PANOECONOMICUS, 2014, 6, pp. 691-707
Received: 24 March 2013; Accepted: 20 May 2014.
Francisco
López-Herrera
Facultad de Contaduría
y Administración,
Universidad Nacional
Autónoma de México,
México
EGADE Business School,
Tecnológico de Monterrey,
México
roberto.santillan@itesm.mx
Edgar Ortiz
Facultad de Ciencias Políticas y
Sociales,
Universidad Nacional
Autónoma de México,
México
Interdependence of NAFTA
Capital Markets: A Minimum
Variance Portfolio Approach
francisco_lopez_herrera@yahoo.com.mx
Roberto J.
Santillán-Salgado
UDC 339.166.2:662.75
DOI: 10.2298/PAN1406691L
Preliminary report
edgaro@unam.mx
This project was partially sponsored by
the Financial Markets, Asset Valuation
and Risk Management Chair, at
EGADE Business School, Tecnológico
de Monterrey.
Summary: We estimate the long-run relationships among NAFTA capital market returns and then calculate the weights of a “time-varying minimum variance
portfolio” that includes the Canadian, Mexican, and USA capital markets between March 2007 and March 2009, a period of intense turbulence in international markets. Our results suggest that the behavior of NAFTA market investors is not consistent with that of a theoretical “risk-averse” agent during periods of high uncertainty and may be either considered as irrational or attributed
to a possible “home country bias”. This finding represents valuable information
for portfolio managers and contributes to a better understanding of the nature
of the markets in which they invest. It also has practical implications in the
design of international portfolio investment policies.
Key words: NAFTA, Stock markets, International diversification, Financial
integration, Optimal portfolios.
JEL: F36, F37, G01, G11, G15.
During the 1970s and 1980s, many empirical studies focused on the determination of
the optimal parameters that portfolios of risky assets should have in the theoretical
context of the mean-variance model, increasingly incorporating the effects of international diversification and reporting (in most cases) encouraging results. By investing
in securities traded in different national markets, an investor enjoys the diversification effects derived not only from imperfectly correlated macro-financial contexts but
also from different market-determined exchange rates.
However, more recently, as a result of free trade and the economic integration
decisions observed in different geographic areas (the European Union, ASEAN
countries, or NAFTA countries), as well as of lesser barriers to international financial
transactions, there has been an increasing similarity in the returns of different national capital markets, and the benefits of international diversification of portfolios have
gradually faded (Claire G. Gilmore and Ginette M. McManus 2004; Raj Aggarwal
and NyoNyo A. Kyaw 2005; Ali F. Darrat and Maosen Zhong 2005). Also, in response to the increasing evidence that the potential benefits of international diversification declined during the 1990s, the focus has gradually shifted to previously overlooked emerging markets (Kuan-Min Wang and Hung-Cheng Lai 2013).
692
Francisco López-Herrera, Roberto J. Santillán-Salgado and Edgar Ortiz
1. Literature Review
During the last two decades, there was an interest in studying the gradual reduction
in diversification benefits across different national capital markets due to increasing
integration and globalization of financial markets.
François Longin and Bruno Solnik (2001) used extreme value theory to study
the structural dependence among national capital markets. They modeled a multivariate distribution for high yields, i.e., those that exceed a certain (positive or negative) limit, and estimated the correlation for increasing limits. They hypothesized that
under the assumption of multivariate normality the correlation of beyond-the-limit
high yields should converge asymptotically to zero as the limit increases. The results
obtained from their estimates over 38 years of monthly observations (January 1959 to
December 1996) of the five largest stock markets (USA, UK, France, Germany, and
Japan) rejected that hypothesis, at least for negative returns. They reported that the
correlation of highly negative returns does not converge to zero but tends to increase
with the limit used, with high statistical significance. Conversely, the correlation of
positive high yields tends to decrease and converge to zero as the limit used increases. Thus, they concluded that the correlation between national markets increases in
bear markets but not in bull markets.
According to Robert-Paul Berben and W. Jos Jansen (2005), the growing economic importance of national capital markets during the last thirty years was accompanied by a greater degree of synchronization (co-movement) in their returns behavior, driven by factors such as the enhanced speed and reliability of electronic communications, the liberalization of capital controls in virtually all nations, and a fast
increase in regional economic integration, which facilitated the transmission of impulses and information between markets. As a result, domestic markets have been
increasingly affected by disturbances originated in foreign markets. Using weekly
market indices and 10 industry indices, with observations between January 1980 and
June 2000, the authors found that correlations between the markets of Germany, the
UK, and the USA more than doubled, increasing from about 0.30 to about 0.65, during that period. However, that result was in stark contrast with the correlations between the Japanese stock market and the other three markets, which were unchanged
at a level of 0.30.
Regarding the behavior of the correlations of industrial indices studied with
the same methodology, Berben and Jansen reported that their results were very similar to those observed at the aggregate level. For Germany, the UK, and the USA, inter-industry cross-country correlations either increased or remained at the same level,
while for Japan they remained mostly unchanged. That means that increased synchronization between international capital markets is determined not only by global
factors but also by specific factors at the country and industry levels, the influence of
which may be substantial.
Cristiana Tudor (2011) studied the common long-run stochastic trends and the
short-term interaction mechanisms among six Central and Eastern European (CEE)
stock markets and the USA stock markets, paying special attention to the effects of
the 2007-2009 global financial crisis. Using Vector autoregressive (VAR) models,
Cointegration tests, and Granger causality tests, the author chose the optimal number
PANOECONOMICUS, 2014, 6, pp. 691-707
Interdependence of NAFTA Capital Markets: A Minimum Variance Portfolio Approach
of lags to be introduced to the models based on Sims’ likelihood ratio test. In order to
study the effects of the 2007-2009 global financial crisis, she split her sample into
two to capture any possible time-variant stock market integration in the CEE area
before and during the crisis, while paying special consideration to the starting point
of the manifestations of the crisis in international stock markets.
All her tests confirmed the strong interrelations between the CEE stock markets during the crisis, meaning that diversification benefits from investing across all
six countries disappeared during the financial turmoil. Nevertheless, the strong interdependencies present during crises do not necessary imply that CEE economies share
the same long-run equilibrium relationship. Hence, the permanence of increased
stock market integration in the long-run at a regional level remains to be seen and,
the author argued, should receive the attention of future research efforts through the
introduction of a post-crisis data subsample in the analysis.
Nikola Gradojević and Eldin Dobardžić (2013) proposed that, from a portfolio
diversification view, short-horizon investors are more interested in the co-movement
of stock returns at higher frequencies, whereas long-term investors focus on the relationship at lower frequencies, i.e., long-term fluctuations; thus, a frequency domain
analysis to obtain a better insight about the co-movement across various investment
horizons is in order. By using a test for causality in the frequency domain, these authors provided a deeper insight into the relationship between the returns on regional
stock market indices in Croatia (CROBEX), Slovenia (SBITOP), Hungary (CETOP),
and Germany (DAX) relative to the returns on the BELEX 15 index in Serbia. Their
results suggest substantial causality interactions at various frequencies.
George L. Ye (2014) studied the interactions between the USA and Chinese
stock markets. Considering that these stock markets have no overlap in their trading
hours, there are grounds to assume the inexistence of correlation between them.
However, Ye examined the ability of the daily returns on the S&P 500 and the Dow
Jones Industrial Average (DJIA) to forecast the direction of the opening of the
Shanghai Composite Index (SSEC) and Shenzhen Component Index (SZCI), two
benchmark indices in the China stock market, and vice versa; he found that the USA
stock market opening did have a significant ability to forecast Chinese stock market
openings but that the daily returns on the Chinese market did not show a similar ability to forecast USA stock market openings. That fact may be explained by the big
difference in the relative size and global economic importance of the two markets
(the USA market being significantly larger, with a listing of many global corporations).
2. Links between the Canadian, Mexican and USA Stock
Exchanges
Several studies have analyzed the relationships between the capital markets of the
NAFTA member countries (Canada, Mexico, and the USA) and confirmed the existence of a process of financial integration between them (see, for example, Aggarwal
and Kyaw 2005; Darrat and Zhong 2005; Cetin Ciner 2006; Francisco López-Herrera
and Edgar Ortiz 2007; López-Herrera, Ortiz, and Alejandra Cabello 2007). Notably,
PANOECONOMICUS, 2014, 6, pp. 691-707
693
694
Francisco López-Herrera, Roberto J. Santillán-Salgado and Edgar Ortiz
and despite the abundant evidence in favor of increasing financial integration among
NAFTA countries, Bradley T. Ewing, James E. Payne, and Clifford Sowell (2001)
reported a nonsignificant relationship between those markets in the long-run. However, in a study using cointegration, López-Herrera, Ortiz, and Cabello (2008) reported evidence of the existence of long-term meaningful relationships between the
NAFTA markets. In a previous study (López-Herrera, Ortiz, and Cabello 2007), the
same authors used cointegration incorporating the presence of structural breaks in the
series and reported the same result. The conflicting evidence reported by Ewing,
Payne, and Sowell may be explained by the presence of structural breaks over the
period of analysis, which the latter authors did not take into consideration.
Although the number of studies focused on the significance of the relationship
between the stock markets of Canada, Mexico, and the USA has increased in recent
years, such studies are still limited; however, research on the relationship of the volatilities of yields is still more scant. For example, López-Herrera, Ortiz, and Cabello
(2009), found evidence of the existence of significant relationships at the level of the
volatility of returns of the NAFTA stock markets and reported that the volatility of
the USA market was the main driver of the volatility of the Mexican market.
More recently, Jesús Cuauhtémoc Téllez Gaytán and Pablo López Sarabia
(2010) used different frequency observations to analyze the correlation between the
Mexican stock exchange index and other Latin American stock market indices and
key market indicators for the USA stock markets. Their results, based on a wavelet
analysis, indicated that the correlation between the Mexican stock market and the
other markets was not very intense, and only in a few cases did the correlations exceed 0.7, which calls into question whether the co-movement is as intense as that
reported by other studies. It should be noted that although the correlation levels estimated by Téllez Gaytán and Sarabia for the stock markets of Mexico and the USA
seem reasonable (compared with the results reported by López-Herrera, Ortiz, and
Cabello 2009), these may attain higher levels during periods in which there are extraordinary market pressures (cracks or financial panics).
3. Methodological Aspects
Finding the minimum variance portfolio has an analytical solution when solving the
following nonlinear (quadratic) optimization problem:
(1)
Only the variances and covariances (correlations) of returns are needed to find
the right combination of risky assets that would minimize the portfolio volatility.
However, as mentioned earlier, there is enough evidence that both statistical parameters are time-varying; thus, if the time-varying feature is not incorporated in our attempt to model the construction of a minimum variance portfolio, we would tacitly
incur a “risk of the model” problem.
PANOECONOMICUS, 2014, 6, pp. 691-707
Interdependence of NAFTA Capital Markets: A Minimum Variance Portfolio Approach
The time-changing volatility of financial asset returns has been widely studied
and documented. To obtain a more precise understanding of that phenomenon, some
authors have proposed the use of Autoregressive conditional heteroskedasticity
(ARCH) models; others have suggested the use of Generalized autoregressive conditional heteroskedasticity (GARCH), and still others different variations of ARCH
(Tim Bollerslev 2008). While most of those studies focused on the dynamics of the
USA capital markets, in a notable effort, Gerald Kohers, Ninon Kohers, and Theodor
Kohers (2004) examined the changing volatility of twenty-four capital markets from
1980 through 2003 and reported common characteristics in their behavior. An interesting finding was that, in practically all the countries included in their sample, the
volatility remained relatively stable between 1980 and 1996 but increased notably
between 1997 and 2003.
Other studies have also rejected the hypothesis of constant correlations (see,
for example, Robert F. Engle and Kevin Sheppard 2001; Longin and Solnik 2001;
William N. Goetzmann, Lingfeng Li, and K. Geert Rouwenhorst 2002; Ryan S. Suleimann Lemand 2003; A. S. K. Wong and Peter J. G. Vlaar 2003; Geert Bekaert,
Campbell R. Harvey, and Angela Ng 2005; Bekaert, Robert J. Hodrick, and Xiaoyan
Zhang 2005; Lorenzo Cappiello, Engle, and Sheppard 2006). However, despite the
profusion of univariate models for specifying the behavior of changing variances, the
use of multivariate models has not progressed as fast. Evidently, the motivation for
the development of such models is associated with the possibility of jointly estimating the volatility and the correlations among different national market returns, which
are required inputs for the analysis.
We estimated a minimum variance portfolio for the NAFTA stock markets
during the financial crisis by following a two-step procedure: we first estimated a
multivariate GARCH model with varying conditional correlations (MGARCHVCC,
originally proposed by Yiu Kuen Tse and Albert K. C. Tsui 2002), and then we used
the MGARCHVCC output to estimate the optimal weights of a minimum general
risk portfolio (MGRP) (i.e., a minimum variance portfolio). Our objective was to
observe the changing composition of the MGRP throughout the period of analysis,
with the intention to discern what effects the market turbulence of the 2007-2009
financial crisis had on its theoretical composition.
To model m variables, the Tse and Tsui model may be specified as follows:
(2)
where rt is an mx1 vector that contains the different market returns as elements; Ht1/2
is the factor that results from performing Cholesky’s decomposition of Ht, the timechanging conditional variances matrix; and Dt is a diagonal matrix with conditional
variances:
PANOECONOMICUS, 2014, 6, pp. 691-707
695
696
Francisco López-Herrera, Roberto J. Santillán-Salgado and Edgar Ortiz
.
Each
GARCH:
,
(3)
changes over time according to a process similar to the univariate
(4)
where ωi, αj, and βj are the parameters for estimation; R is a matrix of mean correlations to which the process of conditionally changing correlations converges:
(5)
is the rolling estimator of the correlations of and et; and
are the parameters that govern the dynamics followed by the conditional correlations. To maintain the stability of the model, the changing correlations must fulfill the restrictions
0,
0;
1.
The likelihood logarithmic function based on the multivariate normal distribution of observation t is:
(6)
Thus, if one has the observations for t = 1, 2,..., T, the estimation of the model
parameters can be done using the maximum likelihood method to optimize the following function:
(7)
4. NAFTA Stock Markets and the Minimum General Risk
Portfolio
The data used to measure the integration among NAFTA markets and to estimate a
minimum general risk portfolio included the Morgan Stanley Capital Indices for
Canada, Mexico, and the USA during the period from March 3, 2007 through March
9, 2012, with 1,306 daily observations. The three indices were expressed in USD,
eliminating potential distortions introduced by exchange rate fluctuations. Table 1
presents the estimation of the MGARCH (1,1) VCC (1,1) econometric model:
PANOECONOMICUS, 2014, 6, pp. 691-707
Interdependence of NAFTA Capital Markets: A Minimum Variance Portfolio Approach
Table 1 The MGARCH (1,1) VCC (1,1) Model
Coefficient
Mexico
Canada
USA
0.063537
0.0527436
0.0371239
(0.0201585)
(0.0138156)
(0.0089242)
0.088642
0.078989
0.1047957
(0.0184807)
(0.0124889)
(0.0128223)
0.8933415
0.09018628
0.8763321
(0.0204187)
(0.0142757)
(0.0122227)
Correlations model
1
0.0096177
(0.0023064)
2
0.9847595
(0.0046121)
Note: The values in parentheses represent robust standard errors.
Source: Authors’ own elaboration.
All the parameters estimated with the changing-variances model were highly
significant and pertinently met the stability requirements. Also, the changingcorrelations model parameters were highly significant, although their sum suggests a
highly persistent correlation process.
Figure 1 presents the estimated conditional volatility obtained from the model
shown in Table 1 for the market indices of the three NAFTA countries from March
12, 2007 through December 30, 2011. During the same period, the OECD reported
different recessionary subperiods for each of the NAFTA countries: Canada was in a
recession from May 2007 through June 2009, Mexico from June 2007 through May
2009, and the USA from December 2007 through May 2009. We highlighted with a
discontinuous box frame the period from May 2007 until June 2009, during which
the recessionary periods in the three countries overlapped.
The conditional volatility started increasing in moderate waves at the end of
July 2007 but saw a decline in April 2008, when it went back to a level comparable
with that in the before-the-crisis period. However, in September 2008 the volatility
increased again, reaching maximum values during October 2008 (36% for the Canadian market, on October 30, 2008; 48% for the Mexican market, on October 14,
2008; 30% for the USA market, on October 16, 2008). During the following months
the conditional variances receded but not completely, and on February 2009 there
was a new escalation of volatility that faded away until August of that year.
Throughout the whole period of analysis, except for a few observations, the
estimations of the conditional volatility for the Mexican market were permanently
higher, showing the greater sensitivity of that market to daily news reports on the
financial problems of all types of firms, as well as to announcements of new policy
measures implemented mainly by the USA authorities (the Federal Reserve and the
Treasury Department) to address the market turbulence.
PANOECONOMICUS, 2014, 6, pp. 691-707
697
698
Francisco López-Herrera, Roberto J. Santillán-Salgado and Edgar Ortiz
Source: Authors’ calculations, with data obtained from MSCI (2013)1.
Figure 1 Conditional Variance of NAFTA Indices
Figure 2 shows the conditional correlations estimated for the NAFTA stock
market indices. The correlation between the Mexican stock exchange and the Canadian stock exchange diminished from March 19 to May 24, 2007. After a brief re1
Morgan Stanley Capital International (MSCI). 2013. MSCI Developed Markets Indexes.
http://www.msci.com/products/indexes/country_and_regional/dm/#/ (accessed March 20, 2013).
PANOECONOMICUS, 2014, 6, pp. 691-707
Interdependence of NAFTA Capital Markets: A Minimum Variance Portfolio Approach
covery starting on February 11, 2008, there was another downward movement, which
reached its lowest level on September 12 of the same year. In brief, there was a decrease in conditional correlations, and the conditional variances reached their peak
precisely at the time when the NAFTA markets were in free fall.
Source: Authors’ own elaboration, with data from MSCI (2013).
Figure 2 Conditional Correlations for NAFTA Stock Markets
When the stock prices recovered from their lows, toward the end of April
2009, the conditional correlations returned to a level comparable with that observed
at the beginning of March 2008. From that moment on, our estimations suggest that
the conditional correlations remained stable, even when, toward the end of the period, there was a new surge that attained a maximum level during the first days of
February 2012.
The conditional correlations between the Canadian and the Mexican markets,
as well as between the Mexican and the USA markets, were higher than the levels
reported by López-Herrera, Ortiz, and Cabello (2009) for a different period of observations (August 23, 1984 to December 22, 2005). In our sample, one can observe that
the conditional correlations for the Mexican and USA markets followed a lateral trajectory within the range of approximately 0.70 and 0.75 during the first months;
however, there was a noticeable decrease in their correlation during the period from
April 18 through May 20, 2008, when the three markets were in freefall. After returning to the levels observed before August 21, 2009, the correlations started to decrease again until March 16, 2010, when they reached their lowest estimated value.
Alternatively, the conditional correlations for the Canadian and Mexican markets
took a deep dive around the Lehman Brothers bankruptcy announcement, in midSeptember 2008, reaching their lowest level at around 0.45. Finally, the conditional
correlation between the Canadian and USA markets also reached its nadir (close to
0.50) at around the same date.
PANOECONOMICUS, 2014, 6, pp. 691-707
699
700
Francisco López-Herrera, Roberto J. Santillán-Salgado and Edgar Ortiz
5. Minimum Global Risk Portfolios for the NAFTA Markets
during the Financial Crisis
The minimum global risk portfolios were estimated according to Mark Grinblatt and
Sheridan Titman (2002)2, by replacing the constant variance and covariance parameters with our daily estimates of conditional volatilities and conditional correlations.
Figures 3 to 6 show the evolution of the MGRP weights over the period of our
analysis, which was characterized by atypical volatility due to the many extraordinary events related to the financial crisis. Figure 3 shows a clear relationship with the
patterns observed in the conditional variances analysis that was presented in a previous section. For example, a structural break was observed during the first months
of 2008, when the crisis started taking its toll, provoking a significant rebalancing of
the MGRP, and reflected in the volatile change in weights of each of the three NAFTA markets (wcan for Canada, wmex for Mexico, and wus for USA). During the first
months of 2009, the weight of the USA market rose dramatically and reached more
than 100% on several occasions. A similar pattern was observed from October 2010
onward in relation to the sovereign debt problems of European countries, as well as
in May 2011 and toward the end of that year, and again in the first months of 2012
for the same reason.
Source: Authors’ own calculations, with data obtained from MSCI (2013).
Figure 3 Weights of the NAFTA Markets in the MGRP
During the first stages of the financial crisis, different government-driven policy measures (by the USA government in particular) to bailout large financial and
nonfinancial firms (that are too big to fail), as well as other extraordinary measures
regarding monetary and fiscal policy, had a clear effect on the behavior of market
2
See Chapter 4, p. 125.
PANOECONOMICUS, 2014, 6, pp. 691-707
Interdependence of NAFTA Capital Markets: A Minimum Variance Portfolio Approach
volatility, which was reflected on the conditional variances and covariances of the
NAFTA market returns. Because conditional variances and correlations are inputs for
the estimation of the weights of the MGRP, their changes affected the composition of
the latter, resulting in a decrease in the weight of the USA market and an increase in
the weights of the other two markets.
After the bankruptcy announcement of Lehman Brothers (on September 15,
2008), the volatility of financial markets reached a maximum level. From that peak
level, the volatility receded, resulting in a new composition of the optimal portfolio,
in which the USA market weight increased again, this time reaching more than
100%, a level above its average weight (i.e., with negative or “short” positions in the
other two NAFTA markets), as previously mentioned.
Finally, toward the third quarter of 2011, and as a consequence of the worsening liquidity condition of Greece, which eventually made a bailout inevitable, turbulence returned to the financial markets, which again affected the composition of the
MGRP, thus increasing the weights of the Canadian and Mexican markets. During
the first weeks of 2012, there were signs of more stability, and the weights of the
Canadian and Mexican markets went back to their average levels.
Figures 4 to 6 provide a graphical representation of the weights of the individual NAFTA markets in the composition of the MGRP. As a reference to facilitate the
interpretation of the graphs, the average weight throughout the observation period is
indicated with a bold line. Figure 4 shows the MGRP weights for the Canadian stock
exchange above their mean value, beginning in mid-2007 and until March 2009, corresponding to the period when the financial crisis hit the USA market. Again, the
weights increased again between March and November 2010 and from July 2011
until the end of the observation period. The latter subperiod coincided with announcements about the inability of some European countries to meet their sovereign
debt compromises, which provoked spikes in the volatility of world markets.
Figure 5 shows the evolution of the Mexican stock exchange weights for the
NAFTA MGRP. There were two periods of increasing weights. The first, from
March 2008 until May 2009, was associated with great uncertainty related to the
many bankruptcies and liquidity problems faced by American companies, as well as
the extraordinary measures that that country’s authorities took to respond to the financial markets instability (TARP, TARP II, etc.). The second, from March 2011
through the end of that year, was closely associated with the surge of new turbulence
linked to the sovereign debt crisis of Portugal, Spain, Ireland, and Greece.
One can also identify two subperiods during which the optimal weights for the
USA market remained consistently below their mean values for the whole period.
The first one, between January and December 2008, was also the most turbulent period of the financial crisis. The second one was during the third quarter of 2011,
when the USA market weights reached their most negative value of the whole analysis, around the 8th of August.
PANOECONOMICUS, 2014, 6, pp. 691-707
701
702
Francisco López-Herrera, Roberto J. Santillán-Salgado and Edgar Ortiz
Source: Authors’ own calculations, with data obtained from MSCI (2013).
Figure 4 Weights of the Canadian Market in the Minimum Variance Portfolio, 2007-2012
Source: Authors’ own calculations, with data obtained from MSCI (2013).
Figure 5 Weights of the Mexican Stock Exchange in the Minimum Variance Portfolio, 2007-2012
PANOECONOMICUS, 2014, 6, pp. 691-707
Interdependence of NAFTA Capital Markets: A Minimum Variance Portfolio Approach
Source: Authors’ own calculations, with data obtained from MSCI (2013).
Figure 6 Weights of the USA Stock Exchange in the Minimum Variance Portfolio, 2007-2012
Compared with the volatility observed in the weights of the USA market, the
volatility in the weights of the other two NAFTA markets was significantly greater
due to their smaller relative economic size.
Our findings suggest that, in the face of extreme market turbulence, the adequate investment strategy for a risk-averse investor in a NAFTA market is to constantly rebalance the MGRP in order to respond to changing domestic market return
variances and cross-country covariances.
6. Conclusion
In this work, we analyzed the evolution of conditional variances and conditional correlations for the three NAFTA market index returns. As expected, we found notorious increases in the volatility of returns in all three NAFTA markets during the
2007-2009 global financial crisis. Those increases happened during a period in
which, according to the OECD, the economies of the NAFTA countries were undergoing a recessive phase. Noticeably, the highest volatility levels were achieved only
toward the second half of those recessive phases, which was closely associated with
the Lehman Brothers bankruptcy and a series of important events, including the announcement of extraordinary measures to bail out the financial system (TARP,
TARP II, etc.).
As reported by previous studies, in the long-run the increasing economic integration of the three NAFTA markets could explain the closer co-movements of their
capital market returns. One can, however, detect some interesting exceptions to that
PANOECONOMICUS, 2014, 6, pp. 691-707
703
704
Francisco López-Herrera, Roberto J. Santillán-Salgado and Edgar Ortiz
trend during the first months of the crisis. For example, the negative impact of the
complex events that were going on in the USA was relatively mild on the Mexican
market and resulted in lesser volatility than expected. Among the most logical explanations, one can mention that not many Mexican financial entities held “toxic assets”
(subprime mortgage-related assets and the derivatives issued to hedge them) in their
balance sheets and that the Mexican macroeconomic fundamentals (fiscal balance,
external debt levels, unemployment levels, etc.) were sound.
Contrary to what has been observed and well documented in former crisis episodes, as the 2007-2009 financial crisis worsened, the levels of changing conditional
correlations estimated for the three NAFTA pairs (Canada-Mexico, Canada-USA,
and Mexico-USA) diminished in free fall, particularly in the Canada-Mexico and
Canada-USA cases. The conditional correlation of the Mexico-USA market returns
also experienced a reduction from its long-term mean values but not in the same
measure as the two former pairs.
According to the optimal weights of the MGRP, investors should have increased their MSE holdings and reduced their holdings of the USA assets during the
first semester of 2008. Based on the foreign portfolio investment flows reported by
the Department of Economic Studies of the Banco de Mexico (Mexico’s Central
Bank), the net capital inflows toward the Mexican market were still positive during
that period. However, the same source revealed that during the period of more intense turbulence, between July and November of that same year, there was finally
some capital flight out of Mexico (net negative foreign portfolio investment flows)
but, paradoxically, at a time when the optimal weights obtained from the MGRP
model (for almost the whole second half of 2008) suggested that the Mexican market
could have performed as a hedge (diversification alternative).
A better understanding of the dynamics of highly integrated capital markets
during turbulence periods can help international portfolio managers improve the
quality of their rebalancing decisions. While portfolio managers need not be riskaverse (actually, they are not expected to be risk-averse during normal periods), during turbulence periods like the one recorded during the second semester of 2008, one
could have expected a preference for less risky (minimum variance) portfolios. By
contrast, our analysis shows that their behavior was not optimal. The question that
naturally arises, and which delineates new research avenues, is whether such seemingly irrational behavior may be explained by “home bias” theory, herd behavior,
self-fulfilling expectations, and strategies of trading based on market sentiment, as
suggested by contrarian investors, or maybe a mix of these principles.
It is a well-documented fact that investors typically invest a large portion of
their assets in domestic security markets because of a “home bias” phenomenon, further strengthened by the extra transaction costs and complexity of investing abroad,
besides unfavorable international relations dynamics between countries, thereby negatively affecting the motivation for international diversification (Rizwan Mushtaq
and Syed Zulfiqar Ali Shah 2014). However, it is a well-known fact that trade openness and easy access to information are two of the main reasons that have contributed
to reducing home equity bias. Therefore, in the case of NAFTA countries, where
trade openness and easy access to information are clearly present, the explanation for
PANOECONOMICUS, 2014, 6, pp. 691-707
Interdependence of NAFTA Capital Markets: A Minimum Variance Portfolio Approach
the seemingly contradictory flow of portfolio investments toward the MSE may lie
elsewhere.
Finally, the governments of emerging countries interested in attracting foreign
portfolio investments into their domestic financial markets must be aware of the instability of investment flows, including frequently observed over reactions and seemingly irrational behaviors during crisis episodes. Thus, an obvious related concern
is how to enhance the ability of a government to preserve investor confidence during
such episodes and minimize panic sales or other irrational conducts. This issue opens
up a related research avenue that is centered on the identification and implementation
of economic policies that can achieve such end, representing a wide diversity of interesting questions, which we intend to explore in future studies.
PANOECONOMICUS, 2014, 6, pp. 691-707
705
706
Francisco López-Herrera, Roberto J. Santillán-Salgado and Edgar Ortiz
References
Aggarwal, Raj, and NyoNyo A. Kyaw. 2005. “Equity Market Integration in the NAFTA
Region: Evidence from Unit Root and Cointegration Tests.” International Review of
Financial Analysis, 14(4): 393-406.
Bekaert, Geert, Campbell R. Harvey, and Angela Ng. 2005. “Market Integration and
Contagion.” Journal of Business, 78(1): 39-69.
Bekaert, Geert, Robert J. Hodrick, and Xiaoyan Zhang. 2005. “International Stock Return
Comovements.” National Bureau of Economic Research Working Paper 11906.
Berben, Robert-Paul, and W. Jos Jansen. 2005. “Comovement in International Equity
Markets: A Sectoral View.” Journal of International Money and Finance, 24(5): 832857.
Bollerslev, Tim. 2008. “Glossary to ARCH (GARCH).” School of Economics and
Management, University of Aarhus, Center for Research in Econometric Analysis of
Time Series Research Papers 2008-49.
Cappiello, Lorenzo, Robert F. Engle, and Kevin Sheppard. 2006. “Asymmetric Dynamics
in the Correlations of Global Equity and Bond Returns.” Journal of Financial
Econometrics, 4(4): 537-572.
Ciner, Cetin. 2006. “A Further Look at Linkages between NAFTA Equity Markets.” The
Quarterly Review of Economics and Finance, 46(3): 338-352.
Darrat, Ali F., and Maosen Zhong. 2005. “Equity Market Integration and Multinational
Agreements: The Case of NAFTA.” Journal of International Money and Finance,
24(5): 793-817.
Engle, Robert F., and Kevin Sheppard. 2001. “Theoretical and Empirical Properties of
Dynamic Conditional Correlation Multivariate GARCH.” National Bureau of
Economic Research Working Paper 8554.
Ewing, Bradley T., James E. Payne, and Clifford Sowell. 2001. “Transmission of
Conditional Stock Return Volatility across North American Markets: Evidence from
Pre- and Post-NAFTA.” The International Trade Journal, 15(4): 409-427.
Gilmore, Claire G., and Ginette M. McManus. 2004. “The Impact of NAFTA on the
Integration of the Canadian, Mexican and U.S. Equity Markets.” Research in Global
Strategic Management, 10: 137-151.
Goetzmann, William N., Lingfeng Li, and K. Geert Rouwenhorst. 2002. “Long-Term
Global Market Correlations.” National Bureau of Economic Research Working Paper
8612.
Gradojević, Nikola, and Eldin Dobardžić. 2013. “Causality between Regional Stock
Markets: A Frequency Domain Approach.” Panoeconomicus, 60(5): 633-647.
Grinblatt, Mark, and Sheridan Titman. 2002. Financial Markets and Corporate Strategy.
2nd ed. Irwin: McGraw-Hill.
Kohers, Gerald, Ninon Kohers, and Theodor Kohers. 2004. “A Comparison of the Risk
and Return Characteristics of Developed and Emerging Stock Markets.” Journal of
International Business Research, 3(1): 59-67.
Longin, François, and Bruno Solnik. 2001. “Extreme Correlation of International Equity
Markets.” Journal of Finance, 56(2): 649-676.
López-Herrera, Francisco, and Edgar Ortiz. 2007. “Patterns of Cointegration at the
NAFTA Capital Markets.” Emerging Markets Letters, 1(1): 84-96.
PANOECONOMICUS, 2014, 6, pp. 691-707
Interdependence of NAFTA Capital Markets: A Minimum Variance Portfolio Approach
López-Herrera, Francisco, Edgar Ortiz, and Alejandra Cabello. 2007. “Las bolsas de
valores en el área del TLCAN: un análisis a largo plazo.” Problemas del desarrollo,
38(151): 37-61.
López-Herrera, Francisco, Edgar Ortiz, and Alejandra Cabello. 2008. “NAFTA, Capital
Markets and Structural Breaks: Mexico and Market Integration.” The International
Journal of Finance, 20(4): 4930-4955.
López-Herrera, Francisco, Edgar Ortiz, and Alejandra Cabello. 2009. “Las
interrelaciones de volatilidad y rendimientos entre los mercados de valores del
TLCAN.” Investigación económica, 68(267): 83-114.
Mushtaq, Rizwan, and Syed Zulfiqar Ali Shah. 2014. “International Portfolio
Diversification: United States and South Asian Equity Markets.” Panoeconomicus,
61(2): 241-252.
Suleimann Lemand, Ryan S. 2003. “New Technology Stock Market Indexes Contagion: A
VAR-dccMVGARCH Approach.” Economics Working Paper Archive, Econometrics
0307003.
Téllez Gaytán, Jesús Cuauhtémoc, and Pablo López Sarabia. 2010. “Comovimiento entre
mercados accionarios de América Latina y Estados Unidos: un enfoque de wavelets.”
Economía: teoría y práctica, 32(1): 55-82.
Tse, Yiu Kuen, and Albert K. C. Tsui. 2002. “A Multivariate Generalized Autoregressive
Conditional Heteroscedasticity Model with Time-Varying Correlations.” Journal of
Business and Economic Statistics, 20(3): 351-362.
Tudor, Cristiana. 2011. “Changes in Stock Markets Interdependencies as a Result of the
Global Financial Crisis: Empirical Investigation on the CEE Region.”
Panoeconomicus, 58(4): 525-543.
Wang, Kuan-Min, and Hung-Cheng Lai. 2013. “Which Global Stock Indices Trigger
Stronger Contagion Risk in the Vietnamese Stock Market? Evidence Using a Bivariate
Analysis.” Panoeconomicus, 60(4): 473-497.
Wong, A. S. K., and Peter J. G. Vlaar. 2003. “Modelling Time-Varying Correlations of
Financial Markets.” Netherlands Central Bank, Econometric Research and Special
Studies Department, WO Research Memoranda 739.
Ye, George L. 2014. “The Interactions between China and US Stock Markets: New
Perspectives.” Journal of International Financial Markets, Institutions and Money,
31(C): 331-342.
PANOECONOMICUS, 2014, 6, pp. 691-707
707