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 12

Investors’ Herding in Frontier Markets: Evidence From Mongolia

A. Erdenetsogt^*

V. Kallinterakis^**
^* ABJYA LLC Brokerage Company, Ulaanbaatar, Mongolia
^** University of Liverpool, Management School, Liverpool, United Kingdom

Abstract

This study investigates investor herding in the Mongolian Stock Exchange, one of the world’s rapidly growing frontier markets. We find that investors in Mongolia herded significantly during the Dec. 1999–May 2012 period, with their herding significance persisting irrespective of the market’s direction, the level and day-to-day change of the market’s volume, and the US market’s dynamics. Sorting market returns by their sizes, we find that herding is found to be significant during extreme positive and extreme negative market returns. Moreover, investors in Mongolia herded in periods outside of, yet not during, the 2008 financial crisis.

Keywords

herding

global financial crisis

illiquidity

market returns

market volume

US market dynamics

Mongolia

1. Introduction

Herd behavior as a trading practice has been studied in the context of both developed and emerging markets for different investor types (individual and institutional) over various periods of time. However, recent years have witnessed the advent of a new category of markets, which have been collectively dubbed “frontier,” and for which very little is known concerning the behavior of their investors in general and herd behavior in particular. We aim at contributing in this area by investigating the presence of herding in the Mongolian Stock Exchange (MSE) to gauge whether it is significant, and whether its significance varies with a series of domestic market conditions related to market returns and market volume. What is more, we will investigate whether US market dynamics confer an effect over herding in Mongolia, and finally, we will test whether the 2008 global financial crisis has affected the presence of herding there.

In the wider context of behavioral finance, herding represents the tendency of individuals to mimic the actions of others following the interactive observation of each other’s actions (Hirshleifer and Teoh, 2003). The practice of herding presupposes that individuals follow the behavior of others with disregard for their own private signals or the prevailing market fundamentals. Holmes et al. (2013) and Gavriilidis et al. (2013) have argued that such a practice could be driven by intent or could be completely spurious in nature. Intentional herding is assumed to be motivated by situations entailing asymmetry among market participants. Such asymmetry can be of an informational nature (Devenow and Welch, 1996; Teraji, 2003), in which case less informed investors choose to imitate the trades of better informed ones in order to free ride on the informational advantage of the latter. Widespread mimicry of this type can erode the value of the public pool of information, eventually leading to cascading phenomena (Banerjee, 1992; Bikhchandani et al., 1992), whereby people simply mimic the actions of others because they consider others to be informed. Another type of asymmetry prompting herding—specifically among investment professionals (eg, fund managers and financial analysts)—is professional (Scharfstein and Stein, 1990; Trueman, 1994; Welch, 2000; Clement and Tse, 2005). Fund managers of below-average skills can be expected to be tempted to mimic the trades of their better-able peers; the reason for this is that fund managers are normally assessed on a relative basis (vs each other), hence imitating “good” peers benefits “bad” managers, as it allows them to improve their professional image when their assessments are due.

While the above asymmetries of their trading environments can prompt investors to herd intentionally (with the purpose of extracting some benefit from herding), unintentional—also known as “spurious” (Bikhchandani and Sharma, 2001)—herding is normally the product of commonalities observed in the trading environment. Relative homogeneity among investors, for example, can lead them to exhibit similarities in their trades with no imitative or interactive observations being present. Such relative homogeneity can manifest itself through investors bearing similar educational backgrounds or professional qualifications, thus receiving similar signals or interpreting them similarly while being subject to a uniform regulatory environment (De Bondt and Teh, 1997; Wermers, 1999; Voronkova and Bohl, 2005). Style investing (Bennett et al., 2003) can also culminate in spurious herding; if several investors, for example, practice the same investment strategy, their trades can be expected to be correlated without, however, this being the result of direct imitation among them.

Herding was the focus of ample empirical research throughout the past two decades, with research indicating the presence of some patterns across markets. Overall, funds in emerging markets tend to exhibit stronger herding behavior than that of their peers in developed ones,^a while retail investors have also been found to herd significantly (Kumar and Lee, 2006; Dorn et al., 2008; Kumar, 2009). Herding further appears to be more pronounced for the largest (Wylie, 2005; Walter and Weber, 2006) and the smallest (Lakonishok et al., 1992; Wermers, 1999; Chang et al., 2000) capitalization stocks (ie, there exists a size effect in herding^b) and has also been documented to exhibit industry effects (see eg, the studies by Gavriilidis et al., 2013; Gebka and Wohar, 2013).

Although researchers have investigated herding in several developed and emerging markets, there has been very little attention to the specific market segment of frontier markets. An interesting feature of frontier markets is the wide heterogeneity characterizing them; indeed, their ranks include high-income countries, such as Qatar and Bahrain, alongside several very poor sub-Saharan countries. Despite the lack of a uniform definition for frontier markets, they tend to entail one or more of the following features: nascent financial development; very small market capitalization and volume; and low but rising per capita income due to rapid economic growth, which is often stimulated by heavy reliance on natural resources (Umland, 2008; Behar and Hest, 2010; De Groot et al., 2012). Most of these markets have restrictions on the entry and trade of foreign investors; however, the relative illiquidity characterizing these markets is a key deterrent to foreign investors’ entry even where such restrictions are absent. In general, frontier markets are treated as high-risk destinations, with many of them being in transitional stages of political and/or economic development, having moved since the 1990s either from central planning to market economies or from illiberal regimes to democratic systems.

The early levels of financial development in most frontier markets can be expected to be reflected through the inadequate institutional designs of their stock exchanges, and this is a fact capable of enhancing informational uncertainty in these markets, giving rise to herding among their investors. If investors in such a market feel concerned about the quality of public information or the enforceability of disclosure rules, it is reasonable to expect that they might start mimicking their peers if they consider their peers’ actions to be informative. This can be expected to be particularly true for these markets’ retail investors, who, due to the relatively young age of most frontier stock exchanges, lack investment experience. However, institutional investors (domestic as well as foreign ones) may also consider herding a viable strategy in order to counter this high information risk; what is more, herding by institutional investors may also be motivated by these markets’ high liquidity risk, since (as mentioned earlier) frontier markets tend to be characterized by rather low capitalization and trading volume.^c

This study contributes to research on herding in frontier markets by examining herding in the context of the MSE during the Dec. 1999–May 2012 period. Our study aims at addressing the following research questions:

• Do investors herd in Mongolia?

• Does herding in Mongolia remain significant, controlling for various domestic market variables (market returns; market volume)?

• Does herding in Mongolia remain significant when controlling for the dynamics of the US market?

• Does the significance of herding in Mongolia vary within as opposed to outside the recent 2008 global financial crisis?

In summary, we report that there is evidence supporting the presence of herding in the Mongolian market, irrespective of the market’s performance (ie, positive/negative market returns), the level of volume, the day-to-day change in volume, and the US market’s dynamics. Controlling for the size of market returns (the market’s performance) presents us with a different pattern, as herding is found to be significant only for extreme positive and extreme negative market returns. Finally, investors in Mongolia herded outside, but not during, the period of the 2008 financial crisis.

The rest of our study is structured as follows: the next section provides a brief summary of the evolution of the MSE since the country’s transition to a market economy in the 1990s, while Section 3 presents the data used in this study and the empirical design employed, along with some descriptive statistics. Section 4 discusses the results and provides some concluding remarks, coupled with the implications of our study for various market participants.

2. Mongolian Stock Exchange: A Brief Overview

The MSE was established in 1991 as part of the government’s privatization policy amid the country’s transition from central planning to a market economy. A total of 96.1 million shares worth a total of $7 million from 475 state-owned enterprises were successfully placed on the primary market during the 1992–95 period; this led to the kick starting of secondary market operations, initially via 29 brokerage firms. By the end of 2011, total market capitalization had expanded to $1.7 billion; at the same time, the value of total transactions on the MSE in that year was $270 million, of which 68% related to government bonds, 31% to equities, and 1% to corporate bonds. Government bonds are not traded regularly; however, when their sale is announced, domestic investors (mainly commercial banks) dominate their subscriptions. Equity transactions grew by 75% in 2011, and their value is almost evenly split between local (52%) and foreign (48%) investors. Even though Mongolia has a bank-based financial system, its capital market has experienced rapid development over the years, supported by steady economic growth that is courtesy of the country’s abundant natural resources and increasing foreign investments. In 2012 a total of 1,519 institutional^d and 563,000 individual investors were registered with the Securities Clearing House and Central Depository, with the bulk of equity trading (70%) being in the hands of institutional investors.

The overall market is characterized by illiquidity and very high concentration. Assuming the 2011 total transaction value of $270 million above, and given that 31% of it is related to equities, this implies that equity trading had an annual value totaling around $84 million for a total of over 300 listed stocks, that is, the average annual value traded per stock was just under $250,000, a notably low value. Most listed firms on the MSE are actually very small in size^e (as of May 2012, about 60% of the listed companies’ market capitalization was less than $100,000, down from 90% in 1999), while the fact that the average monthly salary in Mongolia is around $150 obviously deters retail investors from substantially participating in their market’s trading volume. On top of that, the brokerage industry is also concentrated, with the five largest brokerage firms making up around 71% of total equity transactions and 95% of government and corporate bond transactions.

Fig. 12.1 presents the evolution of the TOP-20 index (the market’s main index, encompassing the twenty largest listed stocks) Dec. 9, 1999–May 8, 2012. As Fig. 12.1 indicates, the index grew slowly during the years up to 2006 (from around 250 units in Dec. 1999 to around 1200 units in Aug. 2006), followed by a dramatic surge afterwards, which (with some fluctuations) was maintained until the spring of 2008 (index value then was around 13,000 units), when it started declining for about a year. By Mar. 2009 the index had begun rallying again, this time reaching a dazzling peak of almost 33,000 units in Feb. 2011, before crashing afterwards to around 20,000 units in May of that year and hovering around that value until the spring of 2012.

Figure 12.1 TOP20 index evolution (9/12/1999–8/5/2012).

3. Data and Methodology

Linking herding to the relationship between the cross-sectional dispersion of equity returns and the return of the market was first empirically calibrated by Christie and Huang (1995), who proposed the following test to detect herding:

$C S S D_{t} = α_{0} + α_{1} D_{t}^{U} + α_{2} D_{t}^{L} + ɛ_{t}$ $C S S D_{t} = α_{0} + α_{1} D_{t}^{U} + α_{2} D_{t}^{L} + ɛ_{t}$

(12.1)

In Eq. 12.1,

D_{t}^{U}

$D_{t}^{U}$

is a dummy variable that assumes a value of 1 if the market return rests in the extreme upper tail of the market-return distribution; otherwise it assumes a value of 0. Similarly,

D_{t}^{L}

$D_{t}^{L}$

is a dummy variable assuming the value of 1 if the market return rests in the extreme lower tail of the market-return distribution; otherwise it assumes the value of 0. CSSD is the cross-sectional standard deviation, calculated as:

$C S S D_{t} = \sqrt{\frac{\sum_{i = 1}^{n} {(r_{i, t} - r_{m, t})}^{2}}{n - 1}}$ $C S S D_{t} = \sqrt{\frac{\sum_{i = 1}^{n} {(r_{i, t} - r_{m, t})}^{2}}{n - 1}}$

(12.2)

In Eq. 12.2, r_i,t is the return of security i in day t, r_m,t is the equal-weighted return of the market in day t, and n represents the number of stocks traded on day t.

The rationale of the Christie and Huang (1995) model was rather straightforward. On the one hand, the presence of herding in the market would be expected to lead stocks to track the overall market’s return, thus leading the cross-sectional dispersion of returns to decline; herding would also be expected to lead to abnormally high absolute market returns. On the other hand, given the differing sensitivity of each stock to market movements, an increase in a market’s absolute returns would be expected to lead to a linear increase in the cross-sectional dispersion of returns. If herding is indeed present during extreme periods, the dummies’ coefficients (α₁ and/or α₂) should be negative and significant, indicating that the CSSD decreases during extreme market periods.

The issue with the aforementioned approach is threefold. First of all, as Economou et al. (2011) noted, outliers can easily introduce biases in the calculated CSSD. Second, the employment of dummies to capture extreme market periods is rather crude (Hwang and Salmon, 2004), as it restricts the sample upon which herding is tested; herding, for example, may be present during periods of mild market returns and this is something the aforementioned model cannot identify. Third, there is little certainty that the relationship between the cross-sectional dispersion of returns and the absolute returns of the market will remain linear in the presence of herding, given evidence (Lux, 1995; Lux and Marchesi, 1999) indicating that herding is capable of introducing nonlinear dynamics in the market. To account for the aforementioned issues, Chang et al. (2000) proposed the following empirical design to test for the presence of herding:

$C S A D_{m, t} = α_{0} + α_{1} | r_{m, t} | + α_{2} r_{m, t}^{2} + ɛ_{t}$ $C S A D_{m, t} = α_{0} + α_{1} | r_{m, t} | + α_{2} r_{m, t}^{2} + ɛ_{t}$

(12.3)

In Eq. 12.3, CSAD is the cross-sectional absolute deviation of returns, calculated as:

$C S A D_{t} = \frac{1}{n} \sum_{i = 1}^{N} | r_{i, t} - r_{m, t} |$ $C S A D_{t} = \frac{1}{n} \sum_{i = 1}^{N} | r_{i, t} - r_{m, t} |$

(12.4)

The definition of the variables in Eqs. 12.3 and 12.4 is the same as in Eq. 12.2. The advantage of the Chang et al. (2000) approach over that of Christie and Huang (1995) is that it utilizes the entire market-return distribution (without resorting to dummies to arbitrarily identify extreme periods), while accounting for both the linear (through coefficient α₁) and the nonlinear (through coefficient α₂) part of the relationship between the CSAD and market returns. Rational asset-pricing assumptions would predict a positive value for α₁ and an insignificant value for α₂; however, in the presence of herding, we would expect α₂ to be significantly negative.

To test for the robustness of our findings from Eq. 12.3 under various domestic market conditions, we repeat our estimates, conditioning this equation upon the following two variables:

Market returns. We first reestimate Eq. 12.3 twice, once for days of positive market returns (upmarket days) and once for days of negative market returns (downmarket days)—the latter is proxied through r_m,t—as follows:

$C S A D_{m, t}^{UP} = α_{0}^{UP} + α_{1}^{UP} | r_{m, t} | + α_{2}^{UP} r_{m, t}^{2} + ɛ_{t}$ $C S A D_{m, t}^{UP} = α_{0}^{UP} + α_{1}^{UP} | r_{m, t} | + α_{2}^{UP} r_{m, t}^{2} + ɛ_{t}$

(12.5)

C S A D_{m, t}^{DOWN} = α_{0}^{DOWN} + α_{1}^{DOWN} | r_{m, t} | + α_{2}^{DOWN} r_{m, t}^{2} + ɛ_{t}

$C S A D_{m, t}^{DOWN} = α_{0}^{DOWN} + α_{1}^{DOWN} | r_{m, t} | + α_{2}^{DOWN} r_{m, t}^{2} + ɛ_{t}$

(12.6)

The superscripts ^UP and ^DOWN are used to denote that the equation is run for up and down market days, respectively. The purpose of this is to gauge whether herding varies in significance with the sign of market returns, given evidence from the literature suggesting that herding is more prevalent during periods of negative market performance.^f This may be due to investors’ greater risk aversion (selling with other investors when the market falls to minimize losses) or due to professional reasons (“bad” managers mimicking the trades of “good” managers during market slumps can argue that they made the correct investment decisions—essentially those they copied from their “good” peers—and blame the adverse market conditions for their losses). We also rank the values of r_m,t in ascending order and split them into five quintiles (Q1–Q5),^g each with an equal number of observations, and reestimate Eq. 12.3 for each of these quintiles; the rationale for this is to gauge whether herding significance, aside from the sign, also varies with the size of market returns.

Market volume. We calculate the total daily market volume as the sum of the daily volumes of trade of all actively traded stocks and reestimate Eq. 12.3 separately for days of increasing and days of decreasing market volume,^h as follows:

C S A D_{m, t}^{UPV} = α_{0}^{UPV} + α_{1}^{UPV} | r_{m, t} | + α_{2}^{UPV} r_{m, t}^{2} + ɛ_{t}

$C S A D_{m, t}^{UPV} = α_{0}^{UPV} + α_{1}^{UPV} | r_{m, t} | + α_{2}^{UPV} r_{m, t}^{2} + ɛ_{t}$

(12.7)

C S A D_{m, t}^{DOWNV} = α_{0}^{DOWNV} + α_{1}^{DOWNV} | r_{m, t} | + α_{2}^{DOWNV} r_{m, t}^{2} + ɛ_{t}

$C S A D_{m, t}^{DOWNV} = α_{0}^{DOWNV} + α_{1}^{DOWNV} | r_{m, t} | + α_{2}^{DOWNV} r_{m, t}^{2} + ɛ_{t}$

(12.8)

The superscripts ^UPV and ^DOWNV are used to denote that the equation is run for increasing or decreasing market volume days, respectively. We also rank this daily aggregate volume series in ascending order, split it into five quintiles (Q1–Q5),ⁱ and estimate Eq. 12.3 for each quintile in order to gauge the impact of different volume sizes on herding.

In view of the established (Chiang and Zheng, 2010) impact of US market dynamics over herding worldwide, we employ the following specification to test whether such an impact holds for the Mongolian market:

$C S A D_{m, t} = α_{0} + α_{1} | r_{m, t} | + α_{2} r_{m, t}^{2} + α_{3} r_{S & P500, t}^{2} + ɛ_{t}$ $C S A D_{m, t} = α_{0} + α_{1} | r_{m, t} | + α_{2} r_{m, t}^{2} + α_{3} r_{S & P500, t}^{2} + ɛ_{t}$

(12.9)

In the previous equation,

r_{S & P500, t}^{2}

$r_{S & P500, t}^{2}$

is the squared daily return of the S&P 500 index, proxying here for US market dynamics. A significant value for α₃ would suggest a significant effect of the US market over the Mongolian CSAD; if the value of α₃ is both significant and negative, this would essentially suggest that the US market motivates herding in the Mongolian one.

Finally, we test for the presence of herding in Mongolia during and outside the 2008 global financial crisis period using the following specification:

$C S A D_{m, t} = α_{0} + α_{1} D^{CRISIS} | r_{m, t} | + α_{2} (1 - D^{CRISIS}) | r_{m, t} | + α_{3} D^{CRISIS} r_{m, t}^{2} + α_{4} (1 - D^{CRISIS}) r_{m, t}^{2} + ɛ_{t}$ $C S A D_{m, t} = α_{0} + α_{1} D^{CRISIS} | r_{m, t} | + α_{2} (1 - D^{CRISIS}) | r_{m, t} | + α_{3} D^{CRISIS} r_{m, t}^{2} + α_{4} (1 - D^{CRISIS}) r_{m, t}^{2} + ɛ_{t}$

(12.10)

Here, D^CRISIS is a dummy that assumes a value of 1 from Aug. 1, 2008 onwards, and 0 otherwise.

Our data covers the period of Dec. 10, 1999–May 8, 2012 and includes daily closing prices and trading volumes of 341 stocks; all stocks—both active and dead/suspended—listed on the MSE during that period are included, thus mitigating the survivorship bias. Data were obtained from the MSE. Table 12.1 presents some descriptive statistics related to CSAD and r_m,t,^j as well as the trading activity of the MSE. If there is one thing that is very clearly manifested through this table, it is the very small fraction of actively traded stocks on average: a mere 18, or 5.3% of the total number of listed stocks during that period (341). This is indicative of a market in which about one in 20 stocks trades actively on average every day, something in line with our expectations from frontier stock exchanges.

Table 12.1

Descriptive Statistics

	Mean	Standard deviation
CSAD_m,t	0.0117	0.0150
r_m,t	0.0268	0.2948
Total number of stocks in sample	341
Average total market trading volume	225,547
Average daily number of actively traded stocks	18

Notes: Table 12.1 presents some descriptive statistics on the variables used in the estimation of the Chang et al. (2000) model and its variants employed in this study. The variable r_m,t refers to the Mongolian market’s average return, calculated as the equal-weighted average of all listed stocks’ returns; the variable CSAD_m,t refers to the cross-sectional absolute deviation of returns for the Mongolian market. All returns are calculated as first differences of logarithmic prices. Total market-trading volume is calculated daily as the sum of the daily volumes of individual shares. All data were obtained from the MSE and cover the period Dec. 10, 1999–May 8, 2012.

4. Results and Conclusion

Table 12.2 presents the estimated coefficients from Eq. 12.3, where herding is estimated unconditionally. As the results indicate, α₁ is significantly^k positive (1.1905), denoting that the cross-sectional absolute deviation of returns bears a linearly increasing relationship with average market returns, in line with rational asset-pricing predictions. The coefficient α₂ is significantly negative (−0.2629), suggesting that investors herd significantly in the Mongolian market.

Table 12.2

Estimates of Herding in Mongolia for the Full Sample Period (Unconditional Herding)

α₀	α₁	α₂	Adjusted R²
0.1107 (0.0000)	1.1905 (0.0000)	−0.2629 (0.0000)	0.5101

Notes: Table 12.2 presents the estimates from the following equation:

$C S A D_{m, t} = α_{0} + α_{1} | r_{m, t} | + α_{2} r_{m, t}^{2} + ɛ_{t}$ $C S A D_{m, t} = α_{0} + α_{1} | r_{m, t} | + α_{2} r_{m, t}^{2} + ɛ_{t}$

All estimates’ p-values are reported in parentheses. The variable r_m,t refers to the Mongolian market’s average return, calculated as the equal-weighted average of all listed stocks’ returns; the variable CSAD_m,t refers to the cross-sectional absolute deviation of returns for the Mongolian market. All returns are calculated as first differences of logarithmic prices.

Controlling for market returns confers little effect over our results, since, as Panel A in Table 12.3 shows, the values of α₂ from Eqs. 12.5 and 12.6 are significantly negative during both up- and downmarket days. Since

| α_{2}^{UP} | > | α_{2}^{DOWN} |

$| α_{2}^{UP} | > | α_{2}^{DOWN} |$

, this implies that herding is stronger during upmarket days compared to downmarket days, with the difference between the two estimates being statistically significant. The presence of stronger herding during up markets may be the product of investors’ optimism, mainly on the part of unsophisticated/inexperienced investors (Grinblatt and Keloharju, 2001; Lamont and Thaler, 2003), whose effect can be quite substantial in frontier markets, as explained earlier. Panel B in Table 12.3 contains the results from estimating Eq. 12.3 for each of the return quintiles discussed in the previous section. As the results indicate, herding is significant only for Quintiles 1 and 5, thus showing that it is only during extreme positive or negative market days that herding appears significant.^l The value of α₂ for Q1 (−0.6481) is—in absolute terms—larger than that for Q2 (−0.1962), in line with what we witnessed in Panel A about herding being stronger during up markets. Overall, results from Table 12.3 suggest that more than the sign, it is the size of market returns that is crucial for the identification of herding in Mongolia. The fact that investors in that market herd when returns grow extreme can perhaps be attributed either to uncertainty (the case of extreme negative market returns prompting investors to liquidate their positions to mitigate further losses) or euphoria (the case of extreme positive market returns driving investors to jump onto the bandwagon).

Table 12.3

Estimates of Herding in Mongolia for the Full Sample Period (Conditioning Herding Upon Market Returns)

Panel A: herding conditioned upon the sign of market returns (positive/negative)
	$α_{0}^{UP}$ $α_{0}^{UP}$	$α_{1}^{UP}$ $α_{1}^{UP}$	$α_{2}^{UP}$ $α_{2}^{UP}$	Adjusted R²
Up markets	0.0679 (0.0000)	1.5743 (0.0000)	−0.6842 (0.0000)	0.5415

	$α_{0}^{DOWN}$ $α_{0}^{DOWN}$	$α_{1}^{DOWN}$ $α_{1}^{DOWN}$	$α_{2}^{DOWN}$ $α_{2}^{DOWN}$	Adjusted R²
Down markets	0.1306 (0.0000)	1.2083 (0.0000)	−0.2508 (0.0000)	0.5112
F₁ (test statistic) ( $α_{1}^{UP} = α_{1}^{DOWN}$ $α_{1}^{UP} = α_{1}^{DOWN}$ )	117.9750 (0.0000)
F₂ (test statistic) ( $α_{2}^{UP} = α_{2}^{DOWN}$ $α_{2}^{UP} = α_{2}^{DOWN}$ )	1573.1214 (0.0000)

Panel B: herding conditioned upon the size of market returns (Q1–Q5)
	α₀	α₁	α₂	Adjusted R²
Q1	0.0730 (0.0188)	1.5268 (0.0000)	−0.6481 (0.0000)	0.3461
Q2	−0.0677 (0.1859)	5.2133 (0.0009)	−17.5771 (0.0862)	0.1448
Q3	0.0875 (0.0000)	−1.8094 (0.3208)	67.4903 (0.2598)	−0.0010
Q4	0.0750 (0.0000)	2.6716 (0.0745)	−26.9633 (0.2944)	0.0124
Q5	0.2467 (0.0000)	0.9362 (0.0000)	−0.1962 (0.0000)	0.3458

Notes: Table 12.3, Panel A presents the estimates from the following equations:

$C S A D_{m, t} = α_{0}^{UP} + α_{1}^{UP} | r_{m, t} | + α_{2}^{UP} r_{m, t}^{2} + ɛ_{t}$ $C S A D_{m, t} = α_{0}^{UP} + α_{1}^{UP} | r_{m, t} | + α_{2}^{UP} r_{m, t}^{2} + ɛ_{t}$

$C S A D_{m, t} = α_{0}^{DOWN} + α_{1}^{DOWN} | r_{m, t} | + α_{2}^{DOWN} r_{m, t}^{2} + ɛ_{t}$ $C S A D_{m, t} = α_{0}^{DOWN} + α_{1}^{DOWN} | r_{m, t} | + α_{2}^{DOWN} r_{m, t}^{2} + ɛ_{t}$

The superscript ^UP (^DOWN) is used to denote that the equation is run for up (down) market days. F₁ and F₂ statistics test, respectively, the following null hypotheses: $α_{1}^{UP} = α_{1}^{DOWN}$ $α_{1}^{UP} = α_{1}^{DOWN}$ and $α_{2}^{UP} = α_{2}^{DOWN}$ $α_{2}^{UP} = α_{2}^{DOWN}$ . Table 12.3, Panel B presents the estimates from the following equation:

$C S A D_{m, t} = α_{0} + α_{1} | r_{m, t} | + α_{2} r_{m, t}^{2} + ɛ_{t}$ $C S A D_{m, t} = α_{0} + α_{1} | r_{m, t} | + α_{2} r_{m, t}^{2} + ɛ_{t}$

The equation is estimated for Quintiles 1–5, which are constructed as follows: having calculated r_m,t, we rank its values in ascending order and split them into five quintiles (Q1–Q5) of equal number of observations.
All estimates’ p-values are reported in parentheses. The variable r_m,t refers to the Mongolian market’s average return, calculated as the equal-weighted average of all listed stocks’ returns; the variable CSAD_m,t refers to the cross-sectional absolute deviation of returns for the Mongolian market. All returns are calculated as first differences of logarithmic prices.

Controlling for market volume demonstrates that herding is significant irrespective of whether volume increases or decreases at the aggregate level; as Table 12.4 (Panel A) shows, the values of α₂ from Eqs. 12.7 and 12.8 are significantly negative during both increasing and decreasing volume days. Since

| α_{2}^{UPV} | < | α_{2}^{DOWNV} |

$| α_{2}^{UPV} | < | α_{2}^{DOWNV} |$

, this implies that herding is stronger during decreasing, compared to increasing, volume days, with the difference between the two estimates being statistically significant. This is possibly the result of investors following each other into and out of the same stocks with greater propensity when the overall volume in the market is lower, as a means of avoiding liquidity risk. When viewing the estimates from the volume quintiles (Panel B in Table 12.4), we notice that α₂ is significantly negative for all of them, thus again confirming that herding significance in Mongolia is not a function of the market’s total trading activity. This is a very interesting finding and is probably due to the fact that volume in that market is not just low, but also concentrated among very few stocks. With the average daily fraction of actively traded stocks hovering around 5% (around 18 stocks; Table 12.1), it is highly likely that investors in that market trade not whenever they want, but whenever the number of investors willing to trade is sufficient enough to allow order execution. Under such circumstances, herding should come as no surprise, given that the already limited volume is allocated among a very small number of stocks.^m

Table 12.4

Estimates of Herding in Mongolia for the Full Sample Period (Conditioning Herding Upon Market Volume)

Panel A: herding conditioned upon the day-to-day change of market volume (increasing/decreasing)
	$α_{0}^{UP}$ $α_{0}^{UP}$	$α_{1}^{UP}$ $α_{1}^{UP}$	$α_{2}^{UP}$ $α_{2}^{UP}$	Adjusted R²
Up-volume markets	0.1120 (0.0000)	1.2745 (0.0000)	−0.2567 (0.0000)	0.5498

	$α_{0}^{DOWN}$ $α_{0}^{DOWN}$	$α_{1}^{DOWN}$ $α_{1}^{DOWN}$	$α_{2}^{DOWN}$ $α_{2}^{DOWN}$	Adjusted R²
Down-volume markets	0.0806 (0.0000)	1.4708 (0.0000)	−0.6190 (0.0000)	0.5124
F₁ (test statistic) ( $α_{1}^{UPV} = α_{1}^{DOWNV}$ $α_{1}^{UPV} = α_{1}^{DOWNV}$ )	19.5318 (0.0000)
F₂ (test statistic) ( $α_{2}^{UPV} = α_{2}^{DOWNV}$ $α_{2}^{UPV} = α_{2}^{DOWNV}$ )	110.1038 (0.0000)

Panel B: herding conditioned upon the size of market volume (Q1–Q5)
	α₀	α₁	α₂	Adjusted R²
Q1	0.0290 (0.0000)	1.6046 (0.0000)	−0.5811 (0.0000)	0.6489
Q2	0.0710 (0.0000)	1.4615 (0.0000)	−0.2796 (0.0000)	0.6458
Q3	0.0787 (0.0000)	1.7696 (0.0000)	−0.9144 (0.0000)	0.5907
Q4	0.1513 (0.0000)	1.3753 (0.0000)	−0.4272 (0.0000)	0.5181
Q5	0.1265 (0.0000)	1.2958 (0.0000)	−0.5601 (0.0000)	0.3815

Notes: Table 12.4, Panel A presents the estimates from the following equations:

$C S A D_{m, t} = α_{0}^{UPV} + α_{1}^{UPV} | r_{m, t} | + α_{2}^{UPV} r_{m, t}^{2} + ɛ_{t}$ $C S A D_{m, t} = α_{0}^{UPV} + α_{1}^{UPV} | r_{m, t} | + α_{2}^{UPV} r_{m, t}^{2} + ɛ_{t}$

$C S A D_{m, t} = α_{0}^{DOWNV} + α_{1}^{DOWNV} | r_{m, t} | + α_{2}^{DOWNV} r_{m, t}^{2} + ɛ_{t}$ $C S A D_{m, t} = α_{0}^{DOWNV} + α_{1}^{DOWNV} | r_{m, t} | + α_{2}^{DOWNV} r_{m, t}^{2} + ɛ_{t}$

The superscript ^UPV (^DOWNV) is used to denote that the equation is run for increasing (decreasing) market volume days. The F₁ and F₂ statistics test, respectively, the following null hypotheses: $α_{1}^{UPV} = α_{1}^{DOWNV}$ $α_{1}^{UPV} = α_{1}^{DOWNV}$ and $α_{2}^{UPV} = α_{2}^{DOWNV}$ $α_{2}^{UPV} = α_{2}^{DOWNV}$

Table 12.4, Panel B presents the estimates from the following equation:

$C S A D_{m, t} = α_{0} + α_{1} | r_{m, t} | + α_{2} r_{m, t}^{2} + ɛ_{t}$ $C S A D_{m, t} = α_{0} + α_{1} | r_{m, t} | + α_{2} r_{m, t}^{2} + ɛ_{t}$

The equation is estimated for Quintiles 1–5, which are constructed as follows: having calculated the aggregate (total) daily market volume by summing up the trading volumes of all listed stocks every day, we rank its values in ascending order and split them into five quintiles (Q1–Q5) of equal number of observations.

Table 12.5 presents the results from Eq. 12.9, from which it is obvious that the US market dynamics bear no effect whatsoever over the Mongolian market. The coefficient α₃ is insignificant, whereas α₂ remains significant and negative, confirming the robustly significant herding in that market. Given that Mongolia is a very small frontier market with limited integration to the global financial system, it should perhaps come as little surprise that the US market’s dynamics do not impact upon it.ⁿ

Table 12.5

Estimates of Herding in Mongolia for the Full Sample Period (Controlling for the Effect of US Market Dynamics)

α₀	α₁	α₂	α₃	Adjusted R²
0.1107 (0.0000)	1.907 (0.0000)	−0.2630 (0.0000)	−0.3754 (0.1843)	0.5102

Notes: Table 12.5 presents the estimates from the following equation:

All estimates’ p-values are reported in parentheses. The variable r_m,t refers to the Mongolian market’s average return, calculated as the equal-weighted average of all listed stocks’ returns; the variable r_S&P500,t is the daily return of the S&P 500 index; the variable CSAD_m,t refers to the cross-sectional absolute deviation of returns for the Mongolian market. All returns are calculated as first differences of logarithmic prices.

The 2008 financial crisis does indeed appear to have affected herding in Mongolia, as Table 12.6 indicates; the estimates from Eq. 12.10 show that, whereas α₄ is significantly negative (suggesting the presence of herding precrisis), α₃ is significantly positive (ie, herding was absent during the crisis). We experimented with alternative windows to proxy for the 2008 financial crisis (eg, Aug.–Dec. 2008), with results in all cases being similar to those of Table 12.6, thus confirming that investors in Mongolia herded outside, but not during, the 2008 financial crisis. As a result, the onset of this crisis appears to have dampened the herding tendencies of investors in Mongolia, possibly due to the new fundamentals revealed by the crisis, leading investors to trade based on them and away from the precrisis consensus (Borio, 2008).^o

Table 12.6

Estimates of Herding in Mongolia for the Full Sample Period (Controlling for the Effect of the 2008 Crisis)

α₀	α₁	α₂	α₃	α₄	Adjusted R²
0.1485 (0.0000)	−2.4681 (0.0000)	1.1144 (0.0000)	3.9059 (0.0000)	−0.2458 (0.0000)	0.5409

Notes: Table 12.6 presents the estimates from the following equation:

All estimates’ p-values are reported in parentheses. The variable r_m,t refers to the Mongolian market’s average return, calculated as the equal-weighted average of all listed stocks’ returns; D^CRISIS is a dummy assuming the value of 1 from Aug. 2008 onwards and 0 otherwise; the variable CSAD_m,t refers to the cross-sectional absolute deviation of returns for the Mongolian market. All returns are calculated as first differences of logarithmic prices.

In summary, we report that there is evidence supporting the presence of herding in the Mongolian market, irrespective of the sign of the market’s performance (ie, positive/negative market returns), the level of volume, the day-to-day change in volume, and the US market’s dynamics. Controlling for the size of the market’s performance presents us with a different pattern, as herding is found to be significant only for extreme positive and extreme negative market returns. Finally, investors in Mongolia herded outside, but not during, the 2008 financial crisis.

The evidence presented in this study is of key relevance to the investment community, in particular for those investors with a global outlook. Frontier markets exhibit little correlation with developed ones, and this naturally implies diversification benefits from investing in them. However, as the case of Mongolia indicates, foreign investors targeting these markets should be ready to embrace their liquidity risk and—most importantly—face extensive herding. Although the presence of herding should theoretically provide foreign investors with arbitrage opportunities (in the spirit of De Long et al., 1990), whether they will be able to take advantage of them in view of such low trading activity remains an open question.

Our results are also of particular interest to regulators and policy makers in Mongolia, as they denote a market where investors resort to imitation in their trades to a great extent, and this needs to be addressed in order to avoid the potential side effects of herding, including a rise in systemic risk and price destabilization. Although no explicit “counter-herding” measures have been prescribed in the relevant literature, one way to discourage herding among investors would be to improve the transparency of the market environment. Key to this attempt is the promotion of investors’ financial education, so that they are able to gain better understanding of the investment process and rely less on the influence of their peers. Improving disclosure rules (such as, eg, raising the reporting standards demanded in terms of financial statements, auditing, and insider dealing/trading) and their enforceability (via the stringent monitoring of their implementation) could lead to an increase in the quality of the market’s informational environment, allowing investors to place greater faith in public information. To this end, regulators should also focus on undertaking measures aimed at encouraging foreign investors’ participation, since the sophistication of overseas investors would help boost the market’s informational efficiency—and liquidity. Given the aforementioned issue of liquidity risk facing overseas traders, the introduction of indexed products, such as exchange-traded funds and index futures/options, would allow foreign investors to enter the market with greater ease, without having to search through a universe of hundreds of listed stocks with little volume or analysts’ coverage. Although the aforementioned observations are related to Mongolia given the focus of our study, they contain useful implications for the rest of the frontier markets, which—to varying degrees—are facing similar issues in their evolutionary process.

..................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 12: Investors’ Herding in Frontier Markets: Evidence From Mongolia

Create new playlist

Sign In

Sign Up

Abstract

Keywords

1. Introduction

2. Mongolian Stock Exchange: A Brief Overview

3. Data and Methodology

4. Results and Conclusion

Table of Contents for
Chapter 12: Investors’ Herding in Frontier Markets: Evidence From Mongolia