Equity Market Integration and Contagion: A Sectoral Perspective

1
Equity Market Integration and Contagion A
Sectoral Perspective

• Kate Phylaktis and Lichuan Xia
• Cass Business School
• London
• 20th May 2008
• Policy Challenges from the Current Crisis
• Brunel University

2
Outline

• Equity market co-movement and contagion
• What do we know about contagion
• Contagion at sector level a different
  perspective
• Data and methodology
• Empirical results
• Summary and conclusions

3
Equity market co-movement and contagion

• There were a series of financial crises in the
  1990s
• 1992 ERM attacks, 1994 Mexican peso collapse,
  1997 Asian crisis, 1998 Russian collapse, 1998
  LTCM crisis, 1999 Brazilian devaluation, 2000 IT
  crisis
• One common feature during the crises
• Markets move more closely together than during
  tranquil times
• The close market co-movement during the crises is
  often referred to as contagion
• Although a precise definition of contagion has
  not been agreed upon, it is generally defined as
  the spread of market shocks mostly from
  downside from one country to another, a process
  observed though co-movement in market prices
  (Dornbusch et al, 2001 Karolyi, 2003)
• The Study contagion is important for
  international diversification
• If contagion prevails in times of crises,
  international diversification is hampered when it
  is mostly needed

4
What do we know about contagion

• The previous literature focuses on market level
  evidence
• They answer the question of whether idiosyncratic
  shocks from one particular market or a group of
  markets are transmitted to the other markets
  during crises.
• The empirical results are mixed
• Contagion was present during crises
• 1987 US market crash King and Wadhwani (1990)
  Bertero and Mayer (1990) Lee and Kim (1993)
• 1994 Mexican crisis Calvo and Reinhart (1996)
• 1997 Asian crisis Baig and Goldfajn (1999)
• No contagion
• The Mexican and Asian crisis no contagion, but
  only interdepedence Forbes and Rigobon
  (2001,2002)
• No contagion during the Mexican crisis, but there
  was contagion during the Asian crisis
• Bekaert, Harvey and Ng (2005)

5
Contagion at sector level a different
perspective (1)

• Studying contagion at sector level is important
• Cross-market contagion study may have bias due to
  the different industrial structures embedded in
  the market returns
• Evidence shows that the global industry factors
  are becoming more important than the country
  specific factors in driving the variation of
  international equity returns (Baca et al, 2000
  Cavaglia, Brightman and Aked, 2000, Phylaktis and
  Xia, 2006)
• Contagion may exist in a group of sectors while
  other sectors are immune from contagion. e.g.
• Banking industry it can even transmit contagion
  (Tai, 2004 Kaminsky and Reinhart, 1999)
• Traded goods sectors are more vulnerable (Forbes,
  2002)
• The importance of industry analysis is also
  highlighted in other studies. e.g.
• the importance of industry volatility relative to
  the market and firm level volatilities (Campbell
  et al, 2001)
• exchange rate movement and industry competition
  (Griffin and Stulz, 2001)
• industry momentum strategies and time-varying
  industry risk premium (Mostowitz and Grinblatt,
  1999 Rouwenhorst, 1999)

6
Contagion at sector level a different
perspective (2)

• The purpose of current paper and its contribution
  to the literature
• The paper intends to explore the equity market
  contagion at the disaggregated sector level, an
  issue which has not been examined in the
  literature. Do unexpected shocks from a
  particular market or group of markets transmit to
  the sectors in other countries?
• The paper also examines the sector level
  integration, thus nesting the empirical studies
  in Carrieri et al (2004), Berben and Jansen
  (2005) and Kaltenhauser (2002, 2003). Our focus
  is on a broader coverage of countries, whereas
  the above studies only deal with several large
  developed countries

7
Data and methodology (1)

• Data
• Weekly sector returns for a set of 29 smaller
  countries in Europe, Asia and Latin America for
  the period of 1/1990 6/2004. 10 broad sectors
  based on FT Actuaries are employed.
• Methodology
• Contagion is defined as excess correlation, i.e.
  correlation over and above what one expects from
  economic fundamentals
• Two-factor (world and regional risks) asset
  pricing model based on Bekaert, Harvey and Ng
  (2005) to account for the fundamentals, and
  contagion is defined by the correlation of model
  residuals.
• The model GARCH (1,1)

8
Data and methodology (2)

• So the betas in the model can test the sector
  level integration, where the integration can be
  either global or regional the residuals of the
  model can test the contagion effect at sector
  level
• Betas (global and regional) we first assume
  betas to be time-invariant to see whether sectors
  are more integrated at the global level or at
  regional level. Then we allow those betas to be
  time-varying by one-year window rolling
  estimation to see whether the integration changes
  over time, particularly during crises
• Residuals any significant correlations between
  residuals indicate the excess correlations beyond
  what is captured in the model, suggest evidence
  of contagion. We run the following regressions
• where
• eij is the sector residuals after the
  time-varying betas have been accounted for,
• eg represents either the US market residuals,
  regional residuals or the residuals of equivalent
  sectors within one region

9
Data and methodology (3)

• When m and n are jointly significantly different
  from zero, it can be interpreted as the overall
  contagion for the full sample period when n is
  significantly different from zero, it can be
  interpreted as additional contagion over a
  particular crisis period
• Sector returns, together with the US and regional
  market returns in GARCH (1,1) model are treated
  as a joint multivariate likelihood function,
  which is estimated in three stages.
• the model for the US market is estimated
• then based on the US estimates, the regional
  market model is examined
• In the final stage, a univariate model is
  estimated sector by sector, conditioning on the
  US and the regional market estimates.
• Knowing that the contagion result is sensitive to
  the correct modelling, we carry out a series of
  specification tests on the model residuals. i.e.
  we test up to the fourth moment of the
  standardized idiosyncratic shocks for each sector
  in each country

10
US and Regional Market Return Model
11
Empirical results Sector level integration (1)

• Sector level integration
• Testing integration hypotheses sector-by-sector
  tests of whether sector returns are globally
  (regionally) priced or locally priced
• Conclusion
• - most sectors are priced at both global and
  regional level, with local information having
  little explanatory power in the return process

12
Empirical results Sector level integration (2)

• Examining whether sectors are more integrated at
  the global or regional level comparison of the
  US and regional betas
• Conclusion
• - Europe and Latin America sectors are more
  strongly integrated at the regional level,
• - Asia sectors are more influenced by the US
  market
• - IT sector is globally integrated, regardless
  of its location

13
Empirical results Sector level integration (3)

• Time-varying integration one-year window rolling
  estimation of global and regional betas
• Conclusion
• - Sector betas in the three regions do change
  over time.
• - The changes of betas dominance from one to the
  other usually occurred during crisis period, a
  possible indication of contagion at sector level
• (Two examples)

14
Empirical results Sector level contagion (1)

• Sector level contagion
• The overall contagion over our entire sample
• Conclusion an overall contagion is found, but
  it varies across regions
• In terms of possible channels, contagion across
  three regions is transmitted via global and
  regional shocks. But in Europe and Asia an
  additional shocks from equivalent sectors within
  the region is identified.
• In terms of magnitude of contagion, the most
  severe contagion comes from regional shocks in
  Europe and Latin America whereas it is mainly
  driven by regional sector shocks in Asia

15
Empirical results Sector level contagion (2)

• Additional contagion during a particular crisis
• Conclusion
• 1994 Mexican crisis nearly half the sectors in
  the three regions contained contagion
• 1997 Asian crisis most sectors in Asia had a
  worsened contagion, whereas the contagion was
  negligible elsewhere

16
Empirical results Sector level contagion (3)

• Additional contagion during a particular crisis
  (cont.)
• Contagion is transmitted in some sectors and not
  in others, explains the mixed results found in
  studies at the market level
• Though contagion might be prevalent at the market
  level, there are still some sectors, which are
  immune from contagion effect implications for
  portfolio diversification
• The Financial Sector (TOTFL) in Latin America
  exhibited additional correlation from the US
  during the Mexican Crisis
• The Financial Sector (TOTFL) in Asia exhibited
  additional correlation from the US during the
  Asian Crisis
• Analysis lends support to the importance of
  financial links through a financial centre, such
  as US, in propagating a crisis
• - Frankel and Schmuckler (1998)
• - Van Rijckeghem and Weder (2001)

17
Summary and conclusions (1)

• Major findings
• Consistent with the rising importance of industry
  effects, sectors have surpassed the country
  borders and become increasingly linked to the
  rest of world.
• Sectors in Europe and Latin America have stronger
  integration at the regional level whereas they
  are more integrated at the global level in Asia.
• However, the integration is time-varying,
  especially during the crisis periods. Many
  sectors are found to switch from regional beta
  dominance to the global one or vice versa during
  the crises
• The heterogeneous performance of sectors across
  regions shows that sectors are less globally
  correlated than we expect and are still subject
  to the regional effects
• The exception is the Information Technology
  sector, which is more globally integrated
  regardless of its geographic location

18
Summary and conclusions (2)

• Contagion does exist at the sector level.
• An overall contagion over our entire sample
  period was found for the majority of sectors in
  Europe, Asia and Latin America. Especially in
  Asia the idiosyncratic shocks are mostly
  transmitted via the equivalent sectors
• During crisis periods contagion is transmitted in
  some sectors and not in others
• During the 1994 Mexican crisis, nearly half the
  sectors in Europe, Asia and Latin America were
  contagious.
• During the 1997 Asian crisis, only sectors in
  Asia were found to contain contagion.
• Established the importance of financial links
  through a financial centre, such as the US, in
  propagating a crisis

19
Summary and conclusions (3)

• Our results confirm the sector heterogeneity of
  contagion and this has implications for
  international diversification
• Selecting portfolios across regions rather than
  within regions is more efficient as regional
  effects remain strong
• There are sectors which are immune from contagion
  during the financial crises and they can still
  provide a tool to diversify risks during the
  crisis periods, e.g. general industries,
  non-cyclical consumer services during the Asian
  crisis.
• In contrast, sectors, such as cyclical consumer
  goods, are contagious worldwide, and choosing
  assets in those contagious sectors should be
  avoided.