O P E R A T I O N S R E S E A R C H A N D D E C I S I O N S

No. 2 2010

Joanna OLBRYŚ*

THREE-FACTOR MARKET-TIMING MODELS

WITH FAMA AND FRENCH’S SPREAD VARIABLES1

The traditional performance measurement literature has attempted to distinguish security selec-
tion, or stock-picking ability, from market-timing, or the ability to predict overall market returns.
However, the literature finds that it is not easy to separate ability into such dichotomous categories.
Some researchers have developed models that allow the decomposition of manager performance into
market-timing and selectivity skills. The main goal of this paper is to present modified versions of
classic market-timing models with Fama and French’s spread variables SMB and HML, in the case of
Polish equity mutual funds.

Keywords: mutual funds, performance evaluation, market-timing, selectivity, mimicking portfolios

1. Introduction

FAMA and FRENCH [6] found that two variables, the market value (MV) and the
book value to market value ratio (BV/MV) capture much of the cross-section of aver-
age stock returns. They report that “(...) firms that have high BV/MV (a low stock price
relative to book value) tend to have low earnings on assets, and the low earnings per-
sist for at least five years before and five years after book-to-market equity is meas-
ured. Conversely, low BV/MV (a high stock price relative to book value) is associated
with persistently high earnings. Size is also related to profitability. Controlling for
book-to-market equity, small firms tend to have lower earnings on assets than big
firms” [7, pp. 7–8].

* Faculty of Computer Science, Bialystok University of Technology, ul. Wiejska 45A, 15-351 Biały-
stok, e-mail: [email protected]
1
Financial support in 2009–2011 from the Polish Committee for Scientific Research within the grant
No. N N113 173237 is gratefully acknowledged.
92 J. OLBRYŚ

In 1993, FAMA and FRENCH [7] formed portfolios meant to mimic the underlying
risk factors in returns related to size and book-to-market equity. These mimicking
portfolios are introduced as explanatory variables into regressions of stock excess
returns. Fama and French examined regressions that use the following explanatory
variables:
• Rm. – RF – the market factor (i.e. the excess return on market portfolio M) or suit-
able orthogonalized market factor,
• SMB (small minus big) – the size factor,
• HML (high minus low) – the book-to-market factor.
In 1972 FAMA [5] suggested that the portfolio manager’s forecasting ability could be
split into two separate activities: microforecasting (which is referred to in the literature as
selectivity) and macroforecasting (which is referred to in literature as market-timing).
Classical performance measurement literature has attempted to distinguish security se-
lection, or stock-picking ability, from market-timing, or the ability to predict overall
market returns. However, the literature indicates that it is not easy to separate ability into
such dichotomous categories. The goal of this paper is to present modified market-timing
models with Fama and French’s spread variables SMB and HML.
The paper is organized as follows. Chapter 2 provides a brief overview of the
Fama and French procedure of constructing a mimicking portfolio based on Polish
companies listed on the Warsaw Stock Exchange. In Chapter 3 the classic parametric
T–M (TREYNOR–MAZUY [27]) and H–M (HENRIKSSON–MERTON [11]) market-
timing models are tested in the case of 15 Polish equity open-end mutual funds. In
Chapter 4 modified three-factor versions of the T–M and H–M models are presented.
The additional factors are Fama and French’s mimicking portfolios SMB and HML.
The market-timing and selectivity abilities of fund managers are evaluated for the period
January 2003–December 2009. In Chapter 5 we compare the regression results of the
models and investigate their statistical properties.

2. Fama and French’s mimicking portfolios SMB and HML

We construct the size (SMB) and book-to-market (HML) mimicking portfolios

using the Fama and French procedure [7]. We sort all the firms listed on the Warsaw
Stock Exchange according to the following conditions:
1. Firms must have existed on the Warsaw Stock Exchange for two years before
we use them, that is from December 31, 2001 (when the end of calendar year book
value per share is available based on the annual report);
2. The database of the stock “closing” prices must not have too many missing val-
ues (for example, Optimus does not meet this condition);
 Three-factor market-timing models with Fama and French’s spread variables 93

3. We do not use negative –BV firms (for example, Atlantis, Bytom, Mostostal
Zabrze, Netia, Polnord, Próchnik, Stalexport, Swarzędz are negative –BV firms);
4. All of the firm’s annual reports must be available (for example, Advadis and
Getin do not meet this condition).
Finally, the 61 Warsaw Stock Exchange companies that complied with conditions
1–4 were entered into the database. The stock “closing” prices were obtained from
http://bossa.pl. Annual reports were obtained from www.bankier.pl (based on Notoria
Service).
We sort all firms according to their market capitalization at the end of June each
year, beginning in June 28, 2002. We take the market capitalization MV to be the
number of shares as of the end of June (per WSE) multiplied by the end of June WSE
share price. We also sort these same firms according to their end of calendar year
book-to-market ratio BV/MV.
In June of each year t from 2002 to 2009, all stocks are ranked according to the
size of MV. The median size is then used to divide these stocks into two groups,
S – Small and B – Big, where the big group includes all the firms with market capitali-
zation greater than or equal to the median.
Next we divide the stocks into three book-to-market equity groups based on the
breakpoints for the bottom 30% (Low), middle 40% (Medium) and top 30% (High) of
the ranked values of BV/MV for stocks.
The decision to sort firms into three groups according to BV/MV and only two ac-
cording to MV follows Fama and French’s observation that book-to-market equity has
a stronger role in average stock returns than size.

Table 1. The number of firms contained in the auxiliary portfolios

(at the end of June each year)

index 2002 2003 2004 2005 2006 2007 2008 2009

BH 0 0 1 1 0 3 2 5
BM 12 13 14 12 15 13 12 12
BL 18 17 15 17 15 14 16 13
SH 18 18 17 17 18 15 16 13
SM 12 11 10 12 9 11 12 12
SL 1 2 4 2 4 5 3 6
∑ 61 61 61 61 61 61 61 61

Source: Author’s calculations.

Proceeding in this way, we construct six portfolios for each year: BH, BM, BL, SH,
SM, SL from the intersections of the two MV and the three BV/MV groups. For exam-
ple, the BH portfolio contains the stocks in the B – Big size group that are also in the
H – High book-to-market value group. The daily value-weighted returns on the six
94 J. OLBRYŚ

portfolios are calculated from July of year t to June of (t + 1), and the firms associated
with each portfolio are updated – according to the data from the end of June of (t + 1).
We calculate returns beginning in July of year t so that book equity for the year (t – 1)
is known [6].
In the next step we form the size mimicking portfolio SMB by taking the difference
between an equally weighted combination of three small market capitalization indices
and three big market capitalization indices. The rate of return on this portfolio is equal
to (1):
1
RSMB = ⋅ ( RSH + RSM + RSL − RBH − RBM − RBL ) . (1)
3
In the last step we form the book-to-market mimicking portfolio HML by taking
the difference between an equally weighted combination of two high book-to-market
indices and two low book-to-market indices. The rate of return on the HML portfolio
is equal to (2):
1
RHML = ⋅ ( RBH + RSH − RBL − RSL ) . (2)
2

Fig. 1. Daily rates of return on the WSE stock index (WIG)

in the period January 2, 2003–December 31, 2009
Source: Author’s calculations.

Figures 1, 2, 3 present daily rates of return on the WSE stock index (WIG), the
mimicking portfolio SMB and the mimicking portfolio HML respectively, in the pe-
riod investigated January 2, 2003–December 31, 2009.
 Three-factor market-timing models with Fama and French’s spread variables 95

Fig. 2. Daily rates of return on the mimicking portfolio SMB

in the period January 2, 2003–December 31, 2009
Source: Author’s calculations.

Fig. 3. Daily rates of return on the mimicking portfolio HML

in the period January 2, 2003–December 31, 2009
Source: Author’s calculations.

We use daily data following BOLLEN and BUSSE’S observation [1] that daily data
provide better opportunity for inference regarding timing ability than monthly data.
This evidence has been examined in the case of Polish equity mutual funds by
OLBRYŚ [20]. Table 2 reports summarized return statistics for the SMB, HML and WIG
portfolios. The SMB and HML portfolios exhibit higher excess kurtosis than the WIG
index.
96 J. OLBRYŚ

Table 2. Summarized statistics for market portfolios’ daily returns (Jan. 2, 2003–Dec. 31, 2009)

Mean Min Max Standard deviation Skewness Kurtosis

SMB 0.0005 –0.201 0.086 0.015 –1.956 28.403
HML 0.0007 –0.086 0.267 0.016 3.904 55.965
WIG 0.0007 –0.080 0.063 0.014 –0.259 2.686
Source: Author’s calculations.

Table 3 presents Pearson coefficients of correlation. We can observe that the Pear-
son correlations between pairs of variables are negative and rather weak.

Table 3. Pearson coefficients of correlation (Jan 2, 2003–Dec 31, 2009)

(X, Y ) (SMB, HML) (SMB, WIG) (HML, WIG)

r(X, Y ) –0.370 –0.165 –0.121
Source: Author’s calculations.

3. Classical parametric market-timing models

TREYNOR and MAZUY [27] (the T–M model) developed a procedure for detecting
timing ability that is based on a regression analysis of the realised returns of managed
portfolio, which includes a quadratic term, as follows:
rP , t = α P + β P ⋅ rM , t + γ P ⋅ (rM , t ) 2 + ε P ,t (3)

where:
rP , t = RP , t − RF , t is the simple excess return on portfolio P in period t,
rM , t = RM , t − RF , t is the simple excess return on portfolio M in period t,
RP , t is the one-period return on portfolio P,
RM ,t is the one-period return on market portfolio M,
RF ,t is the one-period return on riskless securities,
Jensen’s α P [13] measures the selectivity skills of the manager of portfolio P,
β P is the systematic risk measure of portfolio P,
γ P measures the market-timing skills of the manager of portfolio P,
ε P , t is a residual term, with the following standard CAPM conditions: E (ε P , t ) = 0 ,
E (ε P , t | ε P , t −1 ) = 0.
 Three-factor market-timing models with Fama and French’s spread variables 97

If the portfolio manager has the ability to forecast security prices, the intercept α P
in equation (3) will be positive [13, pp. 394]. Indeed, it represents the average incre-
mental rate of return on the portfolio per unit time which is due solely to the man-
ager’s ability to forecast future security prices. On the one hand, a passive strategy
(random buy-and-hold policy) can be expected to yield a zero intercept. On the other
hand, if the manager is doing worse than a random selection buy-and-hold policy, α P
will be negative.
If a mutual fund manager increases (decreases) the market exposure of the portfo-
lio prior to a market increase (decrease) then the portfolio return will be a convex
function of the market return and γ P will be positive. The size of the estimate γˆP
informs us about the manager’s market skills.
The parametric T–M test examined the performance of 15 selected equity open-
end mutual funds using daily data. We have studied daily ordinary excess returns from
Jan. 2003 to Dec. 2009. Daily returns on the main index of Warsaw Stock Exchange
companies are used as the returns on the market portfolio. The daily average of returns
on 52-week Treasury bills are used as returns on riskless assets.

Table 4. Classic T–M model (3) (Jan 2, 2003–Dec 31, 2009)

Equity funds α̂ P βˆP γˆP R2 DW AIC

1 Arka BZ WBK Akcji FIO 0.0007* 0.711* -2.34* 0.617 2.42 -12008.2
2 Aviva Investors FIO Polskich Akcji 0.0005* 0.748* –1.92* 0.690 2.56 –12403.0
3 BPH FIO Akcji 0.0002 0.709* –1.02* 0.715 2.69 –12820.8
4 DWS Polska FIO
0.0005* 0.396* –2.47* 0.240 2.47 –11123.5
Top 25 Małych Spółek
5 DWS Polska FIO Akcji 0.0002 0.621* –1.44 0.372 2.75 –10732.8
6 DWS Polska FIO Akcji Plus 0.0003* 0.534* –1.66* 0.346 2.75 –11055.9
7 ING FIO Akcji 0.0001 0.741* –0.96* 0.694 2.71 –12495.6
8 Legg Mason Akcji FIO 0.0003* 0.690* –1.13* 0.698 2.61 –12774.1
9 Millennium FIO Akcji 0.0001 0.670* –1.22* 0.669 2.64 –12635.1
10 Pioneer Akcji Polskich FIO 0.0001 0.809* –1.53 0.696 2.66 –12196.2
11 PKO/CREDIT SUISSE Akcji FIO 0.0003 0.550* –2.33* 0.420 2.71 –11481.0
12 PZU FIO Akcji KRAKOWIAK 0.0002 0.698* –1.39* 0.686 2.66 –12631.8
13 SEB 3 – Akcji FIO 0.0005* 0.504* –1.20* 0.307 2.68 –10930.9
14 Skarbiec – Akcja FIO 0.0004* 0.443* –0.81 0.273 2.64 –11130.1
15 UniKorona Akcja FIO 0.0005* 0.499* –1.44* 0.300 2.70 –10926.6
*Significant at the 5% level
Source: Author’s calculations (using Gretl 1.8.5).

Table 4 provides details on the estimated market-timing T–M models. The Newey–
West (1987) procedures have been used to correct for both autocorrelated and hetero-
skedastic error terms. Results of the T–M tests show that the estimates of JENSEN’S
98 J. OLBRYŚ

[13] measure of performance ( α̂ P ) are positive, but not significant at the 5% level in
the case of half of the funds. These results suggest that fund managers do not possess
selectivity skills, in this way being consistent with most studies on performance
evaluation which support the hypothesis of a weakly efficient market [1], [3], [12],
[18], [25], [26]. We can observe that the levels of systematic risk ( βˆP ) are signifi-
cantly positive and high. Unfortunately, almost all of the funds (except DWS Polska
FIO Akcji, Pioneer Akcji Polskich FIO and Skarbiec-Akcja FIO) present significantly
negative estimates of market-timing skills ( γˆP < 0 ). These empirical results show no
statistical evidence that Polish equity fund managers have outguessed the market. Ta-
ble 4 documents a negative correlation between regression intercepts and timing coef-
ficients, as in other studies [1], [12], [18]–[20], [25], [26].
HENRIKSSON and MERTON [11] (the H–M model) assume that in each period the
managers forecast whether stocks will outperform riskless bonds or vice-versa. They
can choose between two target levels of systematic risk [12, pp. 76–78]:
• η1 when they predict that riskless securities outperform the market, RM , t ≤ RF , t
• η2 when they predict that the market outperforms riskless securities, RM , t > RF , t
Since the managers’ forecasts are not observable, the risk of the portfolio at time t,
βt, should be a random variable for a market-timer, assuming a value of η1 or η2 de-
pending on whether the manager forecasts a down market or an up market. If the fore-
caster is rational, then η 2 > η1 . Under the assumption that beta is not observable, the
return on the portfolio P in the period t is given by:
RP , t = RF , t + (b + θ t ) ⋅ rM , t + λ + ε P , t (4)
where:
RP , t , RF ,t , rM , t , ε P , t are as in equation (3),
b is the unconditional (on the forecast) expected value of βt,
θ t = βt − b is the unanticipated (dependent on the forecast) expected value of βt,
λ is the expected excess return from selectivity.
In this form,
η1 = b + θ t (5)
is the target level of systematic risk when the forecaster predicts RM , t ≤ RF , t and

η2 = b + θt (6)
is the target level of systematic risk corresponding to a forecast of RM , t > RF , t [12, pp. 77].
Using the return process described in (4), least squares regression analysis can be
used to estimate the separate contributions from security analysis and market-timing,
as follows:
 Three-factor market-timing models with Fama and French’s spread variables 99

rP , t = α P + β P ⋅ rM , t + γ P ⋅ yM , t + ε P , t (7)
where:
rP ,t , rM , t , α P , β P , γ P , ε P , t are as in equation (3),
y M , t = max{0, RF ,t − RM , t } = max{0,−rM , t } .
Equation (7) is motivated by Merton’s analysis of the value of market-timing [17].
He shows that the returns obtained by a timing strategy would be similar to the returns
obtained from an option investment strategy of the put protective type. In this way,
α̂ P measures the contribution of security selection to portfolio performance, which
corresponds to testing the null hypothesis:
H0 :α P = 0 (8)
i.e., the manager does not have any microforecasting ability [26].
The estimate βˆP represents the proportion invested in the market portfolio when fol-
lowing the option investment strategy. The estimate γˆP represents the number of “free”
put options on the market due to the manager’s market-timing skills. In this context, the
evaluation of market-timing skills is carried out by testing the null hypothesis:
H0 : γ P = 0 (9)
i.e., the manager does not possess any timing ability or does not act on his forecast
( η1 = η 2 ) [12, pp. 77–78]. A negative value for the regression estimate γˆP would im-
ply a negative value for market-timing.
Table 5 provides details on the estimated market-timing H–M models. The Newey–
West procedures have been used to correct for both autocorrelated and heteroskedastic
error terms. We can observe that the null hypothesis (8) is rejected for almost all of the
funds (except DWS Polska FIO Akcji), i.e. they present a significant positive estimate of
selectivity. According to Jensen’s interpretation of the value of α̂ P , this measure could
be positive for two reasons: (1) the extra returns earned on the portfolio are actually due
to the manager’s ability, or (2) the positive bias in the estimate of α̂ P resulting from the
negative bias in the estimate of βˆ [13, pp. 396]. Results of the H–M parametric tests
P
show that equity fund managers rather do not possess market-timing skills ( γˆP ). The
null hypothesis (9) is rejected for almost all of the funds (except DWS Polska FIO Akcji
and Skarbiec – Akcja FIO), i.e. they present a significant estimate of timing ability, but
unfortunately none of them exhibit positive estimates of γˆP . We thus find evidence of
negative market-timing. Significant negative estimates of market-timing indicate that,
contrary to what would be expected of rational investors, managers increase the exposure
of their portfolios to the market in down markets and act inversely in up markets [26,
pp. 361]. Therefore, our initial conclusions remain unaltered.
100 J. OLBRYŚ

Table 5. Classic H–M model (7) (Jan. 2, 2003–Dec. 31, 2009)

Equity funds α̂ P βˆP γˆP R2 DW AIC

1 Arka BZ WBK Akcji FIO 0.0013* 0.617* –0.20* 0.617 2.43 –12005.5
2 Aviva Investors FIO Polskich Akcji 0.0010* 0.669* –0.17* 0.689 2.57 –12401.9
3 BPH FIO Akcji 0.0005* 0.664* –0.09* 0.715 2.70 –12821.5
4 DWS Polska FIO
0.0011* 0.297* –0.21* 0.239 2.47 –11121.8
Top 25 Małych Spółek
5 DWS Polska FIO Akcji 0.0006 0.564* –0.12 0.372 2.75 –10732.2
6 DWS Polska FIO Akcji Plus 0.0008* 0.464* –0.15* 0.346 2,75 –11055.9
7 ING FIO Akcji 0.0004* 0.695* –0.09* 0.694 2.71 –12497.0
8 Legg Mason Akcji FIO 0.0006* 0.639* –0.11* 0.698 2.61 –12775.3
9 Millennium FIO Akcji 0.0005* 0.611* –0.12* 0.670 2.64 –12638.0
10 Pioneer Akcji Polskich FIO 0.0005* 0.741* –0.14* 0.697 2.66 –12197.3
11 PKO/CREDIT SUISSE Akcji FIO 0.0009* 0.452* –0.20* 0.420 2.71 –11480.8
12 PZU FIO Akcji KRAKOWIAK 0.0006* 0.636* –0.13* 0.687 2.67 –12633.3
13 SEB 3 – Akcji FIO 0.0009* 0.428* –0.16* 0.307 2.68 –10928.9
14 Skarbiec – Akcja FIO 0.0007* 0.403* –0.08 0.273 2.64 –11130.8
15 UniKorona Akcja FIO 0.0009* 0.439* –0.13* 0.300 2.70 –10926.5
* Significant at the 5% level.
Source: Author’s calculations (using Gretl 1.8.5).

4. Modified three-factor market-timing models

In 1994 GRINBLATT and TITMAN [9] showed that tests of mutual fund performance
are quite sensitive to the chosen benchmark. For this reason, we run three-factor ana-
logs of equations (3) and (7), in which the two new additional spread variables are
Fama and French’s SMB and HML factors. Performance evaluation in terms of modi-
fied three-factor versions of the T–M or H–M models might allow a better assessment
of manager’s selectivity and timing skills.
We express the three-factor modified T–M model as:
rP , t = α P + β P ⋅ rM , t + δ1P ⋅ rSMB , t + δ 2 P ⋅ rHML, t + γ P ⋅ (rM , t ) 2 + ε P ,t (10)
where:
rP,t, rM,t, αP, βP, γP, εP,t are as in equation (3),
rSMB,t = RSMB,t – RF,t is the simple excess return on the mimicking portfolio SMB in
the period t,
rHML,t = RHML,t – RF,t is the simple excess return on the mimicking portfolio HML in
the period t,
δ1P is the sensitivity measure of the returns on portfolio P to changes in the SMB
factor returns,
 Three-factor market-timing models with Fama and French’s spread variables 101

δ 2 P is the sensitivity measure of the returns on portfolio P to changes in the HML

factor returns.

Table 6. Three-factor versions of the T–M models (10) (Jan. 2, 2003–Dec. 31, 2009)

Equity funds α̂ P βˆP δˆ1P δˆ2 P γˆP R2 DW AIC

1 Arka BZ WBK Akcji FIO 0.0006* 0.730* 0.070* 0.046* –2.03* 0.623 2.43 –12030.9
2 Aviva Investors FIO Polskich
0.0005* 0.751* –0.002 0.021 –1.92* 0.691 2.56 –12403.5
Akcji
3 BPH FIO Akcji 0.0001 0.716* 0.011 0.031* –0.96* 0.716 2.69 –12826.1
4 DWS Polska FIO
0.0002 0.445* 0.216* 0.070* –1.54* 0.297 2.61 –11257.6
Top 25 Małych Spółek
5 DWS Polska FIO Akcji 0.0002 0.630* 0.033 0.017 –1.30 0.373 2.74 –10731.4
6 DWS Polska FIO Akcji Plus 0.0002 0.555* 0.092* 0.025 –1.27 0.355 2.75 –11075.4
7 ING FIO Akcji 0.0001 0.746* 0.009 0.027* –0.92* 0.695 2.70 –12497.5
8 Legg Mason Akcji FIO 0.0003* 0.698* 0.017 0.035* –1.05* 0.700 2.60 –12781.8
9 Millennium FIO Akcji 0.0000 0.683* 0.033* 0.050* –1.07* 0.673 2.63 –12653.3
10 Pioneer Akcji Polskich FIO 0.0000 0.814* –0.001 0.035* –1.52* 0.698 2.65 –12202.1
11 PKO/CREDIT SUISSE Akcji
0.0003 0.559* 0.029 0.028 –2.20* 0.422 2.71 –11481.7
FIO
12 PZU FIO Akcji KRAKOWIAK 0.0001 0.702* –0.003 0.035* –1.39* 0.689 2.65 –12641.5
13 SEB 3 – Akcji FIO 0.0004 0.522* 0.085* 0.017 –1.64 0.315 2.68 –10945.9
14 Skarbiec– Akcja FIO 0.0003 0.457* 0.068* 0.013 –0.52 0.279 2.64 –11139.8
15 UniKorona Akcja FIO 0.0004* 0.519* 0.091* 0.019 –1.06 0.309 2.70 –10944.2

* Significant at the 5% level.

Source: Author’s calculations (using Gretl 1.8.5).

We have detected (based on the Augmented Dickey–Fuller test) that the analysed
series rM , t , rSMB , t and rHML, t (see equations (10) and (11)) are stationary. The Newey–
West procedures HAC (heteroskedasticity and autocorrelation consistent covariance
method) have been used to correct for both autocorrelated and heteroskedastic error
terms. We have used the Akaike Information Criterion (AIC) to compare the T–M
models (see Table 4) and their three-factor versions (see Table 6). The lowest value of
the AIC index indicates the preferred model, that is, the one with the fewest parame-
ters that still provides an adequate fit to the data. The evidence is that in the case of
14 funds (all except DWS Polska FIO Akcji), the addition of the regressors caused
a (small) decrease in the AIC index.
The results for the three-factor versions of the T–M models, given in Table 6, reveal
that the determination coefficients are larger than for the classical T–M models (Table 4).
Seven out of the fifteen coefficients of the SMB variable ( δˆ1P ) are positive and statisti-
cally significant. Moreover, eight out of the fifteen coefficients of the HML variable
102 J. OLBRYŚ

( δˆ2 P ) are positive and statistically significant. As for timing, all the funds exhibit nega-
tive timing coefficients ( γˆP ) and ten of them are statistically significant.
In a way analogous to (10), we express the three-factor modified H–M model as:
rP , t = α P + β P ⋅ rM , t + δ1P ⋅ rSMB , t + δ 2 P ⋅ rHML, t + γ P ⋅ yM ,t + ε P , t (11)

where:
rP ,t , rM , t , rSMB , t , rHML, t , yM , t , α P , β P , γ P , δ1P , δ 2 P , ε P, t are as in equations (7)
or (10), as appropriate.

Table 7. Three-factor versions of the H–M models (11) (Jan. 2, 2003–Dec. 31, 2009)

Equity funds α̂ P βˆP δˆ1P δˆ2 P γˆP R2 DW AIC

1 Arka BZ WBK Akcji FIO 0.0010* 0.650* 0.071* 0.046* -0.17* 0.623 2.43 -12028.3
2 Aviva Investors FIO Polskich
0.0010* 0.672* –0.002 0.021 –0.17* 0.690 2.57 –12402.3
Akcji
3 BPH FIO Akcji 0.0004* 0.673* 0.011 0.031* –0.09* 0.716 2.69 –12826.7
4 DWS Polska FIO
0.0005 0.387* 0.217* 0.071* –0.12* 0.297 2.61 –11256.2
Top 25 Małych Spółek
5 DWS Polska FIO Akcji 0.0005 0.579* 0.034 0.017 –0.11 0.373 2.74 –10730.8
6 DWS Polska FIO Akcji Plus 0.0005 0.501* 0.092* 0.025 –0.11 0.355 2.76 –11075.3
7 ING FIO Akcji 0.0004 0.702* 0.008 0.026* –0.09* 0.695 2.71 –12498.7
8 Legg Mason Akcji FIO 0.0006* 0.651* 0.016 0.035* –0.10* 0.700 2.61 –12782.8
9 Millennium FIO Akcji 0.0004 0.631* 0.032* 0.049* –0.11* 0.674 2.64 –12655.8
10 Pioneer Akcji Polskich FIO 0.0005 0.746* –0.001 0.034* –0.14* 0.698 2.65 –12203.2
11 PKO/CREDIT SUISSE
0.0008* 0.467* 0.029 0.028 –0.19* 0.422 2.71 –11481.4
Akcji FIO
12 PZU FIO Akcji KRAKOWIAK 0.0005* 0.640* –0.003 0.035* –0.13* 0.689 2.66 –12642.9
13 SEB 3 – Akcji FIO 0.0007 0.462* 0.085* 0.017 –0.13 0.314 2.68 –10944.1
14 Skarbiec – Akcja FIO 0.0005 0.429* 0.067* 0.012 –0.06 0.279 2.64 –11140.3
15 UniKorona Akcja FIO 0.0007 0.475* 0.091* 0.019 –0.09 0.309 2.70 –10944.1

* Significant at the 5% level.

Source: Author’s calculations (using Gretl 1.8.5).

We have used the Akaike Information Criterion (AIC) to compare the H–M models
(see Table 5) and their three-factor versions (see Table 7). In a manner comparable to the
T–M models, the evidence is that in the case of 14 funds (all except DWS Polska FIO
Akcji), the addition of regressors caused a decrease in the AIC index, i.e. indicate that
the three-factor versions of the H–M models should be preferred.
The results for the three-factor versions of the H–M models, given in Table 7, reveal
that the determination coefficients are larger than for the classical H–M models (Table 5).
By analogy to Table 6, seven out of the fifteen coefficients of the SMB variable ( δˆ1P ) are
 Three-factor market-timing models with Fama and French’s spread variables 103

positive and statistically significant and eight out of the fifteen coefficients of the HML
variable ( δˆ2 P ) are also positive and statistically significant. Furthermore, all the funds
exhibit negative timing coefficients ( γˆP ) and ten of them are statistically significant.

5. Comparison of empirical results

Table 8 presents a brief summary of the results on selectivity and market-timing

for 15 equity open-end mutual funds using daily data from Jan. 2, 2003 to Dec. 31,
2009 (based on Tables 4–7).

Table 8. Summary of selectivity and market-timing (Jan. 2, 2003–Dec. 31, 2009)

T–M H–M 3-factor T–M 3-factor H–M

Equity funds model (3) model (7) model (10) model (11)
α̂ P γˆP α̂ P γˆP α̂ P γˆP δˆ1P δˆ2 P α̂ P γˆP δˆ1P δˆ2 P
1 Arka BZ WBK Akcji FIO ++ –– ++ –– ++ –– ++ ++ ++ –– ++ ++
Aviva Investors FIO
2 ++ –– ++ –– ++ –– – + ++ –– – +
Polskich Akcji
3 BPH FIO Akcji + –– ++ –– + –– + ++ ++ –– + ++
DWS Polska FIO
4 ++ –– ++ –– + –– ++ ++ + –– ++ ++
Top 25 Małych Spółek
5 DWS Polska FIO Akcji + – + – + – + + + – + +
6 DWS Polska FIO Akcji Plus ++ –– ++ –– + – ++ + + – ++ +
7 ING FIO Akcji + –– ++ –– + –– + ++ + –– + ++
8 Legg Mason Akcji FIO ++ –– ++ –– ++ –– + ++ ++ –– + ++
9 Millennium FIO Akcji + –– ++ –– 0 –– ++ ++ + –– ++ ++
10 Pioneer Akcji Polskich FIO + – ++ –– 0 –– – ++ + –– – ++
PKO/CREDIT SUISSE
11 + –– ++ –– + –– + + ++ –– + +
Akcji FIO
12 PZU FIO Akcji KRAKOWIAK + –– ++ –– + –– – ++ ++ –– – ++
13 SEB 3 – Akcji FIO ++ –– ++ –– + – ++ + + – ++ +
14 Skarbiec – Akcja FIO ++ – ++ – + – ++ + + – ++ +
15 UniKorona Akcja FIO ++ –– ++ –– ++ – ++ + + – ++ +

Notes: – negative, – – significantly negative at the 5% level, + positive,

++ significantly positive at the 5% level.
Source: Author’s calculations.

The results presented in Tables 4–8 show that equity fund managers rather do not
posses security selection. The null hypothesis (8) is only rejected for some of the
104 J. OLBRYŚ

funds, i.e. only some of the funds present a significant positive estimate of selectivity
( α̂ P ). As was mentioned above (in Chapter 3), the α̂ P measure could be positive for
two reasons: (1) the extra returns earned on the portfolio are actually due to the man-
ager’s ability, or (2) the positive bias in the estimate of α̂ P . With respect to systematic
risk, all of the funds present significant estimate of βˆ at the 5% level.
P
Another important finding of the analysis in Tables 4–7 is the negative correlation
between selectivity ( α̂ P ) and timing ( γˆP ). In fact, almost all of the funds exhibit op-
posite signs for these two measures. A negative correlation between the measures of
selectivity and timing has also been found in other studies [12], [18], [26]. This type of
evidence is consistent with the majority of studies on timing and selectivity and sup-
ports the hypothesis of a weakly efficient market.

6. Conclusions

In this paper we have presented modified market-timing models with Fama and
French’s spread variables SMB and HML and we have examined the usefulness of
these models in the evaluation of investment managers’ performance. Although some
groups argue, that the Fama and French model is controversial [25, pp. 259], the re-
sults in Tables 6, 7 and 8 suggest, that the SMB and HML variables have significant
explanatory power for our sample of funds.
Following [1], [20], we have studied daily ordinary excess returns on the funds
from Jan. 2003 to Dec. 2009. The daily data present one notable complication com-
pared to monthly data. DIMSON (1979) [4] describes the problem of a nonsynchronous
trading which hampers regression analysis for individual securities. Since they are
typically well-diversified stock portfolios, mutual funds are not as susceptible to
problems associated with a nonsynchronous trading as individual securities are. How-
ever, to study and remedy these problems, we could use Dimson’s correction and in-
clude lagged values of the factors as additional independent variables in the regres-
sions to accommodate infrequent trading. Therefore, further investigation of modified
market-timing models may be needed.

References

[1] BOLLEN N.P.B., BUSSE J.A., On the timing ability of mutual fund managers, The Journal of Finance,
2001, Vol. 56, No. 3, 1075–1094.
[2] CARHART M.M., On persistence in mutual fund performance, The Journal of Finance, 1997, 52, No.
1, 57–82.
 Three-factor market-timing models with Fama and French’s spread variables 105

[3] CZEKAJ J., WOŚ M., ŻARNOWSKI J., Efektywność giełdowego rynku akcji w Polsce, PWN, Warszawa,
2001.
[4] DIMSON E., Risk measurement when shares are subject to infrequent trading, Journal of Financial
Economics, 1979, 7, 197–226.
[5] FAMA E.F., Components of investment performance, The Journal of Finance, 1972, 27, No. 2,
551–567.
[6] FAMA E.F., FRENCH K.R., The cross-section of expected stock returns, The Journal of Finance, 1992,
47, 427–465.
[7] FAMA E.F., FRENCH K.R., Common risk factors in the returns on stocks and bonds, Journal of Fi-
nancial Economics, 1993, 33, 3–56.
[8] FAMA E.F., FRENCH K.R., Size and book-to-market factors in earnings and returns, The Journal of
Finance, 1995, 50, No. 1, 131–155.
[9] GRINBLATT M., TITMAN S., A study of monthly mutual fund returns and performance evaluation
techniques, Journal of Financial and Quantitative Analysis, 1994, 29, 419–444.
[10] HAUGEN R.A., Nowa nauka o finansach. Przeciw efektywności rynku, WIG-Press, Warszawa,
1999.
[11] HENRIKSSON R., MERTON R., On market timing and investment performance. II. Statistical proce-
dures for evaluating forecasting skills, Journal of Business, 1981, 54, No. 4, 513–533.
[12] HENRIKSSON R., Market timing and mutual fund performance: an empirical investigation, Journal of
Business, 1984, 57, 73–96.
[13] JENSEN M., The performance of mutual funds in the period 1945–1964, Journal of Finance, 1968,
23, 389–416.
[14] KUFEL T., Ekonometria. Rozwiązywanie problemów z wykorzystaniem programu Gretl, PWN, War-
szawa, 2004.
[15] LEHMANN B., MODEST D.M., Mutual fund performance evaluation: A comparison of benchmarks
and benchmark comparisons, The Journal of Finance, 1987, 42 (June), 233–265.
[16] MADDALA G.S., Ekonometria, PWN, Warszawa, 2008.
[17] MERTON R., On market timing and investment performance. I. An equilibrium theory of value for
market forecasts, Journal of Business, 1981, 54, No. 3, 363–406.
[18] OLBRYŚ J., Parametric tests for timing and selectivity in Polish mutual fund performance.
Optimum, Studia Ekonomiczne, Wydawnictwo Uniwersytetu w Białymstoku, 2008, 3(39), 107–
118.
[19] OLBRYŚ J., Parametryczne testy umiejętności wyczucia rynku – porównanie wybranych metod na
przykładzie OFI akcji, [in:] Z. Binderman (ed.) Metody ilościowe w badaniach ekonomicznych, IX.
Wydawnictwo SGGW w Warszawie, Warszawa, 2008, 81–88.
[20] OLBRYŚ J., Ocena umiejętności stosowania strategii market-timing przez zarządzających portfelami
funduszy inwestycyjnych a częstotliwość danych, Studia i Prace Wydziału Nauk Ekonomicznych
i Zarządzania Nr 10, Uniwersytet Szczeciński, Szczecin, 2008, 96–105.
[21] OLBRYŚ J., KARPIO A., Market-timing and selectivity abilities of Polish open-end mutual funds
managers, [in:] P.Chrzan, T.Czernik (ed.) Metody matematyczne, ekonometryczne i komputerowe
w finansach i ubezpieczeniach 2007, Wydawnictwo AE im. K. Adamieckiego w Katowicach, Kato-
wice, 2009, 437–443.
[22] OSIŃSKA M., Ekonometria finansowa, PWE, Warszawa, 2006.
[23] OSIŃSKA M. (ed.), Wybrane zagadnienia z ekonometrii. Wydawnictwo WSIiE TWP w Olsztynie,
Olsztyn, 2005.
[24] PETKOVA R., Do the Fama–French factors proxy for innovations in predictive variables? Journal of
Finance, 2006, 61, No. 2, 581–612.
[25] PRATHER L.J., MIDDLETON K.L., Timing and selectivity of mutual fund managers: An empirical test
of the behavioral decision-making theory, Journal of Empirical Finance, 2006, 13, 249–273.
106 J. OLBRYŚ

[26] ROMACHO J.C., CORTEZ M.C., Timing and selectivity in Portuguese mutual fund performance,
Research in International Business and Finance, 2006, 20, 348–368.
[27] TREYNOR J., MAZUY K., Can mutual funds outguess the market? Harvard Business Review, 1966,
44, 131–136.
[28] URBAŃSKI S., Wpływ innowacji wybranych czynników na równowagę cenową walorów notowanych
na Giełdzie Papierów Wartościowych w Warszawie, Bank i Kredyt, NBP, 2008, 7, 37–49.