteria problem; Use ELECTRE III method to select which assets choose to form a certain portfolio, based on financial ratios, in a buy and hold strategy Test if portfolios based on ELECTRE III outperforms the market (PSI-20TR).

The basis of Multicriteria Decision Making Models

Definition of a decision problem consists into a two-element process (C, ) (Roy, 1991, EscobarToledo & Lpez-Garcia, 2005, and Zbigniew & Watrbski (2008):
C represents a set of criteria, describing relations between properties of decision alternatives and preference levels of considered alternatives, and represents a set of meta-data of a decision situation, consisting into the decision makers expectations about a decision situation.

ELECTRE III method main features

ELECTRE from the European school (Roy, 1991) is a method (F(C, ) -> max u) which allows to max u, where u is an indicator of a decision makers

satisfaction, measured by his preferences;

ELECTRE III methods, considers a indifference

threshold q, a preference threshold p, and an

additional binary relation Q, in a ranking problematic.

ELECTRE III method main features

aPbh (a is strongly preferred to bh) g(a) - g(bh)> p

aQbh (a is weakly preferred to bh) q < g(a) - g(bh) <= p

aIbh (a is indifferent to bh, and bh to a) |g(a) - g(bh)|<= q

Using a concordance index and a discordance index for each criteria gj, descending and ascending distillations procedures are conducted (Belton & Stewart, 2001, and Rogers et al., 1999). Being defined the two pre-orders, they are combined to get the final overall ranking of the alternatives.

The importance of financial theory

Allow us to understand why some firms cease growing, discontinue, fail,

or go into bankruptcy (Beaver, 1966, 1968, and Altman, 1968);

The announcement of accounting results influence the behavior of investors in capital markets, for instance, the assets liquidity (Romacho & Cidrais,2007) : the most liquid assets have an increased profitability mainly in the pre-announcement of results: investors can anticipate the results through speculation and rumors, being these forecasts consistent with the results presented by companies.

The importance of financial theory

in the less liquid assets, an increase profitability variability was verified, before and after the announcement of results: although investors have anticipated results, they continued to feel the need to adjust their behavior after the

announcement of results, however this

adjustment is quite fast.

Methodology
Empirical work objectives:
Give to portfolio management a new perspective in an effort
to support decision maker in his investment decision. So, the hypothesis is to transform portfolio management into a multicriteria problem, and using ELECTRE III method, we explore the application of financial theory (financial ratios)

to select which assets choose to form a certain portfolio, and

test if outperform the market (PSI-20TR), in a buy and hold strategy.

Methodology
Financial ratios (Criterias of the model):
Return on Assets (ROA): express how much profit a
company generated compared to its assets.

ROAN = Net IncomeN / Total AssetsN-1

Methodology
Return on Equity (ROE): Give us the ratio between profits and shareholders' equity, and is expected to have a rate of return higher than the rate of return on treasury bonds, to be able to say that the company is really profitable. ROEN = Net Income N / Equity N-1

Methodology
Financial Autonomy (FA): This ratio related to the company's financial structure, express the extent to

which the asset is being financed by equity and debt

capital. FAN = Equity N / Total Assets N

Methodology
General Liquidity (GL): Liquidity refers to the ability to convert the asset into cash, being some items more

liquid than others.

GLN = (Clients + Stocks+ Cash) N / Current liabilities N

Methodology
Reduced Liquidity (RL): Measures a company's ability to meet its short-term liabilities with cash provenience of its net assets, but in a way more demanding than in the

general liquidity ratio, assuming that stocks (stocks of raw

materials and intermediate and finished products) will be difficult to convert into cash quickly. RLN = (Clients + Cash) N / Current liabilities N

Methodology
Hypothesis of the model:
Portfolios composed by assets selected by ELECTRE outperform the market (PSI-20TR)? Portfolio profitability is higher than market profitability?
Is portfolio riskier than the market? Portfolio standard deviation is lower than market standard deviation?

Methodology
Data:
Alternatives: Shares traded in PSI-Geral from 1999 to
2011;

Weekly closing prices relative to shares selected, from

1999 to 2011, collected from Euronext site;

Weekly closing prices from PSI-20TR from 1999 to

2011, collected from Euronext site;

Methodology
Monthly data from interest treasury bonds, 3M, from 1999 to 2011, Institute for the Management of Treasury and Public Credit, IP (IGCP, IP); Annual financial data from financial statements (balance sheet, income statement and annex for each company previously selected) from 1999 to 2011, in order to calculated financial ratios. Specifically we collected: total asset, equity, net income, current liabilities, cash, clients and inventories; Dividends paid by each company selected, from 2000 to 2011.

Methodology
Table 2 Portfolios dates
Portfolio n P1 P2 P3 P4 P5 P6 P7 P8 Initial /Historical Period 2000 2004 2001 2005 2002 2006 2003 2007 2004 2008 2005 2009 2006 2010 2007 2011 Follow-up Period
1y Jan05-Dec05 Jan06-Dec06 Jan07-Dec07 Jan08-Dec08 Jan09-Dec09 Jan10-Dec10 Jan11-Dec11 2y Jan05-Dec06 Jan06-Dec07 Jan07-Dec08 Jan08-Dec09 Jan10-Dec10 Jan11-Dec11 3y Jan05-Dec07 Jan06-Dec08 Jan07-Dec09 Jan08-Dec10 Jan10-Dec11

Source: Own elaboration, May, 2012.

Methodology
Thresholds of the model: For ROA and ROE: P 1 (p > 3,42%; q 2,63%) P 2 (p > 2,88%; q 2,21%) P 3 (p > 2,80%; q 2,15%) P 4 (p > 2,79%; q 2,14%) P 5 (p > 3,14%; q 2,41%) P 6 (p > 3,24%; q 2,50%) P 7 (p > 3,07%; q 2,36%) P 8 (p > 2,90%; q 2,23%) q threshold (indifference) correspond to the interest treasury bonds, 3M, annual average, for the periods defined; p threshold (preference) corresponds to q threshold plus 30%.

Methodology
For FA: p > 30%; q 25% For GL and RL: p > 1,50; q 1,00

where, q threshold (indifference); p threshold (preference).

Results
Table 3 Shares selected by ELECTRE
P1 (00-04) 1 Orey 2 3 4 5 6 Ibersol Brisa Cimpor Semapa BPI Cofina 7 PT Soane BCP 8 9 Portucel Reditus 9 6 7 P2 (01-05) 1 Cimpor 2 3 Orey Ibersol Brisa Semapa BPI Cofina Teixeira Duarte PT BCP Mota Engil Portucel Reditus BES 8 7 P3 (02-06) 1 Cimpor 2 3 Ibersol PT Semapa Teixeira Duarte Orey Cofina 5 6 Brisa Zon BPI Sonae Portucel BCP Jernimo Martins BES 7 P4 (03-07) 1 Cimpor 2 3 Zon Ibersol PT Teixeira Duarte Sonae Reditus Semapa Cofina BPI Orey Jernimo Martins Portucel 8 9 Brisa Mota Engil 5 4 P5 (04-08) 1 Cimpor 2 Zon PT Reditus 3 BPI Ibersol Jernimo Martins Orey Semapa Sonae Portucel Brisa 6 7 P6 (05-09) 1 Cimpor 2 3 4 PT Jernimo Martins Zon BPI Portucel 5 6 7 Ibersol Reditus Mota Engil EDP Teixeira Duarte Brisa 7 P7 (06-10) 1 PT 2 3 4 5 6 Brisa Jernimo Martins Ibersol Zon Portucel Cimpor Mota Engil EDP Semapa BPI Reditus P8 (07-11) 1 PT Jernimo 2 Martins 3 4 Brisa Ibersol Portucel Zon Cimpor EDP Mota Engil Semapa

4 5

5 6 7 8 9

8 9

10 BPI Orey Corticeira 11 Amorim 12 Cofina 13 Soane

10 Toyota BES

Mota Engil 9 Semapa Soares da Soares da Costa 10 Costa

Source: Own elaboration based on ELECTRE III outputs, May 2012.

Results
Profitability by portfolio
Chart 1 Profitability: P1 vs PSI

2000-2004

2005

2005-2006

2005-2007

Source: Own elaboration, September 2012

Results Profitability by portfolio

Chart 2 Profitability: P2 vs PSI

2001-2005

2006

2006-2007

2006-2008

Source: Own elaboration, September 2012

Results Profitability by portfolio

Chart 3 Profitability: P3 vs PSI

2002-2006

2007

2007-2008

2007-2009

Source: Own elaboration, September 2012

Results Profitability by portfolio

Chart 4 Profitability: P4 vs PSI

2003-2007

2008

2008-2009

2008-2010

Source: Own elaboration, September 2012

Results Profitability by portfolio

Chart 5 Profitability: P5 vs PSI

2004-2008

2009

2009-2010

2009-2011

Source: Own elaboration, September 2012

Results Profitability by portfolio

Chart 6 Profitability: P6 vs PSI

2005-2009

2010

2010-2011

Source: Own elaboration, September 2012

Results Profitability by portfolio

Chart 7 Profitability: P7 vs PSI

2006-2010

2011

Source: Own elaboration, September 2012

Results Profitability by portfolio

Chart 8 Profitability: P8 vs PSI

2007-2011

Source: Own elaboration, September 2012

Results
Profitability by holding period
Table 4 Follow-up results, by group (1 year)
Follow-up Results - 1 y PSI ELECTRE MEAN -0,2060% 0,1093% SHARPE'S INDEX -6,77% -1,27% N OBSERVATIONS 84 84 RISK FREE 0,1832%
Source: Own elaboration, September 2012

Results Profitability by holding period

Table 5 Follow-up results, by group (2 years)
Follow-up - 2 y PSI ELECTRE MEAN -0,2060% 0,2264% SHARPE'S INDEX -6,8551% 0,6226% N OBSERVATIONS 84 144 RISK FREE 0,1879%
Source: Own elaboration, September 2012

Results Profitability by holding period

Table 6 Follow-up results, by group (3 years)
Follow-up - 3 y PSI ELECTRE -0,2060% 0,3902% -6,9733% 3,0272% 84 180 0,1947%

MEAN SHARPE'S INDEX N OBSERVATIONS RISK FREE

Source: Own elaboration, September 2012

Results
Parametric and Non-parametric tests:
Test the significance of assumptions (methods of the
model) that may influence the behaviour of variable

of reference (profitability), by comparing means. In

our work:
Which method (ELECTRE or PSI) allows to achieve better
profitability?

Results Parametric tests

Normality Test: Parametric tests demands two conditions:
(1) depend variable have normal distribution (KolmogorovSmirnov test or Shapiro-Wilk test ), and (2) population variance is homogeneous, when we are comparing two or more variables (Levene test).

From Normality test results, only ELECTRE has a normal distribution for 1st and 2nd follow-up periods. In 3rd followup period any method has a normal distribution.

Results Parametric tests

t-student test for individual method: Tests whether the mean of a single

variable differs from a specified constant:

Ho : = 0% (average profitability of Methodi isnt significant different from zero, with i = PSI and ELECTRE). H1 : , < , > 0% (p average profitability of Methodi is significant different from zero, with i = PSI and ELECTRE).

Generally, we conclude that in every method our average profitability isnt significantly different from 0%.

Results Parametric tests

t-student test for compare two method: Compares means for two groups of cases, PSI

versus ELECTRE, (95% confidence interval):

Ho : i k = 0 (average profitability of Methodi isnt significant different from average profitability of Methodk, with i, k = PSI and ELECTRE). H1 : i k 0 (average profitability of Methodi is significant different from average profitability of Methodk, with i, k = PSI and ELECTRE).

Differences observed between average profitability for these two groups arent statistically different in every follow-up period (despite values are different, means cannot be differentiated, probably because sample is to small), but ELECTRE achieved higher average profitability than the market.

Results Non-parametric tests

Mann Whitney test: Test if distribution of one method is greater than

distribution obtained with other method (error probability of 5%):

H0: F(Xi) F(Xj) (F(Xj and F(Xj)) are method function distribution,

with i, j =PSI and ELECTRE)

H1: F(Xi) < F(Xj) (i, j = PSI and ELECTRE)

For second and third follow-up period ELECTRE achieved higher average
profitability than the market PSI. In a three years holding period, ELECTRE performance better.

Conclusions
ELECTRE methodology proved to be a good tool to select assets to invest in a buy and hold perspective, within shares traded in PSI; Compared to PSI-20 TR, and in the long term (holding period of three years), ELECTRE III, in average, achieved higher profitability than the market itself;

Conclusions
Findings left by this empirical work, leaves us open other lines of future research:
comparing ELECTRE results with results obtained by traditional methods, we would keep this conclusion?

which is better, this new line of interpreting portfolio

management or keep the traditional line of investigation?

OBRIGADA

a.lima@doc.isvouga.pt