Financial crises and the global value premium: Revisiting Fama and French

Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

Contents lists available at ScienceDirect

Journal of International Financial Markets, Institutions & Money j o ur na l ho me pa ge : w w w . e l s e v i e r . c o m / l o c a t e / i n t f i n

Financial crises and the global value premium: Revisiting Fama and French Ehab A. Yamani a,b,∗, Peggy E. Swanson c a b c

DeVry University (Dallas Campus), College of Business and Management, Irving, TX, United States Tanta University, College of Business, Tanta, Egypt University of Texas at Arlington, Arlington, TX, United States

a r t i c l e

i n f o

Article history: Received 28 April 2013 Accepted 19 July 2014 Available online 27 July 2014 Keywords: Financial crisis Contagion Financial integration Global value premium GARCH JEL classification: G01 G12 G15

a b s t r a c t This paper examines the impact of financial contagion resulting from global financial crises based on analyses of the global value premium as represented by thirteen countries. We propose a new model that is a composite of the asymmetric GARCH model and the Fama–French two factor model. Then we investigate behavior of the value premium within crisis periods as well as behavior for pre-crisis, crisis and post-crisis periods. Results show that equity markets become more integrated after financial crises that exhibit global effects but less integrated after crises that exhibit regional effects. Overall findings support the risk story of the global value premium. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Financial crises occur with recurring patterns as evidenced by at least one severe global financial crisis per recent decadethe 1987 stock market crash, the 1997 Asian financial crisis, and the 2007 credit meltdown. A common feature of the crises of the last few decades has been the rapid spread

∗ Corresponding author. Tel.: +1 817 673 6883. E-mail addresses: [email protected] (E.A. Yamani), [email protected] (P.E. Swanson). http://dx.doi.org/10.1016/j.intfin.2014.07.012 1042-4431/© 2014 Elsevier B.V. All rights reserved.

116

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

from one country to others in a process that has come to be known as “contagion”. Starting with the 1987 stock market crisis that began in Hong Kong, global stock markets plummeted one after another in Europe and in the United States. The 1997–1998 Asian crisis began in Thailand with the collapse of the Thai Baht and spread rapidly into neighboring countries. The 2007–2008 financial crisis that hit the world as a result of the implosion of the US mortgage market was followed by a series of collapses in major world markets. The surprising frequency has stimulated extensive research related to crises1 . In this study, our objective is to investigate the effects of financial market crises from two perspectives—the global value premium and equity market integration. The primary purpose is to investigate whether or not the global value premium is a risk factor affecting equity market integration. For many years, there has been a continuing debate about whether the value premium is a risk premium or an anomaly. On one hand, the risk story is based on the rational asset pricing explanation that views the value premium (HML) as a risk premium for a state-variable risk related to ‘relative distress’ (Fama and French, 1993, 1995). On the other hand, the behavioral story views the value premium as an anomaly that might result from overreaction (DeBondt and Thaler, 1985). Following the work of Lakonishok et al. (1994), if global value stocks are fundamentally riskier than global growth stocks, one would expect value stocks to perform more poorly than growth stocks during crises since risk-averse investors should replace risky value stocks with less risky glamour stocks during economic recessions when the marginal utility of consumption is especially high. The rationale is that when a financial crisis occurs in any region in the world with fears of contagion, risk-averse investors rush to quality and liquidity by trying to get rid of high-risk, illiquid securities and replacing them with low-risk liquid securities2 . Based on data from January 1975 to December 2007 for thirteen countries and covering four crisis periods, our results support the risk story for a global value premium. The monthly returns on the international value strategy (i.e., longing value stocks and shorting growth stocks) are negative for most of the sample countries during the four crises. This result means that value stocks consistently perform more poorly than growth stocks during bad times. Given that the monthly returns on the global value strategy reflect the price of the global value premium, our finding that value stocks underperform growth stocks during crises indicates that crises led to a change in the price of the risk factor and therefore countries share “distress risk factors” (proxied by the global value premium) during crisis periods. For the effects of the financial crises on the integration of financial markets, a co-movement of returns investigation of the impact of the Latin American debt crisis and the Asian crisis indicates that international financial markets became less integrated during and after the Latin American debt crisis relative to before the crisis, while financial markets became more integrated during and after the Asian crisis relative to the period before the crisis. The paper contributes to both integration literature and asset pricing literature. To the best of our knowledge, this paper is the first attempt to investigate the impact of financial crises along two dimensions: global value premium and financial integration. Our argument on relating global value premium to financial integration follows from the line of thought in international finance literature that documents the importance of global risk factors in the predictability of international equity returns (e.g., Campbell and Hamao, 1992; Ferson and Harvey, 1993; Nitschka, 2010). If financial

1 The recurring financial crises during the last few decades motivated three strands of crisis literature. One strand investigates the crisis transmission in terms of how financial crises lead to comovements among the international stock markets (e.g., Arghyrou and Kontonikas, 2012; Philippas and Siriopoulos, 2013; Mierau and Mink, 2013). The second strand examines the reasons behind the occurrence of the financial crises (e.g., Melvin and Taylor, 2009; Demyanyk and Hemert, 2011; Bordo and Meissner, 2012). The third category focuses on the relation between crises and market efficiency (e.g., Saffi and Kari Sigurdsson, 2011; Ahmad et al., 2012; Driessen and Hemert, 2012). 2 A striking example that supports the rationale of such an approach is the collapse of Long-Term Capital Management (LTCM). In January 1998, the spread between high-yield corporate bonds and US treasuries was 4 percentage points. LTCM believed that such spread was excessively wide as a consequence of the Asian crisis, and that this spread would narrow when the crisis ended. Consequently, LTCM engaged in ‘market neutral arbitrage’ by longing high-yielding, less liquid bonds and shorting low-yielding, more liquid bonds. By August 1998, the crisis continued and fear spread over the world because of the collapse of the Russian market. Consequently, the spread between US B-rated bonds and high-rated corporate bonds rose from 2 percentage points before the crisis to 5.7 percentage points. This wide spread led to the collapse of the LTCM by September 1998 (Edwards, 1999).

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

117

markets are integrated, the predictability of a country’s excess returns is determined by its world risk exposure and should reflect common movement in international excess returns3 . Thus, the finding of common movement of international value-growth strategies during financial crises could suggest integration of international equity markets because this finding implies that international equity markets are subject to common shocks that move returns on value strategies in similar ways. This common force is the “distress risk factors” (proxied by the global value premium) which reflects changes in risk aversion of a representative agent resulting from the occurrences of financial crises, in the sense of Marcus Alan, 19894 . The finding that value stocks consistently perform more poorly than growth stocks during crises is expected if investors rush to quality during periods of uncertainty, replacing value stocks with growth stocks. In addition, this paper proposes a new international asset pricing model that takes into account market risk, foreign exchange risk, a value premium, a time-varying risk premium, and an asymmetric volatility effect. These are well-known phenomena in the literature that refer to asymmetric responses of return volatility series to bad news and good news. The model is named GJR-GARCH-FF because it is a composite of the asymmetric Sign-GARCH model developed by Glosten et al. (1993) (GJR-GARCH model) and the international version of the Fama and French model (1998). The original FF model is based on two crucial assumptions—that purchasing power parity holds and that the price of risk is constant. The merit of our newly introduced model is that it relaxes these two assumptions by adding the exchange risk premium to the FF model and then incorporating it as the mean equation in the GJR-GARCH model. The results show that the new model provides an attractive representation of countries’ returns since the average intercept of the model is significantly lower than the FF two factor model. The remainder of the paper is organized as follows: data and variables are explained in Section 2, methodology is set forth in Section 3, empirical results are reported and interpreted in Sections 4 and 5, respectively, while Section 6 concludes. 2. Data and crisis periods To measure global crisis impacts, data from thirteen developed financial market countries serve as proxies for global markets, independent of the origination of the crisis. The countries are: Australia, Belgium, France, Germany, Hong Kong, Italy, Japan, Netherlands, Singapore, Sweden, Switzerland, the UK and the US. The study period is January 1975 to December 2007. This time span is long enough to study the effects of the risk explanation for the global value premium surrounding four global crisis periods. A persistent challenge in examining financial crises is to identify the precise beginning and ending of a crisis, that is to pinpoint the source of the crisis (identify the major triggering event). According to the Federal Deposit Insurance Corporation (FDIC), 1997, the Latin American debt crisis officially started on August 15, 1982 when Mexico declared its inability to pay its debt obligations. By the end of 1983, the crisis management efforts could be called successful, since Mexico had fulfilled its interest payments and the large international banks did not collapse. Therefore, we define the ‘debt crisis period’ from August 1982 to December 1983. According to the International Monetary Fund (1998), the Asian crisis began July 2, 1997 when the bank of Thailand devalued the baht by roughly 20%. On March, 1999, the Asian crisis had ended when the IMF approved a $1 billion increase in its emergency loan package for Indonesia, and the DJIA closed above the 10,000 level for the first time in its history. Therefore, we

3 Financial integration can be defined in terms of either a global market portfolio or Purchasing Power Parity (PPP). The global market portfolio argues that international markets are integrated if all financial assets yield the same risk-adjusted expected returns to investors the world over (assuming that investors do not hedge exchange rate risks and the single relevant source of risk is the market portfolio). According to PPP, two financial markets are integrated if securities are priced identically and the exchange risk premium is zero. 4 We do not want to overstate the strength of this evidence since risk aversion is not the only force that makes international returns move is similar ways. Other potential common factors include changes in volatility (e.g., Fama and French, 1989) or exogenous shifts in the noise traders demand (e.g., Shiller, 1984). However, examining these competing explanations is beyond the scope of this paper.

118

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

define the ‘Asian crisis period’ from July 1997 to March 1999. The within crisis period analysis covers the Latin American debt crisis in 1982–1983 and the Asian crisis in 1997–1998, plus the European Exchange Rate Mechanism (ERM) crisis in 1992–1993 and the September 11, 2001 attack. However, pre-crisis and post-crisis periods can be studied for only two of the four crises because each sub-period must have a sufficient number of observations for analysis while avoiding overlapping observations. In order to meet these criteria, only sub-periods of two of the crises can be studied: the Latin American debt crisis in 1982–1983 and the Asian crisis in 1997–1998. The resulting decomposition of the sample period of the study for each crisis and each sub-period is: Crisis

Pre-crisis period

Crisis period

Post-crisis period

Latin American debt crisis Asian crisis

Aug 1979–July 1982 July 1994–June 1997

Aug 1982–Dec 1983 July 1997–Mar 1999

Jan 1984–Dec 1986 April 1999–Mar 2002

Three important currencies—the British pound, German mark/euro5 and Japanese yen proxy for foreign exchange risk while the monthly market return is the value weighted average of returns for firms in each country6 . The return on one-month US Treasury Bills is used as the risk-free rate, because the analysis is from the perspective of a US investor. All of the exchange rate data is obtained from COMPUSTAT, while the returns on individual countries’ portfolios, the global market portfolio, value and growth portfolio, and the risk-free rate are obtained from Kenneth French’s homepage7 . 3. Methodology 3.1. Supporting rationale Two primary approaches for measuring financial market integration are market beta and purchasing power parity (PPP). The more popular is beta as measured by the CAPM that assumes investors do not hedge against exchange rate risks and the single relevant source of risk is the market portfolio. However, studies differ in terms of the definition of the market portfolio as to whether it is a global, local or hybrid portfolio. The global CAPM (the integration model) is based on the logic that the international markets are perfectly integrated, such that all financial assets yield the same riskadjusted expected returns to investors the world over. In this case, the global portfolio risk is the only risk considered. The domestic CAPM (the segmentation model) assumes that the local market is perfectly segmented from the world market and, therefore, the capital markets may demonstrate substantially different risk-return trade-offs. In this case, the local market portfolio will be the source of risk and should replace the returns for the global index in the CAPM formulation. Many studies find evidence of market segmentation (e.g., You et al., 2006; Bruner et al., 2008; Koedijk et al., 2002; Harris et al., 2003). A more realistic approach than the two above versions is a two-factor Hybrid CAPM (Mid-Segmentation Model) that assumes capital markets are neither perfectly integrated nor perfectly segmented, so that global and domestic risk factors should be priced separately. This approach was originally launched by Errunza and Losq (1985) and Errunza et al. (1992), and empirically supported by Bruner et al. (2008). Finally, the Fama–French two factor model (1998) is an alternative version of CAPM that says that international returns are explained by the value premium and the global market

5 On January 1, 1999, the euro was adopted by eleven countries – including Germany – as their official currency. Therefore, we use DM during the first part of the period and we then use the Euro as a proxy for the DM during the latter part of the study. 6 As a robustness test, we applied the trade weighted US dollar index as another proxy for currency exchange risk, and we found similar results as those using the three currencies. The traded weighted US dollar index is a weighted average of the foreign exchange value of the U.S. dollar against a set of index currencies that circulate widely in the Euro Area, Canada, Japan, United Kingdom, Switzerland, Australia, and Sweden. Detailed results are available from the authors upon request. 7 http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data library.html.

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

119

premium. They empirically test their model and find that for thirteen countries the model fits the data well. Rather than being defined in terms of the market portfolio, financial integration also can be defined in terms of PPP based on the ‘Law of One Price’. The idea behind PPP is very simple: once converted to a common currency at current exchange rates, national prices should be equal. According to PPP, therefore, two financial markets are integrated if securities are priced identically and the exchange risk premium is zero. Any deviations from PPP cause national investors to perceive different exchange rate adjusted returns from the same security and thus the exchange risk premium should be considered. Solnik (1983) was the first to explore this issue by arguing that the CAPM should contain an exchange risk premium in addition to the market risk premium when PPP is violated. Wu (2008) finds that the international version of the FF model does not fit the data well, and he documents that the international CAPM with exchange risk is the best international model for forecasting compared to the CAPM and the FF model.

3.2. Model development The cited literature reveals two primary weaknesses in modeling the first and second moments. First, there is no clear-cut determination of the best international asset pricing model for explaining international returns. Asset pricing models that use beta as an integration measure differ in many regards including whether to include the home country or the global index as the mean-variance efficient portfolio in the regression, whether to include exchange rate risk as an additional source of risk, and whether to incorporate the value premium as captured by Fama and French (1998) in the CAPM regression. Second, a common weakness in studies that apply CAPM in the international setting is that they are based on the constant price of risk (CRP) assumption inherent in CAPM. However, a time-varying risk premium on equities is widely documented in the literature. This means that investors’ expectations about future security returns are conditional on all available information, and that modeling integration of financial markets without taking into account this time variation will yield imprecise results. Therefore, the CRP assumption should be relaxed to allow the price of market risk to vary over time. In light of these two methodological observations in the international CAPM literature, we propose a new model that takes into account market risk, foreign exchange rate risk, the value premium, and a time-varying risk premium. In particular, the FF two factor model assumes that there is no deviation from PPP and the price of risk is constant. This paper relaxes these two assumptions. To this end, we add two modifications to the original FF two-factor model: first, we assume that PPP does not hold by adding the foreign exchange risk premium as another risk factor in addition to the market risk premium and the value premium. Second, in order to relax the constant risk price (CRP) assumption, we incorporate the new model as the mean equation in the asymmetric GJR-GARCH model. Although conditional heteroscedasticity models appear to be among the best that are currently available, there is a major drawback of using first generation GARCH models because they are said to be symmetrical, due to the quadratic specification used for the conditional variance (i.e., the error term is squared). Therefore, volatility will be a function only of the innovation’s magnitude, since the lagged shock will have the same effect on the present volatility whether the lagged shock is positive or negative (i.e., neutral impact). This symmetrical nature of the traditional GARCH models makes them not well suited for capturing asymmetric volatility effects8 (e.g., Glosten et al., 1993). To handle the asymmetries in the conditional variance, we use the asymmetric Sign-GARCH model of Glosten et al., 1993 (GJR-GARCH model) that allows for different reactions of volatility to the sign of past innovations.

8 The asymmetric volatility phenomenon (AVP) is a well-known phenomenon where periods of crisis (when residuals are negative) cause the level of market volatility to increase more than it increases in periods of relative calm (when residuals are positive).

120

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

We call the newly created model GJR-GARCH-FF, because it is a composite of both the GJR-GARCH model with the FF model. Mathematically, it can be represented as follows:





ri,t − rf,t = ˇ0 + ˇ1MKT rg,t − rf,t + ˇ2HML (H − L) JY

+ˇ3BP (Pound) + ˇ4DM (Mark) + ˇ5 (Yen) + εi,t







2 εi,t  εi,t−1 , εi,t−2 , . . . ∼N 0, i,t 2 =˛ + i,t 0

q 

˛1 ε2i,t−1 +

i=1

where :

Si,t−1 =



p 



2 ı1 i,t−1 +

i=1

1, if εi,t−1 < 0 0,



r 

(1) 1 Si,t−1 ε2i,t−1

i=1

otherwise

The model consists of two equations: the mean equation and the variance equation. In the mean equation, the excess monthly market returns of country  i are  computed by subtracting the risk-free rate (rf ) from the monthly market return for that country ri,t . The excess return of a country is a function of three risk factors: global market risk premium (rg,t − rf ) where (rg,t ) is the return on the global equity market portfolio, the value premium (H − L) measured by the return difference between the high and low BE/ME portfolio for each country, representing the international version of the distress factor in the FF two-factor model, and the exchange risk premium proxied by three important currencies—the British pound, German mark, and Japanese yen, using the US dollar as the reference currency. The ˇ0 intercept is the mean   of the excess monthly market returns of country i conditional on past information εi,t−1 , εi,t−2 , . . . . If our model successfully explains international returns, the intercept ˇ0 should be statistically indistinguishable from zero. The ˇ1MKT coefficient measures the sensitivity of the market return of country i to “global market risk”, proxied by the return of a global market index. The size and significance of ˇ1MKT implies the extent of integration of country i. The ˇ2HML coefficient shows the sensitivity of the returns of country i to “distress risk”, proxied by the difference between the returns of global high and low book-to-market portfolios as estimated by Fama and French (1998). The factor JY loadings ˇ3BP , ˇ4DM , and ˇ5 measure the sensitivity of the returns of country i to the “foreign exchange risk”, proxied by the British pound, German mark/euro and Japanese yen. Following Fama and French (1998), we do not include size premium (the difference between the returns on diversified portfolios of small and big stocks) because the data for the twelve countries (other than US) is from Morgan Stanley’s Capital International (MSCI) which primarily includes large firms that represents 80 percent of a market’s invested wealth. It is clear from the above formulation that the mean equation is the international Fama and French two-factor model with currency risk. By incorporating such equation into the GJR-GARCH model, the new model becomes the time-varying empirical counterpart to the FF model (with currency risk). 2 )) as a function of The variance equation expresses the current volatility (measured by variance (i,t four factors: the mean volatility (˛0 ), news about volatility from   the previous period measured as the lag of the squared residual from the mean equation ε2i,t−1 (ARCH term), the last period’s variance





2 i,t−1 (GARCH term), and the asymmetric volatility term (Si,t−1 ε2i,t−1 ). Si,t−1 is a dummy variable that equals one when the residual is negative (i.e., bad news) and zero otherwise. This specification allows the conditional variance to differ on crash days, since the effect of lagged shock on current only, as in standard volatility is a function of its magnitude and its sign rather than  its magnitude  GARCH models. Specifically, volatility is affected by two terms ˛ + ı when the residual is positive





(i.e., good news), while it is affected by three terms ˛ + ı +  when the residual is negative (i.e., bad news). If  > 0 and significant, volatility will be higher for a given negative shock than for a positive shock of equal magnitude. The size and significance of ˛, ı, and  imply the existence of ARCH,  2 GARCH,  and volatility asymmetry effects, respectively. A model with ‘q’ lags of (ε2i,t−1 ) ‘p’ lags of i,t−1 and ‘r’ lags of (Si,t−1 ε2i,t−1 ) is labeled GJR (p, q, r), with lag structure in the conditional variance equation based on Akaike (AIC) and Bayesian (BIC) criteria.

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

121

4. Empirical results 4.1. GJR-GARCH-FF model estimation results The objective of this subsection is to evaluate the ability of the GJR-GARCH-FF model to explain international returns. As a preliminary test, we run the ARCH LM Test, which is a Lagrange multiplier (LM) test for autoregressive conditional heteroskedasticity (ARCH) in the residuals since ignoring ARCH effects may result in a loss of efficiency. The results show that the values of Lagrange Multiplier (LM) are insignificant for all countries (except France). This result leads to rejection of the null hypothesis (i.e., the residuals are homoscedastic), indicating that the variance equation in the GJR-GARCH model has been correctly formulated9 . In the second test, we use the standard GJR-GARCH (p, q, r) model (under the restriction that ˇ1 = ˇ2 = ˇ3 = ˇ4 = ˇ5 = 0) to model the monthly excess return for individual countries to examine whether ARCH is a manifestation of monthly time dependence in the rate of information arrival to the country’s stock market. We test several GARCH (p) terms, ARCH (q), and asymmetric (r) terms and find that the most appropriate model is GJR-GARCH (3, 1. 1): ri,t − rf,t = ˇ0 + εi,t ;







2 εi,t  εi,t−1 , εi,t−2 ., . . . ∼N 0, i,t



2 = ˛ + ˛ ε2 2 2 2 2 i,t 0 1 i,t−1 + ı1 i,t−1 + ı2 i,t−2 + ı3 i,t−3 + 1 Si,t−1 εi,t−1



where : Si,t−1 =

1, if εi,t−1 < 0 0,



(2)

otherwise

Table 1 sets forth results of estimating Eq. (2) and shows that most of the estimated coefficients of the ARCH term (˛1 ) and the three GARCH terms (ı1 ,ı2 ,ı3 ) are highly significant at 1% for almost all the countries. Moreover, the asymmetry parameters are statistically significant for eight of the thirteen included countries, validating the presence of the asymmetric volatility effect. This indicates that GJR-GARCH is an acceptable representation of the time-varying nature of return volatility for our sample countries. Table 2 sets forth the intercepts and estimated coefficients for the model (Eq. (1)) based on the entire sample period (January 1975–December 2007). The results show that the average intercept ˇ0 of the GJR-GARCH-FF model is low in magnitude and insignificant. The ˇ1MKT coefficient measures the sensitivity of the market returns of country i to the “global market risk” and is proxied by the return of a global market index. The size and significance of ˇ1MKT indicates the extent of integration of country i. As mentioned in Section 3, the global CAPM is based on the logic that international markets are perfectly integrated, such that all financial assets yield the same risk-adjusted expected returns to investors the world over. Table 2 shows that the ˇ1MKT coefficients are statistically significant at the 1% level for all thirteen countries. That suggests integration of international equity markets during the sample period. The ˇ2HML coefficient shows the sensitivity of the monthly returns of country i to the “distress risk”, proxied by the difference between the returns of global high and low book-tomarket portfolios (H − L). The ˇ2HML coefficients are significant in six of thirteen countries—Australia, JY

Belgium, Germany, Singapore, Switzerland, and UK. The factor loadings ˇ3BP , ˇ4DM , and ˇ5 measure the sensitivity of the returns of country i to the “foreign exchange risk”, as proxied  by the British pound,  JY German mark/euro and Japanese yen. The foreign exchange factor loadings ˇ3BP , ˇ4DM , and ˇ5 are statistically significant in four, four, and seven out of thirteen countries, respectively.

4.2. Fundamental riskiness of the global value premium Although there is ample research on the US value premium, less is known about the global value premium and its source. In this section, we first examine the performance of the global value

9

Complete results are available from authors upon request.

122

Table 1 Maximum likelihood estimates of GJR-GARCH (3, 1, 1) model—without risk factors entire period (Jan 1975–December 2007). ˛1

ı1

ı2

ı3

1

Australia Belgium France Germany Hong Kong Italy Japan Netherlands Singapore Sweden Switzerland UK USA

0.0404*** (0.0114) 0.1913*** (0.0506) 0.0426*** (0.0184) 0.1621** (0.0688) 0.2192*** (0.0662) 0.0177*** (0.0002) 0.03112* (0.0192) −0.0379 (0.0255) 0.0844*** (0.0177) 0.1602*** (0.0552) 0.3084*** (0.0938) 0.0788*** (0.0247) −0.0620 (0.0637)

2.3619*** (0.0404) 0.0942** (0.0446) 2.0433*** (0.0443) 0.2898 (0.2432) 0.5077*** (0.1672) −0.7249*** (0.0107) 1.1706*** (0.0375) 0.7928*** (0.0800) 0.8196*** (0.0261) 0.7983*** (0.0995) −0.3907*** (0.1207) 1.7593*** (0.046262) 0.0852 (0.1927)

−2.2783*** (0.0706) −0.0830** (0.0448) −1.8961*** (0.0770) 0.4931* (0.2876) 0.8461*** (0.0373) 1.0052*** (0.0017) −1.1580** (0.0525) −0.8175*** (0.1008) 0.8737*** (0.0261) −0.7519*** (0.1049) 0.6247*** (0.0701) −1.6134*** (0.0951) 0.3766** (0.1690)

0.8838*** (0.0417) 0.7013*** (0.0826) 0.7690*** (0.0498) −0.0511 (0.3052) −0.5112*** (0.1542) 0.7341*** (0.1086) 0.8761*** (0.0453) 0.6258*** (0.0898) −0.7652*** (0.0143) 0.5746*** (0.1133) 0.3168** (0.1677) 0.7090*** (0.0550) 0.1134 (0.1802)

−0.0356*** (0.0110) 0.0703 (0.0693) 0.0630** (0.0259) −0.0107 (0.0674) −0.1616*** (0.0595) 0.0035 (0.0077) 0.0768** (0.0309) 0.1665*** (0.0555) −0.0597*** (0.0186) 0.0931 (0.0791) −0.1953** (0.0852) 0.0400 (0.0288) 0.3508*** (0.1300)



ri,t − rf,t = ˇ0 + εi,t ; εi,t  εi,t−1 , εi,t−2 , . . . The table presents coefficient estimates of the GJR-GARCH (3, 1, 1) model:





2 ∼N 0, i,t



2 2 2 2 i,t = ˛0 + ˛1 ε2i,t−1 + ı1 i,t−1 + ı2 i,t−2 + ı3 i,t−3 + 1 Si,t−1 ε2i,t−1





1, if εi,t−1 < 0 0, otherwise The model consists the excess monthly market returns of country i are computed by subtracting  of two equations: the mean equation and the varianceequation.  In the mean equation,  where :

the risk-free rate



rf,t

from the monthly market return for that country

information εi,t-1 , εi,t−2 , . . . GARCH terms





Si,t−1 =

ri,t . The intercept

ˇ

is the mean of the excess monthly market returns of country i conditional on past

0  2   . The variance equation expresses the current volatility i,t of country i as a function of four factors: mean volatility (˛0 ), one ARCH term ε2i,t−1 , three      2 2

2 2 i,t−1 , i,t−2 , i.t−3 , and the asymmetric volatility term

Si,t−1 εi,t−1 .The size and significance of (˛1 ) and

ı1 , ı2 , ı3

implies the existence of the ARCH and GARCH

effects, respectively. Si,t−1 is a dummy variable that is equal one when the residual is negative (i.e., bad news), and zero otherwise. If 1 > 0, this indicates that volatility will be higher for a given negative shock than for a positive shock of equal magnitude. * Significant at 10% level. **

Significant at 5% level.

***

Significant at 1% level.

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

Country name

Table 2 Maximum likelihood estimates of the GJR-GARCH-FF model: entire period (Jan 1975 December 2007). ˇ0

ˇ1MKT

[Australia Belgium France Germany Hong Kong Italy Japan Netherlands Singapore Sweden Switzerland UK USA

0.0780 (0.25665) 0.3583* (0.1942) 0.1062 (0.1509) 0.2472 (0.1981) 0.3869 (0.3411) 0.0329 (0.2709) −0.2254 (0.1673) 0.2165 (0.1616) 0.1539 (0.2690) 0.3766 (0.2439) 0.1671 (0.1895) 0.0661 (0.1289) 0.2332 (0.1799)

0.9643*** 0.7842*** 1.0488*** 0.8841*** 0.9744*** 0.9720*** 1.1148*** 0.8727*** 0.7769*** 0.9607*** 0.6878*** 0.8400*** 0.6063***

(0.0475) (0.0354) (0.0369) (0.0394) (0.0684) (0.0645) (0.0302) (0.0339) (0.0535) (0.0521) (0.0336) (0.0307) (0.0345)

The table presents coefficient estimates of the GJR-GARCH-FF model:   ri,t − rf,t = ˇ0 + ˇ1MKT +ˇ3BP

(Pound) +

ˇ4DM

rg,t − rf,t

(Mark) +

JY ˇ5

+ ˇ2HML (H − L)



(Yen) + εi,t ; εi,t  εi,t−1 , εi,t−2 , . . .

2 2 2 2 i,t = ˛0 + ˛1 ε2i,t−1 + ı1 i,t−1 + ı2 i,t−2 + ı3 i,t−3 + 1 Si,t−1 ε2i,t−1 ;





ˇ3BP

ˇ4DM

ˇ5

0.3035*** (0.0874) 0.1209** (0.0605) −0.0085 (0.0487) 0.1217** (0.0605) 0.1623 (0.1124) 0.0887 (0.0882) −0.0484 (0.0486) 0.0550 (0.0428) 0.4051*** (0.1007) 0.0879 (0.0786) 0.1009* (0.0619) 0.1142*** (0.0411) −0.0324 (0.0404))

−13.2677 (14.9505) 11.4240 (8.8093) −15.8766 (10.6198) −2.7082 (8.2640) −2.7809 (16.6076) −0.7660 (13.1864) −26.4086*** (9.2015) 1.9586 (6.6565) −11.3974 (15.1652) −24.3429* (12.7325) 6.6354 (8.6565) 38.8402*** (8.9922) −21.8383** (10.5614)

12.1657 (9.4389) 4.4949 (2.9505) 12.7131*** (2.2681) 14.4222*** (2.0880) −9.9128 (15.0074) 2.9652 (9.6852) −6.0782* (3.4905) −4.0104 (3.6234) −8.1434 (13.2926) 13.6049 (11.4274) 5.4241** (2.7425) 1.2024 (5.1374) 5.2890 (8.4082)

−8.4258 (10.0956) −7.1480 (6.2745) −18.6552*** (6.4500) −18.0845** (7.2382) 8.6519 (15.6419) −4.0444 (10.7835) 32.1272*** (6.5921) −20.3498*** (5.9884) 7.9714 (10.9933) −25.6233*** (10.0470) −0.5114 (7.0024) −13.7862** (5.4624) −22.2785*** (8.4195)

2 ∼N 0, i,t



Si,t−1 =

where :

JY

ˇ2HML



1, if εi,t−1 < 0 0, otherwise



The model consists of two equations: the mean equation and the variance equation. In the mean equation, the excess monthly market returns of country i



ri,t − rf,t



are a function of

three risk factors: global market risk premium (rg,t − rf ) where (rg,t ) is the return on the global equity market portfolio, the value premium (High BE/ME–Low BE/ME) for each country, and the exchange risk premium proxied by the British pound, German mark, and Japanese   yen, with the US dollar as the reference currency. We use the German markuntil January 1, 1999

2 and the euro thereafter. The variance equation estimates the current volatility i,t









of country i as a function of four factors: mean volatility (˛0 ), one ARCH term ε2i,t−1 , three GARCH

2 2 2 terms i,t−1 , and the asymmetric volatility term Si,t−1 ε2i,t−1 . The standard errors are in parentheses. , i,t−2 , i.t−3 * ** ***

Significant at 10% level. Significant at 5% level. Significant at 1% level.

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

Country name

123

124

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

strategies (HML) during  financial crises. Then we examine the country’s risk exposure to the global value premium ˇ2HML before, during and after financial crises. 4.2.1. Global value premium analysis: Short-term effects For many years, there has been a continuing debate about whether the value premium is a risk premium or an anomaly. On one hand, the risk story is based on the rational asset pricing explanation that views the value premium (HML) as a risk premium for a state-variable risk related to ‘relative distress’. This means that low (high) BE/ME is typical of strong (weak) firms with persistently high (low) earnings (Fama and French, 1993, 1995). On the other hand, the behavioral story views the value premium as an anomaly that might result from overreaction (DeBondt and Thaler, 1985; Lakonishok et al., 1994). Following the approach of Lakonishok et al. (1994) where a fear of market breakdown causes riskaverse investors to rush to quality and liquidity, we examine the performance of the global value premium (high BE/ME–low BE/ME) during financial crises for the thirteen countries in the sample. The rationale is that if value stocks are fundamentally riskier than glamour stocks, then risk-averse investors should replace risky value stocks with less risky glamour stocks during economic recessions when the marginal utility of consumption is especially high. If global value stocks are fundamentally riskier, therefore, one would expect a negative return for the value strategies during financial crises. A common problem in empirical research of financial crises is that results may be affected by the choice of window length. That is, our choice of duration of three years before and three years after the crises may have an impact on the results. Therefore, we report month-by-month returns on the highest BE/ME (value) relative to the lowest BE/ME (growth) portfolios for four crises: the Latin American debt crisis (1982–1983), the ERM crisis (1992–1993), the Asian crisis (1998–1999), and the September 11, 2001 attack. 4.2.1.1. Latin American debt crisis. Although several of the world’s largest banks faced the prospect of major loan defaults and failure as a consequence of the crisis, the fear of a widespread banking collapse in creditor countries was concentrated primarily in the USA and Japan10 . Table 3 reveals that monthly returns on the US value strategy are negative in the three months following August 1982, the starting month of the crisis. In addition, the average value premium during the entire crisis period (August 1982–December 1983) in the USA and Japan is zero and −0.68%, respectively. The value premium of −1.15 for Hong Kong may relate to Hong Kong’s close relationship to Japan. Although the European banks had less exposure to third world lending than did the U.S. and Japanese banks, the value strategy in August and September of 1982 yielded negative returns in the UK, Germany, and Italy where the major European banks are located. 4.2.1.2. ERM crisis. Table 4 shows the month-by-month returns of the value strategy during the ERM crisis11 from September 1992 until December 1993, and reflects the fear of investors during the period. For September 1992, Italy earned the highest negative return (−9.98%) followed by Sweden with a return of −9.20% for the value strategy, reflecting speculation against the Italian lira. On September 11, 1992 the Italian lira was the first victim of the crisis as it was subject to severe speculative attacks.

10 The US banks’ exposure was the highest because the major debtors were concentrated in Latin America. The largest Latin American countries (Mexico, Brazil, Venezuela, and Argentina) owed the eight largest US banks $37 billion that constituted 147% of their capital and reserves at the end of 1983. The size of such debt coupled with the overexposure (lending in excess of capital assets) gave rise to a fear of financial collapse to the largest American banks such as Citibank, Bank of America, Chase Manhattan, and Morgan Guaranty (Federal Deposit Insurance Corporation (FDIC), 1997). The Japanese banks also had a similar rate of debt exposure. The exposure level of the Bank of Tokyo (that handled $753 million as the leading international bank in forming syndicate loans by 1981 in Mexico) was 83% of its capital, and the exposure of Long-Term Credit Bank of Japan and Mitsui Bank was 53.6% and 28.3%, respectively (Katada, 2001). 11 The ERM crisis of 1992–1993 is a crisis of the exchange rate system. A primary reason for the crisis is the difference in economic strength of members of ERM. Germany witnessed a strong economy and was concerned about domestic inflation and raised its interest rate to 8.7%. The UK and Italy, however, were suffering from recession and severe budget deficits, forcing them to adopt stimulative policies by reducing interest rates. Such contradiction dramatically widened the interest rate differentials between Germany and other ERM members.

08/82 09/82 10/82 11/82 12/82 01/83 02/83 03/83 04/83 05/83 06/83 07/83 08/83 09/83 10/83 11/83 12/83 Average

Australia

Belgium

France

Germany

HK

Italy

Japan

Neth.

Sing.

Sweden

Switz.

UK

USA

Australia

−1.46 −1.09 −2.84 −0.95 0.96 −5.26 3.89 2.44 5.44 −2.17 1.89 6.61 4.48 7.52 1.89 3.65 −3.45 1.27

−1.58 −0.39 3.2 −2.76 0.92 2.75 3.7 1.54 −4.43 2.44 −3.31 5.85 1.61 5.16 5.19 −8.37 8.59 1.18

1.65 −2.97 5.27 3 4.23 9.87 −3.55 −1.26 7.55 4.07 3.5 −6.42 15.83 −8.62 −1.29 0.46 −0.03 1.84

−0.24 −4.55 −4.94 −3.47 5.69 0.97 4.02 5.7 −1.1 −4.39 0.72 4.69 0.42 2.36 1.58 2.62 −2.11 0.47

6.93 6.04 −1.19 2.6 5.25 −2.64 4.97 −8.18 −1.54 −8.81 0.45 −3.56 −8.15 −7.46 −6.15 −0.98 2.82 −1.15

−4.58 −5.86 −3.83 2.77 −5.51 −7.99 9.65 19.62 −2.13 −2.33 −7.83 2.02 2.56 1.59 −4.44 6.54 4.28 0.27

−0.73 −1.92 −2.34 −2.22 2.76 3.16 4.82 −2.61 −3.44 −6.34 −4.04 −4.76 2.6 0.42 2.36 −2.71 3.42 −0.68

−6.31 6.21 5.56 5.96 −7.07 −8.21 12.94 −3.96 −7.02 6.93 0.74 2.18 −3.84 −1.45 3.12 1.62 −1.19 0.37

1.59 12.15 9.61 −0.49 −0.95 11.61 3.24 2.45 −2.74 0.17 −1.2 −1.23 −2.93 4.15 −2.51 −0.6 −1.36 1.82

2.68 −4.07 −7.99 −16.97 −0.59 15.48 −0.89 −1.48 −10.46 −1.61 5.69 3.33 6.35 0.33 0.45 2.24 6.61 −0.05

0.29 −0.64 0.79 −5.31 1.46 4.44 1.58 0.8 4.71 −4.1 1.72 2.74 −2.74 2.51 5.64 −3.84 −2.28 0.46

−4.18 −12.65 9.49 0.79 1.89 8.76 3.3 3.94 1.42 1.41 3.62 2.23 3.55 −0.12 2.25 3.4 6.92 2.12

1.1 −1.55 −3.52 −3.55 0.11 1.79 −1.62 0.67 1.07 −0.87 −4.79 4.41 5.81 −1.38 3.65 −1.4 0.12 0.00

−1.46 −1.09 −2.84 −0.95 0.96 −5.26 3.89 2.44 5.44 −2.17 1.89 6.61 4.48 7.52 1.89 3.65 −3.45 1.27

The table reports the difference in monthly returns between the highest BE/ME (value) and the lowest BE/ME (growth) portfolios during the Latin American debt crisis period (August 1983–December 1983). The negative return means that value stocks performed worse than growth stocks during the crisis period. The last row reports the arithmetic mean across the entire period for each country. Data are obtained directly from French’s website.

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

Table 3 Month-by-month returns during the Latin American debt crisis period (August 1982–December 1983): value–growth.

125

126

09/92 10/92 11/92 12/92 01/93 02/93 03/93 04/93 05/93 06/93 07/93 08/93 09/93 10/93 11/93 12/93 Average

Australia

Belgium

France

Germany

HK

Italy

Japan

Neth.

Sing.

Sweden

Switz.

UK

USA

3.48 6.48 6.28 −6.82 1.31 −2.31 1.66 5.06 2.47 1.45 6.89 3.12 3.07 −3.3 −2.6 −4.84 1.34

0.43 −5.07 −2.77 5.87 2.55 6.51 2.81 −1.34 −0.02 −1.4 4.53 4.66 1.66 1.54 2.65 −0.33 1.39

1.13 1.65 −2.08 3.27 0.75 4.04 −4.91 1.25 0.35 3.71 5.11 −1.67 −2.72 −1.04 −2.35 0.02 0.41

−4.61 −3.1 0.54 2.84 3.04 −0.62 0.16 −1.83 −0.7 5.4 −1 1.62 0.84 0.19 −0.32 5.34 0.49

5.74 −2.91 2.2 −7.16 2.22 1.53 1.38 2.21 −2.73 4.88 −7.58 −8.43 −1.22 4.15 5.57 13.48 0.83

−9.98 2.13 2.22 1.97 0.92 12.46 −0.03 −0.54 −8.18 0.7 3.44 −1.19 −7.35 −1.55 2.98 5.57 0.22

−1.13 3.48 1.33 1.38 −1.29 −2.39 6.06 3.07 −2.69 −2.61 3.12 0.79 −0.6 3.31 −1.81 7.2 1.08

−5.08 −16.93 0.59 −1.6 8.78 −11.8 11.76 2.78 7.82 3.05 7.59 4.46 0.96 −2.79 −1.34 −0.47 0.49

−0.76 −2.99 2.03 0.22 −4.63 8.9 0.17 7.34 −2.2 −3.88 2.59 −5.4 8.93 0.21 −6.47 29.4 2.09

−9.2 −14.94 15.38 −6.54 15.09 −5.13 −3.87 −5.78 4.36 −1.39 7.67 −0.34 −17.15 12.41 15.98 7.21 0.86

−8.81 −2.77 −8.87 −4.88 15.81 5.22 1.17 1.21 −2.23 −3.32 9.62 −3.15 −5.91 −1.55 0.4 4.5 −0.22

0.13 2.4 0.09 5.55 2.84 4.42 5.17 1.84 −0.62 2.31 2.6 −1.11 2.66 −1.84 1.41 2.02 1.87

−0.43 −1.67 −0.36 2.76 6.37 5.28 1.07 3.6 −3.29 4.02 3.62 −0.39 −0.19 −1.91 −1.4 0.9 1.12

The table reports the difference in monthly returns between the highest BE/ME (value) and the lowest BE/ME (growth) portfolios during the ERM crisis period (September 1992–December 1993). The negative return means that value stocks performed worse than growth stocks during the crisis period. The last row reports the arithmetic mean across the entire period for each country. Data are obtained directly from French’s website.

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

Table 4 Month-by-month returns during the ERM crisis (September 1992–December 1993): value–growth.

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

127

On September 17, the lira was withdrawn from the ERM after it fell below its ERM floor. The speculative attacks turned to Sweden’s currency, forcing the Riksbank to raise its lending rate on September 8 to 24% and then 75% the next day (Sevilla, 1995). In contrast to the Italian lira, the French franc remained a strong currency. However, it was subject to massive speculative sales in July–August 1993, but by the end of 1993 it returned to levels close to those prevailing before the crisis. These facts are in line with the results in Table 4 where French data reveals a positive return for the first two months of the crisis, but the return on the value premium turns negative from August 1993 until November 1993, the same period as the franc devaluation due to massive speculative sales in July–August 1993. Giavazzi and Giovannini (1989) point out that the US dollar tends to be weak in foreign exchange markets when the EMS is unstable. The ERM crisis mirrored such regularity, since the US dollar reached its lowest value in September 2, 1992 which is the official starting month of the crisis (Edison and Kole, 1995). As Table 4 shows, it seems that fear spread from the European countries at that time to the US because the value strategy in the US yields negative returns during the early stage (September–November 1992) and the final stage (August–November 1993) of the ERM crisis. 4.2.1.3. Asian crisis. Although data from the Latin American debt crisis and the ERM crisis supports the risk story of the value premium, the strongest supporting evidence is from the Asian crisis. The Asian crisis began as a currency crisis where the Asian countries were subject to a series of speculative currency attacks starting with the Thai baht, followed by the Malaysian ringgit, Philippine peso, and Singapore dollar. These events raised fears of worldwide economic meltdown due to contagion effects. Such fear is in line with the results in Table 5 that presents monthly returns of the value strategy relative to the growth strategy during the Asian crisis and mirror chronological events. The table indicates a negative return on the value premium in July 1997 (the starting month of the crisis) for nine of thirteen countries, and in October 1997 for Hong Kong (when the Hong Kong dollar came under attack and the Hong Kong stock index lost roughly 30% of its value Kaminsky and Schmukler, 1999). Also, the November returns for the same strategy in Japan is −5.65%, reflecting the collapse of Hokkaido Takushoku Bank Ltd. which was one of the biggest Japanese banks at that time. On January, 1998 another catastrophic financial event occurred–the collapse of the Peregrine Investment Holdings, the then largest private investment bank in Asia. Peregrine was based in Hong Kong, and this might explain the extremely poor performance of value stocks relative to growth stocks on January 1998 for Hong Kong (−21.86%) and Singapore (−27.83%). Table 5 data for June 1998 is in line with the crash of the Russian stock market because of Russia’s inability to pay its debts. Final evidence comes from the collapse of the LTCM in September 1998 that raised fears of a breakdown of the entire international banking system. This is again consistent with the negative return on the US (−2.49%) and especially the Japanese (−3.62%) value premium. Overall, ten out of thirteen countries had average negative returns on the value strategy over July 1997 to March 1999. 4.2.1.4. September 11, 2001 attack. The last crisis examined is the terrorist attack on the US on September 11, 2001. Unlike the previous three examined crises that covered more than 12 months, the analysis of the September 11, 2001 crisis covers only a few months. This gives a more clear-cut explanation of the global value premium. The figures in Table 6 reflect the performance of the value strategy relative to the growth strategy surrounding the attack period (April 2001–February 2002). The most striking observation is that the value stocks underperformed the growth stocks on September 2001 for ten of thirteen countries. In particular, the returns on the value strategy on September 2001 were negative for these ten countries. 4.2.2. Analysis of country’s exposure to global value risk: Long-term effects Table 2 shows  thatthe coefficients from regressing each country’s portfolio return on the global value premium ˇ2HML from Eq. (1) are significant in six of thirteen countries. Tables 3–6 show that there is a change in the price of risk exposure that is reflected in the change in value of the global value premium during financial crises. It is interesting, therefore, to examine whether individual country’s risk exposure to the global value premium  changes  during financial crises. To this end, we examine coefficients of the global value premium ˇ2HML from Eq. (1) for the pre-crisis period, crisis period,

128

07/97 08/97 09/97 10/97 11/97 12/97 01/98 02/98 03/98 04/98 05/98 06/98 07/98 08/98 09/98 10/98 11/98 12/98 01/99 02/99 03/99 Average

Australia

Belgium

France

Germany

HK

Italy

Japan

Neth.

Sing.

Sweden

Switz.

UK

USA

2.59 1.68 5.26 2.92 6.61 −1.34 −2.23 −8.83 2.13 4.15 −4.31 9.98 −9.57 −6.67 −0.04 2.94 −6.37 −5.35 −8.68 7.37 −0.64 −0.40

−4.54 6.69 −3.36 −1.16 0.84 −2.71 −3.73 1.87 3.56 4.98 −1.02 −9.33 −6.29 3.68 −6.91 1.88 −0.37 −8.89 −1.15 5.69 11.57 −0.41

5.81 4.64 7.52 5.1 −4.4 2.2 −1.75 −1.78 9.5 −0.74 3.83 −5.92 −3.89 −9.72 −3.93 −1.33 −4.59 −9.54 −0.7 4.46 2.12 −0.15

−5.57 2.59 −4.14 −0.74 4.28 −2.28 −13.7 8.01 5.13 4.67 −5.48 −5.67 1.32 6.86 15.27 −8.76 −4.87 7.95 −2.78 5.48 9.65 0.82

−12.73 12.24 −8.78 −9.73 5.62 3.32 −21.86 24.92 −2.84 −6.07 −11.75 −14.99 −10.65 10.22 −6.89 37.74 10.7 −2.3 −15.08 3.03 4.91 −0.52

−3.16 0.31 1.86 7.84 −8.18 1.8 11.02 11.08 14.53 −8.67 −0.02 −6.53 −7.2 −9.02 −4.11 3.43 −0.3 −0.01 −9.93 −5.3 4.79 −0.28

−7.69 0.33 −7.74 2.00 −5.65 −5.92 21.46 4.66 −2.76 −6.06 1.31 3.87 −3.01 −2.62 −3.26 9.2 0.77 −1.69 2.77 −2.4 2.01 −0.02

−0.13 1.27 4.34 −6.64 −4.75 −0.22 1.83 −2.86 1.61 0.67 −7.07 −1.3 −0.18 −6.78 −15.6 0.67 −7.21 −4.08 −7.1 4.92 9.84 −1.85

−1.21 24.23 −13.98 −10.04 −12.69 −2.55 −27.83 27.56 15.01 −7.98 −5.3 −3.97 −17.92 −25.86 7.86 59.6 12.39 −5.33 7.73 −0.86 12.98 1.51

−1.34 2.04 −2.24 −0.88 −2.59 0.88 −3.26 −1.29 3.78 −1.01 0.32 −8.81 −2.7 −2.33 −2.05 −0.15 −13.6 −2.74 4.17 −1.75 5.47 −1.43

0.42 7.8 −2.83 −4.09 −0.41 −4.33 −0.82 −4.48 18.62 6.66 0.01 −4.71 3.53 −14.09 2.97 −2.3 6.06 −5.82 −5.77 1.87 9.28 0.36

2.32 3.59 −4.91 1.83 0.62 −3.76 −5.33 0.37 5.33 −0.07 3.94 −3.53 0.29 −6.38 7.29 −8.64 0.85 −8.1 −4.41 6.15 3.9 −0.41

−1.38 3.03 0.56 0.99 −1.53 3 −3.2 −0.48 2.49 −0.59 3.26 −3.45 −3.48 2.51 −2.49 −3.71 −3.02 −5.73 −6.5 1.49 −3.54 −1.04

The table reports the difference in monthly returns between the highest BE/ME (value) and the lowest BE/ME (growth) portfolios during the Asian crisis period (July 1997–March 1999). The negative return means that value stocks performed worse than growth stocks during the crisis period. The last row reports the arithmetic mean across the entire period for each country. Data are obtained directly from French’s website.

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

Table 5 Month-by-month returns during the Asian crisis (July 1997–March 1999): value–growth.

04/01 05/01 06/01 07/01 08/01 09/01 10/01 11/01 12/01 01/02 02/02

Australia

Belgium

France

Germany

HK

Italy

Japan

Neth.

Sing.

Sweden

Switz.

UK

USA

−1.26 2.09 9.46 3.4 −0.59 −1.26 3.45 1.21 2.21 −4.91 −4.3

6.97 4.24 −10.82 −0.74 −1.6 −9.15 17.48 9.33 9 −0.84 9.17

−6.56 9.31 1.14 2.85 9.16 −7.51 −1.57 0.38 0.19 −6.49 −0.32

−8.17 1.1 −0.19 4.85 −4.79 −6.79 5.29 4.58 0.71 −7.19 0.12

−10.24 0.73 2.99 5.41 2.68 −12.9 6.74 3.53 13.95 −4.78 −2.65

−1.15 13.21 −3.41 2.67 11.32 −10.1 −11.8 −0.27 −2.38 5.8 −8.88

5.31 2.83 6.26 4.04 8.77 −6.48 −4.3 −3.2 1.53 3.48 1.18

12.06 −10.37 −24.53 −5.58 −15.8 −4.85 18.63 7.4 4.27 1.74 −0.67

4.16 8.36 2.94 −0.6 −4.34 5.27 4.92 −3.69 −1.53 8.98 0.83

−10.62 4.12 2.65 2.97 8.91 3.42 −3.75 −6.18 3.95 4.98 10.86

2.39 −4.42 1.56 −1.04 1.95 −16.52 4.13 1.78 −1.78 4.02 −3.33

2 −7.62 −5.88 −1.22 −0.1 4.09 3.12 8.09 0.23 −8.05 −6.51

−3.95 3.74 0.75 2.54 1.5 −0.87 −5.58 1.49 1.7 0.44 0.85

The table reports the difference in monthly returns between the highest BE/ME (value) and the lowest BE/ME (growth) portfolios during the period of September 11, 2001 terrorist attack. The negative return means that value stocks performed worse than growth stocks during the crisis period. The last row reports the arithmetic mean across the entire period for each country. Data are obtained directly from French’s website.

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

Table 6 Month-by-month returns around the September 11, 2001 attack: (April 2001–February 2002): value–growth.

129

130

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

Table 7 Global value premium coefficient estimates of GJR-GARCH-FF model: Latin American debt crisis (1982–1983). Country name

HML ˇ2,Pre-crisis

HML ˇ2,Crisis

HML ˇ2,Post-crisis

Australia Belgium France Germany Hong Kong Italy Japan Netherlands Singapore Sweden Switzerland UK USA

−0.4453* (0.2621) −0.5381 (0.5294) −0.7886** (0.3969) 0.2683 (0.2054) −0.2520 (0.6464) −0.0442 (0.7555) 0.3290 (0.2533) −0.7365** (0.3645) 0.6844 (0.5894) −0.2599 (0.4127) −0.0473 (0.3105) −0.2452 (0.4686) −0.2296 (0.3489)

−0.0813 (0.6037) −0.1569 (0.8401) 0.6853 (0.6592) 0.1199 (1.0788) 2.4366* (1.3502) 0.1242 (0.9172) −0.3371 (0.4127) 0.4856 (0.8571) 1.6578** (0.4698) 0.3164 (1.7647) 0.2666 (0.7027) −0.0155 (0.3465) −0.1010 (0.5670)

−0.2510 (0.4317) 0.0155 (0.2979) −0.1085 (0.3112) 0.5183 (0.3172) −0.1873 (0.6179) 1.1072** (0.4665) −0.0728 (0.2117) −0.1452 (0.2397) 0.4164** (0.2322) −0.0732 (0.3203) 0.3016 (0.2157) −0.0789 (0.5196) 0.0448 (0.1820)

The table shows the values of the



ˇ2HML



coefficient in the GJR-GARCH-FF model that measures the loading of the monthly

return on the dollar global value premium. Mathematically: excess return for the country’s   ri,t − rf,t = ˇ0 + ˇ1MKT +ˇ3BP

(Pound) +

ˇ4DM

rg,t − rf,t

(Mark) +

JY ˇ5

+ ˇ2HML (H − L)



(Yen) + εi,t ; εi,t  εi,t−1 , εi,t−2 , . . .





2 ∼N 0, i,t







1, ifεi,t−1 < 0 0, otherwise The model consists of two equations:  the mean equation and the variance equation. In the mean equation, the excess monthly 2 2 2 2 i,t = ˛0 + ˛1 ε2i,t−1 + ı1 i,t−1 + ı2 i,t−2 + ı3 i,t−3 + 1 Si,t−1 ε2i,t−1 ;

market returns of country i ri,t − rf,t

where :

Si,t−1 =

are a function of three risk factors: global market risk premium (rg,f − rf ) where (rg,t ) is the

return on the global equity market portfolio, the value premium (High BE/ME–Low BE/ME) for each country, and the exchange risk premium proxied by the British pound, German mark, and Japanese yen, with the US dollar as the reference currency. We  use 2 the German mark before January 1, 1999 and the euro thereafter. The variance equation estimates the current volatility i,t









2 2 2 of country i as a function of four factors: mean volatility (˛0 ), one ARCH term ε2i,t−1 , three GARCH terms i,t−1 , i,t−2 , i.t−3 ,





and the asymmetric volatility term Si,t−1 ε2i,t−1 . The entire sample period is broken into three sub-periods to test the impact of the Latin American debt crisis on the integration of financial markets: pre-crisis period (August 1979–July 1982), crisis period (August 1982–December 1983), and post-crisis period (January 1984–December 1986). The standard errors are in parentheses. * Significant at 10% level. ** Significant at 5% level. *** Significant at 1% level.

and post-crisis period. In order to avoid overlapping periods between different crises, the analysis of the impact of the crises on the global value premium focuses on two major crises: the Latin American debt crisis in 1982–1983 and the Asian crisis in 1997–1998.   Table 7 reports results of the estimated coefficients of the global value premium ˇ2HML from Eq. (1) for pre-crisis, crisis, and post-crisis periods for the Latin American debt crisis. As mentioned earlier, US and Japanese banks’ exposure was the greatest and they faced the prospect of major loan defaults and failure as a consequence of the crisis. It is easy to understand, therefore, that Japan is the only country in Table 7 whose global premium coefficientis lower  during the crisis period than during the precrisis and post-crisis periods. Specifically, the ˇ2HML coefficient is positive for Japan for the pre-crisis period (0.3290) and turned negative during the crisis period (−0.3371). Similarly, the estimated global premium coefficient for the US during the crisis period was negative (−0.1010) before turning positive after the crisis (0.0448). However, the HML coefficient is statistically insignificant in magnitude for Japan and the US during the crisis period,  although they are both negative in sign. Overall, there is a lack of statistical significance in the ˇ2HML coefficients for the three sub-periods.

  HML



Table 8 sets forth the values of the ˇ2HML coefficient estimates before, during, and after the Asian



crisis. The results show that the ˇ2 coefficients during the crisis period are lower than those during the pre-crisis period for most of the thirteen countries. In addition, the HML coefficient is negative for nine out of thirteen countries. However, the HML coefficient is again insignificant for most of our sample countries in the three sub-periods.

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

131

Table 8 Global value premium coefficient estimates of GJR-GARCH-FF model: Asian crisis (1997–1999). Country name

HML ˇ2,Pre-crisis

HML ˇ2,Crisis

HML ˇ2,Post-crisis

Australia Belgium France Germany Hong Kong Italy Japan Netherlands Singapore Sweden Switzerland UK USA

0.1805 (0.4098) 0.4891*** (0.3244) 0.4811 (0.4432) 0.8470** (0.4286) 0.5641 (1.0583) 0.9196 (0.9246) −0.8629*** (0.3077) 0.7653** (0.3350) 1.8167*** (0.4452) 0.3529 (0.7445) −0.9893** (0.4824) 0.2333 (0.2988) 0.1535 (0.4239)

0.1108 (0.4847) −0.5747* (0.3172) −0.0526 (0.2416) −0.1182 (0.3106) −0.4689 (0.8817) −0.0107 (0.4164) −0.0874 (0.4965) −0.4296 (0.3171) −0.2188 (0.7141) 0.4417 (0.4403) −0.1612 (0.2379) 0.1233 (0.1884) 0.1593 (0.4444)

0.3172** (0.1339) 0.3030* (0.1872) 0.0456 (0.0741) −0.1212 (0.1306) −0.0110 (0.1842) −0.2834** (0.1229) −0.1214*** (0.1030) 0.1409** (0.0830) 0.2307 (0.3108) −0.3953 (0.3260) 0.1940** (0.1079) 0.0853* (0.0503) 0.0118 (0.1581)



The table shows the values of the ˇ2HML



coefficient in GJR-GARCH-FF model that measure the loading of the monthly excess

return for the country’s  portfolio on the dollar global value premium. Mathematically: ri,t − rf,t = ˇ0 + ˇ1MKT +ˇ3BP

(Pound) +

ˇ4DM

rg,t − rf,t

(Mark) +

JY ˇ5

+ ˇ2HML (H − L)



(Yen) + εi,t ; εi,t  εi,t−1 , εi,t−2 , . . .





2 ∼N 0, i,t







1, if εi,t−1 < 0 0, otherwise The model consists of two equations:  the mean equation and the variance equation. In the mean equation, the excess monthly 2 2 2 2 i,t = ˛0 + ˛1 ε2i,t−1 + ı1 i,t−1 + ı2 i,t−2 + ı3 i,t−3 + 1 Si,t−1 ε2i,t−1 ;

market returns of country i ri,t − rf,t

where :

Si,t−1 =

are a function of three risk factors: global market risk premium (rg,t − rf ) where (rg,t ) is the

return on the global equity market portfolio, the value premium (High BE/ME–Low BE/ME) for each country, and the exchange risk premium proxied by the British pound, German mark, and Japanese yen, with the US dollar as the reference currency.  We 2 use the German mark until January 1, 1999 and the euro thereafter. The variance equation estimates the current volatility i,t









2 2 2 of country i as a function of four factors: mean volatility (˛0 ), one ARCH term ε2i,t−1 , three GARCH terms i,t−1 , i,t−2 , i.t−3 ,





and the asymmetric volatility term Si,t−1 ε2i,t−1 . The entire sample period is broken into three sub-periods to test the impact of the Asian crisis on the integration of financial markets: pre-crisis period (July 1994–June 1997), crisis period (July 1997–March 1999), and post-crisis period (April 1999–March 2002). The standard errors are in parentheses. * Significant at 10% level. ** Significant at 5% level. *** Significant at 1% level.

The results from Tables 3, 5, 7 and 8 provide an interesting story of the relation between global value strategies and financial crises. Tables 3 and 5 reveal the change in the price of the value premium during the Latin American debt crisis and the Asian crisis, respectively, while Tables 7 and 8 show each country’s risk exposure to the global value premium risk factor (i.e., the coefficient from the crosssectional regression of individual country’s return on the global value premium) during the Latin American debt crisis and the Asian crisis, respectively. The figures in Tables 3 and 5 show that the value premium is negative (i.e., value stocks consistently perform more poorly than growth stocks) for most of the sample countries during the crisis period. Tables 7 and 8 reveal that the coefficients on the global value risk factors are insignificant for most of the countries. Taken together, these findings indicate that both crises led to a change in the price of the risk factor but not in the individual country’s risk exposure. 4.3. Financial crises and integration This subsection investigates the impact of financial crises – via contagion – on the integration of international financial markets. As a preliminary test, we calculate the coefficient of correlation between each country’s portfolio return and global market portfolio return for each of the pre-crisis, crisis, and post-crisis periods. Next, we measure the impact of the financial crises on the integration of financial markets by comparing  the coefficients from regressing each country’s portfolio return on the global market premium ˇ1MKT estimated from Eq. (1).

132

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

Table 9 Integration results: Latin American debt crisis (1982–1983). Panel A: correlation coefficients results Pre-crisis period Australia 0.7868 0.51185 Belgium 0.5708 France Germany 0.6521 0.7518 Hong Kong 0.5427 Italy 0.8312 Japan 0.7029 Netherlands 0.5625 Singapore 0.4592 Sweden 0.6932 Switzerland UK 0.8359 USA 0.5144 Panel B: global market premium coefficient estimates MKT Country name ˇ1,Pre-crisis Australia 1.1118*** (0.1645) 0.7081*** (0.1977) Belgium France 1.4178*** (0.1582) 0.6757*** (0.1207) Germany Hong Kong 1.7771*** (0.3810) 0.8516** (0.3708) Italy 0.8794** (0.1042) Japan 0.9569*** (0.2528) Netherlands 0.9028*** (0.3008) Singapore 0.5523* (0.2107) Sweden 0.9842*** (0.1539) Switzerland 1.1464*** (0.2078) UK 0.6064*** (0.1236) USA

Crisis period

Post-crisis period

−0.0456 0.5590 0.2672 0.6024 0.1762 0.0896 0.8909 0.3126 0.2333 0.3411 0.6915 0.2285 0.01606

0.3539 0.6762 0.7112 0.6564 0.2496 0.5818 0.9024 0.7306 −0.1261 0.6059 0.6081 0.6499 0.3386

MKT ˇ1,Crisis −0.8127** (0.6037) 1.2394** (0.5000) 0.3092*** (0.2866) 0.9942** (0.5261) −0.0610 (0.9762) −0.1291 (0.3584) 1.5407*** (0.2443) 0.7272 (0.6111) 0.7586 (0.5511) 0.2716 (1.1124) 0.8704** (0.4140) 0.6680*** (0.2245) 0.4640 (0.3973)

MKT ˇ1,Post-crisis 0.9756*** (0.2682) 0.7800*** (0.2083) 0.8442*** (0.1412) 0.8924*** (0.2111) 0.4731 (0.4257) 0.8011*** (0.2779) 1.2304*** (0.1254) 0.8119*** (0.1915) 0.3864** (0.1693) 0.6633*** (0.1918) 0.6158*** (0.1450) 0.7515*** (0.2333) 0.2636 (0.1943)

The table shows the results of market integration during the Latin American debt crisis. The entire sample period is broken into three sub-periods: pre-crisis period (August 1979–July 1982), crisis period (August 1982–December 1983), and post-crisis period (January 1984–December 1986). Panel A shows the results of coefficient of correlation between the country’smarket 

excess return and global market return. Panel B presents the values of the global market premium coefficients estimates ˇ1MKT

in GJR-GARCH-FF model that measure the loading of the monthly excess return for the country’s market return on the dollar Mathematically: global market return.     ri,t − rf,t = ˇ0 + ˇ1MKT ∼N



2 0, i,t

rg,t − rf,t

+ ˇ2HML (H − L) + ˇ3BP (Pound) + ˇ4DM (Mark) + ˇ5 (Yen) + εi,t ; εi,t  εi,t−1 , εi,t−2 , . . . JY





1, if εi,t−1 < 0 0, otherwise where ri,t is the market returns of country i, rf,t is the risk-free rate, rg,t is the return on the global equity market portfolio, and H − L is the global value premium (High BE/ME–Low BE/ME) for each country. The pound, mark, and yen are the exchange rates of British pound, German mark, and Japanese yen, respectively, with the US dollar as the reference currency. The mark becomes the euro after January 1, 1999. The standard errors are in parentheses. * Significant at 10% level. ** Significant at 5% level. *** Significant at 1% level. 2 i,t

2 2 2 = ˛0 + ˛1 ε2i,t−1 + ı1 i,t−1 + ı2 i,t−2 + ı3 i,t−3 + 1 Si,t−1 ε2i,t−1 ;

where :

Si,t−1 =

Tables 9 and 10 show the results of our integration test during the Latin American debt crisis and the Asian crisis, respectively. Panel A shows that the correlation coefficient between a country’s return and the global market portfolio is lower in the crisis period than  inthe pre-crisis period for eleven of thirteen countries. Further, Panel B presents the values of ˇ1MKT from Eq. (1) for the three subperiods of the Latin American debt crisis. Comparison of results from the pre-crisis period  and crisis  period reveals that the size of the country’s loading on the US dollar global market return ˇ1MKT is greater in the pre-crisis period for ten out of thirteen countries—Australia, France, Hong Kong, Italy, Netherlands, Singapore, Sweden, Switzerland, the UK, and finally the USA. Surprisingly, some countries (such as Hong Kong and Australia) were positively correlated with the global market portfolio before

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

133

Table 10 Integration results: Asian crisis (1997–1999). Panel A: correlation coefficients results Pre-crisis period Australia 0.6087 Belgium 0.5851 0.6344 France Germany 0.5006 0.4957 Hong Kong 0.2802 Italy 0.8799 Japan 0.7348 Netherlands 0.2870 Singapore 0.3662 Sweden 0.3378 Switzerland UK 0.7318 USA 0.5144 Panel B: global market premium coefficient estimates MKT Country name ˇ1,Pre-crisis Australia 0.8989*** (0.1553) 0.4622*** (0.1150) Belgium France 0.7650*** (0.2547) 0.4787*** (0.1880) Germany Hong Kong 0.9859*** (0.2522) 0.6643** (0.3455) Italy 1.5024*** (0.1284) Japan 0.7934*** (0.1282) Netherlands 0.4034** (0.2047) Singapore 0.6191** (0.3642) Sweden 0.5049** (0.2189) Switzerland 0.6108*** (0.1374) UK 0.4899*** (0.1415) USA

Crisis period

Post-crisis period

0.7595 0.6624 0.8847 0.8019 0.6056 0.6069 0.7563 0.8403 0.5825 0.7830 0.7872 0.8634 0.8371

0.7443 0.5335 0.9123 0.8513 0.7576 0.7101 0.8002 0.8784 0.5237 0.8241 0.5005 0.8376 0.8406

MKT ˇ1,Crisis 0.9518*** (0.2587) 0.7729*** (0.2746) 1.2847*** (0.2119) 1.1914*** (0.1613) 1.0754** (0.5796) 1.3755*** (0.3069) 0.9548** (0.5770) 1.1156*** (0.2226) 1.1090 (0.7783) 1.5156*** (0.2108) 0.7889*** (0.2141) 0.7577*** (0.1011) 0.9140*** (0.2371)

MKT ˇ1,Post-crisis 1.0722*** (0.1907) 0.6847*** (0.1642) 1.3981*** (0.1167) 1.1999*** (0.1302) 1.1230*** (0.1627) 0.9783*** (0.1523) 1.1388*** (0.1487) 1.0179*** (0.0728) 0.8577*** (0.2638) 1.7028*** (0.2346) 0.5832*** (0.1156) 0.7841*** (0.0573) 0.9726*** (0.1816)

The table shows the results of market integration during the Asian crisis. The entire sample period is broken into three subperiods: pre-crisis period (July 1994–June 1997), crisis period (July 1997–March 1999), and post-crisis period (April 1999–March 2002). Panel A shows the results of coefficient of correlation between the country’s marketexcess  return and global market

return. Panel B presents the values of the global market premium coefficients estimates

ˇ1MKT

in GJR-GARCH-FF model

that measure the loading of the monthly excess return for the country’s market return on the dollar global market return. Mathematically:     ri,t − rf,t = ˇ0 + ˇ1MKT ∼N



2 0, i,t

rg,t − rf,t

+ ˇ2HML (H − L) + ˇ3BP (Pound) + ˇ4DM (Mark) + ˇ5 (Yen) + εi,t ; εi,t  εi,t−1 , εi,t−2 , . . . JY





1, if εi,t−1 < 0 0, otherwise where ri,t is the market returns of country i, rf,t is the risk-free rate, rg,t is the return on the global equity market portfolio, and H − L is the global value premium (High BE/ME–Low BE/ME) for each country. The pound, mark, and yen are the exchange rates of British pound, German mark, and Japanese yen, respectively, with the US dollar as the reference currency. The mark euro is substituted on for the mark after January 1, 1999. The standard errors are in parentheses. ** Significant at 5% level. *** Significant at 1% level. 2 i,t

2 2 2 = ˛0 + ˛1 ε2i,t−1 + ı1 i,t−1 + ı2 i,t−2 + ı3 i,t−3 + 1 Si,t−1 ε2i,t−1 ;

where :

Si,t−1 =





the crisis but become negatively correlated during the crisis. Comparison of the ˇ1MKT coefficient from the pre-crisis period and the post-crisis   period leads to similar conclusions. Except for Belgium, Germany, Japan, and Sweden, the ˇ1MKT coefficient is higher in the pre-crisis period than the post crisis-period. Thus, it appears that international financial markets became less integrated after the Latin American debt crisis. Table 10 shows the results of the above tests for the Asian crisis. Its contents are contradictory to those in Table 9 since twelve out of thirteen became more integrated (reveal greater  countries  correlation coefficients in panel A and greater ˇ1MKT coefficients in panel B) during the Asian crisis relative to the pre-crisis  period.  Comparison of the pre-crisis period and the post crisis period confirms the results since the ˇ1MKT coefficients estimated using the post crisis period are higher than those

134

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

calculated in the pre-crisis period for all of the sample countries except Japan. Thus international financial markets became more integrated after the Asian crisis. Results in Tables 9 and 10 reflect the regional versus global distinction. The Latin American debt crisis is considered regional while the Asian crisis is considered global, based on the geographic impact of the individual crisis. The debt crisis in 1982–1983 is considered a regional crisis since it affected primarily the Latin American indebted countries and did not spread to the rest of the world. The important roles played by the US and Japan as major international creditors contributed to the concentration of the debt crisis in the region (Katada, 2001). We found no empirical evidence in the literature on the impact of the crisis on global equity markets. In contrast, the Asian crisis started in Asia and spread to non-Asian countries. In the US, for example, the DJIA, NYSE, AMEX, and NASDAQ indices declined by 7.2%, 6.6%, 5.8%, and 7%, respectively on October 27, 1997 (Michayluk and Neuhauser, 2006). In addition, there are several studies that document the impact of the Asian crisis on the global equity returns in North America, South America, and Europe (e.g., Tuluca and Zwick, 2001; Sander and Kleimeier, 2003; Fernandez-Izquierdo and Lafuente, 2004; Kenourgios et al., 2011). Such effects define the Asian crisis as a global crisis. Our results suggest that not all crises have the same financial market integration impact. In other words, financial crises with regional effects decrease financial integration while crises with global effects increase integration. Our integration results can be framed from the perspective of news demands as follows. During regional crises (such as the Latin American debt crisis), investors pay little attention to international news and focus primarily on country or regional news. Thus, the Latin American countries adopted restrictive capital controls for self-protection, further segmenting their markets. This implies an increase in gains from global market portfolio diversification, since the Latin American countries become less integrated with the world. On the other hand, global crises (such as the Asian crisis) are headline public news in the international media. Consequently, international investors tend to interpret any adverse news about one country as information affecting the entire world. This creates higher cross market correlations between different countries. As a result, the Asian countries relied on the rest of the world to help share their crisis, further integrating their financial markets. 5. Conclusion This paper investigates the impact of financial crises along two dimensions: the global value premium and financial market integration. For risk aversion effects of the crises on the global value premium, the estimated coefficients of regressing the monthly excess return for the country’s portfolio on the dollar global value premium during the Latin American debt crisis and the Asian crisis period are insignificant for most of the sample countries. In addition, the month-to-month performance of value stocks relative to growth stocks is examined. Results show that value stocks consistently perform less well than growth stocks during four selected financial crises—the Latin American debt crisis, the ERM crisis, the Asian crisis, and the terrorist attack on September 11, 2001. These findings support the risk story for a global value premium, because they indicate that international investors tend to rush to quality and liquidity by getting rid of high risk (high BE/ME) stocks and replacing them with low risk (low BE/ME) stocks during financial crises. Taken together, these findings indicate that both crises led to a change in the price of the risk factor but not in the individual country’s risk exposure. For the effects of the financial crises on the integration of financial markets, investigation of the impact of the Latin American debt crisis (1982–1983) and the Asian crisis (1997–1998) indicates that international financial markets became less integrated during and after the Latin American debt crisis relative to before the crisis, while financial markets became more integrated during and after the Asian crisis relative to the period before the crisis. Therefore, financial crises with regional effects (such as the less developed debt crisis in 1982–1983 that affected primarily Latin American countries) decrease financial integration, while crises with global effects (such as the Asian crisis in 1997 and 1998) increase integration. This suggests that the Latin American countries adopted restrictive capital controls for self-protection, further segmenting their markets. However, the Asian markets, with their vast global capital flows, rode out their crises with increased cooperation not only with each other but with the remainder of the world.

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

135

In summary, we conclude that (1) the proposed GJR-GARCH-FF model is superior to popular existing models for explaining international returns data, (2) the global value premium is a risk factor and not an anomaly, (3) value premiums change as a result of financial market crises, and (4) the global value premium contributes to measuring integration. Findings open directions for further research. First, identifying links between the global value premium and integration allows creation of new and better methodologies to incorporate the global value premium. Second, results support use of the new GJR-GARCH-FF model in future studies because the average intercept of the GJR-GARCH-FF model is lower than that of competing models. Therefore, incorporating cross-sectional asset pricing models (e.g., CAPM; Fama–French three factor model, 1996; Carhart four-factor model, 1997) as the mean equation in GARCH models could yield improved empirical results. References Ahmad, R., Rhee, S., Wong, Y., 2012. Foreign exchange market efficiency under recent crises: Asia-Pacific focus. Journal of International Money and Finance 31, 1574–1592. Arghyrou, M., Kontonikas, A., 2012. The EMU Sovereign-debt crisis: fundamentals, expectations and contagion. Journal of International Financial Markets, Institutions and Money 22 (4), 658–677. Bordo, M., Meissner, C., 2012. Does inequality lead to a financial crisis? Journal of International Money and Finance 31, 2147–2161. Bruner, R.F., Li, W., Kritzman, M., Myrgren, S., Page, S., 2008. Market integration in developed and emerging markets: evidence from CAPM. Emerging Markets Review 9, 89–103. Campbell, J.Y., Hamao, Y., 1992. Predictable stock returns in the United States and Japan: a study of long-term capital market integration. Journal of Finance 47, 43–69. Carhart, M., 1997. On persistence in mutual fund performance. Journal of Finance 52, 57–82. DeBondt, W., Thaler, R., 1985. Does the stock market overreact? The Journal of Finance 40 (3), 793–805. Demyanyk, Y., Hemert, O., 2011. Understanding the subprime mortgage crisis. Review of Financial Studies 24 (6), 1848–1880. Driessen, J., Hemert, O., 2012. Pricing of commercial real estate securities during the 2007–2009 financial crisis. Journal of Financial Economics 105, 37–61. Edison, H., Kole, L., 1995. European monetary arrangements: implications for the Dollar, exchange rate variability and credibility. European Financial Management 1, 61–86. Edwards, F., 1999. Hedge funds and the collapse of long-term capital management. Journal of Economic Perspectives 13, 189–210. Errunza, V., Losq, E., 1985. International asset pricing under mild segmentation: theory and test. Journal of Finance 40, 105–124. Errunza, V.R., Losq, E., Padmanabhan, P., 1992. Tests of integration, mild segmentation and segmentation hypotheses. Journal of Banking and Finance 16, 949–972. Fama, E., French, K., 1989. Business conditions and expected returns on stocks and bonds. Journal of Financial Economics 25, 23–49. Fama, E., French, K., 1993. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33, 3–56. Fama, E., French, K., 1995. Size and book-to-market factors in earnings and returns. Journal of Finance 50, 131–155. Fama, E., French, K., 1998. Value versus growth: the international evidence. Journal of Finance 1, 1975–1999. Federal Deposit Insurance Corporation (FDIC), 1997. An Examination of the Banking Crises of the 1980 and Early 1990, Volume I., pp. 191–210, Chapter 5. Fernandez-Izquierdo, A., Lafuente, J.A., 2004. International transmission of stock exchange volatility: empirical evidence from the Asian crisis. Global Finance Journal 15, 125–137. Ferson, W.E., Harvey, C.R., 1993. The risk and predictability of international stock returns. Review of Financial Studies 6, 527–566. Giavazzi, F., Giovannini, A., 1989. Limiting Exchange Rate Flexibility: The European Monetary System. MIT Press, Cambridge, MA. Glosten, L.R., Jagannathan, R., Runkle, D.E., 1993. On the relation between the expected value and the volatility of the nominal excess return on stocks. Journal of Finance 48, 1779–1801. Harris, S., Marston, F., Mishra, D., O’Brien, T., 2003. Ex ante cost of equity estimates of S&P 500 Firms: the choice between global and domestic CAPM. Financial Management 32 (Autumn (3)), 51–66. International Monetary Fund, 1998. World Economic and Financial Surveys: International Capital Markets Developments, Prospects and Key Policy Issues. International Monetary Fund, Washington, D.C. Kaminsky, G., Schmukler, S., 1999. What triggers market jitters? A Chronicle of the Asian crisis. Journal of International Money and Finance 18, 537–560. Katada, S., 2001. Japan and the Latin American debt crisis. In: Banking on Stability: Japan and The Cross-Pacific Dynamics of International Financial Crisis Management. University of Michigan Press, Michigan, pp. 123–144. Kenourgios, D., Samitas, A., Paltalidis, N., 2011. Financial crises and stock market contagion in a multivariate time-varying asymmetric framework. Journal of International Financial Markets, Institutions and Money 21, 92–106. Koedijk, C., Kool, C., Schotman, P., Van Dijk, M., 2002. The cost of capital in international financial markets: local or global? Journal of International Money and Finance 21 (6), 905–929. Lakonishok, J., Shleifer, A., Vishny, W., 1994. Contrarian investment, extrapolation, and risk. The Journal of Finance 49 (5), 1541–1578. Marcus Alan, J., 1989. An equilibrium theory of excess volatility and mean reversion in stock market prices. In: National Bureau of Economic Research Working Paper No. 3106. National Bureau of Economic Research.

136

E.A. Yamani, P.E. Swanson / Int. Fin. Markets, Inst. and Money 33 (2014) 115–136

Melvin, M., Taylor, M., 2009. The global financial crisis: causes, threats and opportunities. introduction and overview. Journal of International Money and Finance 28, 1243–1245. Mierau, J., Mink, M., 2013. Are stock market crises contagious? The role of crisis definitions. Journal of Banking and Finance 37, 4765–4776. Michayluk, D., Neuhauser, K.L., 2006. Investor overreaction during market declines: evidence from the 1997 Asian financial crisis. Journal of Financial Research XXIX, 217–234. Nitschka, T., 2010. Cashflow news, the value premium and an asset pricing view on European stock market integration. Journal of International Money and Finance 29, 1406–1423. Philippas, D., Siriopoulos, C., 2013. Putting the C into crisis: contagion, correlations and copulas on EMU bond markets. Journal of International Financial Markets, Institutions and Money 27, 161–176. Saffi, P., Kari Sigurdsson, K., 2011. Price efficiency and short selling. Review of Financial Studies 24 (3), 821–852. Sander, H., Kleimeier, S., 2003. Contagion and causality: an empirical investigation of four Asian crisis episodes. Journal of International Financial Markets, Institutions and Money 13, 171–186. Sevilla, C., 1995. Explaining the September 1992 ERM crisis: The Maastricht bargain and domestic politics in Germany, France, and Britain. In: Working Paper. Harvard University. Shiller, R.J., 1984. Stock prices and social dynamics. Brookings Papers on Economic Activity 2, 457–498. Solnik, B., 1983. International arbitrage pricing theory. Journal of Finance 38, 449–457. Tuluca, S.A., Zwick, B., 2001. The effects of the Asian crisis on global equity markets. The Financial Review 36, 125–142. Wu, H., 2008. International asset pricing models: a forecasting evaluation. International Research Journal of Finance and Economics 15, 175–184. You, L., Ghai, G., Welch, W., 2006. Estimation of global beta and tests of capital asset pricing models. The International Journal of Finance 18, 4161–4182.