-
*E-mail: [email protected]
Global Asset Pricing and Financial Intermediary Leverage
Leandro de Miranda Gomes*
PUC-RIO
March 2015
Abstract
We propose a 2-factor global asset pricing model based on prices
of risk of financial
intermediary firms leverage and US market returns. We show that
our model present
considerably better statistical and economic properties, as
small intercept and
economically meaningful prices of risk, than benchmarks like US
Fama-French 3-
factor, Global US Fama-French and single factor leverage model.
There are also
considerably improvements when simultaneously pricing US and
International
portfolios in comparison to US-only portfolios, indicating a
global presence of US
financial intermediary system on asset pricing through different
markets. We use
conventional 25 size-book to market sorted US and International
portfolios, but our
results are robust to the inclusion of momentum portfolios and
maturity sorted bonds
portfolios.
-
1. Introduction
At the Federal Reserve Bank of Bostons 56th
Economic Conference on October
2011, the President & Chief Executive Officer of the
institution, Eric S. Rosegreen,
addressed the long term effects of the Great Recession,
highlighting the role of large
global financial intermediaries as they suffered severe distress
on the events of the crisis
and had a decisive role transmitting the shocks to the wider
global financial system
through their interconnectedness with other financial
institutions.
Given its relevant size on the financial system, the
intermediary sector is not
only an important global financial link, but also plays a
relevant role on the US
Financial Markets. Within the intermediary sector, security
dealers and brokers and
institutions from the shadow banking system appear to be the
most prominent ones,
presenting strong explanatory power for many financial asset
returns, as suggested by
Adrian, Moench and Shin (2010), including a relevant role on the
explanation of the
cross sectional US portfolios returns through a single US
Security Dealers and Brokers
Leverage Factor model proposed by Adrian, Etula and Muir (2010,
2015).
Based on the global nature of US Security Dealers and Brokers,
we argue that
the US Security Dealers and Brokers Leverage Factor is also
useful in explaining the
cross section of Global Portfolios, following an intermediary
financial Stochastic
Discount Factor framework as in Adrian, Etula and Muir
(2010,2015). Using different
combinations of US and International Stocks Portfolios sorted by
Size, Book to Market
and Momentum, as well as different maturities for Bonds
Portfolios, we show that a US
2-factor model with the US Security Dealers and Brokers Leverage
and the US Excess
Market Return as factors, is the most stable and robust among
all models tested, like
Adrian, Etula and Muir(2010,2015) single leverage model,
International 2-Factors
Model1 and both International and US Fama French 3 Factors
Model.
Although no model can consistently price the cross-section of US
or
International excess returns, we find that adding International
portfolios to the analysis
1 International 2-Factors are US Security Dealers and Brokers
Leverage Factor and International Excess
Market Return. In the absence of Global Leverage data, USs may
be seen as a proxy.
-
does improve our cross-sectional results as pricing errors
become gerenally smaller and
prices of risk become consistent with theory. The US 2-Factor
Models present the most
significant improvements, even when challenging Momentum
Portfolios are added.
While the addition of Global portfolios lead to the most
noteworthy improvements, the
model is also robust to the inclusion of European-only or even
UK-only portfolios.
While our goal is to show that the leverage factor is priced in
the cross-section of
Global portfolios, a minor contribution to the literature is to
address the robustness of
Adrian, Etula and Muir(2010,2015) single leverage factor model
to changes in
definition of leverage and to sample period. On the June 5, 2014
publication of the
Flow of Funds, there was an important change on one of the
variables that were used to
construct the leverage factor in Adrian, Etula and
Muir(2010,2015). Before, the
Security Repurchase Agreements appeared only as a net variable
on the liabilities
section, while on the updated version it appears both on the
assets and liabilities. Table
1 shows the implications of this change on the construction of
the Leverage in both
scenarios for the fourth quarter of 2013.
The change, in absolute terms, is non-negligible. Unfortunately,
because of data
availability constraints, theres no way to fully replicate the
old methodology to look for
possible implications on the dynamics of the transformation.
However, in order to
investigate imaginable problems arising, the old data is
artificially reconstructed
subtracting the security REPO from both the total financial
assets and total liabilities,
recreating the net variable on the liabilities side.
Intuitively, theres no economic
explanation for the use of the old variable, nonetheless the
following analyses will
report both versions. Results report a considerable sensibility
towards this methodology
change. In addition, results show that when we consider the
1981-2014 or 1991-2014
windows, where the latter is the same as the used on global
portfolios, the single
leverage factor models performance is considerably worse,
indicating a lack of
robustness to shorter time periods. Because data is quarterly,
this can be attributed to the
lack of betas variability, which may prevent a precise estimate
of leverage price of risk.
-
2. Related Literature
Leverage started to gain attention primarily after the Great
Recession, even
though leverage cycle theory existed prior to the crisis, as
seen in Geanokoplos
(1997,2003) and Fostel and Geanokoplos (2008). Contrarily to
standard economic
theory, the Leverage Cycle Theory suggests that asset prices and
consequently the real
economic activity, are driven not by interest rate, but by
endogenous leverage and
volatility, through the role of collateral. Additionally, as
presented in Fostel and
Geanokoplos (2013), the cycle component of this theory focus on
how low volatility
and safe economic perspective can lead the real economic to a
boom and a highly
leveraged economy, as well as a confluence of factors, primarily
a bad news leading to
increased volatility and uncertainty can lead to credit
tightening and a severe
deleverage, and thus, a market crash.
Empirically, Adrian, Moench and Shin (2010) documented the high
pro-
cyclicality of the leverage cycle, where low asset prices are
associated to an also low
Leverage, and proposed that this cyclical characteristic may be
due to regulatory or
other constraints, such as Value at Risk limits. Another
approach to rationalize the
stylized fact emphasizes how personal experience of agents can
shape risk preferences,
as theoretically proposed by Krishnamuty (2009) and empirically
suggested by
Malmendier and Nagel (2011) that look to the role of the Great
Depression on
individuals willingness to invest in equities, and Koudijs and
Both (2014) that
investigate the collateralized Amsterdam loan market in the
18th
century.
Given the relevance of financial intermediary firms, Adrian,
Etula and Muir
(2010, 2015) suggest a shift on the traditional asset pricing
theory emphasis on
household Stochastic Discount Factor (SDE) to an intermediary
SDE. They argue that
the assumptions of a household SDE are easily violated, like
full market participation of
all households, absence of transaction costs, and the need of
continuous optimized
expectations, while the financial intermediary firms fit the
hypothesis more naturally.
Furthermore, they construct a single factor model based on a
non-seasonal log
innovations of US Security Dealers and Brokers Leverage that
outperforms US Fama
French 3-factor and CAPM models on the 1968-2009 time window.
Adrian, Moench
and Shin (2014) also extends these results to a dynamic
framework using the Dynamic
-
Asset Pricing Model (Adrian, Crump and Moench, 2014) that allows
for time-varying
price of risk and time-varying betas.
Nonetheless, US security dealers and brokers are genuine global
companies,
incorporating shocks from all parts of the world through its
subsidiaries, or foreign
holders. Since there is no availability of data for security
dealers and brokers from
others countries, it is impossible to calculate a true global
leverage factor.
The US Leverage Factor has been used inthe literature as a proxy
for the
unavailable global variable. Stressing the global nature of the
financial system and the
leverage cycles, Bruno and Shin (2014a) propose a global banking
model, where
regional banks borrow in US dollars from global banks and then
lend to local corporate
owners, emphasizing that the global intermediary financial
system is responsible for the
aggregate credit risk, and consequently, global liquidity.
Testing theory predictions,
they also construct a panel including 46 countries using global
factors, as the US
Security Dealers and Brokers leverage as a proxy, to support
their predictions.
Credit conditions are also determined by global factors in Bruno
and Shin
(2014b), that also synchronizes the corporate risk-taking of
different regions and
sectors. Empirically, they show that US Security Dealers and
Brokers as a proxy for
Global Liquidity is indeed a relevant global risk-taking factor.
Bruno and Shin (2014c)
also present the monetary policy risk-taking channel of capital
flows, where US
monetary policy is relevant to both international banks leverage
and cross-borders
capital flows.
Finally, the Global Asset Pricing literature so far has
struggled to propose a
global-factors only model, and Lewis (2011), explore possible
explanations that make
this a hard task to be accomplished, highlighting the domestic
nature of monetary and
fiscal policies and the home equity bias, as well as the
exchange risk included on global
portfolios. However, Lewis points out that a world CAPM still is
the most common
benchmark when looking to Global Portfolios.
-
3.Data
The US Security Dealers and Brokers Leverage is defined as:
Both Total Financial Assets and Total Liabilities are from the
L.128 Table from
Flow of Funds Report, now Z1 Release, as December 11, 2014. Data
is quarterly. To
approximate the old methodology, the approach is to recreate the
NET Repo Agreement
variable on the liabilities side. Precisely, Old Leverage is
defined as:
To construct the Leverage Factor, we first take the Leverage Log
Innovation,
i.e., log(Leverage), and make it seasonally adjusted using
expanding windows starting
in 19652. Following Adrian, Etula and Shin(2010,2015), as the
data before mid-60s
doesnt seem reliable because of negative or too low leverage,
the time series of the
factor will start at 1968.
The Portfolios used for the US analysis are the 25 Size-Book to
Market and 10
Momentum from Frenchs library. US Treasury Bonds Portfolios are
sorted by
maturities, 0-6 months, 12-18months, 24-30 months, 36-42 months,
48-54 months and
5-10 years. The data is from CRSP.
International Portfolios, both 25-Size Book to Market and
25-Size Momentum,
are constructed using data from 22 Developed Countries excluding
the US, and are
available only starting at 1991. European 25-Size Book to Market
and 25-Size
Momentum are from 16 countries and available at the same source.
Finally, UK 25
2 Following the same methodology, the season adjustment is
proceeded using expanding windows, i.e.,
using all the data up to date, starting three years before the
initial starting date (1968). Therefore, through
the regressions below, the leverage factor is adjusted using
data starting in 1965.
-
Size-Book to Market are from Xfi Centre For Finance and
Investment by University of
Exeter and are available since 1981.
When we refer to Global portfolios, both International and
US-only portfolios
are being simultaneously used, unless further noticed. In only
one situation another set
of Global portfolios is used, which are constructed very
similarly of International
portfolios, but also includes US portfolios. This data is also
available at Frenchs
library.
For the excess return calculations, all Portfolios above use the
3-Month T-Bill as
the risk-free rate, with the exception to the UKs, that use the
Risk-Free rate from
Exeter database. As the leverage is only available at a
quarterly frequency, all excess
returns are time-aggregated into quarterly excess returns, and
then multiplied by four to
transform them in annual excess returns.
Three different sets of factors are used. US, Global3 and
European Fama-French
Factors are from Frenchs library, while UKs Factors are from
Exeter Univeristy
library.
For USs Portfolios, four different dates combinations will be
used to allow us
compare the modes performance through time: 1968-2009, the time
window used by
Adrian, Etula, Shin (2010,2015), 1968-2014, 1981-2014 and
1991-2014. For UKs
Portfolios, only 1981-2014 and 1991-2014, while for the others,
1991-2014.
US 25, International 25, Europe 25 and UK 25 portfolios are
sorted by size-book
to market only, while US 31 and 35 include, respectively, 6
Bonds and 10 Momentum
portfolios. US 41 includes both of them and International,
Europe and UK 50 Portfolios
both 25 size-book to market and 25 size-momentum portfolios.
4. Methodology
For all different portfolios, i=1,,N, each factors beta is
estimated through
conventional time-series regressions, i.e.,
Ri,te = ai + i,f ft + ei,t , t=1,...,T.
3 The Global Factors, differently from International Portfolios,
take the United States into account.
-
Then, as usual, the factors price of risk, f, is calculated
through the cross
section of Portfolios Expected Returns, where, vi + c is the
pricing error.
E[Ri,te] = c + i,f f + vi , i=1,...,N
Good model properties are a small intercept, as well as stable
and economically
significant prices of risk. Although the conventional
t-statistic is presented (in italic), it
is prudent to take into consideration that the betas are
estimated and thus already have
own estimation errors, hence we also report the t-statistic
correction as proposed by
Shanken (1992). We also hope for a small Mean Absolute Pricing
Error (MAPE), i.e.,
, and non-rejection of the null hypothesis that all pricing
errors are
jointly equals to zero, using the Gibbons-Ross-Shanken (1989),
GRS statistic.
5. Results
Since Adrian, Etula and Muir (2010,2015) proposed the Leverage
Factor, the
definition of the variables used in its construction has
changed. As one of ours main
goals is to compare models performance in the pricing of
International portfolios, first
we look at the US-only case, taking a closer attention to the
robustness over the
methodology change and to changes in the sample period as
International data is
considerably shorter. Tables 2-4 provide the comparison between
old and new Leverage
Factor in four different time windows, 1968-2009, 1968-2014,
1981-2014 and 1991-
2014. We also compared their performance to an alternative US
2-factor that also
includes the market return as a factor, and to the US Fama
French 3-factor models, . We
test these model using US 25, US 31 and US 41 Portfolios.
Next, we take a Global perspective, pricing simultaneously US
and International
Portfolios4. In addition to using International portfolios, we
also test the effect of adding
European-only and UK-only portfolios for robustness.. Tables 5-9
present the
comparison between a range of models (Leverage Single Factor, US
2-Factors, US
Fama French 3-factor, Global 2-factor and Global 3-factor) using
different portfolio
combinations: US 25 + International 25, International 25, US 31
+ Interntional 25, US
4 Data for 25 Size-Book to Market and 25 Size-Momentum sorted
portfolios are also available at Frenchs
library. However, due to a small time frame, we prefer to
include International portfolios to estimate
leverage price of risk more precisely.
-
35 + International 50 and US 41 + International 50.CAPM performs
considerable
worse, so results are not reported.
We also show, a horse race between the 2-US Factor and 4-US
Factor, which is
a combination of the leverage factor and the three factors on
Fama French model, on
table 10, and the use of a different global portfolios set on
table 11 to assess robustness.
5.1 US-only Portfolios
First, we address the possible problems concerning the
methodology change of
the leverage calculation, as well as the model performance over
shorter time periods.
The results shown in tables 2-4 imply that the old leverage
calculation methodology
performs considerably better than the new-definition leverage.
Nonetheless, there are
some shortcomings on the older methodology, as Both GRS
statistic and pricing errors
increase considerably on shorter time-windows. For example, when
all 41 US
portfolios are analyzed, the intercept goes from 0.36 on the
1968-2009 time frame to
7.97 on 1991-2014. This pattern can be seen in all other models,
but the increase is not
nearly as steep as for the old Leverage 1-factor model. Also,
both new and old leverage
price of risk remains stable through different portfolios sets,
time-frames and
methodologies approach. When we add the 6 bonds or 10 momentum
portfolios,
however, US 2-factor outperform both old and new leverage
1-factor, presenting
smaller and non-significant intercepts, at the 1.5-3.25 range,
while maintaining the price
of risk stability for both leverage and market.
Finally, while GRS tests reject all the tested models, the
relevant leverage price
of risk in all the models, as well as higher than 60% Adjusted
R2 estimates, provide
some evidence on the role of the US Security Dealers and Brokers
on the cross section
of asset markets.
5. 2 International and Global Portfolios Pricing
In this section, we test our main hypothesis that the US
leverage factor can price
the cross-section of global portfolios, i.e., international and
US portfolios. We test the
conjecture that the addition of international portfolios can
improve the overall
-
performance of the model due to a more precise estimate of the
price of leverage risk,
relying on the hypothesis that the US leverage factor is a good
proxy for an
unobservable global leverage factor. Overall, tables 5-9 show
that the inclusion of
international portfolios turn all models better, but mainly US
2-factor and US FF.
The inclusion of only 25 International portfolios on US original
25 portfolios is
already enough to improve the results, as seen in table 5. Our
US 2-factor model, for
example, presents a smaller intercept of 2.61, against a 9.11
with US-only 25 portfolios,
while maintaining relevant and positive leverage and market
prices of risk. However
US Fama-French model intercept also declines from 10.18, when
US-only portfolios are
used, to 0.24, when international and US portfolios are priced
simultaneously. Although
all models present better economic and statistical properties,
the steepest improvements
happen with the US 2-factors and US FFs models. We also show on
table 6 that models
also improve from International-only portfolios when US
portfolios are included.
However, the US Fama-French model is not as robust as the US
2-factor model
to the inclusion of different types of portfolios. Although
adding bonds portfolios seems
make all models better, US 2-factor presents considerable
improvements when
compared to US Fama-French model. US 2-factor is the only model
to present non-
significant intercepts, and present stable coefficients when
comparing the US-only 25 to
US 25 + International 25 portfolios, which is a highly desirable
economic feature. We
also show on table 8, when momentum portfolios are included, and
on table 9, when
both momentum and bond portfolios are added, that the US-factor
clearly presents the
best results. While US Fama-French model presents a high
intercept of 7.88 and a
negative price of risk on the first case, US 2-factor model
presents a sign of market
price of risk that is consistent with economic theory, whereas
maintaining a stable
leverage price of risk and low intercept.
GRS tests interpretation does not permit us to statistically
reject the null
hypothesis of jointly zero jointly prices errors, however, US
2-factor model intercepts
are all between 2.49 and 4.67 and are all insignificant., which
is an unique feature
among all models. Also, from the economic theory stand point,
the fact that leverage
price of risk always lies between 67-73 and market price of risk
between 3.17-5.36,
suggest an interesting stability over different cross-sectional
results. Both statistical and
economic properties across portfolios can be visualized in the
figures 1-5.
-
When we add Europe and UK-only portfolios, as seen in tables
5-9, results
suggests the US 2-factor as the most consistent model, albeit
not performing well in
some cases, as we can see on the significant intercept on UK 25
+ US 41 and Europe 50
+ US 35 besides a negative market price of risk on Europe 50 +
US 35. Also, Tables 5
and 6 also provide evidence on the improvements when US 25
Portfolios are added to
Europe or UK 25 portfolios. US 2-Factors present significant
16.22 and 8.96 intercepts
on Europe-only and UK-only 25, correspondingly, while when
including US 25, turn to
respectively a not significant 5.35 and 6.88.
On a horse race between US 2-factor, US 4-factor and Global
4-factor models,
table 10 shows that the results to US models are similar, and
considerably superior to
the Global model. Indeed, US 4-factor model also present a
similar low intercept, but
also shows a higher adjusted R2. However, this model lack
robustness when we include
of momentum portfolios. Thus, as US 2-factor is a more
parsimoniously model,
presenting generally better results, we do not find added to
value to the extra factors
considered on the 4-factor model.
Finally, we assess the performance of US-2 factor model when US
portfolios are
already used on the construction of Global portfolios. The main
shortcoming to this
approach is the potential lack of variability that can hurt the
estimated betas precision.
Nonetheless, table 11 suggest that although US 2-factor model
maintains the relative
best economic and statistical properties, the performance with
Global 25 portfolios and
Global 50 portfolios are much worse than the benchmark case.
6. Conclusion
We show that US 2-factor model can consistently price
cross-sectional returns of
global portfolios, presenting small pricing errors and prices of
risk compatible to
economic theory. The results are also considerable stable across
different sets of
portfolios, including momentum and bonds. Also, this is the only
model that present
non-significant intercepts on all different global portfolios
combinations.
Other robustness tests include the addition of other non-US
portfolios, as
European and UKs, a horse race with 4-factor model, and a
different construction of
-
global portfolios. In all these scenarios, US 2-factor perform
substantially better than
single factor leverage, or global and US Fama-French models.
One question that emerges when looking to these results why a
Global 2-Factor
model doesnt seem to be the most consistent, if indeed this
economic story of a
relevant intermediary financial system is true. One speculative
answer is that financial
intermediary firms base their decision on a US framework, since
most of them are from
there, or because its the most important benchmark/market
available. Another
speculative answer is that this only happened due to the absence
of a true Global
Leverage Factor and if we could use this variable along with the
Global Market Excess
Return, the results could be improved. One last explanation
draws the attention to the
Risk-Free rate, which even for Global and European Portfolios,
is a common 3 month
T-Bill. Perhaps a better proxy for a global risk-free rate could
lead to significantly
different results.
7. Appendix: tables and figures
Figure 1 Actual vs predicted results from US 2-factor for US
25
-
Figures 2-3 : Actual vs predicted results from US 2-factor for
US 25 + International 25
and US 31 + International 25
-
Figures 4- 5: Actual vs predicted results from US 2-factor for
US 35 + International 50
and US 41 + International 50
-
Table 1: US security dealers and brokers leverage using
September 18 and March 6
versions of Flow of Funds for the fourth quarter of 2013
4Q 2013 L.128 as of September 18 (new)
total financial assets 3,408,371 total liabilities 3,324,558
checkable deposits and currency 111,720 security repurchase
agreements 1,768,587
security repurchase agreements 1,320,966 credit market
instruments 112,398
credit market instruments 476,894 corporate and foreign bonds
112,398
commercial paper 28,654 depository institution loans n.e.c.
-
Treasury securities 136,043 trade payables 14,540
agency- and GSE-backed securities 114,164 security credit
897,052
municipal securities and loans; 18,637 Households and nonprofit
organizations; security
credit; asset 815,484
corporate and foreign bonds 128,949 U.S.-chartered depository
institutions; security credit;
asset 52,885
syndicated loans to nonfinancial corporate
business 50,447
Foreign banking offices in the U.S.; security credit;
asset 28,683
corporate equities 172,386 taxes payable 3,723
security credit 339,184 foreign direct investment in U.S.
119,540
U.S. direct investment abroad 225,786 total miscellaneous
liabilities 408,718
total miscellaneous assets 761,435 equity investment by parent
companies 1,671,859
unidentified miscellaneous liabilities -1,263,141
Leverage = 40.7
4Q 2013 L.128 as of March 6 (old)
total financial assets 2,087,400 total liabilities 1,992,300
checkable deposits and currency 111,700 security repurchase
agreements (net) 135,300
security repurchase agreements - credit market instruments
112,400
credit market instruments 476,900 corporate and foreign bonds
122,025
open market paper 28,700 depository institution loans n.e.c.
112,400
Treasury securities 136,000 trade payables 14,500
agency- and GSE-backed securities 114,200 security credit
1,209,300
municipal securities and loans; 18,600 Customer credit balances
833,800
corporate and foreign bonds 133,100 U.S.-chartered depository
institutions; security credit;
asset 187,100
others loans and advances 46,300 Foreign banking offices in the
U.S.; security credit;
asset 188,400
corporate equities 172,400 taxes payable 3,700
security credit 339,200 foreign direct investment in U.S.
108,300
total miscellaneous liabilities 517,000
total miscellaneous assets 987,200 equity investment by parent
companies 1,659,100
unidentified miscellaneous liabilities -1,250,400
Leverage = 22
-
Table 2. US 25: On this table we present the cross-sectional
results for US single leverage based on old and new methodologies
(Lev and Lev_old, respectively), US 2-factor, US Fama-French
models,, where Lev and Lev_old
are respective, new and old leverage factors, for 25 US
Size-bookt to market sorted portfolios.We present intercept,
leverage factor (LevFac) and the standard 3-factors from Fama
French models coefficients, along with the respectives
t-statistics, below coefficients. T-shanken statistics are
reported beneath standard t-statistics. We also report adjusted
R2, adjR2, Mean Asset Pricing Error (MAPE) and Gibbons, Ross,
Shanken statistic, along with its respective p-value.
Periods analysed are 1968-2009, 1968-2014, 1981-2014 and
1991-2014.
Lev Lev_old US 2-
Factor US FF Lev Lev_old
US 2-
Factor US FF
1968-2009 1968-2014
Intercept 3.12 0.71 5.18 13.09 4.67 2.15 6.68 13.74
4.31 0.71 2.6 2.45 6.83 2.13 3.28 2.46
0.72 0.13 1 3.02 1.19 0.45 1.36 3.2
LevFac 47.39 61.67 45.88
42.47 55.22 41.2
7.26 7.5 6.91
6.66 6.84 6.36
2.59 2.39 2.5
2.6 2.41 2.51
Market-
RF 0.49 -7.81
0.11 -7.28
0.3 -1.48
0.06 -1.32
0.08 -1.52
0.02 -1.45
SMB
1.98
1.73
3.22
2.79
1.1
1.04
HML
5.5
4.94
7.16
6.41
2.77 2.73
Adj R2 0.68 0.7 0.69 0.71 0.64 0.66 0.65 0.66
MAPE 3.12 1.20 5.18 13.09 4.67 2.18 6.68 13.74
GRS 44.18 20.79 43.82 54.35 27.26 23.02 53.11 64.42
1% 65% 0% 0% 29% 52% 0% 0%
1981-2014 1991-2014
Intercept 8.71 5.85 11.292 22.62 10.35 9.78 9.11 18.79
19.17 6.62 3.51 3.86 20.98 15.83 2.9 4.44
2.07 1.1 1.9 4.72 2.26 2.05 1.42 4.02
LevFac 43.82 56.58 38.84
38.17 38.82 40.63
5.16 4.97 3.69
3.98 3.58 3.52
2.08 1.85 1.89
1.42 1.33 1.61
Market-
RF -2.58 -14.68
0.89 -10.18
-0.91 -2.56
0.32 -2.46
-0.38 -2.64
0.12 -1.76
SMB
0.78
2.42
1.04
3.42
0.46
1.16
HML
4.85
3.67
5.17
4.15
2.25 1.37
Adj R2 0.52 0.5 0.51 0.59 0.38 0.33 0.36 0.6
MAPE 8.71 5.85 11.29 22.62 10.35 9.78 9.11 18.79
GRS 49.06 41.74 67.62 68.76 56.75 56.13 52.69 54.95
0% 1% 0% 0% 0% 0% 0% 0%
-
Table 3. US 31: On this table we present the cross-sectional
results for US single leverage based on old and new methodologies
(Lev and Lev_old, respectively), US 2-factor, US Fama-French
models, where Lev and Lev_old
are respective, new and old leverage factors, for 25 US
Size-bookt to market sorted portfolios and 6 bond portfolios
sorted by maturity. We present intercept, leverage factor
(LevFac) and the standard 3-factors from Fama French
models coefficients, along with the respective t-statistics,
located below coefficients. T-shanken statistics are
reported beneath standard t-statistics. We also report adjusted
R2, adjR2, Mean Asset Pricing Error (MAPE) and
Gibbons, Ross, Shanken statistic, along with its respective
p-value. Periods analyzed are 1968-2009, 1968-2014,
1981-2014 and 1991-2014.
Lev Lev_old US 2-
Factor US FF Lev Lev_old
US 2-
Factor US FF
1968-2009 1968-2014
Intercept 2.04 0.59 1.50 1.49 2.84 1.06 1.65 1.58
4.10 1.11 2.42 2.28 5.05 1.94 2.57 2.44
0.92 0.34 1.40 2.18 1.30 0.67 1.73 2.55
LevFac 55.92 62.62 51.85
57.24 63.40 48.37
11.20 12.89 9.10
9.80 13.07 8.05
2.36 2.18 2.87
2.48 2.30 2.91
Market-
RF 3.49 3.59
4.25 4.64
5.79 4.26
6.74 5.54
1.07 1.25
1.40 1.73
SMB
2.04
1.84
3.37
3.02
1.13
1.11
HML
5.74
5.18
7.64
6.84
2.89 2.86
Adj R2 0.81 0.85 0.81 0.82 0.76 0.66 0.81 0.83
MAPE 2.19 1.06 1.70 1.64 2.92 1.34 1.81 1.69
GRS 49.91 25.67 52.10 77.37 2.92 25.84 62.43 89.79
1% 69% 1% 0% 0% 68% 0% 0%
1981-2014 1991-2014
Intercept 7.27 3.65 2.94 2.93 8.38 7.38 3.00 3.06
11.59 4.79 3.43 3.41 10.83 8.76 3.62 3.82
2.05 1.11 3.07 4.64 2.17 1.89 3.80 5.70
LevFac 54.35 80.35 55.62
55.70 66.20 53.67
4.17 7.40 6.24
3.32 4.03 3.62
2.48 2.31 2.49
1.90 1.94 1.85
Market-
RF 4.69 4.49
6.17 5.07
5.42 3.96
7.31 4.71
1.37 1.48
1.58 1.44
SMB
1.23
2.71
1.46
3.19
0.73
1.31
HML
5.46
4.87
5.21
4.89
2.53 1.80
Adj R2 0.35 0.64 0.70 0.71 0.25 0.34 0.75 0.78
MAPE 7.27 3.73 3.04 2.99 8.38 7.38 3.08 3.09
GRS 113.98 54.49 106.90 172.34 109.72 94.49 106.98 158.99
0% 0% 0% 0% 0% 0% 0% 0%
-
Table 4. US 35: On this table we present the cross-sectional
results for US single leverage based on old and new methodologies
(Lev and Lev_old, respectively), US 2-factor, US Fama-French
models, for 25 US Size-book to
market sorted portfolios and 6 bond portfolios sorted by
maturity and 10 momentum sorted portfolios. We present
intercept, leverage factor (LevFac) and the standard 3-factors
from Fama French models coefficients, along with the
respective t-statistics, located below coefficients. T-shanken
statistics are reported beneath standard t-statistics. We
also report adjusted R2, adjR2, Mean Asset Pricing Error (MAPE)
and Gibbons, Ross, Shanken statistic, along with its
respective p-value. Periods analyzed are 1968-2009, 1968-2014,
1981-2014 and 1991-2014.
Lev Lev_old US 2-
Factor US FF Lev Lev_old
US 2-
Factor US FF
1968-2009 1968-2014
Intercept 2.19 0.36 1.9 2.47 3.2 0.99 2.07 2.46
3.41 0.5 2.05 1.96 5 1.42 2.27 2.07
0.83 0.16 1.55 3.38 1.23 0.48 1.9 3.73
LevFac 53.83 63.64 52.46
53.99 63.69 49.1
7.68 9.19 6.81
7.47 9.74 6.44
2.63 2.56 3.18
2.8 2.67 3.26
Market-
RF 2.97 2.62
3.75 3.72
3.38 1.77
4.33 2.67
0.9 0.91
1.23 1.39
SMB
1.6
1.45
1.52
1.44
0.88
0.87
HML
4.6
4.26
3.34
3.21
2.3 2.34
Adj R2 0.59 0.68 0.58 0.34 0.58 0.7 0.6 0.4
MAPE 2.61 1.36 2.37 2.79 3.49 1.62 2.47 2.74
GRS 85.82 43.81 86.37 144.87 90.20 40.60 95.47 153.11
0% 31% 0% 0% 0% 44% 0% 0%
1981-2014 1991-2014
Intercept 7.55 4.38 3.07 3.65 8.81 7.97 3.25 4.04
15.33 7.06 3.51 3.17 14.91 12.35 3.78 3.93
2.1 1.14 2.79 5.19 2.3 1.98 3.38 5.92
LevFac 49.43 69.75 56.97
43.54 52.07 53.48
4.68 7.66 7.13
3.35 4.03 6.18
2.69 3.3 2.97
2.3 2.16 2.21
Market-
RF 4.54 3.84
5.84 4.12
5.27 2.81
6.93 3.38
1.31 1.26
1.49 1.16
SMB
0.82
2.49
0.82
2.57
0.48
1.2
HML
4.43
4.04
3.37
3.31
2.04 1.49
Adj R2 0.34 0.59 0.63 0.4 0.2 0.28 0.66 0.53
MAPE 7.55 4.49 3.29 3.79 8.81 7.97 3.47 4.20
GRS 152.70 80.15 129.57 221.95 176.40 157.11 146.85 222.71
0% 0% 0% 0% 0% 0% 0% 0%
-
Table 5. US 25 + 25: On this table we present the
cross-sectional results for US single leverage based on new
methodology (Lev), US 2-factor, US Fama-French, Global 2-factor and
Global Fama-French models, for 25 US
Size-book sorted portfolios along International, Europe or UK 25
Size-book to market portfolios. We present
intercept, leverage factor (LevFac) and the standard 3-factors
from Fama French models coefficients, along with the
respective t-statistics, located below coefficients. T-shanken
statistics are reported beneath standard t-statistics. We
also report adjusted R2, adjR2, Mean Asset Pricing Error (MAPE)
and Gibbons, Ross, Shanken statistic, along with its
respective p-value. Periods is 1991-2014.
Lev US 2-
Factor
Global
2-
Factor
US FF Global
FF Lev
US 2-
Factor
Global
2-
Factor
US FF Global
FF
US 25 US 25 + International 25
Intercept 10.35 9.11 8.99 18.79 17.40 8.00 2.61 8.20 0.24
11.08
20.98 2.90 2.81 4.44 3.73 20.49 1.43 2.49 0.10 2.04
2.26 1.42 1.45 4.02 3.91 1.53 0.28 1.12 0.03 3.64
LevFac 38.17 40.63 40.48
69.98 73.76 69.64
3.98 3.52 3.63
7.03 7.94 6.06
1.42 1.61 1.61
1.94 1.89 1.87
Market-
RF 0.89 0.72 -10.18 -8.14
5.24 -1.19 5.36 -3.01
0.32 0.24 -2.46 -1.69
2.79 -0.39 1.97 -0.56
0.12 0.09 -1.76 -1.42
0.60 -0.15 0.68 -0.64
SMB
2.42 3.63
4.23 -0.37
3.42 3.71
4.74 -0.36
1.16 1.68
1.88 -0.23
HML
3.67 3.41
6.86 3.58
4.15 4.02
6.56 2.78
1.37 1.39 2.32 1.59
Adj R2 0.38 0.36 0.36 0.60 0.55 0.52 0.57 0.49 0.58 0.14
MAPE 10.35 9.11 8.99 18.79 17.40 8.00 2.95 8.20 1.85 11.08
GRS 56.75 52.69 53.21 54.95 55.58 105.06 98.11 105.36 181.31
203.12
0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
US 25 + Europe 25 US 25 + UK 25
Intercept 8.83 5.35 12.15 8.09 19.46 9.33 6.88 6.95 5.34
4.74
25.43 2.51 4.23 2.42 5.53 26.00 1.21 3.22 2.22 1.68
1.82 0.54 2.15 1.51 2.67 2.03 1.21 1.46 1.32 1.42
LevFac 54.07 58.52 48.89
46.12 48.42 48.00
6.40 6.70 5.13
5.60 5.78 5.73
1.95 1.84 1.93
1.75 1.87 1.88
Market-
RF 3.14 -3.82 -1.78 -11.14
2.17 1.70 1.66 4.28
1.51 -1.46 -0.54 -3.29
1.14 0.80 0.66 1.42
0.34 -0.57 -0.34 -1.44
0.36 0.27 0.34 0.85
SMB
3.93 2.17
3.15 0.62
5.65 2.53
3.58 0.71
1.54 1.24
1.47 0.32
HML
5.12 3.26
4.69 3.36
5.58 4.00
5.25 4.41
1.84 1.37 1.58 1.41
Adj R2 0.45 0.47 0.45 0.55 0.35 0.38 0.39 0.39 0.43 0.28
MAPE 8.83 5.35 12.15 8.09 191.04 9.33 6.88 6.95 5.34 4.75
GRS 139.62 130.18 149.32 201.41 19.46 123.32 114.56 116.38
155.45 164.72
-
Table 6. US/International/Europe/UK-only 25: On this table we
present the cross-sectional results for US single leverage based on
new methodology (Lev), US 2-factor, US Fama-French, Global 2-factor
and Global
Fama-French models, for 25 US/International/Europe/UK-only
portfolios sorted by size-book to market. We present
intercept, leverage factor (LevFac) and the standard 3-factors
from Fama French models coefficients, along with the
respective t-statistics, located below coefficients. T-shanken
statistics are reported beneath standard t-statistics. We
also report adjusted R2, adjR2, Mean Asset Pricing Error (MAPE)
and Gibbons, Ross, Shanken statistic, along with its
respective p-value. Periods is 1991-2014.
Lev US 2-
Factor
Global
2-Factor US FF
Global
FF Lev
US 2-
Factor
Global
2-Factor US FF
Global
FF
US 25 International 25
Intercept 10.35 9.11 8.99 18.79 17.40 6.73 11.79 13.29 6.96
6.04
20.98 2.90 2.81 4.44 3.73 17.07 4.31 3.65 2.16 1.46
2.26 1.42 1.45 4.02 3.91 1.00 1.73 1.59 1.52 1.20
LevFac 38.17 40.63 40.48
97.80 72.81 80.70
3.98 3.52 3.63
5.20 3.26 3.98
1.42 1.61 1.61
1.46 1.60 1.59
Market-
RF 0.89 0.72 -10.18 -8.14
-6.72 -7.62 -3.48 -2.07
0.32 0.24 -2.46 -1.69
-1.88 -2.21 -1.03 -0.56
0.12 0.09 -1.76 -1.42
-0.93 -0.94 -0.61 -0.36
SMB
2.42 3.63
0.42 0.88
3.42 3.71
0.18 1.13
1.16 1.68
0.08 0.52
HML
3.67 3.41
9.05 5.73
4.15 4.02
5.35 6.04
1.37 1.39 2.33 2.42
Adj R2 0.38 0.36 0.36 0.60 0.55 0.52 0.57 0.56 0.63 0.63
MAPE 10.35 9.11 8.99 18.79 17.40 6.73 11.79 13.29 6.96 6.04
GRS 56.75 52.69 53.21 54.95 55.58 19.98 28.03 25.04 48.86
51.71
0.00 0.00 0.00 0.00 0.00 0.70 0.26 0.40 0.00 0.00
Europe 25 UK 25
Intercept 7.72 16.22 18.26 9.76 9.38 8.49 8.96 6.32 5.10
3.92
19.57 4.41 4.18 4.39 3.40 16.74 2.95 2.06 1.66 1.43
1.44 2.57 2.30 2.08 1.91 1.79 1.57 1.16 1.15 0.93
LevFac 54.16 34.73 49.80
39.43 39.08 38.44
3.72 2.21 3.73
2.65 2.54 2.54
1.46 1.10 1.38
1.08 1.09 1.05
Market-
RF -9.48 -10.39 -4.92 -3.06
-0.73 1.65 0.58 2.40
-2.40 -2.61 -2.26 -1.24
-0.23 0.53 0.19 0.79
-1.36 -1.31 -0.86 -0.54
-0.10 0.24 0.10 0.41
SMB
-5.29 -0.91
3.93 2.50
-3.74 -1.68
1.44 1.86
-0.95 -0.46
0.87 1.07
HML
9.20 5.54
4.13 2.48
11.52 10.65
2.73 2.73
2.15 2.22 0.93 0.86
Adj R2 0.35 0.45 0.46 0.86 0.82 0.20 0.17 0.18 0.30 0.39
MAPE 7.72 16.22 18.26 9.76 9.38 8.49 8.96 6.32 5.17 4.05
GRS 25.01 30.50 25.12 21.54 32.86 22.71 21.10 22.10 29.53
31.88
41% 17% 40% 0% 11% 54% 63% 51% 20% 0%
-
Table 7. US 31 + 25: On this table we present the
cross-sectional results for US single leverage based on new
methodology (Lev), US 2-factor, US Fama-French, Global 2-factor and
Global Fama-French models, for 25 US
Size-book sorted portfolios along 6 bonds portfolios sorted by
maturity and International, Europe or UK 25 Size-book
to market portfolios. We present intercept, leverage factor
(LevFac) and the standard 3-factors from Fama French
models coefficients, along with the respective t-statistics,
located below coefficients. T-shanken statistics are
reported beneath standard t-statistics. We also report adjusted
R2, adjR2, Mean Asset Pricing Error (MAPE) and
Gibbons, Ross, Shanken statistic, along with its respective
p-value. Periods is 1991-2014.
Lev US 2-
Factor
Global
2-
Factor
US FF Global
FF Lev
US 2-
Factor
Global
2-
Factor
US FF Global
FF
US 31 International 25 Size + US 31
Intercept 8.38 3.00 3.03 3.06 3.11 7.34 2.49 2.95 2.30 2.99
10.83 3.62 3.69 3.82 3.88 16.85 3.05 3.00 2.67 2.33
2.17 3.80 4.01 5.70 5.70 1.56 1.18 2.76 2.03 5.65
LevFac 55.70 53.67 51.60
72.32 73.75 78.59
3.32 3.62 5.36
6.16 8.34 7.88
1.90 1.85 1.80
1.99 1.99 2.06
Market-
RF 6.17 6.14 5.07 6.58
5.36 3.69 3.16 4.90
7.31 6.78 4.71 6.41
6.02 3.71 2.85 3.30
1.58 1.44 1.44 1.75
1.34 0.94 0.88 1.36
SMB
2.71 1.90
4.24 -0.51
3.19 2.05
4.91 -0.52
1.31 0.96
1.85 -0.31
HML
4.87 4.45
6.71 4.33
4.89 5.02
6.71 3.72
1.80 1.85 2.38 1.91
Adj R2 0.25 0.75 0.75 0.78 0.78 0.40 0.66 0.58 0.66 0.31
MAPE 8.38 3.08 3.11 3.09 3.13 7.35 2.82 3.35 2.72 3.77
GRS 109.72 106.98 111.65 158.99 158.74 155.74 148.06 139.61
282.59 296.12
0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
Europe 25 + US 31 UK 25 + US 31
Intercept 8.07 2.93 3.36 2.94 3.84 8.51 3.36 3.27 2.98 2.88
18.73 3.63 3.67 3.69 3.49 18.86 3.98 3.81 3.58 3.12
1.83 1.76 3.73 5.32 5.50 2.05 2.78 3.40 4.65 5.11
LevFac 58.12 61.76 63.00
49.87 52.06 51.20
5.24 7.60 7.04
4.55 6.37 6.27
2.06 2.11 2.15
1.86 1.90 1.88
Market-
RF 5.47 4.13 3.23 3.71
5.55 5.24 4.05 6.23
6.52 4.57 3.53 3.07
6.31 5.81 3.86 5.43
1.40 1.07 0.85 1.03
1.40 1.24 1.10 1.65
SMB
4.04 1.40
3.23 0.46
5.90 1.44
3.80 0.56
1.62 0.85
1.53 0.24
HML
5.64 3.70
4.84 3.41
6.61 3.97
5.64 4.68
2.00 1.60 1.62 1.43
Adj R2 0.33 0.64 0.56 0.68 0.39 0.26 0.59 0.59 0.63 0.55
MAPE 8.07 3.04 3.52 3.11 4.10 8.51 3.53 3.42 3.19 3.24
GRS 224.28 206.01 203.44 340.76 356.38 318.36 299.39 302.32
443.45 442.39
0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
-
Table 5. US 35 + 50: On this table we present the
cross-sectional results for US single leverage based on new
methodology (Lev), US 2-factor, US Fama-French, Global 2-factor and
Global Fama-French models, for 25 US
Size-book sorted portfolios along 10 US portfolios sorted by
maturity, International, Europe or UK 25 Size-book to
market portfolios and International, Europe Size-momentum
portfolios. We present intercept, leverage factor
(LevFac) and the standard 3-factors from Fama French models
coefficients, along with the respective t-statistics,
located below coefficients. T-shanken statistics are reported
beneath standard t-statistics. We also report adjusted R2,
adjR2, Mean Asset Pricing Error (MAPE) and Gibbons, Ross,
Shanken statistic, along with its respective p-value.
Periods is 1991-2014.
Lev US 2-
Factor
Global
2-
Factor
US FF Global
FF Lev
US 2-
Factor
Global
2-
Factor
US FF Global
FF
US 35 International 50 + US 35
Intercept 10.08 8.47 8.64 18.54 20.41 7.93 4.67 11.10 7.88
17.98
24.53 3.45 3.60 7.38 6.64 22.01 2.77 4.32 3.69 6.31
2.29 1.58 1.73 3.51 3.61 1.64 0.62 2.13 1.40 4.70
LevFac 36.33 40.29 39.79
61.45 67.36 53.01
4.35 3.91 3.92
6.02 6.44 4.34
1.57 1.66 1.64
2.02 1.83 1.65
Market-
RF 1.25 0.83 -9.81 -11.34
3.17 -3.88 -2.28 -9.86
0.57 0.38 -3.97 -3.49
1.78 -1.61 -0.96 -3.44
0.19 0.13 -1.53 -1.68
0.44 -0.61 -0.40 -1.88
SMB
2.62 4.02
3.74 -0.44
3.74 3.89
3.27 -0.55
1.26 1.68
1.74 -0.25
HML
2.73 2.26
3.80 0.64
3.01 2.58
2.71 0.57
0.99 0.92 1.28 0.27
Adj R2 0.34 0.33 0.33 0.50 0.39 0.30 0.32 0.30 0.16 0.13
MAPE 10.08 8.47 8.64 18.54 20.41 8.03 5.01 11.10 7.97 17.98
GRS 90.89 84.44 84.54 92.39 88.03 1038.44 934.47 1153.65 1709.91
1704.43
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Europe 50 + US 35 UK 25 + US 35
Intercept 8.87 9.64 15.58 21.61 24.02 9.38 7.07 7.25 9.15
8.78
23.77 4.18 5.84 8.46 9.36 28.84 4.11 4.00 4.37 3.68
1.89 1.26 3.07 3.95 4.74 2.13 1.66 1.88 2.88 2.90
LevFac 48.20 46.36 31.77
41.24 45.17 44.55
4.93 4.13 2.77
5.52 5.68 5.61
2.15 1.64 1.42
1.81 1.99 1.99
Market-
RF -1.04 -6.77 -13.90 -15.03
2.03 1.49 -1.62 0.24
-0.45 -2.82 -5.44 -5.97
1.22 0.85 -0.76 0.10
-0.14 -1.07 -2.29 -2.50
0.38 0.26 -0.36 0.05
SMB
3.28 2.26
2.48 0.56
3.50 2.51
2.90 0.68
1.33 1.35
1.18 0.29
HML
0.81 1.06
3.73 2.77
0.69 1.19
4.00 3.59
0.26 0.43 1.26 1.17
Adj R2 0.22 0.21 0.27 0.30 0.28 0.33 0.34 0.34 0.27 0.15
MAPE 8.99 9.72 15.58 21.61 24.02 9.38 7.07 7.25 9.15 8.78
GRS 1323.52 1353.77 1581.04 1520.13 1572.83 204.80 188.72 191.97
254.64 270.19
0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
-
Table 9. US 41 + 50: On this table we present the
cross-sectional results for US single leverage based on new
methodology (Lev), US 2-factor, US Fama-French, Global 2-factor and
Global Fama-French models, for 25 US
Size-book sorted portfolios along 10 US portfolios sorted by
maturity, 6 bond maturity sorted portfolios,
International, Europe or UK 25 Size-book to market portfolios
and International, Europe Size-momentum portfolios.
We present intercept, leverage factor (LevFac) and the standard
3-factors from Fama French models coefficients,
along with the respective t-statistics, located below
coefficients. T-shanken statistics are reported beneath standard
t-
statistics. We also report adjusted R2, adjR2, Mean Asset
Pricing Error (MAPE) and Gibbons, Ross, Shanken statistic,
along with its respective p-value. Periods is 1991-2014.
Lev US 2-
Factor
Global
2-Factor US FF
Global
FF Lev
US 2-
Factor
Global 2-
Factor US FF
Global
FF
US 41 International 50 + US 41
Intercept 8.81 3.25 3.34 4.04 3.96 7.54 3.24 4.13 4.18 5.53
14.91 3.78 3.88 3.93 3.61 20.27 3.31 3.54 3.59 3.96
2.30 3.38 3.58 5.92 6.13 1.68 1.16 2.96 2.42 6.35
LevFac 43.54 53.48 52.88
60.96 69.67 70.97
3.35 6.18 6.07
5.60 6.97 6.50
2.30 2.21 2.19
2.00 2.13 2.26
Market-
RF 5.84 5.63 4.12 5.83
4.63 2.59 1.66 2.38
6.93 6.37 3.38 4.43
4.33 2.27 1.17 1.55
1.49 1.35 1.16 1.58
1.14 0.65 0.46 0.65
SMB
2.49 0.62
3.75 -0.94
2.57 0.56
3.29 -1.09
1.20 0.32
1.75 -0.55
HML
4.04 3.33
4.20 1.83
3.31 2.94
3.04 1.51
1.49 1.40 1.46 0.79
Adj R2 0.20 0.66 0.65 0.53 0.47 0.25 0.39 0.39 0.22 0.02
MAPE 8.81 3.47 3.54 4.20 4.13 7.67 3.85 4.58 4.58 5.93
GRS 176.40 146.85 150.40 222.71 227.40 1770.23 1539.75 1519.80
3129.49 3278.19
0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
Europe 50 + US 41 UK 25 + US 41
Intercept 8.42 4.26 4.91 6.30 6.73 8.70 3.60 3.58 3.83 3.65
21.37 3.68 3.96 4.65 4.70 21.74 4.29 4.22 4.13 3.68
1.92 1.93 3.75 6.07 6.91 2.17 3.14 3.49 5.56 5.72
LevFac 49.07 59.14 57.76
43.22 51.16 50.37
4.60 5.74 5.43
4.48 6.74 6.63
2.20 2.56 2.61
1.90 2.17 2.15
Market-
RF 4.22 2.73 1.03 1.59
5.31 4.97 3.63 5.64
3.55 2.34 0.69 1.05
6.26 5.73 3.33 4.78
1.04 0.67 0.27 0.44
1.35 1.18 1.02 1.51
SMB
3.17 0.21
2.63 -0.07
2.81 0.20
3.00 -0.09
1.33 0.13
1.25 -0.04
HML
2.25 1.11
4.17 2.97
1.62 0.99
4.39 3.85
0.76 0.46 1.41 1.26
Adj R2 0.18 0.29 0.25 0.10 0.00 0.23 0.53 0.53 0.46 0.40
MAPE 8.57 4.81 5.36 6.59 6.97 8.70 3.81 3.77 4.02 3.98
GRS 5493.39 4706.07 4823.46 8014.77 8276.72 589.02 515.19 519.13
767.20 762.56
0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
-
Table 10. 4-factor and 2-factor horse race: On this table we
present the cross-sectional results for, US 2-factor, US 4-factors
and Global 4-factors models, for 25 US Size-book sorted portfolios
along 10 US
portfolios sorted by maturity, International, Europe or UK 25
Size-book to market portfolios and International,
Europe Size-momentum portfolios. We present intercept, leverage
factor (LevFac) and the standard 3-factors from
Fama French models coefficients, along with the respective
t-statistics, located below coefficients. T-shanken
statistics are reported beneath standard t-statistics. We also
report adjusted R2, adjR2, Mean Asset Pricing Error
(MAPE) and Gibbons, Ross, Shanken statistic, along with its
respective p-value. Periods is 1991-2014.
US 2-
Factor
US 4-
Factors
Global 4-
Factors
US 2-
Factor
US 4-
Factors
Global 4-
Factors
US 25 + International 25 US 31 + International 25
Intercept 2.61 2.20 12.27 2.49 2.50 2.96
1.43 0.78 2.98 3.05 2.93 2.94
0.28 0.29 2.51 1.18 2.52 3.56
LevFac 73.76 41.07 88.00 73.75 42.01 87.21
7.94 1.46 6.15 8.34 2.15 6.03
1.89 1.21 1.28 1.99 0.74 1.28
Market-
RF 5.24 4.13 -4.93 5.36 3.84 4.18
2.79 3.76 -1.21 6.02 3.32 3.56
0.60 0.53 -0.83 1.34 1.08 1.14
SMB
3.65 -0.12
3.61 -0.29
6.19 -0.16
3.92 -0.37
1.69 -0.07
1.70 -0.15
HML
7.94 4.55
7.96 5.38
1.79 4.59
6.54 5.77
2.24 1.90 2.07 2.22
Adj R2 0.57 0.59 0.51 0.66 0.68 0.57
MAPE 2.95 2.65 12.27 2.82 2.83 3.30
GRS 98.11 148.43 77.78 148.06 225.02 120.51
0% 0.00 0.01 0.00 0.00 0.00
US 35 + International 50 US 41 + International 50
Intercept 4.67 7.14 12.73 3.24 3.87 4.27
2.77 3.86 5.04 3.31 3.83 3.80
0.62 0.78 1.83 1.16 1.34 2.57
LevFac 67.36 85.32 80.64 69.67 87.63 96.56
6.44 5.64 6.06 6.97 5.82 7.31
1.83 1.37 1.50 2.13 1.46 1.79
Market-
RF 3.17 0.80 -5.17 4.63 4.32 3.07
1.78 0.38 -2.06 4.33 3.27 2.50
0.44 0.09 -0.63 1.14 1.09 0.74
SMB
2.87 0.26
2.85 0.08
2.87 0.39
2.86 0.12
1.24 0.15
1.22 0.05
HML
7.06 2.45
7.48 3.51
5.22 2.48
5.59 3.53
1.58 0.95 1.86 1.41
Adj R2 0.32 0.37 0.37 0.39 0.42 0.38
MAPE 5.01 7.21 12.73 3.85 4.27 4.62
GRS 934.47 704.72 749.21 1539.75 1162.28 1027.14
0.00 0.00 0.00 0 0.00 0.00
-
Table 11. Other Global Portfolios: On this table we present the
cross-sectional results for US single leverage based on new
methodology (Lev), US 2-factor, US Fama-French, Global 2-factor and
Global Fama-French
models, for Global constructed portfolios, which already use US
assets. Portfolios are, Global 25 is constructed using
25 size-book to market global portfolios, Global 50, adding 25
size-momentum portfolios to Global 25, Global 25 + 6
bonds, which includes 6 bond portfolios on Global 25 and Global
50 + 6, which include 6 bond portfolios to Global
50. We present intercept, leverage factor (LevFac) and the
standard 3-factors from Fama French models coefficients,
along with the respective t-statistics, located below
coefficients. T-shanken statistics are reported beneath standard
t-
statistics. We also report adjusted R2, adjR2, Mean Asset
Pricing Error (MAPE) and Gibbons, Ross, Shanken statistic,
along with its respective p-value. Periods is 1991-2014.
LevFac US 2-
Factor
Global
2-Factor US FF
Global
FF LevFac
US 2-
Factor
Global
2-Factor US FF
Global
FF
Global 25 Global 50
Intercept 7.65 14.70 14.16 13.58 6.04 7.97 15.82 15.35 20.14
18.53
0.34 4.56 3.94 3.67 1.46 18.68 4.51 4.40 7.71 5.78
1.57 2.95 2.42 3.38 1.20 1.73 3.42 2.87 4.01 3.80
LevFac 61.10 38.80 47.34
55.16 16.31 26.53
11.72 2.62 3.51
3.20 0.68 1.24
1.57 1.39 1.53
1.55 0.45 0.74
Market-
RF -7.85 -7.03 -8.25 -2.07
-8.76 -7.69 -14.89 -12.30
-2.30 -2.08 -2.17 -0.56
-2.34 -2.37 -5.36 -3.93
-1.21 -1.00 -1.52 -0.36
-1.32 -1.18 -2.38 -2.15
SMB
0.87 0.88
1.87 1.12
0.73 1.13
1.40 1.20
0.30 0.52
0.64 0.73
HML
5.17 5.73
0.13 0.44
4.66 6.04
0.08 0.30
1.78 2.42 0.03 0.16
Adj R2 0.52 0.59 0.57 0.66 0.63 0.16 0.22 0.22 0.38 0.21
MAPE 7.65 14.70 14.16 13.58 6.04 8.15 15.82 15.35 20.14
18.52
GRS 48.82 62.96 56.66 73.79 51.71 188.59 277.18 259.62 244.87
210.65
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Global 25 + 6 Bonds Portfolios Global 50 + 6 Bonds
Portfolios
Intercept 6.51 2.83 2.74 2.76 2.41 7.30 3.63 3.59 4.73 4.39
12.94 4.05 3.99 4.57 4.23 15.92 3.24 3.20 4.02 3.45
1.84 3.61 3.60 5.15 4.34 1.90 3.87 3.74 6.37 5.90
LevFac 50.28 74.29 69.81
47.30 73.56 69.30
2.58 5.43 5.31
2.42 3.82 3.69
1.31 1.73 1.71
1.30 1.88 1.87
Market-
RF 4.66 3.68 2.79 1.16
4.19 3.18 1.04 1.08
5.66 5.08 3.27 1.51
3.31 2.86 0.69 0.72
1.05 0.98 0.75 0.33
0.91 0.82 0.28 0.31
SMB
2.71 1.21
3.98 1.46
2.38 1.86
2.37 1.36
0.99 0.74
1.48 0.97
HML
6.81 6.00
2.26 1.13
6.15 6.81
1.17 0.68
2.26 2.47 0.68 0.44
Adj R2 0.16 0.62 0.64 0.72 0.70 0.08 0.24 0.24 0.13
MAPE 6.51 2.95 2.87 2.87 2.53 7.50 4.25 4.22 5.07 4.96
GRS 103.06 71.01 75.52 137.03 78.27 436.63 296.41 314.17 606.65
437.87
0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
-
8. Bibliography
Adrian, T. Etula, E. Muir, T. 2010. Financial Intermediaries and
the Cross-Section of
Asset Returns. . Federal Reserve Bank of New York Staff Reports,
No. 464. Revised
September 2013.
Adrian, T. Moench, E., Shin, H.S. 2010. Financial
Intermediation, Asset Prices, and
Macroeconomics Dynamics. Federal Reserve Bank of New York Staff
Reports, No.
422.
Adrian, T., Shin, H.S. 2010. Liquidity and Leverage . Journal of
Financial
Intermediation 19, 418-437.
Adrian, T. Crump, R.K. Moench, E. 2011. Regression-Based
Estimation of Dynamic
Asset Pricing Models. . Federal Reserve Bank of New York Staff
Reports, No. 464.
Revised December 2014.
Adrian, T. Moench, E., Shin, H.S. 2013. Dynamic Leverage Asset
Pricing.. Federal
Reserve Bank of New York Staff Reports, No. 625. Revised
December 2014.
Adrian, T. Etula, E. Muir, T. 2015. Financial Intermediaries and
the Cross-Section of
Asset Returns. . Forth coming at Journal Of Finance 2015.
Bruno, V., Shin, H.S. 2014a. Cross-Border Banking and Global
Liquidity.
Forthcoming in the Review of Economic Studies.
Bruno, V., Shin, H.S. 2014b. Globalization of Corporate Risk
Taking. Journal of
International Business Studies.
Bruno, V., Shin, H.S. 2014c. Capital Flows and the risk-taking
Channel of monetary
Policy. Forthcoming to Journal of Monetary Economics.
Fostel A, Geanakoplos J. 2008. Lever Cycles and Anxious Economy.
American
Economic Review 2008, 98:4, 1211-1244.
Fostel A, Geanakoplos J. 2013. Reviewing the Leverage Cycle.
Cowles Foundation
Discussion Paper, No. 1918.
-
Geanakoplos J. 1997. Promises, Promises. In the Economy as an
Evolving Complex
System II, ed. W.Brian Arthur, Steven Durlauf, and David Lane,
285-320. Reading,
MA: Addison-Wesley.
Geanakoplos J. 2003. Liquidity, Default, and Crashes: Endogenous
Contracts in
General Equilibrium. In Advances in Economic and Econometrics:
Theory and
Application, Eight World Conference, Vol. 2, 170-205.
Econometric Society
Monographs.
Gibbons, M. Stephen, A. R. Shanken, J. 1989. A Test of the
Efficiency of a Given
Portfolio. Econometrica.
Koudijs, P. Voth, H-S. 2014. Leverage and Beliefs: Personal
Experience and Risk
Taking in Margin Lending. . National Bureau of Economic
Research, Working Paper
no 19957.
Krishnamurthy, A. 2009. Amplifications Mechanism in Liquidity
Crisis. National
Bureau of Economic Research.
Lewis, K.K. 2011. Global Asset Pricing. National Bureau of
Economic Research,
Working Paper no 17261.
Malmendier,U. Nagel, S. 2011. Depression Babies: Do
Macroeconomic Experiences
Affect Risk Taking? The Quarterly Journal of Economics.
Shanken, J. 1992. On the Estimation of Beta Pricing Models.
Review of Financial
Studies.