Search in book...
Toggle Font Controls
Create new playlist

Name your new playlist

Playlist description (optional)
Sign In

Email address

Password

Forgot Password?

or

Continue with Facebook

Continue with Google
Sign Up

Full Name

Email address

Confirm Email Address

Password

or

Continue with Facebook

Continue with Google

Chapter 5

Are Frontier Markets Worth the Risk?

B.K. Uludag^*

H. Ezzat^**
^* Dokuz Eylül University, Faculty of Business, Izmir, Turkey
^** Maastricht University, School of Business, Maastricht, The Netherlands

Abstract

In this study we examine long memory properties in the returns and volatility of the major European frontier stock markets of Slovenia, Slovakia, Romania, Croatia, Estonia, and Lithuania. The sample period is between 2012 and 2014. We test for the long memory property using the Geweke and Porter-Hudak (GPH) and the Gaussian semiparametric (GSP) methods. The findings show that while there is long memory in stock returns for only Romania and Slovenia, the long memory property in the volatility series is found to be highly significant for Estonia, Romania, Slovakia, and Slovenia. The findings further show that the ARFIMA–FIGARCH model does not capture the long memory property in stock returns; however, there is strong evidence of long memory in the volatility series for Slovenia and Romania. The evidence of long memory in these countries implies that investors can exploit predictability and earn speculative returns by using past stock return information.

Keywords

frontier markets

long memory

ARFIMA–FIGARCH

JEL Classification

C22, C50

1. Introduction

In their quest for more rewarding investment opportunities, international investors are continually seeking markets that promise higher returns and more efficient diversification. However, increasing market linkages and interdependence among global equity markets have made diversification opportunities more difficult to uncover, and higher returns are invariably associated with higher risk. Emerging markets once provided adequate diversification and higher returns; however, such benefits have gradually diminished as emerging markets have evolved into developed and advanced emerging markets. This transition was achieved as countries upgraded their market microstructure—which in turn increased market efficiency—and developed their economies. Currently, frontier markets have caught the cautious attention of international investors, replacing emerging markets as a possible venue for achieving the illusive goal of greater diversification with greater returns.

It is widely accepted that frontier markets are less liquid and thinner than most equity markets, and so tend to be less efficient. Indeed, the efficiency of financial markets is related to the long-term dependence of price returns. The existence of long-term dependency among price movements indicates the property of long memory. The presence of long memory in stock returns suggests that current returns are dependent on past returns. This violates the efficient market hypothesis (EMH) and martingale processes which are assumed in most financial asset pricing models (Fama, 1970). Detecting the long memory property in stock markets is important for investors, regulators, and policymakers in their attempts to mitigate market inefficiency.

There is a vast literature of the long memory properties of developed and developing stock markets (Lo, 1991; Cheung and Lai, 1995; Caporale and Gil-Alana, 2002). However, little is known about the long memory processes in frontier markets, and the existing literature on frontier markets generally focuses on Africa and the Gulf region. This paper is motivated by the fact that there are relatively very few papers about the frontier markets in Europe. In addition to this, frontier stock markets in Europe are of particular research interest due to their growth potential and EU membership. Since EU members are subject to EU regulations and economic reforms due to euro zone criteria, investors may find attractive investment opportunities in these countries.

The objective of this paper is to investigate the long memory property in stock returns and volatility, using data from the major frontier markets of Slovenia, Slovakia, Romania, Croatia, Estonia, and Lithuania. The sample period is 2012–14. We use the daily closing prices of stock market indices. Since frontier markets grapple with market thinness, rapid changes in regulatory framework, and unpredictable market responses to information flow, stock returns in frontier markets have distinct properties compared to other markets. Therefore, modeling long memory in return and volatility becomes important in measuring risk in these markets. To test the long memory property, we use the estimation of GPH in conjunction with the GSP method. We also estimate ARFIMA–FIGARCH models to examine the presence of long memory in stock returns and volatility.

The rest of the paper is organized as follows: Section 2 presents the relevant literature review, Section 3 presents the data, Section 4 outlines the methodology used for the study, Section 5 shows the empirical findings and analysis, and Section 6 summarizes and concludes the study.

2. Literature Review

Frontier markets are at the furthest edge of the acceptable investment horizon, beyond which markets are no longer suitable for investment. As such, investments in frontier markets are plagued by persistent risks and hurdles, making navigation through such markets like walking through a minefield. However, the promise of great profits and growth potential make frontier markets all the more attractive to international investors. Investors are keenly aware of the concept that high profits cannot be separated from the possibility of higher risks, and frontier markets demonstrate this concept to the fullest. In order to allocate large amounts of funds to frontier markets, investors must be confident that frontier markets are worth the risk. This requires investigating the diversification efficiency of frontier markets as well as the risk–return relationship and overall profitability compared to other market groups.

There are limited studies on frontier markets. The existing literature examines different aspects of frontier markets. Some researchers present evidence of the significant diversification benefits of frontier markets (Speidell and Krohne, 2007; Berger et al., 2011; Marshall et al., 2013; Jayasuriya and Shambora, 2009; Speidell, 2008; Girard and Sinha, 2008; Gupta, 2014). They demonstrate that frontier markets are less correlated with world markets and have a lower level of integration and interdependence with other market groups, giving them high diversification potential. On the other hand, Samarakoon (2011) reports that although frontier markets have very low correlations with the US market, during the subprime financial crisis the correlations of frontier markets in Asia, Africa, the Middle East, and Europe increased more than the correlations of emerging markets. This suggests that frontier markets may not be the best alternative for diversification during a crisis period.

Some of the existing studies focus on the integration of the frontier markets with other markets. Nikkinen et al. (2011) investigate the financial integration of European frontier markets before and during the subprime financial crisis. They show that the markets of Croatia, Estonia, and Slovenia have substantial financial integration with world markets and the three largest European markets. They report a strengthening of linkages during the crisis period with considerable variation across markets. Overall, the prevailing evidence suggests that frontier markets do provide more efficient diversification than emerging markets; however, it is a premise that needs to be continually evaluated, considering the evolving nature of frontier markets. In their paper, Balcilar et al. (2013) examine herding behavior in the oil-rich frontier stock markets of the Gulf Cooperation Council (GCC). They report that GCC frontier markets are strongly integrated with other financial and oil markets, with global shocks having a significant impact on the volatility of GCC markets.

The prospect of higher profits being generated from frontier markets has motivated researchers to evaluate frontier markets’ profitability compared to developed and emerging markets’ profitability. In one study de Groot et al. (2012) investigate a cross-section of stock returns of more than 1400 stocks in frontier markets. They conclude that portfolios based on value and momentum in frontier markets generate excess returns at least as large as returns in developed and emerging markets. Girard and Sinha (2008) examine the risk premiums of 360 stocks from 19 frontier markets. They observe that frontier markets have greater return potential than both emerging and developed markets. They also report that small and value stocks are less risky than large and growth stocks, and that risk premiums are more greatly affected by political, economic, and financial factors.

Another influential aspect of frontier markets is the problem of insufficient liquidity, which has also attracted the attention of researchers. Benić and Franić (2008) report that the frontier markets of Croatia, Serbia, Bulgaria, and Slovenia are more illiquid than the markets of Germany, Poland, and Hungary. Minović and Živković (2010) document that for Serbia, liquidity risk significantly affects price formation. Marshall et al. (2013) show that frontier markets spreads are on average over 2.5 times greater than spreads in the US market, which can be attributed to the significantly lower liquidity of frontier markets.

There are very limited studies which focus on the long memory process in frontier markets. Using both daily and weekly data, Rambaccussing (2010) investigates four of Africa’s frontier markets: Botswana, Ghana, Mauritius, and Namibia. The results suggest less predictability in daily returns compared to weekly returns. There is strong evidence of long memory in the weekly returns for all four countries under investigation. The daily returns predictability is also present for all countries except Namibia. Anoruo and Gil-Alana (2011) examine long memory using fractionally integrated techniques in 10 African countries: Egypt, Morocco, Tunisia, Nigeria, Mauritius, Kenya, South Africa, Zimbabwe, Botswana, and Namibia. They failed to find evidence of mean reversion for all of the stock market price series. The evidence of long memory in returns is obtained in the cases of Egypt and Nigeria, and to a lesser extent in Tunisia, Morocco, and Kenya. Jayasuriya (2009) examines the long run persistence of stock return volatility for 23 developing markets that include frontier markets for the period from Jan. 2000 to Oct. 2007. The findings suggest that the long memory property is not observed in many developing and frontier markets. An intuitive explanation is that as emerging markets develop and grow, there would be greater market participation that would result in greater efficiency.

3. Data Set

In this paper we’ve selected the frontier markets of Slovenia, Slovakia, Romania, Croatia, Estonia, and Lithuania for study. These chosen frontier markets are located in Europe. Out of the six countries, four are in Eastern Europe and two are Baltic countries. One of the common characteristics of these markets is the presence of a transition economy. All of these countries experienced transition from a communist system to a market economy. The transition process included privatization of state-owned enterprises, an opening to international trade and investment, and a set of economic reforms and regulations to liberalize the financial sector. The Baltic stock exchanges differ from the Eastern European stock exchanges in several respects. The Baltic countries are unified under a joint Baltic exchange. Therefore, these markets are subject to similar market regulations and environment. The Tallinn Stock Exchange merged with the OMX Group in 2004 and the Vilnius Stock Exchange in 2005. In 2007, the NASDAQ took over the OMX Group with the purpose of increasing cross-border trading and attracting more investments to the region.

Table 5.1 presents the summary of characteristics of the major European frontier markets that have been selected for study. It is important to note that with the exception of Slovenia’s market, all of the markets were established after the collapse of Soviet Union. Slovenia’s Ljubljana Stock Exchange (LJSE) has the unique characteristic of being established while Slovenia was a part of Yugoslavia. Table 5.1 shows that Croatia has the highest market capitalization, followed by Romania. The Baltic stock exchanges have lower market capitalization in comparison with that of Eastern European countries.

Table 5.1

Summary of European Frontier Markets

	Index	Stock exchange	Date of establishment	Market capitalization (in 2012) ($)
Slovenia	INDEXDJX:DWSID	Ljubljana	1989	6,474,886,528
Slovakia	INDEXDJX:DWSKD	Bratislava	1991	4,610,591,442
Romania	INDEXDJX:DWROD	Bucharest	1995	15,925,220,857
Croatia	INDEXVIE:CRU	Zagreb	1991	21,559,647,510
Estonia	INDEXDJX:DWEED	Tallinn	1995	2,331,962,196
Lithuania	INDEXDJX:DWLUD	Vilnius	1993	3,963,704,823

Source: http://data.worldbank.org/indicator/CM.MKT.LCAP.CD

For the markets of Slovenia, Slovakia, Romania, Estonia, and Lithuania, we’ve used the Dow Jones Total Market Index in US dollars for each country. For Croatia, the Croatian Traded Index in US dollars was used. All index data were gathered from Google Finance.

The sample period is between 2012 and 2014. Daily stock index returns are calculated as follows:

$R_{t} = ln (\frac{P_{t}}{P_{t - 1}})$ $R_{t} = ln (\frac{P_{t}}{P_{t - 1}})$

where R_t is the index return at time t and P_t and P_t−1 are closing prices of an index at time t and t − 1, respectively.

4. Methodology

4.1. Long Memory

The long memory properties in returns and volatility of stock market indices are estimated by using the GPH and GSP methods. The long memory parameter d, which can capture the slope of the sample spectral density through a simple OLS regression based on the periodogram, is calculated as follows:

$\log [I (w_{j})] = β_{0} + β_{1} \log [4 \sin^{2} (\frac{w_{j}}{2})] + ɛ_{j}$ $\log [I (w_{j})] = β_{0} + β_{1} \log [4 \sin^{2} (\frac{w_{j}}{2})] + ɛ_{j}$

(5.1)

where w_j = 2πj/T, j = 1, 2, …, n; ɛ_j is the residual term; and w_j represents the

n = \sqrt{T}

$n = \sqrt{T}$

Fourier frequencies. I(w_j) denotes the sample periodogram as defined as

$I (w_{j}) = \frac{1}{2 π T} {|\sum_{t = 1}^{T} r_{t} e^{- w_{j} t}|}^{2}$ $I (w_{j}) = \frac{1}{2 π T} {|\sum_{t = 1}^{T} r_{t} e^{- w_{j} t}|}^{2}$

(5.2)

where r_t is covariance stationary time series. The estimate of d, say

{\hat{d}}_{G P H}

${\hat{d}}_{G P H}$

, is

- {\hat{β}}_{1}

$- {\hat{β}}_{1}$

Robinson and Henry (1999) developed the GSP method to measure the persistence, and the GSP estimator is based on the periodogram regression used to estimate the long memory parameter for a covariance stationary series.

It is given as:

$f (w) = G w^{1 - 2 H} a s w \to 0^{+}$ $f (w) = G w^{1 - 2 H} a s w \to 0^{+}$

where

\frac{1}{2} < H < 1, 0 < G < \infty

$\frac{1}{2} < H < 1, 0 < G < \infty$

, and f(w) is the spectral density of r_t. The periodogram with respect to the observations of r_t, t = 1, …, T is defined as

I (w_{j} = \frac{1}{2 π n} {|\sum_{i = 1}^{n} r_{t} e^{i t w_{j}}|}^{2}

$I (w_{j} = \frac{1}{2 π n} {|\sum_{i = 1}^{n} r_{t} e^{i t w_{j}}|}^{2}$

Consequently, the long memory parameter H is determined by

$\begin{array}{l} H = \arg \min_{∆ \leq H \leq ∆_{2}} R (H), \\ where \{\begin{array}{c} 0 < ∆_{1} < ∆_{2} < 1 \\ R (H) = log \{\frac{1}{m} \sum_{j = 1}^{m} \frac{I (w_{j})}{w_{j}^{1 - 2 H}}\} - (2 H - 1) \frac{1}{m} \sum_{j = 1}^{m} \log (w_{j}) \\ m \in (0, [n / 2]) \\ w_{j} = 2 π j / n \end{array} \end{array}$ $\begin{array}{l} H = \arg \min_{∆ \leq H \leq ∆_{2}} R (H), \\ where \{\begin{array}{c} 0 < ∆_{1} < ∆_{2} < 1 \\ R (H) = log \{\frac{1}{m} \sum_{j = 1}^{m} \frac{I (w_{j})}{w_{j}^{1 - 2 H}}\} - (2 H - 1) \frac{1}{m} \sum_{j = 1}^{m} \log (w_{j}) \\ m \in (0, [n / 2]) \\ w_{j} = 2 π j / n \end{array} \end{array}$

(5.3)

4.2. The ARFIMA–FIGARCH Model

The ARFIMA–FIGARCH model is used to estimate the long-term dependence in returns and volatility in time series data. The first component of the model, known as an ARFIMA process, was introduced by Granger and Joyeux (1980) and Hosking (1981). The model considers the fractionally integrated process I (d) in the conditional mean. The ARFIMA(p_m,d_m,q_m) model for a time series process y_t can be expressed as follows:

$Φ (L) {(1 - L)}^{d} y_{t} = Θ (L) ɛ_{t}^{2},$ $Φ (L) {(1 - L)}^{d} y_{t} = Θ (L) ɛ_{t}^{2},$

(5.4)

where d is the fractional integrated parameter; L is a lag operator; Φ(L) and Θ(L) are the lag operator polynomials of order p and q, respectively; and ɛ_t is the random error. Following Hosking (1981), an ARFIMA process is nonstationary when d_m ≥ 0.5. When d = 1, the process has a unitary root with infinite variance. For 0 < d_m < 0.5, the process is said to exhibit long memory. For −0.5 < d_m < 0 the process is said to have short memory, referred to as antipersistence. If d = −0.5, then the series is stationary, but not invertible.

Baillie et al. (1996) introduced long memory in the conditional variance of a GARCH model called FIGARCH(p_v,d_v,q_v). The FIGARCH process is defined as:

$Φ (L) {(1 - L)}^{d} ɛ_{t}^{2} = ω + [1 - β (L)] (ɛ_{t}^{2} - σ_{t}^{2})$ $Φ (L) {(1 - L)}^{d} ɛ_{t}^{2} = ω + [1 - β (L)] (ɛ_{t}^{2} - σ_{t}^{2})$

(5.5)

where 0 ≤ d ≤ 1 designates the parameter of fractional integration with the Φ(L) and [1 − β(L)] roots located outside the unit circle, so that the stationarity of the covariance can be ensured. Hence, reordering the terms of equation:

$[1 - β (L)] σ_{t}^{2} = ω + [1 - β (L) - ϕ (L) {(1 - L)}^{d}] ɛ_{t}^{2},$ $[1 - β (L)] σ_{t}^{2} = ω + [1 - β (L) - ϕ (L) {(1 - L)}^{d}] ɛ_{t}^{2},$

The conditional variance of

ɛ_{t}^{2}

$ɛ_{t}^{2}$

can then be rewritten as:

$\begin{array}{l} σ_{t}^{2} = ω {[1 - β (L)]}^{- 1} + \{1 - {[1 - β (L)]}^{- 1} ϕ (L) {(1 - L)}^{d}\} ɛ_{t}^{2} \\ σ_{t}^{2} \equiv ω {[1 - β (L)]}^{- 1} + λ (L) ɛ_{t}^{2} \end{array}$ $\begin{array}{l} σ_{t}^{2} = ω {[1 - β (L)]}^{- 1} + \{1 - {[1 - β (L)]}^{- 1} ϕ (L) {(1 - L)}^{d}\} ɛ_{t}^{2} \\ σ_{t}^{2} \equiv ω {[1 - β (L)]}^{- 1} + λ (L) ɛ_{t}^{2} \end{array}$

(5.6)

5. Empirical Results

In this section we report the descriptive statistics and the results of the long memory tests (GPH and GSP) in stock returns and volatility. We also report the results of the ARFIMA–FIGARCH specification for the stock return series.

Table 5.2 documents descriptive statistics for the six major European frontier markets in the study. The findings show that with the exception of Croatia, the average stock returns for these are positive. Among the stock markets, Romania has the highest (0.000451) and Croatia has the lowest (−0.0001) average stock returns. Slovenia has the highest standard deviation, indicating a high volatility in the stock market. The skewness is negative for the countries of Slovenia, Slovakia, and Lithuania. The values of kurtosis for all stock markets are above three, indicating a leptokurtic distribution. The null hypothesis of Jarque–Bera is rejected for each stock market. All of the return series exhibit significant deviations from normality.

Table 5.2

Descriptive Statistics for Major European Frontier Markets

	Slovenia	Slovakia	Romania	Croatia	Estonia	Lithuania
Mean	0.000301	0.000105	0.000451	−0.0001	0.000202	0.000211
Median	−0.000064	0	0.000562	−0.00016	7.96E−05	0.000195
Maximum	0.035956	0.024014	0.039526	0.039305	0.055088	0.030656
Minimum	−0.0515384	−0.02674	−0.05083	−0.03087	−0.028031	−0.04540
Std. dev.	0.01027	0.003919	0.009384	0.007511	0.00748	0.005769
Skewness	−0.32342	−0.23986	−0.2943	0.228572	0.879216	−0.74801
Kurtosis	4.890188	13.64306	6.214264	5.712883	10.39224	11.59607
J–B	120.7338*,	2993.691*,	330.1267*,	231.1611*,	1773.017*,	2337.84*,

Notes: The symbols ***, **, * denote significance at the 1, 5, and 10% levels, respectively.

Fig. 5.1 shows the plots of daily returns for Slovenia, Slovakia, Romania, Croatia, Estonia, and Lithuania. It can be observed that all series display volatility clustering effects that large price changes of either sign tend to group together. Furthermore, there is also evidence of synchronized behavior among all return series, with the great majority of peaks and troughs occurring in 2014 in particular. Most of the stock indices experienced the lowest returns in 2014. While Slovenia, Estonia, and Lithuania had their lowest returns in 2014, Slovakia had the highest returns in the same year.

Figure 5.1 Plots of daily returns for major European frontier markets.

Table 5.3 shows the results of unit root tests with and without trend. The stationary of the return series is tested using the ADF and PP unit root tests. Both test results reveal that the return series for all stock market indices are stationary, indicating that there no unit roots in the stock returns. The null hypothesis was rejected, indicating that the return series follow a stationary process.

Table 5.3

Unit Root Test Results

	ADF		PP
	Intercept	Intercept and trend	Intercept	Intercept and trend
Croatia	−27.95947	−27.94037*,	−27.97141	−27.95374*,
Estonia	−25.80732	−26.05522*,	−25.84943	−26.04402*,
Lithuania	−27.78425	−20.84233*,	−25.71520	−25.75275*,
Romania	−22.65379	−22.63905*,	−22.73218	−22.71768*,
Slovakia	−23.70213	−23.69650*,	−36.00700	−36.01518*,
Slovenia	−24.92783	−24.91869*,	−24.87554	−24.86609*,

Notes: The symbols ***, **, * denote significance at the 1, 5, and 10% levels, respectively.

Table 5.4 presents the GPH and GSP estimates of the d parameter, when m is T^0.5. There is no long memory property observed in the stock index returns for Croatia, Estonia, and Lithuania. Both GHP and GSP tests provide consistent results of long memory for the countries of Romania and Slovenia. The estimate of the d parameter for Romania is strong and significant at the 1% level and the d parameter for Slovenia is weak and significant at 10%. The existence of long memory in the return series suggests that investment in stock market indices for Romania and Slovenia might not be a good hedge to achieve portfolio diversification. The long memory estimate for Slovakia is negative and significant at a 1% level. The negative sign of the d parameter shows antipersistence. The process is said to exhibit intermediate memory.

Table 5.4

Long Memory Tests

	Returns		Squared returns
	GPH; T^0.5	GSP; T^0.5	GPH; T^0.5	GSP; T^0.5
Croatia	0.00815062 [0.8440]	−0.00506162 [0.8570]	0.0665235 [0.1083]	0.0535853]* [0.0564]
Estonia	0.0423171 [0.3069]	0.0309044 [0.2711]	0.117067]*, [0.0047]	0.0857316]*, [0.0023]
Lithuania	−0.0042532 [0.9182]	0.024926 [0.3748]	−0.0042532 [0.9182]	0.024926 [0.3748]
Romania	0.14036]*, [0.0007]	0.133948]*, [0.0000]	0.148339]*, [0.0004]	0.12473]*, [0.0000]
Slovakia	−0.298807]*, [0.0000]	−0.224616]*, [0.0000]	0.149623]*, [0.0003]	0.193929]*, [0.0000]
Slovenia	0.0739497]* [0.0741]	0.0391964 [0.1628]	0.147158]*, [0.0004]	0.11258]*, [0.0001]

Notes: The t-value is in brackets [ ]. The symbols ***, **, * denote significance at the 1, 5, and 10% levels, respectively.

Furthermore, the findings show evidence of long memory property in the volatility of stock index returns for Estonia, Romania, Slovakia, and Slovenia. The existence of long memory contradicts the EMH that the future return and volatility values are unpredictable. Overall, no long memory property is observed for Croatia.

Table 5.5 shows the estimates of ARFIMA–FIGARCH results. With respect to the results of AIC and BIC information criteria, we select one lag for all the specifications. The findings show that the long memory parameter d_m in the mean equation is insignificant for all countries. Furthermore, the findings suggest that with the exception of Slovakia, the long memory parameters in the conditional volatility processes (d_v) are positive. However, the coefficients of long memory parameters are significant only for Slovenia and Romania. The magnitude of significance level is high for Slovenia (significant at a 1% level) and low for Romania (significant at a 10% level). The positive long memory evidence for Slovenia and Romania suggests that volatility processes display little tendency to revert toward the volatility mean. The evidence of long memory in volatility indicates that uncertainty or risk is an important determinant of the behavior of daily stock data in the Slovenian and Romanian stock markets. The results of the ARFIMA model support long memory behavior in conditional volatility for Slovenia and Romania. This is consistent with the evidence presented in Table 5.3.

Table 5.5

ARFIMA–FIGARCH(1,d,1) Class Model

	Slovenia	Slovakia	Romania	Croatia	Estonia	Lithuania
Conditional mean
C	0.000395 (0.1769)	0.0001456** (0.0247)	0.000465* (0.0717)	−0.000269 (0.4137)	−0.000017 (0.9508)	0.000211 (0.3102)
d_m	−0.068791 (0.1604)	−0.100080 (0.1735)	−0.114411 (0.3868)	0.027604 (0.4976)	0.007621 (0.9024)	−0.011860 (0.8014)
AR(1)	−0.035997 (0.9028)	0.096879 (0.4567)	0.574559** (0.0130)	−0.846473*** (0.0000)	−0.238688 (0.5153)	−0.294939 (0.1528)
MA(1)	0.158404 (0.5846)	−0.330578** (0.0156)	−0.328688** (0.0382)	0.794569*** (0.0000)	0.188425 (0.6836)	0.408056** (0.0258)
Conditional variance
C(10⁴)	0.574119*** (0.0008)	13.744155** (0.0336)	19.895173 (0.1100)	3.649363 (0.5573)	12.391449 (0.1403)	1.564410 (0.4429)
d_v	0.119174*** (0.0079)	−0.133288 (0.1781)	0.225115* (0.0599)	0.238110 (0.4423)	0.131641 (0.2392)	0.157808 (0.5879)
α	−0.230295 (0.1930)	0.632763*** (0.0000)	−0.221531 (0.2172)	0.651666*** (0.0002)	0.459273*** (0.0063)	0.851163*** (0.0000)
β	−0.323818** (0.0174)	0.326963** (0.0102)	−0.063580 (0.7626)	0.758495*** (0.0000)	0.284746* (0.0635)	0.877624*** (0.0000)
AIC	−6.293397	−8.420499	−6.549453	−6.919513	−7.044397	−7.494659
BIC	−6.237220	−8.364321	−6.493276	−6.863336	−6.988219	−7.438482
ARCH (10)	1.5102	0.066072	0.42443	0.66391	1.4950	0.25127

Notes: Table 5.5 reports the results of the ARFIMA–FIGARCH class model for daily index returns. C, C(10⁴), d_m, and d_v refer to the constants and LM parameters of the mean and variance equations, respectively. Robust standard errors are given in parenthesis. The symbol *** indicates significance at 1%, ** indicates significance at 5%, and * indicates significance at 10%. The Gaussian distribution is reported.

6. Summary and Conclusions

In recent years, frontier markets have been attracting increasing interest due to their diversification potential for investors. In particular, frontier markets in Europe are attractive, as they have EU membership. Since these markets are subject to economic reforms and EU regulations, understanding these markets would help investors, portfolio managers, and policymakers benefit from portfolio diversification.

This paper investigates the long memory property in stock returns and volatility of six European frontier markets—Slovenia, Slovakia, Romania, Croatia, Estonia, and Lithuania—for the period 2012–14. The findings show evidence of synchronized behavior among all return series. Most of the stock indices experienced the lowest returns in 2014.

The findings show long memory property in stock returns for only Romania and Slovenia. There is no long memory property in the stock returns for Croatia, Estonia, and Lithuania. For the volatility series, the results present evidence of a long memory property for Estonia, Romania, Slovakia, and Slovenia. Furthermore, the ARFIMA–FIGARCH model is applied to model the volatility persistence. The results show that the ARFIMA–FIGARCH model captures the long memory property in the conditional volatility for Romania and Slovenia. This result is consistent with the findings of the parametric and semiparametric long memory test results. The evidence of long memory implies that investors can exploit predictability and earn speculative returns through using past stock return information in these two stock markets.

Overall, the findings of this paper have important implications for understanding the frontier markets in Europe, which will be of great interest to investors, policymakers and regulators, as these markets have diversification potential and are subject to EU regulations.

..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.

Table of Contents for Chapter 5: Are Frontier Markets Worth the Risk?

Create new playlist

Sign In

Sign Up