Are the fama and french factors global or country specific?
1. Are the Fama and French Factors
Global or Country Specific ?
John M. Griffin / Arizona State University
Speakers: 張博能/ 歐哲源/ 姚博⽂文
The Review of financial Studies Summer 2002 Vol. 15, No. 3, pp. 783-803
ⓒ 2002 The Society for Financial Studies
1st group, 2nd presentation, 2015
Financial Management @ NCCU
4. Key Result
This paper proposed
1. Portfolio and individual stock
2. In-sample & out-of-sample pricing
3. Two practical applications
# Capital Caculation
# Performance Evaluation
# Additional Foreign Factor lead to pricing error.
# Country-specific model explains the return better.
5. Agenda
Background introduction |
Related literature | Basic research idea
Today’s presentation
Data & Descriptive statistics |
Sorted portfolios | Individual securities
Out-of-sample evidence |
Additional Evidence | Conclusion
1
2
3
7. Three-factor model
Related Works
Expect return = Excess market return
+ Size factor (SMB)
+ Book-to-market factor (HMB)
Fama & French (1992,1993,1995,1996,1998)
1
Fama & French (1998)
Global Model = World market return
+ WorldBook-to-market factor (WHMB)
2
8. Global or Domestic
Important ?
In US stock market,
consider expected return estimates
Global Factors’
model
Country-Specific Factors’
model
8.41%
Difference!
11. BE / ME Effect
Related Works
1
Book-to-the-market equity is spurious.
# Black(1993), Mackinlay(1995)
- Data Mining Bias / Data Snooping Fallacy
# Kothari(1995)
- Selection Biases
12. BE / ME Effect
Related Works
Book-to-the-market equity captures the
“Risk Factor. ”
# Fama & French (1992,1993,1996)
2
# Chan(1991), Davis(1994), Fama&French(1998)
In addition to US stock’s Market, B/M is
significant in other market in the long run.
14. Other critics
Two main groups
1
Size and Book-to-market equity both are due to
inverstors’ overreaction. They [Lakonishok,
Shleifer, and Vishny(1994), Hauge (1995)] argue
that investor systematically overreact to
corporates’ news.
# Underpricing of the Value
# Overpricing of the growth
15. Other critics
Two main groups
2
Daniel and Titman (1997) challenge the risk-
based interpretation.
Firm characteristics(fundamentals) explain
returns better than Fama & French’s risk factor’s
idea.
# Davis, Fama, French(2000), challenge
Daniel’s and Titman’s idea again.
17. One way to further examine the
empirical validity is to use the
international data.
18. International data
Related Work
J
apanese stock return has powerful explanation
in characteristic instead of factor loading.
[Daniel, Titman, and Wei(2011)]
Liew and Vassalou (2000) shows that Fama & French’s
model still works in several international markets.
Fama & French (1998) implies that using world two-
factor model lead to explain international expected
return better.
19. The key point
This paper try to present
World-factor model and country-specific model
would be compared in this research.
B/M-sorted portfolio ,size-sorted portfolio and
individual security data should examined.
1
2
20. Three empirical models
Our objective work
1
World-factor regression-
rit = αi + bi(WMRFt) + si(WSMBt) + hi(WHMLt) + εi
WMRF: World market return in excess risk-free rate
WSMB: The difference between the returns on
small and large capitalization portfolio.
(SMB, small minus big)
WHML: The difference between the return on high
and low B/M portfolios. (HML, high minus low)
22. We should know,
Something details
egressions’ return is dollars-denominated.
RThe WDt-1 shows that the market capitalisation
of these countries in the sample attributable to the
domestic market in the previous month.
Also, WFt-1 indicates the similar reason that the
previous period foreign capitalisation would affect
the foreign capitalisation in this month.
24. The “Big Three” in our work
To sum up,
World-factors regression model
International regression model
Domestic regression model
# Return = World factors
# Return = Domestic Factors + Foreign Factors
# Return = Domestic Factors
55. Out-of-sample evaluation
( )
International(6,factors) world domestic
US 9.9% 9.89% 9.87%
Japan 8.89% 8.9% 8.86%
UK 8.16% 8.22% 8.13%
Canada 7.3% 7.33% 7.25%
◇ 結果:三種模型在預測報酬的誤差上都相當⼤大,
但domestic models 誤差較world & international models ⼩小
56. Additional Evidence
Formation of the world factors
Other International model
Usefulness of factors
Cross-sectional tests
} Domestic
模型較好
} Foreign Factor
解釋⼒力低
57. Formation of the world factors
➜ 如此⼀一來,的確可以提⾼高world
model的解釋能⼒力
但是不管組成國家為哪些,domes&c模型表現⽐比較好。
➜
前者分析是透過兩種⽅方式value-‐weighted和
equal-‐weighted各形成的三因⼦子模型
➜ 另⼀一種建構size和BE/ME的⽅方法是忽略各國間的差異
➜
foreign
factor也是以類似的⽅方法建構