Foreign Portfolio Capital Flows and the Cost of Capital: a study
of Brazilian listed firms
ABSTRACT
This study analyzed the effect of foreign portfolio capital flows on the cost of capital of
Brazilian listed firms. Capital flows increase stock returns, the so-called revaluation effect,
which in turn lowers the cost of equity capital, as an appreciation of stock prices reduces the
dividend yield. It was estimated the effect of foreign portfolio capital flows on the returns of
portfolios of stocks using a modified CAPM, in which a parameter for foreign portfolio
capital flows was included. Stocks were sorted by sector of economic activity and level of
risk. For a considerable number of sectors of the economy, foreign portfolio capitals caused
increases in returns (for sectors related to consumption, commodities, public utilities and
logistics). For the portfolios sorted by risk (in which the stocks’ betas were used as a risk
parameter for sorting), foreign capitals increased the returns for the portfolio in which highbeta stocks were grouped.
Keywords: Foreign Portfolio Capital Flows. Cost of Capital. CAPM
RESUMO
Este estudo analisou o efeito dos fluxos de capital de portfolio no custo do capital de
empresas brasileiras listadas em bolsa. Fluxos de capital geram aumentos no retorno das ações
(revaluation effect), que por sua vez reduz o custo do capital próprio, dado que o incremento
do preço das ações diminui o dividend yield. Foi estimado o efeito dos fluxos de capitais de
portfolio no retorno de portfolios de ações utilizando um CAPM modificado, no qual foi
adicionado um parâmetro para os fluxos de capital de portfolio. Os portfolios de ações foram
formados a partir do setor de atividade econômica e pelo nível de risco. Para um número
considerável de setores da economia, fluxos de portfolio causaram aumentos nos retornos
(setores relacionados ao consumo, commodities, utilidades públicas e logística). Para os
portfolios classificados pelo nível de risco (nos quais o coeficiente Beta das ações foi usado
como um parâmetro de classificação de risco), fluxos de portfolio aumentaram os retornos do
portfolio no qual ações de Beta elevado foram agrupadas.
Palavras-chave: Fluxos de capital de portfolio. Custo de Capital. CAPM
JEL Codes: F320; G120; G150; G320
2
1 INTRODUCTION
In the context of the globalization of financial markets, the Brazilian economy is an
important destination for foreign portfolio capital flows. The economic growth experienced
by Brazil, together with legal reforms in the financial markets and the macroeconomic
stabilization, especially in the last decade, attracted considerable flows of foreign portfolio
capitals to the Brazilian financial market. The graph below illustrates the magnitude of the
total assets held by foreign investors as a share of total Bovespa (São Paulo Stock Exchange)
market capitalization:
Graph 1 – Value of Foreign Investors’ Portfolio as % of Bovespa market capitalization – 2000-2013
Source: Comissão de Valores Mobiliários (Securities and Exchange Commission), 2014.
As the graph shows, the market value of the portfolio of equities held by foreign
investors as a share of total Bovespa market capitalization has grown substantially in the
period between January of 2000 and December of 2013. At the beginning of the period, the
market value of equities held by international investors accounted for roughly 10% of the
market value of all equities traded in Bovespa. Thirteen years after, this amount has doubled,
reaching 20%.
The first graph showed that the stock of equities held by foreign investors grew
steadily over time. When looking at foreign portfolio capital as a flow, instead of as a stock,
the data also shows that there was a remarkable increase both in foreign portfolio capital
inflows and outflows. Net flows (subtracting outflows from inflows) were very volatile,
3
swinging from negative and positive over time. The graph next reports the evolution of
foreign portfolio capitals inflows, outflows, and net flows:
Graph 2 – Foreign Portfolio Capitals as % of stock market traded volume - 2000 – 2013.
Source: Central Bank of Brazil , 2014
The graph above shows inflows, outflows and net foreign portfolio capital flows as a
share of total stock market traded volume. The first important characteristic of the series is
that both inflows (red line) and outflows (green line) showed a growing trend in the period,
rising from a 5% threshold in year 2000 to reach over 25% of total stock market traded
volume by the end of 2013. The average net flow (blue line) was close to zero, what shows
that foreign investors buys (inflows) and sells (outflows) tend to cancel on average. Summing
inflows and outflows, total foreign trading activity reached high thresholds like counting for
up to 75% of total stock market traded volume for some months in the data. The highest net
inflow was around 25%, whereas the highest net outflow reached 15%.
Thus, foreign
portfolio capital flows are responsible for an important share of stock market liquidity.
One important effect of foreign portfolio flows documented in the literature pertains to
the reduction in the cost of equity capital experienced by firms following large waves of
capital inflows. Following the stock market liberalization (considering a previously
segmented equity market), the inflow of international capitals tends to generate a risk sharing
between domestic and international investors, as before liberalization, only domestic investors
could trade on these stocks. Henry (2000) argues that this risk sharing would lower the equity
premium (which is the rate of return that investors claim over investments). The buying
pressure over the stocks following the liberalization of the equity market drives stocks’
returns upwards, which in turn lowers future expected returns (Errunza and Miller , 2000).
4
The Dividend Yield is considered as a proxy for the cost of equity capital (Graham
and Harvey, 2000). The next graph shows the evolution of the Dividend Yield for the stocks
that constitute the Ibovespa index (the main index from the Bovespa Exchange)
Graph 3 –Average Dividend Yield (Ibovespa) – 2001 - 2013
Source: Datastream, 2014
The average Ibovespa’s Dividend Yield has declined between January 2001 and
December 2013 (shown by the downward slopped linear trend). Assuming that the growth
rate of dividends is constant, as suggested in the main valuation models, the decline in the
dividend yield would reflect the appreciation of stock prices in the period (as the dividend
yield is the ratio between dividends paid and the stock price). Thus, the data provides
indication that the cost of capital declined in the period.
Considering the time series presented for foreign portfolio capitals and for the cost of
equity capital, the evidence is that foreign capitals flows increased and the cost of equity
decreased over time. The purpose of this paper is to study the effect of foreign portfolio
capital flows on the cost of equity capital of Brazilian listed firms. Specifically, the main
objective of the paper is to analyze whether foreign portfolio capital flows reduce the cost of
the equity capital of firms, as it is contended in the literature.
Previous studies documented that foreign portfolio capital flows are associated to
increases in stock market returns for the Brazilian stock market (Meurer, 2006; Franzen et al,
2009; Sanvicente, 2014). These studies evaluated the effect of foreign portfolio capitals on
aggregate stock market returns only. The present study brings the following contributions
when compared to prior studies: (i) we estimated the effect of foreign portfolio capital flows
on the stock market returns on a disaggregate level, by using portfolios sorted considering the
5
sectors of economic activity; (ii) risk-sorted portfolios were also employed (using the stocks’
betas as a risk sorting parameter), comparing how foreign portfolio capital flows affected the
portfolio’s returns for portfolios exposed differently to systematic risk; (iii) we used a proper
asset pricing model (a modified CAPM) to evaluate the effect of foreign capitals on stock
returns, whereas previous studies used only VAR or VECM models; (iv) our study brings a
panel structure approach (SURE and fixed-effects estimation), whereas prior papers only used
time series analysis.
The sequence of the paper is structured as follows: Section 2 presents a literature
review related to the effects of foreign portfolio capital flows on the cost of capital. Section 3
describes the data and the variables used in the study. Section 4 brings an asset pricing
analysis, in which regressions were used to estimate the impact of foreign capitals on excess
returns of sector-sorted and risk-sorted portfolios of stocks. In section 5 the results are
discussed, and finally section 6 concludes.
2 LITERATURE REVIEW
2.1 Effects of Foreign Portfolio Capitals on the Cost of Equity Capital
2.2.1 International empirical evidence
The majority of papers that aimed to study the effects of foreign portfolio capital flows
on the cost of equity capital at firm level evaluated how the inflow of foreign capitals that
followed key liberalization events affected stock market returns (Errunza and Miller, 2000,
Bekaert and Harvey, 2000; Henry, 2000; Chari and Henry, 2004, 2006; Patro and Wald,
2005; Christoffer, Chung and Errunza, 2006, among others). The main methodology used on
these papers was the event study: stock market returns are compared before and after
liberalization dates or events. The general conclusion is that the liberalization of equity
markets is associated to an increase in stock market returns, which ultimately leads to a
decrease in the cost of equity capital.
Graham and Harvey (2000) analyzed the change in the dividend yield before and after
the liberalization of the equity market for various emerging economies. The dividend yield
was obtained by dividing the dividends paid by the price of the stock. They also assumed that
that the dividend growth process does not change after liberalization, besides relying on the
usual assumption of constant growth rate of dividends. The authors estimated a regression
6
between the dividend yield and the measure of market liberalization (they used various
measures as the stock of equities held by foreign investors, a dummy variable for
liberalization events, among others). They found that the dividend yield declined as a function
of the liberalization process, providing support for the hypothesis that foreign capital flows
reduce the cost of equity capital.
Errunza and Miller (2000) studied the impact of foreign capitals on the cost of equity
capital analyzing whether firms from developing and developed countries that issued ADRs
(American Depositary Receipts) in the U.S. stock market experienced reductions on the cost
of equity capital. They found that after firms issued ADRs in the US stock market, the firms’
stocks experienced higher returns. This finding supports the hypothesis of equity cost
reduction due to risk-sharing between domestic and international investors.
Foerster and Karolyi (2000) also tested the key prediction of the market segmentation
hypothesis, which claims that firms involved in Global Equity Offerings (GEOs) should
benefit from the access to globally integrated markets, by gaining access to capital at a lower
cost. They also studied ADR offerings, adding to the study the possibility of another type of
depositary receipt offering which is the private placement orders under the Rule 144A (only
available to specific “sophisticated” investors), which differs from normal public ADR
offerings in the sense that private placement equity offers are less subject to complying to US
regulations, and hence is a simplified way for foreign firms to access the US equity market.
ADR offerings were associated to positive long run returns, but that ADRs under rule 144A
(private placement) were associated to negative returns. The main findings were that listing
abroad increased stock market returns, and also that investors rewarded the firms that made
efforts to comply with the more rigorous US accounting standards (as the public offering
resulted in higher returns than the private placement).
Cheri and Henry (2004, 2006) studied whether the revaluation effect could be split
between firm specific characteristics and the common shock to the whole equity market. The
common shock is the risk sharing that affects all firms, regardless of the status of the firm in
the stock market as being of an investible firm (a firm that foreign investors would like to add
to their portfolio after liberalization) or non-investible firms .The idea is that after
liberalization, there is a common shock to expected returns, because as the country moves
from being a segmented market to an integrated market, the risk-free rate falls. Authors found
that investible firms experienced an average stock price revaluation of 15%, out of which 40%
of this effect is due to firm-specific risk-sharing characteristics. Firm level characteristics
7
mediated the process of equity cost reduction, suggesting that the equity cost reduction benefit
did not spillover equally across all listed companies.
Patro and Wald (2005) analyzed the impact of stock market liberation in 18 emerging
markets. The authors found that after liberalization process, the inflow of portfolio capitals
caused expected returns to fall, implying a lower cost of equity capital, according to
predictions of risk sharing between domestic and international investors. Dividend yields
dropped substantially after liberalization. During the liberalization process, smaller firms
showed higher returns, suggesting that the revaluation effect for small firms is higher.
Christoffersen, Chung and Errunza (2006) also estimated the revaluation effect that
stocks would go through after the liberalization process. The authors checked whether the
returns yielded by these stocks were higher when compared to a global returns benchmark,
and also if the returns were comparatively higher for larger firms. The authors found that
there is a positive revaluation effect, being the returns on stocks higher than the global
benchmark during the liberalization of capital flows. Also, their results support the hypothesis
that large firms experience higher revaluation effects than small firms, the opposite result
found by Patro and Wald (2005).
2.2.2 Brazilian empirical evidence
Most of the Brazilian studies evaluated the impact of foreign portfolio capitals on the
aggregate stock market return. The main conclusion of the studies is that foreign portfolio
capital flows are associated to increases in stock market returns, giving support to the
hypothesis of the revaluation effect and consequently equity cost reduction.
Tabak (2003) used VAR and cointegration analysis to study the relationship between
foreign capitals and stock market returns. He showed that inflows of foreign portfolio capitals
to the Brazilian stock market had a huge increase after 1994 (the date that marks
macroeconomic stabilization), and that these inflows seem to be induced by increases in the
stock market index. Also, these flows helped to increase the efficiency of the Brazilian stock
market. He found evidence of cointegration between the IBOVESPA index and international
portfolio flows, concluding that the variables have a long-run equilibrium relationship. As for
the direction of the causality, he found evidence that higher stock returns induce foreign
capital flows.
Meurer (2006) studied the influence of net inflow of resources from international
investors on the Brazilian stock market (BOVESPA), using VAR analysis. The author found
8
that variations in the IBOVESPA index precede a variation in the total participation of foreign
investors in the total market capitalization of Brazilian stock market, directly or indirectly.
The direct effect occurs when variations in the index precede variations in the share of
international investors in the market. The indirect effect materializes from variation in the
index to a higher liquidity in the market, and finally higher liquidity attracting more
international investors. Hence, investors observe alterations in the index, and based on these
alterations they decide whether to invest in stocks or not.
As a description of the behavior shown by international investors in the Brazilian
stock market, Meurer (2006) found that portfolio flows are higher when the index is low, and
the outflows are higher when the index is high, showing that portfolio investors are trying to
operate in the opposite way with respect to the market, buying stocks when prices are low and
selling when prices are high, chasing profitable opportunities.
But overall, the
contemporaneous correlation between stock returns and foreign capitals was positive,
suggesting that foreign portfolio capital flows are contemporaneous associated to higher
returns.
Franzen et al (2009) studied how the portfolio investment flows to the Brazilian stock
market are affected by the return of IBOVESPA index, exchange rate variation, basic interest
rate (SELIC) and country risk, between 1995 and 2005, in a monthly base. The methodology
employed was VAR and VECM analysis. Their results were in synergy with the ones reported
by Meurer (2006) with respect to the rational behavior of international investors, as they also
found that investors join the market after falls and withdraw after rises of IBOVESPA index.
Authors found that portfolio investments are positively related to lag returns of the index,
showing that international investors consider past returns when building their investment
strategies.
Reis, Meurer e Da Silva (2010) studied the relationship between Brazilian stock
market returns and foreign portfolio investments using a linear regression model. They
regressed the IBOVESPA stock market returns on the first difference of the value of equities
held by foreign investors as a share of BOVESPA market capitalization. They also added
other controlling variables, as the exchange rate between the Brazilian real and the US dollar,
the returns to Global equity markets, country’s risk measured by the spread to the Brazilian
bonds (EMBI+), considering also lags for the independent variables. Their result supported a
positive contemporaneous relationship between stock market returns and foreign portfolio
capital inflows. However, they did not find evidence that foreign capitals Granger-cause
returns, but instead found that past returns granger cause contemporaneous inflows. They
9
concluded that foreign investors are feedback traders: they observe returns first, and next
decide on their investment strategy.
Sanvicente (2014) contends that the relationship between the Brazilian stock market
returns and foreign portfolio capitals can be described in three different forms. The first
possibility is in synergy with the argument of Reis, Meurer and da Silva (2010), as foreign
investors would be “trend chasers”, buying Brazilian stocks after past returns were found to
be attractive (feedback trading). The second possibility is that actually portfolio capital flows
cause returns, which is called the information contribution argument. The main rationale for
this argument is that foreign institutional investors are better informed than domestic
investors, so they anticipate better profit opportunities, buying stocks first, driving prices up,
and profiting from capital gains.
The third theoretical rationale to describe the relationship
between the variables is the mutual feedback.
Sanvicente
(2014)
employed
a
simultaneous equation test to study the relationship between stock market returns and net
foreign portfolio capital flows. Foreign capital flows caused excess stock market returns, as
there is positive contemporaneous correlation between the two variables, but also because the
first, second and third lag of foreign capital flows were positively correlated to stock market
returns. On the other hand, for the equation modeling foreign capital inflows, he found only
contemporaneous correlation between the two variables, and no effect of lags of returns on
foreign capitals. Therefore, the results supported the information contribution argument
3 DATA DESCRIPTION
3.1 Stock Market data
The sample consists of firms’ stocks traded at the BOVESPA stock exchange. As of
December of 2013 (the last month comprised by the sample), 363 firms were listed in the
stock exchange. After excluding stocks of financial firms, and also after removing outliers,
the final sample consists of 191 stocks. Outliers were removed when a given stock had more
than 5 consecutive months of missing data for returns (returns were collected from the
Datastream and Economatica databases).
The sampling period is from January 2000 to December 2013. Not all 191 stocks were
listed contemporaneously, so firms were added to the analysis as they got listed in the stock
market. Only firms that were active in December 2013 were included in the sample, so firms
that were de-listed between 2000 and 2013 are excluded from the analysis.
10
The approach used in the study was to sort the 191 stocks in the sample by portfolios.
First, the sorting was made according to the sector of the economy that each stock belongs,
being the stocks grouped in 9 different sectors’ portfolios (following the BMF&Bovespa –
São Paulo Stock Exchange classification criteria). Next, the firms were sorted according to
their level of risk, using the stock’s Beta coefficient as a measure of riskiness. Four portfolios
were formed based on the distribution of Betas for the 191 firms. The portfolios’ returns were
valued-weighted using the stock’s market capitalization as a weighting parameter.
3.2 Variables
The study used country-level time series variables (for stock market returns, risk free
rates and foreign portfolio capital flows) and portfolio-level time series variables (excess
returns). The next table summarizes the variables used in the study:
Table 1 –Variables
Variable
Legend
Rm
Ibovespa Returns
Rf
CDI
Rm-Rf
Excess Ibovespa
Returns
Fk
Net Foreign
Portfolio Capital
flows
FII
Foreign
Institutional
Investors
Ownership
Foreign Capitals
to firms
(portfolios)
FkII
Rit
Excess Returns
for portfolio i
Beta
i
Calculation
Interpretation
Source
Datastream
Monthly returns for all Equity premium for the
and
stocks of the index
Brazilian market
Economatica
Central
Monthly spread on the
Brazilian Risk-free rate Bank of
interbank deposit rate
Brazil
Monthly difference
Excess return on the
between market returns
Brazilian stock market
and risk free rate
Monthly net foreign
Central
portfolio capital flows
Bank of
Measure of foreign
(inflows minus
Brazil
portfolio investor’s
outflows) as a
activity on the stock
percentage of total
market
stock market traded
volume
Minimum percentage A direct measure of the Securities
of equity of a given
minimum stake of
and
firm / portfolio held by foreign institutional Exchange
foreign institutional
investors on firms’ Commission
investors
equity
(CVM)
A firm-level (portfolio)
Fk x FII
measure of foreign
capitals
Datastream
 P 
Monthly Returns for
it 
and
 Rf
Rit  ln 
t
portfolio i in time t
 Pi 
Economatica
t1
CovR m, Ri 
Proxy for firms’
i 
(portfolios’) risk
Var R m
11
The motivation to estimate the effect of foreign capitals using sectors’ portfolios
comes from three rationales: first, Damodaran (2007) argues that usually firms from the same
sector have similar exposure to macroeconomic shocks. Thus, if we consider that the inflow
or outflow of foreign capitals is a macroeconomic shock, firms from the same sectors should
be influenced by these shocks in a similar way. Second, usually firms from the same sector of
activity share similar characteristics, like size of assets’ base, capital structure, profitability,
investment, among others (Smart et al, 2007). Third, the economic growth experienced by the
Brazilian economy in the recent years was much fostered by some sectors of economic
activity, like: (i) consumption due to public policies of income transfers to low-income agents
and a substantial increase in credit availability, and (ii) commodity prices, as the decade
between 2000 and 2010 was marked by a considerable appreciation of commodity prices.
Thus it is interesting to study whether the exposure of stocks’ returns to foreign capitals was
different across various sectors of economic activity.
In the sectors’ portfolio analysis, initially the portfolios were sorted using the broad
sector of activity. Next, they were once again divided as per the sub-sector classification. The
sub-sector classification narrows the differences between each firms’ stocks in a given broad
sector, as we group more similar firms in smaller portfolios. In the risk-sorted portfolio
analysis, the grouping of stocks in different portfolios was done based on the level of
exposure to the systematic risk (betas).
The main independent variable used in the study is net portfolio capital flows as a
percentage of total stock market traded volume (Fk). It is important to remark that only the
foreign capitals directed to the equity market are captured by this variable, or in other words,
it measures how much foreign portfolio capitals were used to buy or sell stocks. The measure
does not include FDI flows, fixed income flows, whether private or public debt bonds,
derivatives flows and finally loans from international organizations like the IMF.
The choice for using net foreign portfolio capitals (inflows minus outflows), instead of
using portfolio inflows and outflows, considered two factors. First, net flows are stationary in
level. On the other hand, both inflows and outflows grew steadily over time, so the variables
are non-stationary in level. As stock returns are stationary, using the first-difference of
inflows or outflows on a regression having returns as a dependent variable would yield a
marginal effect of foreign capitals on excess returns of difficult interpretation, instead of a
direct elasticity as we can have using a variable in level.
The second reason has to do with the characteristics of this study. Net inflows provide
a measure of the resulting effect of buying and selling stocks from foreign investors, whereas
12
inflows measure only the buying pressure whereas outflows measure only the selling
pressure. If we want to evaluate the ultimate effect of foreign capitals on stocks’ returns, it is
important to allow for both buying and selling pressure. Since inflows and outflows are highly
correlated (like in this sample), they cannot be used in the same equation (for multicollinearity
reasons), so the best solution was to use net flows.
The variable FII reported in the table measures the minimum equity ownership held by
institutional investors on a given firm. It was collected from the CVM fillings (Comissão de
Valores Mobiliarios – Securities and Exchange Comission). The purpose of this variable is to
allow for a firm-specific measure of capital flows, since the main variable available to
measure foreign capital inflows is a macro aggregated variable (net foreign capitals directed
to the whole equity market). The variable FII was collected as follows: the Reference form
(Formulario de Referência) of each firm was analyzed as of the end of the 2013 fiscal year.
This is a document that all listed firms must publish mandatorily as per the Brazilian capital
market regulation, in which information about the equity ownership is provided. Typically
investors communicate firms when they reached a 5% stake on the firms’ ownership. The
variable captures only the minimum amount of the firms’ equity that is held by foreign
institutional investors, because foreign investors that hold less than 5% of equity are not
forced to disclose information on this ownership. Thus, it is an imperfect measure of foreign
institutional investors ownership, but at least it allows differentiating firms with respect to
how much foreign capitals each firm is likely to receive.
Since this variable is published yearly, and the regulation that forces the disclosure of
ownership by investors is being enforced relatively recently, only one observation for each
firm was collected, and not for all firms. Many firms in the sample did not report any foreign
ownership on their ownership structure. But still could be the case that some institutional
investors hold equities on these firms, but just below the 5% threshold. Thus, when the
analysis considered this variable (a fixed-effects regression), only firms that reported some
foreign investors ownership on their ownership structure were considered. Next, FII was
interacted with the variable Fk, which gave cross-sectional variability to the macro variable
used for capturing the foreign capitals flows, which separately has only time series variability.
The minimum foreign institutional investors ownership was averaged for each sectors’
portfolio (only for the portfolios sorted based on the sector of economic activity) using the
market value of equity of each firm inside each portfolio as a weighting parameter.
13
3.3 Descriptive statistics for time series variables
Table 2 –Descriptive Statistics for time series variables – Jan/2000 to Dec/2013
Rm
Rf
Rm-Rf
Fk
Mean
Variance
Standard
deviation
Min
Max
N
.0092
.0110
-.0017
.0251
.0053
.0000
.0053
.0020
.0733
.0035
.0731
.0525
-.248
.004
-.259
-.1091
.1792
.0208
.1622
.2429
168
168
168
168
Source: Economatica, Datastream, CVM and Central Bank of Brazil
The average monthly returns for the Brazilian stock market (Rm) were centered on zero
in the period between 2000 and 2013. The excess return (Rm-Rf) was slightly negative in the
period, probably affected by the strong devaluations that the index went through in the 2008
and 2011 crises, and also considering that the Brazilian risk free rate is substantially high.
Net foreign porfolio capital flows (Fk) were positive in the period, as the period was
marked by an average inflow of portfolio capitals corresponding to 2.5% of total stock market
traded volume. In the month in which the highest inflow was observed, it accounted for 24%
of total stock market traded volume. For the month in which the highest outflow was reported,
it accounted for 10.9% of total stock market traded volume. Therefore, the participation of
foreign investors in the Brazilian stock market is indeed substantial.
Table 3 – Correlation Matrix for time series variables – Jan/2000 to Dec/2013
Rm
Rf
Fk
Rm
Rf
Fk
1.00
0.08
0.26***
1.00
-0.19**
1.00
*** Significant at 0.01 level; ** Significant at 0.05 level; *Significant at 0.1 level
The correlation matrix shows positive association between the Brazilian stock market
returns, Rm, and net foreign portfolio capital flows, Fk (0.26). The interpretation is that net
foreign portfolio capital flows are associated to a contemporaneous increase in aggregate
stock market returns. This finding is a first indication that net foreign portfolio capitals could
be associated to higher stock returns. As the correlation is positive, the behavior of foreign
investors in the period between 2000 and 2013 seems to be of exerting a buying pressure on
Brazilian stocks, pushing returns upwards. The risk-free rate is negatively correlated to
14
foreign capitals (-0.19). It could suggest that net foreign capital flows could be driving the
risk-free rate downwards.
3.4 Descriptive statistics for portfolios
Table 4 – Descriptive Statistics for Sectors and Sub-Sectors Portfolios
Returns’
Variance
.007
.009
.006
Avg.
Beta
.740
.752
.731
Var.Beta
FII
#stocks
168
168
168
Returns’
Mean
.003
.001
.001
.112
.090
.142
.154
16
7
9
Basic Materials
Chemicals
Diverse Materials
Wood & Paper
Siderurgy & Metallurgy
Mining
168
168
168
168
168
168
.003
.000
-.013
-.000
.002
.001
.006
.010
.026
.006
.011
.007
1.04
.819
1.01
.699
.871
1.82
.783
.330
.106
.131
.139
3.50
.173
27
4
3
4
11
5
Construction & Transp.
Construction & Eng.
Transport
168
168
142
.004
.004
.008
.007
.007
.010
.815
.666
.901
.437
.194
.438
.046
19
7
12
Public utilities
Energy
Water, Sanitation & Gas
168
168
168
-.004
-.001
-.001
.008
.004
.004
.564
.556
.601
.106
.120
.054
.090
27
22
5
Non-cyclical consumpt.
Drugstores & Supermkts
Agriculture & Meat
Healthcare plans
Personal Goods
Insurance plans
168
168
122
109
168
122
.009
-.000
.013
.001
-.007
.011
.003
.006
.008
.007
.019
.006
.642
.752
.706
.493
.845
.538
.106
.114
.147
.041
.043
.085
.082
29
3
13
5
2
5
Cyclical consumpt.
Retail stores
Education
Textiles
Domestic utilities
Services
168
168
81
168
168
168
-.001
.014
.008
.000
.024
-.018
.008
.011
.017
.006
.013
.038
.757
1.06
.314
.517
.594
.873
.216
.132
.503
.026
.023
.428
.137
33
10
4
8
6
3
Telecomunications
168
-.009
.005
.793
.085
.090
4
Oil & Gas
168
-.003
.010
.890
.279
.053
4
Real Estate
168
.002
.009
.783
.213
.193
30
N
Sector
Industrial Goods
Transport Goods
Machines and Equip.
Mean returns for all broad sectors and sub-sectors were around zero between January
2000 and December 2013. Some sectors presented on average negative returns, like Public
15
Utilities, Cyclical Consumption, Telecommunications and Oil & Gas. For the other sectors,
the average returns on the period were positive.
The largest Beta was for Basic Materials
(1.04), followed by Oil & Gas (.890). There is some heterogeneity in the exposure to
systematic risk within portfolios, as the variance of Betas is non-negligible for most of the
broad sectors. By grouping stocks by their sub-sectors, the dispersion in Betas tend to be
smaller, reflecting a more similar exposure to shocks when the returns are computed
considering firms with more similar business activities.
The variable FII reports the minimum foreign investors’ ownership on the total equity
of each broad sectors portfolio. It was calculated by averaging the minimum equity stake of
all firms inside the portfolio, using the market value of equity as a weighting parameter. It
does not measure all the foreign ownership of each portfolio, but only the minimum stake that
is reportedly held by foreign institutional investors. As the table shows, the sector that
showed the largest minimum stake of equity held by foreign investors was Real Estate (19%),
followed by Basic Materials (17%), Industry (15%) and Cyclical Consumption (13%).
For the portfolios sorted by the level of risk, using stocks’ betas, a total of 4 portfolios
were sorted, according to the distribution of stocks’ beta, ranging from low beta (0 to 25%
deciles), low to mid beta (26% to 50% deciles), mid to high beta (51% to 74% deciles) and
finally high beta (75% to 99% deciles). It was estimated a modified CAPM for each firm’s
stock in the sample, in which besides the traditional parameter for market’s excess return, one
extra parameter for the Net Foreign Capitals was included (Fk). Therefore, each stock’s beta
is net of the effect of foreign capitals. The next table brings descriptive statistics for the risksorted portfolios.
Table 5 – Descriptive Statistics for Risk-sorted Portfolios
Returns’
Variance
.002
Avg. Beta
Var.Beta
#stocks
168
Returns’
Mean
.011
.219
.062
47
Beta 2 (low to mid)
168
.000
.003
.562
.006
46
Beta 3 (mid to high)
168
-.000
.005
.810
.007
50
Beta 4 (high)
168
-.000
.010
1.35
.135
48
N
Portfolio
Beta 1 (low)
Mean returns for the five portfolios are also centered on zero. For the mid-to high and
high beta portfolios, average returns are slightly negative. This is probably due to the fact that
excess returns for the market were also slightly negative in the period, as shown in Table 2.
16
As the higher the portfolio’s beta, higher is the sensitivity of the portfolio’s return to the
market’s return, the negative average for these two portfolios is reflecting this sensitivity.
Table 6– Sector’s Constitution of Risk-sorted Portfolios
Sectors
Cyclical Consumption
Construction & Transp.
Public Utilities
Non-Cyclical Consump.
Basic Materials
Real Estate
Oil & Gás
Telecom
Industrial Goods
Total
Beta 1
8
5
9
10
6
4
1
0
4
47
Beta 2
9
4
9
7
3
6
0
1
7
46
Beta 3
9
7
8
7
7
6
2
2
2
50
Beta 4
8
3
1
5
12
14
1
1
3
48
The Table 7 shows the constitution of each risk-sorted portfolio in terms of the
number of firms’ stocks from each broad sector of activity. The composition of each risksorted portfolio is well balanced with respect to the sector of activity of firms’ stocks in each
portfolio, suggesting that the two portfolio classifications are indeed sorting the stocks
according to two different criteria.
4 REGRESSION ANALYSIS
4.1 Methodology
A modified version of the CAPM was estimated. In addition to the beta coefficient,
one extra parameter (gamma) was included, to capture the effects of foreign portfolio capital
flows on stocks’ returns.
Amihud (2002) also employed a modified CAPM, with an
additional parameter, in a paper studying the effects of stock’s liquidity on returns. In that
paper, the author also added an extra parameter to account for the partial effect of stocks’
liquidity on returns.
The widely used method was time series analysis using Vector Autoregressive
Analysis, but this method does not capture properly the effect of foreign portfolio capital
flows on the excess returns of portfolios controlling for the systematic risk of each portfolio.
In this study an appropriate asset- pricing model is employed. By estimating a modified
version of the CAPM, the effects of systematic risk are controlled for precisely by the beta
coefficient. Moreover, the CAPM allows to estimate the effects of foreign capitals at portfolio
17
level, which adds an interesting feature to the study that was neglected in previous Brazilian
studies: how firm characteristics, like sector of activity and level of risk, affect the
relationship between foreign capitals and stock returns.
The CAPM predicts that returns are proportional to the stocks’ beta coefficient. Thus,
the higher is the exposure to systematic risk; higher should be the returns (Cochrane, 2001).
So the expected relationship between the stocks’ beta and returns is that returns should be
increasing in beta. For the interpretation of the gamma coefficient, which measures the partial
effect of net foreign capitals on firms’ (portfolios) excess returns, the literature suggests that
foreign capitals are associated to increases in stocks’ returns. As a previously segmented
market integrates to the global equity market, the flow of foreign portfolio investments
pressures stock prices up, the so-called revaluation effect (Henry, 2000; Patro and Wald,
2005). Therefore, the expected relationship between excess returns and net foreign capitals is
of an increasing effect.
The equation has a panel structure on the left-hand side (excess returns for portfolio i
at time t), but the same regressors for all the portfolios, which are excess market returns (RmRf) and net foreign capitals (Fk). To identify the correct Panel method to be employed in the
regression, a “poolability” test was conducted before. The idea of the test is to verify whether
there are common parameters for the portfolios, or in other words, if there is a single beta and
a single gamma that captures the partial effects of systematic risk (for beta) and foreign
capitals (for gamma) on the portfolios’ excess returns (Rit). If the data is “poolable”, fixed
effects or random effects panel regressions are appropriate. If the data is not “poolable”,
separate equations should be estimated for each portfolio, which implies that each portfolio
would have its own beta and gamma parameters.
If the model to be estimated was a simple CAPM, the pooling test would be
unnecessary, because asset pricing theory predicts that portfolios exposed to different levels
of systematic risk should have different betas, suggesting the CAPM is not likely to be
poolable. However, as we are adding a second parameter (gamma, for foreign portfolio
capitals), the test becomes relevant, because nothing is known beforehand about the behavior
of this parameter across different portfolios.
The test is performed as follows: first, the equation 9 is estimated assuming that all
betas are equal across portfolios and also all gammas are equal across portfolios. The
estimation is conducted using the Pooled OLS regression with robust standard errors:
Ri t  r f t    


R mt  r f t   F k t   it
(9)
18
Next, the same equation is estimated separately for each portfolio using OLS
regression with robust standard errors:




Ri t  r f t   i   i R mt  r f t   i F k t   i t,
Ri j  r f t   j   j R mt  r f t   j F k t   j t
(10)
...


Ri n  r f t   n   n R mt  r f t   nF k t   nt
The null hypothesis of the test is that:
 i   j   n,
i j  n
Next, the Chow test of poolability, comparing the restricted (pooled) and the
unrestricted (unpooled) models is computed, based on the difference of the sum of squared
residuals between the two models:
SSR pooled   SSR unpooled
F i1 k1, iN k1  
i 1  
k1
 SSR unpooled
(11)
N  k 1i
In the equation above, SSRpooled denotes the sum of square residuals of the pooled
regression, whereas  SSR unpooled denotes the sum of square residuals summed across the
individual equations (one equation for each portfolio). Additionally, i denotes the number of
groups (portfolios in our case), k is the number of regressors in the model and N is the total
number of observations for each unpooled equation. A high F-statistic rejects the null
hypothesis of poolability, suggesting that separate equations must be estimated. The intuition
is that by allowing for separate parameters across groups, the model fit is improved.
The results of the poolability tests are reported together with the output of each
regression in the results’ section. Anticipating the results, both for the sector-sorted portfolios
and for the risk-sorted portfolios, the data was found to be not poolable (the null hypothesis of
poolability is rejected in both cases). This finding implies that a system of equations is more
appropriate, allowing for a different beta and gamma parameter for each portfolio. Therefore,
the following equation was estimated:
19


R i t  r f t   i   i R m t  r f t   i F k t   i C 2008 t   i t
(12)
R i t denotes returns to portfolio i in time t;
rf
t
denotes the risk free rate in time t;
 i denotes the ith portfolio’s intercept;
 i denotes the beta coefficient between the ith portfolios’ returns and the market’s returns;
 i denotes the regression coefficient measuring the partial effect of foreign capitals (Fk) on
portfolio i’s excess returns;
F k t denotes the net foreign portfolio capital flows as a ratio of total stock market traded
volume at time t;
i
denotes the regression coefficient measuring the partial effect of the 2008 financial
instability on portfolio i’s excess returns;
C 2008 t
is a dummy variable for the months of the year of 2008, controlling for the effects of
the financial instability of the 2008 crisis on the excess returns. For each month of the year
2008 the dummy takes the value of 1, whereas it takes the value of 0 for all other months;
 i t denotes the residuals.
4.1.1 Seemingly Unrelated regression
Given the data should not be pooled, the Seemingly Unrelated Regression estimator
was employed. Greene (2003) argues that the CAPM estimation requires especial attention
mainly due to two reasons: first, the regressor is the same for all stocks (portfolios), which is
the excess market’s return. Second, since each stock’s (portfolio’s) returns are likely to be
correlated to the excess market return, they are exposed to similar shocks, what makes the
disturbances in the regression equation (the residuals) very likely to be correlated. Therefore,
it is advised to model the returns of all stocks (portfolios) considering the correlation between
each equation’s residual, as actually the regressions are related through the disturbance error.
The SURE estimator uses the GLS (generalized least squares) structure for the
variance-covariance matrix. So, the estimator is robust to non-spherical variance of the error
20
term (heteroskedasticity and autocorrelation). As the true variance-covariance matrix is
unknown, the Feasible Generalized Least Squares (FGLS) estimator is devised using the
conventional least-squares residuals (Greene, 2003).
Moreover, maximum likelihood
estimators are obtained by iterating the FGLS estimator. Therefore, the SURE method is
robust to distributional assumptions and consistent (maximum likelihood) and also robust to
non-spherical variance of residuals as it applies the FGLS variance-covariance matrix.
However, the SURE estimator is preferable if and only if the residuals across
equations are indeed correlated. In case they are not, the OLS estimator is preferable. To test
the suitability of the SURE estimator, the Breusch-Pagan test of independency of residuals is
presented for each system of equations estimated. The null hypothesis of the test is that the
residuals are not correlated. Thus, rejecting the null hypothesis gives support to the utilization
of the Seemingly Unrelated Regression estimator.
4.1.2 Fixed Effects regression
The variable FII measures the minimum stake held by foreign institutional investors
on each broad sectors’ portfolio, as explained before. By interacting this variable with the net
foreign portfolio capital flows (Fk), a new variable was obtained (FkII), which has both time
series and cross-sectional variability, becoming a panel variable. Most importantly, this new
variable allows for measuring (although imperfectly) the exposure to foreign capitals of each
sectors’ portfolio.
However, as it was shown by the pooling test, the CAPM should not be pooled in
principle, because it is not realistic to impose that all portfolios have the same beta parameter
(what would imply that their returns are exposed similarly to systematic risk). But on the
other hand, the Seemingly Unrelated Regression (SURE) estimator requires that all regressors
are the same across equations (excess market returns and net foreign capitals in our case).
Therefore, it is not possible to use the FkII variable in the SURE estimation, because it has a
different cross-sectional variation across different portfolios.
Thus, if we want to allow for some portfolio-level differentiated exposure to foreign
capitals, we must whether assume data could be pooled, or estimate completely separated
equations for each portfolio. But as argued by Greene (2003), since the returns are highly
correlated, residuals are likely to be correlated too, so the separate estimation of each equation
comes with a cost of not considering the correlation across each equations’ residuals.
21
Having to choose between these two options, we relaxed the “unpoolability”
assumption, estimating a fixed-effects panel (acknowledging the strong assumption imposed
of equal betas as a limitation). In this formulation, we are imposing that all portfolios have the
same beta and the same gamma parameters. To account for any problems with non-spherical
variance of residuals, robust standard errors were used (Arellano robust errors for panel). The
fixed-effects model estimated is as follows:
Ri t  r f t   i  


R m t  r f t   F kII it   C 2008 t   i t
(13)
The main difference of equation 13 with respect to equation 12 is that now all
portfolios were imposed to have the same beta and gamma parameters, and that now each
portfolio has its own exposure to foreign capitals captured by the variable FkIIit, which has
both time and cross-sectional variability.
4.2 Results
4.2.1 Sectors-sorted portfolios
Three equations were estimated for the sectors-sorted portfolios: the first equation had
168 observations for each portfolio, and includes all broad sectors. The second equation
included most of the sub-sectors, precisely the ones that had 168 time series observations. The
equations for the broad sectors and for the sub-sectors were estimated separately because
otherwise the returns to the same stocks would be considered twice (at the broad sector
portfolio and at the sub-sector portfolio, and clearly the correlation between residuals would
be inflated).
The third equation includes stocks that were listed from a shorter period, namely the
following sub-sectors: Agriculture and Meat, Healthcare Plans and Insurance Plans from Noncyclical consumption. It also included in the same equation the sub-sector of Education,
belonging to Cyclical consumption. The equation for these four sub-sectors had 81
observations. For one sub-sector, Transport & Logistics from the Construction & Transport
sector, because it had 142 observations, a bit less than 168 but much more than the 81
observations used to estimate the second equation, a single OLS time series equation was
estimated for this specific sub-sector, using robust standard errors, to account for any
problems with non-spherical variance of residuals.
22
Two tests are reported. First, the Chow test of poolability is shown. The null
hypothesis of the test is that data can be pooled, what suggest that fixed effects or random
effects panel models should be used. If the null is rejected, the SURE estimator is appropriate.
Second, after each regression, the Breusch-Pagan test of independence of residuals is
reported. The idea of the test is to verify whether the residuals of the equations estimated are
correlated. The null hypothesis of the test is that the residuals are not correlated. The test first
generates a correlation matrix between the residuals of all equations in the system, and next
tests for the statistical significance of these correlations. Therefore, if the null hypothesis is
rejected, we can conclude that the Seemingly Unrelated Regression estimator is appropriate.
If, instead, the null hypothesis is accepted, then the OLS estimator is more efficient (being
advisable to estimate the OLS equation with robust standard errors.
Following Cochrane (2001), when estimating the CAPM, the intercept of the
regression corresponds to the pricing errors. He argues that when the explanatory variables
are excess returns, the constant should rather equal zero or not be statistically significant from
zero. But in the context of the CAPM, the constant does not give much information. For all
the equations estimated, the constant was rather statistically not different from zero, or
centered on zero. As it does not convey important information for the study, the constant was
not reported in the output of the regressions.
Table 7 – Chow Test (Poolability)
Model
Sector-sorted Portfolios
Risk-sorted Portfolios
i 1 k 1  i N  k 1 
SSR pool
 SSR i
5.973
5.662
24
1485
3.39***
2.216
1.767
12
825
17.46***
F stat
*** Significant at 0.01 level;
As the Chow Test provides evidence, the data should not be pooled, neither for the
sector-sorted portfolios nor for the Risk-sorted portfolios, as the null hypothesis of the test is
rejected. The result was expected for the Betas, as each portfolio was expected to have its own
parameter, but it is informative with respect to the gammas, as it suggests foreign capitals
affect sectors through different parameters.
23
Table 8 – Equation 12 regression Output - SURE regression – Sector Portfolios
Dep. Var: Rit-Rft
Industrial Goods
Transport Goods
Machines & Equip.
N RMSE “R-sq” chi-2
Rm-Rf std. err
0.37 100.8*** .632***
.074
168 .066
168 .080
0.29 69.2*** . .622*** .089
168 .068
.076
0.30 74.6*** .530***
Basic Materials
Chemicals
Diverse Materials
Wood & Paper
Siderurgy & Metal.
Mining
168
168
168
168
168
168
.049
.091
.148
.060
.062
.070
0.62 282.0*** .825 ***
0.23 50.5*** .556***
0.18 37.4*** .900***
0.46 144.5*** .738***
0.67 341.8*** 1.18***
0.33 84.3*** .588***
.055
.101
.164
.067
.069
.078
.129 *
.145
-.036
-.103
.124
.186*
.077
.142
.231
.094
.096
.110
-.028*
-.068**
-.064
-.051**
-.000
-.042*
.015
.028
.045
.018
.019
.021
Constr. & Transp.
Construction
Transport & Logist1.
168
168
142
.069
.074
.079
0.38 103.2*** .692***
0.25 56.2*** .558***
0.40 18.8*** .804***
.077
.082
.150
.151
-.026
.332*
.108
.115
.197
-.020
-.035
-.010
.021
.022
.024
168
Public utilities
168
Energy
Water, Sanitation & Gas 168
.070
.045
.042
0.38 106.5*** .722 ***
0.50 168.8*** .609***
0.55 212.2*** .628***
.078
.050
.047
.188*
.094
.112*
.109
.071
.066
.005
.008
.008
.021
.014
.013
Non-cyc. consumpt2.
Drugs & Supermark.
Personal Goods
Healthcare plans
Insurance plans
Agriculture & Meat
168
168
168
81
81
81
.047
.070
.124
.066
.054
.051
0.39 110.0*** .521***
0.26 60.0*** .584***
0.18 38.1*** .824***
0.30 35.3*** .434***
0.47 74.5*** .698***
0.53 94.4*** .834***
.053
.078
.138
.127
.105
.097
-.155**
-.097
-.092
.386*
.181
-.010
.074
.109
.194
.198
.164
.153
-.032**
-.008
-.006
-.030
-.019
-.002
.014
.021
.038
.022
.018
.017
Cyclical consumpt.
Retail Stores
Textiles
Domestic Utilities
Services
Education
168
168
168
168
168
81
.059
.074
.070
.109
.152
.105
0.56 221.4*** .884***
0.48 159.9*** .851 ***
0.28 69.6*** .583***
0.12 23.8*** .456***
0.39 110.9*** 1.59***
0.35 45.4*** .643***
.066
.083
.078
.122
.169
.207
.130
.309***
-.098
.328*
.418*
.812**
.092
.116
.109
.171
.237
.316
-.021
-.057**
-.044**
.005
.011
-.074**
.018
.023
.021
.033
.046
.035
Telecomunications
168
.054
0.40 115.1*** .654***
.061
-.230***
.085
.020
.016
Oil & Gas
168
.070
0.51 175.8*** .952 ***
.078
.117
.110
-.012
.021
Real Estate
168
.076
0.35
90.9***
.669***
.085
.071
.119
-.069**
.023
-
-
151.8***
-
-
-
-
-
-
-
-
Breusch-Pagan test – eq 1 168
Breusch-Pagan test – eq 2 168
Breusch-Pagan test – eq 3 81
Fk
.076
.060
.174
std. err C2008 std. err
.103 -.050** .020
.125 -.059** .024
.107 -.044** .021
525.4***
-
-
72.6***
*** Significant at 0.01 level; ** Significant at 0.05 level; *Significant at 0.1 level
1
The equation for sub-sector Transport & Logistic was estimated separately, using OLS robust standard errors.
The chi-2 statistic is instead an F-test..
2
The sub-sectors Healthcare plans, Insurance plans and Agriculture & meat were estimated in a different SURE,
as explained before.
24
The Breusch-Pagan tests of independence of residuals, reported in the end of the table,
reject the null hypothesis that the residuals of the equations in the system are independent (for
the two equations with 168 observations for the broad sectors and subsectors, and also for the
third equation with 81 observations for the sub-sectors with shorter time series).
The seventh column on the table reports the coefficient for the Net Foreign Portfolio
Capitals (Fk). The variable was statistically significant for a considerable quantity of sectors
and sub-sectors, what is a first indication that indeed foreign capitals affected portfolios’
returns. For some broad sectors the impact was positive (Basic Materials and Public Utilities),
whereas
for
others
the
impact
was
negative
(Non-Cyclical
Consumption
and
Telecommunications). Thus, it also provides a first indication that foreign capitals do not
unambiguously affect stock returns positively, as some sectors experienced an increase in the
cost of capital (decrease in returns) due to foreign capital inflows.
The sub-sectors Mining, Transport & Logistics, Water, Sanitation & Gas, Healthcare
Plans, Retail Stores, Domestic Utilities, Services and Education had all a positive effect of
foreign portfolio capital flows on returns. It is interesting to point that the highest elasticities
of excess returns with respect to foreign capitals were reported for sectors linked to
consumption (Education: 0.81; Healthcare plans: 0.38; Retail Stores: 0.30; Domestic Utilities:
0.32, Services: 0.41). It may suggest that foreign institutional investors allocated capitals
trying to take advantage of the consumption-based business cycle that marked the Brazilian
economy in the last decade.
Also, the excess returns of the Basic Materials sector showed to be increasing in
foreign capitals, especially for the Mining sector. Maybe investors allocated capital in this
sector to profit from the average-high commodity prices during the 2000-2013 period.
The negative coefficient of foreign capitals for some sectors suggest that these foreign
capitals are associated to decreases in excess returns, like for Non-Cyclical Consumption and
Telecommunications. Could be that institutional investors reallocated capital from these two
sectors towards other sectors in which the marginal product of capital was expected to be
higher.
The next analysis considered the fixed-effects estimation as described by equation 13.
The 9 broad sectors portfolios were included (industrial goods, basic materials, construction
and
transport,
public
utilities,
non-cyclical
consumption,
cyclical
consumption,
telecommunications, oil and gas, real estate). Because for many firms inside each portfolio
there was no information about the minimum foreign stake on the firms’ ownership, the
returns to each portfolio were re-estimated, using only the returns for the firms who reported
25
foreign investors ownership. Overall 9 portfolios were included, but since some portfolios did
not have listed firms since the beginning of the time series (oil and gas and non cyclical
consumption, specifically), the panel is unbalanced.
Table 9 – Equation 13 regression Output - Fixed Effects regression – Sectors Portfolios
Dep.Var:
Rit-Rft
Coefficients
Robust std
errors
constant
0.000
.000
Rm-Rf
.779***
.049
FkII
.743***
.278
C2008
−.022**
.008
F(11,1326)
N
i
average t
adj. R-squared
DWatson
F-test for F.E
63.32***
1338
9
148
.338
1.705
4.70***
*** Significant at 0.01 level; ** Significant at 0.05 level; *Significant at 0.1 level
When pooling all portfolios together, the coefficient obtained for the beta (Rm-Rf) is a
measure of the average exposure to systematic risk across the 9 portfolios (0.77). It is not the
best measure, clearly, because we are imposing that all betas are equal, whereas the pooling
test presented before showed this is not exactly the case. But at least it is controlling for the
systematic risk on each portfolios’ excess returns, allowing for a ceteris paribus effect of our
main variable of interest in this regression, which is FkII. The dummy used to control for the
financial instability caused by the 2008 financial crisis is statistically significant and negative.
The coefficient for FkIII is statistically significant and positive (0.74), suggesting that
net foreign capitals are associated to an increase in excess returns. For each 1% of net foreign
portfolio flows (weighted for the minimum stake of foreign institutional investors on each
portfolios’ total equity), excess returns increase by 0.74%. The main advantage of this
estimation with respect to the previous one is that now the foreign capital flows had some
portfolio-specific variation, as the variable Fk (a macro variable) was interacted with the
variable FII (a cross-sectional variable).
The utilization of this variable yields an imperfect measure of the amount of foreign
capitals going to each portfolio, as it considers just the minimum stake held by foreign
26
investors in each portfolio (it does not proxy for the total stake, as explained before). Also, it
assumes that this minimum stake is constant, which is a strong assumption. In other words,
the main assumption is that the higher is the minimum stake held by foreign institutional
investors on the portfolios’ total equity, higher is the exposure of this portfolio to foreign
portfolio capital flows. At least this measure allows us to separate in each sector’s portfolio
the firms that were more likely to have received foreign portfolio investments from other
firms that were unlikely to benefit from these flows.
This could be one explanation for the considerably high coefficient we found for the
variable FkII (0.74). If we look at the previous estimations using only aggregate flows (Fk),
for the broad sectors for which foreign capitals affected returns positively, the elasticity was
around 0.20. For the sub-sectors, the elasticities were around 0.30 on average, and 0.81 for the
highest sub-sector (education). Since in the fixed-effects analysis only firms that indeed are
exposed to foreign portfolio capital flows were included, there is no dilution of the effect of
foreign capitals due to firms that actually were not exposed to foreign investors trading. So it
is likely to be a refined measure of the effects of foreign portfolio capital flows on returns.
4.2.2 Risk-sorted portfolios
The table below brings the results for the regressions using risk-sorted portfolios. The
portfolios were sorted by the proximity of the Beta coefficient between stocks. Again the
Breusch-Pagan test for independence of residuals supports the utilization of the SURE
estimator. All equations were statistically significant too. As expected, the higher the
portfolios Beta, higher the r-squared, as a higher fraction of excess returns are explained by
the stocks’ sensitivity to market returns.
Table 10 – Equation 12 regression Output - SURE regression – Risk-sorted (beta) Portfolios
RMSE “R-sq”
Dep. Var: Excess
Returns to portfolios
Beta 1 (low)
168
.047
0.23
Beta 2 (low-mid).
168
.035
Beta 3 (mid-high)
168
Beta 4 (high)
168
N
std. err
Fk
std. err
C2008
std. err
.052
-.023
.073
-.030**
.014
0.64 299.8*** .629***
.039
.004
.055
-.028***
.011
.039
0.73 461.9*** .862***
.043
.069
.060
-.012
.012
.045
0.80 711.7*** 1.25***
.050
.119*
.071
-.004
.014
chi-2
Rm-Rf
51.6*** . .328***
14.7**
Breusch-Pagan Test
*** Significant at 0.01 level; ** Significant at 0.05 level; *Significant at 0.1 level
27
Foreign capital flows are associated to increased excess returns only for the high-beta
portfolio. The elasticity of excess returns with respect to foreign capitals was .119, implying
that a 1% inflow of foreign portfolio capitals increased the excess returns of the high-beta
portfolio by 0.12%. It seems that foreign investors are buying stocks whose returns follow
closely the market index, or whose returns have an even higher sensibility to the market
index.
An important discussion regarding this finding is whether foreign investors are buying
riskier stocks (high-beta stocks), or if foreign investors trading make stocks riskier. Or, in
other words, the results call for a discussion on the direction of the causality effect.
Considering that a stock with a low beta is a stock whose returns are less sensible to
systematic risks, foreign investors should avoid holding such a stock. If institutional investors
invest globally, and use different equity markets for hedging against systematic risk, there is
no point for holding stocks insensible to the Brazilian stock market, because these stocks bear
much idiosyncratic risk, which can’t be diversified away.
The second possibility is that foreign investors’ trading induces higher volatility to
stocks, which in turn make these stocks riskier. A simple way to investigate whether foreign
capitals induce a higher volatility is looking at the correlation matrix between the standard
deviation of the risk-sorted portfolios (a measure of the portfolios’ volatility or risk) and
foreign portfolio capitals.
Consider sigma i as the standard deviation of ith portfolio, computed in the period
between t-4 and t. Thus, it measures the risk of the portfolio as a function of the standard
deviation of returns between t and the four preceding periods (t included). Fk is the average
net foreign capital flows between t and t-4 (t included).
Accordingly, in the table shown
below, sigma1 is the standard deviation for the low beta portfolio, sigma2 is the standard
deviation of the low-mid beta portfolio, sigma3 is the standard deviation of the mid-high beta
portfolio, and finally sigma4 is the standard deviation of the high beta portfolio.
28
Table 11 – Correlations: Risk-sorted Portfolios’ volatility and Foreign Capitals – May/2000 Dec/2013
1
2
3
1
1.00
2
0.53***
1.00
3
0.34***
0.62***
4
0.37***
0.73*** 0.61***
Fk
0.00
-0.08
4
Fk
1.00
1.00
-0.26*** -0.33***
1.00
*** Significant at 0.01 level; ** Significant at 0.05 level; *Significant at 0.1 level
For the low-beta and low-mid beta portfolio, the average foreign capital flow is not
correlated to the standard deviations of excess returns. For the mid-high and high-beta
portfolios, average foreign capital flow is negatively correlated to the riskiness of these two
portfolios (-0.26 and -0.33 respectively). The insight from the correlations is that for the midhigh and high-beta portfolios, the flow of foreign capitals apparently reduces the volatility of
excess returns. Foreign capitals may actually reduce the riskiness of portfolios, what is in line
with the risk-sharing hypothesis.
5 DISCUSSION OF RESULTS
The capital asset pricing model estimated for sectors-sorted portfolios supported that
net foreign portfolio capital flows are associated to excess returns to a considerable number of
portfolios. Considering the broad portfolio sorting, Basic Materials and Public Utilities were
the two sectors in which foreign capital flows caused excess returns to increase. When
gauging the relationship more deeply using the sub-sectors, Mining, Retail Stores, Domestic
Utilities, Education, Services, Healthcare Plans, Water, Gas and Sanitation and finally
Transport & Logistics were the sub-sectors for which net foreign capitals caused an increase
in excess returns. On the other hand, for two broad sectors net foreign capital flows caused
decreases in excess returns, namely Telecommunications and Non-cyclical consumption.
As discussed in the literature review, prior studies provided evidence that the
liberalization of a previously segmented equity market generally leads to a reduction in the
cost of equity capital (Henry, 2000; Erunza and Miller, 2000; Chari and Henry, 2006; Stulz,
1999). After the liberalization occurs, stocks in the liberalized market experience the so-
29
called revaluation effect (Chari and Henry, 2004). However, most of these studies employed
the event-study methodology, in which the impact of foreign capitals was assessed
immediately after a liberalization event or set of events. However, building on the argument
by Stulz (1999), the liberalization of an equity market is not best described as a one-shot event
or a set of important regulatory marks, but is rather a process.
Foreign investors may start to trade on a previously segmented market after a given
regulatory mark, but sustained investment depends on other factors, like macroeconomic
stability and growth on the country level, and on corporate governance improvements and
adoption of accounting best practices (internationally recognized) on the firm level (Stulz,
2006).
Therefore, although event studies are useful to assess the short-term impact of
liberalization on the cost of equity, given that the liberalization process takes time to occur, it
is not likely that the revaluation effect on stock prices lasts only for such short periods, as
contended by these event studies.
Considering the Brazilian stock market, a large number of companies went public
during the 2000-2010 decade, much influenced by the inflows of foreign capitals that became
more abundant after the macroeconomic stabilization and regulatory improvements from the
end of the 1990-decade onwards. The 2000-2010 decade also marks the consolidation of
differentiated segments of the stock market for firms following international patterns of
corporate governance (like the Novo Mercado – New Market segment). However, the
Brazilian stock market is still very small in number of listed firms and liquidity, if contrasted
to US or European markets. Thus it is reasonable to consider that as a broad process, the
liberalization of the Brazilian equity market is still an ongoing process, but that this process
had an important development in the period between 2000-2013, exactly the period covered
by this study.
Therefore, in light of the theoretical predictions on the effects of foreign capital flows
on stock returns, and ultimately on the cost of equity, also relating these predictions to
previous empirical studies, the positive effect of net foreign capitals on the excess returns of
some of the sectors’ portfolios can be interpreted as the revaluation effect that these stocks
went through during the 2000-2013 period. As there was an appreciation of stock prices due
to net foreign capital flows, keeping the growth rate of dividends constant, as per the Gordon
valuation model, the cost of equity capital declined (Stulz, 1999; Graham and Harvey, 2000;
Henry, 2000). The finding that stock market returns increase following net foreign portfolio
capitals in Brazil is in line with the results found by Sanvicente (2014) for the broad Brazilian
stock market index.
30
Considering the regression models estimated, the gamma parameter for net foreign
capitals fitted an approximation for the average revaluation effect for the stocks from each
sector for the period between January of 2000 and December of 2013. In the first estimation
of the sectors portfolios (separate equations using the SURE method), each sector and next
each sub-sector had its own gamma parameter, which was on average around 0.25 for the
statistically significant elasticities.
For the second estimation (using fixed-effects), a common gamma was imposed to all
broad portfolios, yielding an elasticity of 0.74, a much higher number. Considering that the
fixed-effects estimation allowed differentiating the flow of foreign capitals across portfolios,
using only firms indeed exposed to foreign capitals in the constitution of the portfolios, it can
be considered a refined measure of the revaluation effect. For this measure, a 1% net flow of
foreign portfolio capitals (weighted for the average equity stake of foreign institutional
investors in each portfolio) increased excess returns in 0.74%. This result is in synergy with
the findings of positive revaluation effects as reported by Patro and Wald (2005), Chari and
Henry (2004) and Christoffersen, Chung and Errunza (2006).
Considering the model estimated using separate equations (SURE), the highest
elasticities of excess returns with respect to net foreign capitals were found for consumptionrelated sub-sectors (retail stores, domestic utilities, services, education, healthcare plans).
Also for sectors linked to commodities (mining), public utilities (water, gas and sanitation),
and transport and logistics, the econometric evidence suggests that the revaluation effect was
positive after foreign capitals inflows. These evidences suggest that foreign investors
allocated capital on the business sectors more benefited from the business cycle that the
Brazilian economy experienced in the last decade, which was strongly influenced by the
growth in consumption and the appreciation of commodity prices.
Also, the increase in income due to income transfers from public policies is likely to
have fostered private investments in energy and especially sanitation, what could possibly
explain the positive effect of foreign capitals on the excess returns of the Public Utilities
portfolio (one must have in mind that the provision of public services in Brazil is still an
incipient process, especially in rural areas). The transport and logistic sector may have
attracted foreign capitals due to the investments undertaken to supply the demand for
transport solutions from the commodity-exporting sector (agro-business and basic materials).
Differently from prior studies, the results obtained from the sectors’ regressions
suggest that some firms, belonging from Telecommunications and Non-cyclical consumption,
experienced an increase in the cost of capital, instead of a decrease, as the marginal effect of
31
net foreign capitals on excess returns was negative. Thus, there is new evidence that foreign
capital flows do not unambiguously reduce the cost of capital.
Although it is difficult to
point out concrete reasons to account for this devaluation of stocks, one possibility is that
foreign investors reallocated capital to other sectors, in which the marginal product of capital
was higher (or expected to be higher) due to the economic momentum. Or rather, they just
sold their equity positions on the firms from these sectors, investing somewhere else.
Prior studies provided evidence that firm characteristics are important factors in the
process of equity cost reduction following foreign capital flows. The results from the
regressions in this study provide new insights about the importance of the sector of activity as
a relevant factor for the reduction in the cost of capital. As it was argued, sectors of activity
more related to the business cycle could have attracted more foreign capitals, thus having a
higher impact on the cost of capital. It also suggests that foreign investors observe the
exposition of firms to macroeconomic conditions when choosing their investing strategy.
The second part of the analysis evaluated how foreign capital flows affected excess
returns for risk-sorted portfolios, by grouping stocks for the proximity of their Beta
coefficient. Foreign investors are likely to be investing in such ways to follow the Ibovespa
index, as the high-beta sector was the only portfolio with positive elasticity of excess returns
with respect to net foreign capitals. This finding is reasonable if we consider the hedging
behavior of foreign investors, attempting to diversify the risk from their portfolios by buying
stocks that follow the Brazilian index, instead of buying low-beta stocks, which are likely to
bear idiosyncratic risk. Also, net foreign capitals were correlated to reductions in the volatility
of risk-sorted portfolios.
On the other hand, net foreign capitals seem to do little or nothing for decreasing the
cost of capital of low and mid beta stocks, as for these two risk-sorted portfolios, returns do
not respond to foreign capital inflows, what can suggest that the risk-sharing effect of net
foreign capitals do not spillover equally across all firms (portfolios) in the stock market.
Foreign capitals may reduce the cost of capital only for firms benefitting from the current
business cycle, and firms whose stocks’ can be a good investment to hedge foreign investors’
portfolios (high beta stocks).
Franzen et al (2009), Meurer (2006) and Sanvicente (2014) discussed the causality
relationship between foreign capitals and stock market returns. The results presented in this
study are not directly informative with respect of the direction of the causality, because the
asset-pricing model used to evaluate the effect of foreign capital flows on excess returns
considers only the contemporaneous correlation between the two variables (no lags of returns
32
or foreign capitals were included). However, regardless of the fact that higher returns may
induce foreign investment, and foreign investments can in turn induce continuity in higher
returns, since foreign capitals must enter or leave the stock market exerting a buying or selling
pressure on the stocks’ prices, the contemporaneous correlation is likely to be always in place.
Moreover, the feedback-trading hypothesis assumes in some extent that institutional
investors believe that past returns are good predictors for future returns. If we analyze it under
the perspective of the Market Efficiency Hypothesis, stock returns actually follow a random
walk (Fama, 1970). As foreign investors are usually well-informed institutional investors, it
is reasonable to expect that they gauge information on future corporate profits using more
complex information sources then past returns. Thus it is more likely that they are better
informed than local investors, and just anticipate good investing opportunities, buying stocks
and driving returns upwards. As the results found in this study are supportive of this last
possibility, they corroborate in some extent the information contribution argument.
7 CONCLUSIONS, LIMITATIONS AND FURTHER RESEARCH
The objective of this study was to analyze the impact of foreign portfolio capital flows
on the cost of equity capital of a sample of Brazilian listed firms. The regression models
estimated for the modified CAPM (in which portfolios’ excess returns were modeled as a
function of portfolios’ betas plus a parameter for net foreign portfolio capital flows) provided
evidence that foreign portfolio capital flows affect the returns of portfolios sorted by sector of
activity and by risk.
The main conclusion from the sector-sorted portfolios regressions is
that foreign capitals are associated to an increase in returns, which accounts for the
revaluation effect that stocks went through during the 2000-2013 period. As stock prices
increased in the period, the cost of equity capital declined. Sectors and sub-sectors directly
related to consumption, commodities, logistics and public utilities were the ones that showed
positive correlation with net foreign capitals. Telecommunications and non-cyclical
consumption were the sectors for which foreign capitals caused decreases in returns.
The main conclusion from the risk-sorted portfolios regressions is that foreign
investors trade more on stocks that follow closely the index (high beta stocks), probably to
hedge their portfolios against systematic risks. Also, evidence suggested that foreign capitals
are negatively correlated to the volatility of returns for mid and high beta portfolios,
corroborating the risk sharing argument.
33
The study had some important limitations.
The main variable used to measure
foreign capitals was an aggregate measure (total flows to the stock market), instead of a firmlevel measure, because data on the flow of foreign investment that goes to each firm
specifically was not publicly available. A second version of this variable was also
implemented, using the minimum stake that foreign institutional investors held on firms’
(portfolios) equity to allow for cross-sectional variability of foreign capitals. But as argued
before, this is an imperfect measure of firm-level (portfolio level) foreign ownership, because
it captures only the minimum stake of a firms’ equity held by foreign institutional investors,
and because it was assumed to be constant. Further studies, maybe with access to more
sophisticated databases, could use specific firm-level data on foreign ownership, overcoming
this limitation.
The study did not account for any potential differentiation between the direct effect of
foreign capital flows on stock returns, and the effect that can be transmitted through increased
market liquidity. Further studies can try to disentangle these effects, maybe employing
different measures of foreign capital flows, or using high-frequency data to separate the effect
of liquidity on stock prices. Finally, this study did not investigate in detail the feedbacktrading hypothesis for the relationship between foreign capitals and stock returns. Further
studies could blend time series analysis (as VAR and VECM) with asset pricing models to
give a better treatment for this possible relationship.
34
REFERENCES
Amihud, Y. (2002). Illiquidity and stock returns: cross-section and time-series
effects. Journal of financial markets, 5(1), 31-56.
Bekaert, G., & Harvey, C. R. (2000). Foreign speculators and emerging equity markets. The
Journal of Finance, 55(2), 565-613.
Bekaert, G., Harvey, C. R., & Lumsdaine, R. L. (2002). The dynamics of emerging market
equity flows. Journal of International money and Finance,21(3), 295-350.
Chari, A., & Henry, P. B. (2004). Risk sharing and asset prices: evidence from a natural
experiment. The Journal of Finance, 59(3), 1295-1324.
Chari, A., & Blair Henry, P. (2008). Firm-specific information and the efficiency of
investment. Journal of Financial Economics, 87(3), 636-655.
Christoffersen, P., Chung, H., & Errunza, V. (2006). Size matters: The impact of financial
liberalization on individual firms. Journal of International Money and Finance, 25(8), 12961318.
Cochrane, J. H. (2001). Asset Pricing. Princeton University Press
Damodaran, A. (2007). Avaliação de empresas. Pearson Prentice Hall.
Errunza, V., & Losq, E. (1985). International asset pricing under mild segmentation: Theory
and test. The Journal of Finance, 40(1), 105-124.
Errunza, V., & Losq, E. (1989). Capital Flow Controls, International Asset Pricing, and
Investors' Welfare: A Multi‐ Country Framework. The Journal of Finance, 44(4), 1025-1037.
Errunza, V. R., & Miller, D. P. (2000). Market segmentation and the cost of the capital in
international equity markets. Journal of Financial and Quantitative analysis, 35(04), 577-600.
Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work*. The
journal of Finance, 25(2), 383-417.
Foerster, S. R., & Karolyi, G. A. (2000). The long-run performance of global equity
offerings. Journal of Financial and Quantitative Analysis, 35(04), 499-528.
Franzen, A., Meurer, R., Gonçalves, C. E. S., & Seabra, F. (2009). Determinantes do fluxo de
investimentos de portfólio para o mercado acionário brasileiro. Estudos Econômicos (São
Paulo), 39(2), 301-328.
Fratzscher, M. (2012). Capital flows, push versus pull factors and the global financial
crisis. Journal of International Economics, 88(2), 341-356.
Greene, W. H. (2003). Econometric analysis. Pearson Education India.
35
Henry, P. B. (2000). Stock market liberalization, economic reform, and emerging market
equity prices. The Journal of Finance, 55(2), 529-564.
Henry, P. B. (2003). Capital account liberalization, the cost of capital, and economic
growth (No. w9488). National Bureau of Economic Research.
Meurer, R. (2006). Fluxo de capital estrangeiro e desempenho do IBOVESPA.Revista
Brasileira de Finanças, 4(1), 345-361.
Patro, D. K., & Wald, J. K. (2005). Firm characteristics and the impact of emerging market
liberalizations. Journal of banking & finance, 29(7), 1671-1695.
Rogoff, K., Wei, S. J., & Kose, M. A. (2003). Effects of financial globalization on developing
countries: some empirical evidence (Vol. 17). Washington, DC: International Monetary Fund.
Reis, L., Meurer, R., & Da Silva, S. (2010). Stock returns and foreign investment in
Brazil. Applied Financial Economics, 20(17), 1351-1361.
Sanvicente, A. Z. (2014). The foreign capital flows and the behavior of stock prices at
BM&FBovespa. BAR-Brazilian Administration Review, 11(1), 86-106.
Singh, A. (2003). Capital account liberalization, free long-term capital flows, financial crises
and economic development. Eastern Economic Journal, 191-216.
Smart, S. B., Gitman, L. J., & Megginson, W. L. (2007). Corporate finance. Thomson SouthWestern.
Stulz, R. M. (1999). Globalization, corporate finance, and the cost of capital. Journal of
applied corporate finance, 12(3), 8-25.
Stulz, R. M. (2005). The limits of financial globalization. The Journal of Finance, 60(4),
1595-1638.
Tabak, B. (2003). The random walk hypothesis and the behaviour of foreign capital portfolio
flows: the Brazilian stock market case. Applied Financial Economics, 13(5), 369-378.