The stock price–volume relationship in emerging stock markets: the case of Latin America

International Journal of Forecasting 14 (1998) 215–225

The stock price–volume relationship in emerging stock markets: the case of Latin America Kemal Saatcioglu*, Laura T. Starks Department of Finance University of Texas at Austin, Austin, TX 78712, USA

Abstract This paper examines the stock price–volume relation in a set of Latin American markets. Using monthly index data, we first document a positive relation between volume and both the magnitude of price change and price change itself, a finding reported by many for developed markets. However, using a vector autoregression (VAR) analysis to test for Granger causality, we fail to find strong evidence on stock price changes leading volume. This is contrary to evidence reported by studies on developed markets. In fact, we find that in four of the six markets we look at, volume seems to lead stock price changes. Thus, we conclude that this set of emerging markets with different institutions and information flows than the developed markets, do not present similar stock price–volume lead–lag relation to the preponderance of studies employing U.S. data. The implication of these results is that differences in institutions and information flows in this set of emerging markets are important enough to affect the valuation process of equity securities and warrant further analysis.  1998 Elsevier Science B.V. Keywords: Capital markets; Causality; Emerging markets; Granger causality; Latin American equity markets; Price–volume relation; VAR; Vector autoregression

1. Introduction The relation between stock prices and trading volume in financial markets has received considerable attention over the past two decades. Although numerous studies have attempted to establish the empirical and theoretical structure of this relation, a consensus is yet to be reached. First, the empirical studies have employed individual stock and index data across varying time intervals and sample periods to reach divergent conclusions. For example, conclusions have ranged from no relation between weekly

*Corresponding author. Tel.: 11 512 4714368; fax: 11 512 4715073; e-mail: [email protected]

price changes and volume (Granger and Morgenstern, 1963)) to a positive correlation between monthly price changes and volume (Rogalski, 1978) to a positive relation between lagged absolute transaction price change and volume but not vice versa (Smirlock and Starks, 1988) to bidirectional nonlinear causality between returns and volume (Hiemstra and Jones, 1995). Second, the theoretical studies have attempted to set forth frameworks whose predictions would be in line with the observed relations.1 Although there has been extensive research into the empirical and theoretical aspects of the stock 1

See Karpoff (1987) for a review of the early literature on the stock price–volume relation.

0169-2070 / 98 / $19.00  1998 Elsevier Science B.V. All rights reserved. PII: S0169-2070( 98 )00028-4

216

K. Saatcioglu, L.T. Starks / International Journal of Forecasting 14 (1998) 215 – 225

price–volume relation, this research has focused almost exclusively on the well-developed financial markets, usually the U.S. markets. Given the divergent conclusions of this research, further insights should be obtainable through an investigation of an alternative set of financial markets, in particular, a set of emerging markets. The advantages of employing emerging markets for such a study are severalfold. Because of their generally low correlations with the more developed markets, the emerging markets present a separate data source, so that any datasnooping biases are lessened. In addition, the information flows in emerging markets are not equivalent to the information flows in the more developed markets and there are significant institutional differences across the markets.2 Information flows and institutions have been previously conjectured to have important implications on the stock price–volume relation. There are theoretical models that hypothesize a stock price–volume relation based on information flows and the existence of market institutions (see, for example, Copeland, 1976 and Jennings et al., 1981). Given these hypotheses, an empirical study using alternative markets should provide new insights into the relation. In this study, we focus on a set of markets (Latin American) that have a commonality in that they are all in the same hemisphere and all but one (Brazil) have the same language. The paper proceeds as follows. We first provide an overview of the previous research on the relation between price changes and volume in Section 2. After that, we describe the emerging markets data set used in this study in Section 3. In Section 4 we discuss the methodology and present the empirical results. Finally, we offer our conclusions in Section 5.

2. Research on the stock price–volume relation An understanding of the relation between stock prices and volume is important for a number of reasons as pointed out by Karpoff (1987). First, the empirical relation between returns and volume helps 2

For recent research on emerging markets and discussions of some of the differences between emerging and developed markets, see Aggarwal et al. (1993); Barry and Lockwood (1995); Divecha et al. (1992); Errunza (1994); Harvey (1995).

discriminate between competing theories on how information is disseminated in financial markets. Second, for event studies that use combinations of return and volume data to infer the information content of the event in question, the construction of the tests and the validity of the inferences depend on the joint distribution of returns and volume. Third, the return–volume relation is critical in assessing the distribution of returns themselves. For example, the mixture of distributions hypothesis has been employed to view the distribution of price changes (i.e., returns) as a finite-variance mixture of normal distributions where volume is the mixing variable (e.g., Epps and Epps, 1976). Fourth, a better understanding of the statistical structure of volume and return can help explain technical analysis (see Blume et al., 1994). Beyond these rationales, an early and continual motivator of empirical research on the stock price– volume relation has been a desire to determine whether two Wall Street adages are valid: (i) volume is relatively heavy in bull markets and light in bear markets, and (ii) it takes volume to make prices move. The first adage implies a positive correlation between volume and returns while the second one implies a positive correlation between volume and the magnitude of returns. In addition, the first adage implies that returns cause volume, whereas the second one implies the opposite. The early empirical research on the stock price– volume relation in financial markets primarily focuses on two of the empirical relations implied by these adages: (i) the correlation between volume (V ) and price change (DP) and (ii) the correlation between volume (V ) and the absolute value of the price change (uDPu). A couple of early studies use spectral analysis on weekly index data, and daily and transactions individual stock data. Both studies conclude that prices and volume are virtually unrelated and that price changes follow a random walk (Granger and Morgenstern, 1963; Godfrey et al., 1964). In contrast, using daily and hourly price changes for both market indices and individual stocks Crouch (1970a), (1970b) finds a positive correlation between volume and the magnitude of returns. Examining the relation between volume and returns, a positive contemporaneous correlation has

K. Saatcioglu, L.T. Starks / International Journal of Forecasting 14 (1998) 215 – 225

been found by Rogalski (1978) using monthly stock and warrant data and by Epps (1975), (1977) using transactions data. To explain such results, Epps proposes a theoretical framework consistent with his findings. This framework implies the ratio of volume to returns should be greater for price increases than for price decreases, which has been supported by empirical evidence in Smirlock and Starks (1985). More recent empirical work has investigated the lagged relation between price changes and volume. For example, Smirlock and Starks (1988), employing individual stock transactions data, document a strong positive lagged relation between volume and absolute price changes. Similarly, using daily data, Bhagat and Bhatia (1996) test for causality in both mean and variance and provide evidence that price changes lead volume, but no evidence that volume leads price changes. In addition, Hiemstra and Jones (1995) find a new result through the use of nonlinear Granger causality. They find a significant positive relation going in both directions between returns and volume. Two studies have examined the price–volume relation in markets outside of the United States. Using daily, weekly, and monthly series of different indices in the Tokyo Stock Exchange, Tse (1991) has mixed results for the relations between volume and returns (both DP and uDPu). He finds significant positive correlation in some series and not in others. He concludes that ‘‘the relationship between price changes and volumes in the market, if there is any, is weak.’’ Chan and Tse (1993) employ the multiple time series approach of Tiao and Box (1981) and show that ‘‘there is implicit positive correlation between price and volume through their residuals.’’ Given this mix of results, more information is needed about other financial markets outside of the United States.

3. Data Our Latin American stock market data derives from the 1995 Emerging Markets Database (EMDB) prepared and maintained by the International Finance Corporation (IFC). The IFC defines a stock market as ‘‘emerging’’ if it is located in a developing country, as defined by the World Bank’s GNP per

217

capita criterion for a developing country.3 For example, the cutoff rate applied to 1995 data was $8625. The 1995 EMDB includes data for over 1400 individual stocks in 25 countries. For ease of comparison and interpretation, we restrict the analysis to a set of six Latin American stock markets with at least $5 billion in market capitalization: Argentina, Brazil, Chile, Colombia, Mexico, and Venezuela. For each of the sample markets, EMDB reports the weekly and monthly value-weighted total return indices in local currency and U.S. dollars for the January 1986 to April 1995 time period. In our empirical tests, we employ the monthly valueweighted total return index in both U.S. dollars and local currency. We prefer U.S. dollar return series over the local currency return series for two reasons. First, a common currency across the six markets allows for easier comparisons of the results. Second, several of these markets have experienced inflation at extremely high levels, leading to problems with employing returns denominated in the local currency. Therefore, we concentrate on the results obtained using the U.S. dollar return series. However, when we think they are materially different, we also report the results obtained using the local currency return series. Our empirical tests employ market turnover, the ratio of the value of shares traded to market capitalization, as our measure of volume rather than raw volume (i.e., the number of shares traded). Using the number of shares traded directly would require controlling for events such as stock splits, rights issues, and stock dividends. Because such events increase the number of shares outstanding, without adjustments, trading volume would become noncomparable before and after the event occurrence. Further, in our tests such events could have significant effects on the raw volume series because we employ monthly data on indices with relatively few firms. In contrast, our turnover measure is not biased by events that change the number of outstanding shares.4 Panel A of Table 1 provides summary statistics on 3 For a description, see International Finance Corporation (1994), (1995). 4 Bhagat and Bhatia (1996) employ both raw volume and turnover in their study and report that the results are essentially the same with either measure.

K. Saatcioglu, L.T. Starks / International Journal of Forecasting 14 (1998) 215 – 225

218

each of the six markets with regard to the number of firms in the index and in the market and the U.S. dollar value capitalization of the index and of the total market. From the table it can be seen that as of April 1995, Brazil, Chile and Mexico had the largest market capitalizations and number of firms traded. The stock markets of these three countries represent 85% of the total market value of all six countries. Examining the indices in each of the individual countries as a proportion of their total markets indicates that although the indices tend to have a low proportion of the number firms outstanding, the firms they contain are the largest, representing over 50% of the market value in each country. In fact, a comparison of the capitalizations of each index and its market indicates that the indices appear to be roughly in line with the IFC’s goal of including

approximately 60% of the total market capitalization in their index. Panel B of Table 1 reports the mean monthly return in the local currency and in U.S. dollars along with the latter‘s standard deviation, skewness, and kurtosis. Also included for each market is the mean turnover and its standard deviation. The return distributions reflect the generally large returns in these markets over the January 1986 to April 1995 sample period. The statistics on the returns also indicate that each of the markets has had large volatility over the time period. For comparison, U.S. markets had a monthly mean return of 0.0108 and standard deviation of 0.0431 using the value-weighted CRSP index over the same period. In addition, the skewness and kurtosis statistics can be compared to those of the U.S. markets which were 2 1.3 and 6.6,

Table 1 Summary statistics for Latin American stock markets a Panel A: Representativeness of the indices Country

Number of firms in index

Number of firms in market

% of firms in market included in index

Market capitalization of index (US$ billion)

Market capitalization (US$ billion)

% of market capitalization included in index

Argentina Brazil Chile Colombia Mexico Venezuela

34 89 47 25 83 16

156 541 315 190 202 90

21.8 16.5 14.9 13.2 41.1 17.8

18.5 96.8 49.8 10.0 58.3 3.0

33.0 150.7 71.3 17.1 90.1 4.7

56.0 64.3 69.9 58.4 64.7 63.0

Panel B: Summary statistics for return and turnover Country

Argentina Brazil Chile Colombia Mexico Venezuela a

Return

Turnover

Mean (local currency)

Mean (US$)

Standard deviation (US$)

Skewness (US$)

Kurtosis (US$)

Mean

Standard Deviation

0.1644 0.2309 0.0438 0.0510 0.0565 0.0417

0.0462 0.0297 0.0371 0.0358 0.0334 0.0215

0.2662 0.2062 0.0823 0.0935 0.1394 0.1307

3.1269 0.3639 0.0649 1.4437 21.0134 0.6346

17.8979 0.4252 20.1772 3.3885 4.4760 1.6112

0.0282 0.0419 0.0074 0.0076 0.0044 0.0194

0.0190 0.0205 0.0031 0.0042 0.0023 0.0183

This table provides descriptive statistics for the IFC Global indices, markets, index returns and turnover for six Latin American stock markets over the period January 1986 through April 1995. IFC Global indices are market capitalization weighted. Index returns are total returns including dividends. Turnover is percentage of total market capitalization traded in a given period. All summary statistics are for monthly data series. The number of firms and capitalization of the indices and markets are provided as of April 1995.

K. Saatcioglu, L.T. Starks / International Journal of Forecasting 14 (1998) 215 – 225

respectively, again for the value-weighted CRSP index over the same time period.

4. Empirical results The first issue we investigate is whether the two Wall Street adages: ‘‘volume is relatively heavy in bull markets and light in bear markets’’ and ‘‘it takes volume to make prices move’’ are relevant for the Latin American markets. We first start by plotting volume against stock price changes which we present in Fig. 1. While it is difficult to make definitive statements using only the approximately 100 data points we have for each market, with the exception of Venezuela, these Latin American markets yield price–volume plots that do not contradict the V-shape reported by most previous studies. (see, for example, Karpoff, 1987; Gallant et al., 1992; Blume et al., 1994) In general, large price changes do seem to be associated with relatively large volumes. Next, we move on to examine whether the above mentioned stylized facts fit the Latin American markets by testing for contemporaneous correlation. To do so, we employ two alternative forms of the stock price change (return)5 as shown in the following two regressions: V 5 a0 1 a1 ln(Pt /Pt21 ).

(1)

V 5 b0 1 b1 uln(Pt /Pt 21 )u.

(2)

where the dependent variable (V ) is volume measured by monthly turnover, the percentage of market capitalization traded in a given month, and the independent variable is the natural logarithm of the price relative (or its absolute value) for a given month. The results of these regressions are shown in Table 2 where Eq. (1) and Eq. (2) are displayed in

5 In the relevant literature, the terms ‘‘price’’ and ‘‘return’’ are used interchangeably. In both cases, what the authors mean is some form of change in prices. The most commonly used measures are the simple change in prices (DP), the magnitude of the change in prices (uDPu), and the natural logarithm of returns (ln(Pt /Pt 21 )). Throughout the remainder of this paper ‘‘return’’ will be used to represent all forms of these changes in prices.

219

Panels A and B, respectively, for U.S. dollar returns, and in Panels C and D, respectively, for local currency returns. The results in Panel A indicate that when the signed price change is used as the measure of return, the contemporaneous correlation between monthly return and volume is significantly positive (at least at the 10% level) for five of the six markets. Only Mexico has an insignificant relation. The results in Panel B indicate that this relation also holds when the absolute price change is employed as the measure of return. Again, five of the six markets show significant positive correlation between monthly return and volume, only they are not the same five markets. In this case, Mexico has a significant coefficient and Brazil is the one country with an insignificant coefficient. Using local currency returns instead of U.S. dollar returns, Panels C and D reveal that four of the six markets show a significant positive correlation between monthly return and volume. Argentina and Brazil which showed mostly significant contemporaneous correlation between monthly return and volume when U.S. dollar return was used, now do not exhibit any significant correlation. We attribute this difference between U.S. dollar and local currency results for Argentina and Brazil to these two countries’ unstable currencies over our sample period. Finally, the regression results in Table 2 also indicate that contemporaneous return explains a relatively small portion of volume in these markets as evidenced by the low adjusted R-squares.

4.1. Granger causality tests A more important test of the relation between return and volume takes into account whether there is a relation between the lagged values of the two series. Consequently, to test whether volume leads return or return leads volume we employ Granger causality tests as has been done in previous research on developed markets (e.g., Smirlock and Starks, 1988). Granger causality tests whether variable X ‘‘Granger causes’’ variable Y, that is, whether X leads Y after controlling for past values of Y. In this case, we want to test whether returns lead volume or volume leads returns, but we also need to control for any serial correlation in the dependent variable itself. Our Granger causality regressions are as follows:

220

K. Saatcioglu, L.T. Starks / International Journal of Forecasting 14 (1998) 215 – 225

Fig. 1. Stock price–volume plots. These plots show the stock price–volume relation in six Latin American stock markets over the period January 1986 through April 1995. Return is calculated as the natural logarithm of the price relative using the IFC Global total return index and is plotted on the horizontal axis. Returns in U.S. dollar are presented in Panel A, while returns in local currency are presented in Panel B. Turnover, the percentage of total market capitalization traded in a given period, represents volume and is plotted on the vertical axis.

K. Saatcioglu, L.T. Starks / International Journal of Forecasting 14 (1998) 215 – 225

Fig. 1. (continued)

221

222

K. Saatcioglu, L.T. Starks / International Journal of Forecasting 14 (1998) 215 – 225

Table 2 Stock price–volume relation Panel A: Regression results for Vol 5 a0 1 a1 ln(Pt /Pt 21 ) using U.S. dollar returns Country

Number of observations

a0

a1

F-statistic

Adjusted R-square

Argentina

111

0.0331

96

6.29**

0.0528

Chile

111

3.45*

0.0218

Colombia

111

11.44***

0.0867

Mexico

111

0.94

20.0006

Venezuela

88

0.0173** (2.18) 0.0251** (2.51) 0.0067* (1.86) 0.0150*** (3.38) 0.0135 (0.97) 0.0473*** (3.37)

4.76**

Brazil

0.0280*** (15.69) 0.0418*** (20.53) 0.0072*** (23.23) 0.0071*** (17.62) 0.0431*** (19.64) 0.0191*** (10.38)

11.35***

0.1063

Panel B: Regression results for Vol5 b0 1 b1 uln(Pt /Pt 21 )u using U.S. dollar returns Country

Number of observations

b0

b1

F-statistic

Adjusted R-square

Argentina

111

0.0350

96

0.21

20.0084

Chile

111

5.48**

0.0391

Colombia

111

17.78***

0.1323

Mexico

111

11.49***

0.0871

Venezuela

88

0.0226** (2.23) 20.0070 (20.46) 0.0130** (2.34) 0.0245*** (4.22) 0.0596*** (3.39) 0.0372* (1.67)

4.99**

Brazil

0.0251*** (10.99) 0.0429*** (13.76) 0.0065*** (13.60) 0.0060*** (11.37) 0.0372*** (13.47) 0.0158*** (5.40)

2.79*

0.0202

Panel C: Regression results for Vol5 a0 1 a1 ln(Pt /Pt 21 ) using local currency returns Country

Number of observations

a0

a1

F-statistic

Adjusted R-square

Argentina

111

20.0074

96

0.97

20.0003

Chile

111

3.87*

0.0254

Colombia

111

7.32***

0.0543

Mexico

111

6.80***

0.0501

Venezuela

88

0.0028 (0.43) 0.0091 (0.98) 0.0074* (1.97) 0.0122*** (2.70) 0.0407*** (2.61) 0.0531*** (3.67)

0.19

Brazil

0.0280*** (14.47) 0.0401*** (14.47) 0.0071*** (22.13) 0.0071*** (16.08) 0.0415*** (18.56) 0.0181*** (9.70)

13.45***

0.1251

K. Saatcioglu, L.T. Starks / International Journal of Forecasting 14 (1998) 215 – 225

223

Table 2 (continued) Panel D: Regression results for Vol5 b0 1 b1 uln(Pt /Pt 21 )u using local currency returns Country

Number of observations

b0

b1

F-statistic

Adjusted R-square

Argentina

111

20.0039

96

0.00

20.0106

Chile

111

6.74**

0.0496

Colombia

111

14.83***

0.1117

Mexico

111

30.84***

0.2134

Venezuela

88

0.0058 (0.76) 0.0002 (0.02) 0.0141** (2.60) 0.0211*** (3.85) 0.1113*** (5.55) 0.0513** (2.25)

0.57

Brazil

0.0273*** (12.09) 0.0418*** (11.19) 0.0065*** (13.83) 0.0061*** (11.50) 0.0316*** (11.01) 0.0144*** (4.93)

5.08**

0.0448

a

This table provides the coefficient estimates from regressions of volume against price changes (returns) for six Latin American stock markets over the period January 1986 through April 1995. Turnover, the percentage of market capitalization traded in a given period, is used as a measure of volume. Return is calculated as the natural logarithm of the price relative using the IFC Global total return index. Both signed and absolute return is used as well as returns in US$ and local currency. Panels A and B present the results for signed return and absolute return, respectively, using U.S. dollar returns. Panels C and D present the same results for local currency returns. Also shown for each regression are the F-statistic and the adjusted R-square. t-statistics are in parentheses. * Statistical significance at the 10% level. ** Statistical significance at the 5% level. *** Statistical significance at the 1% level.

Vol t 5 a0 1

O

i 51212

ai Volt 2i 1

O

j 51212

bj Rett 2j (3)

Ret t 5 g0 1

O

g Rett 2i 1

i51212 i

O

d Vol t 2j

j 51212 j

(4) where (Vol t ) is the turnover ratio, defined as the percentage of market capitalization traded in month t, and (Ret t ) is the natural logarithm of the month t price relative. Granger causality test is in effect an F-test for block exogeneity, and as such is vulnerable to serial correlation. (see, for example, Kennedy, 1993, p. 68) Therefore, before running the Granger causality tests, we correct the data series for first-order autocorrelation. Summary results of Eqs. (3) and (4) using monthly returns and turnover across the January 1986 through April 1995 time period are shown in Table 3. The table provides the intercept and the first two lags of the volume and return variables along with an F-statistic for block exogeneity and the adjusted R-square statistic. In the bivariate case, the F-test for block exogeneity is equivalent to a test for Granger causality.

Panel A provides the results for Eq. (3). In a test of the null hypothesis that return does not Granger cause volume, the F-statistic is significant at the 10% level for three of the six countries, and is not significant at the 1% level in any of them. Hence, for the Latin American markets in general, we cannot reject the null, and therefore do not find strong evidence for returns ‘‘causing’’ volume. The results of Eq. (4) are shown in Panel B. The Granger causality tests are quite different from those of Eq. (3). In the test of the null hypothesis that volume does not Granger cause returns, the F-statistic is significant at the 5% level in four of the six cases, the exceptions being Argentina and Chile. Thus, for the Latin American markets in general, we find some evidence supporting volume ‘‘causing’’ returns. The results of our Granger causality tests are not consistent with previous evidence from the U.S. markets which have mostly shown that returns ‘‘cause’’ volume but not vice versa. Our results for the Latin American markets imply the opposite, volume ‘‘causing’’ returns with causality not running in the other direction. We also run our Granger causality tests using local

K. Saatcioglu, L.T. Starks / International Journal of Forecasting 14 (1998) 215 – 225

224

Table 3 Granger causality test results Panel A: Summary results for Vol t 5 a0 1o i 51212 ai Vol t 2 i 1o j 51 212 bj Ret t 2 j Country

Number of observations

a0

a1

a2

b1

b2

F-statistic

Adjusted R-square

Argentina

98

1.82*

0.3356

Colombia

98

1.94**

0.7734

Mexico

98

1.61

0.2759

Venezuela

75

20.0095 (20.87) 20.0079 (20.30) 0.0014 (0.30) 20.0101 (21.05) 0.0145 (0.90) 0.0168 (0.73)

0.5789

98

0.0171** (2.19) 0.0153 (1.57) 0.0091** (2.28) 0.0116** (2.02) 0.0240** (2.03) 0.0132 (0.95)

0.85

Chile

20.1027 (20.17) 0.2102 (0.24) 20.5423*** (23.11) 20.6676*** (23.21) 20.0576 (20.28) 20.4124 (20.26)

0.6978

62

0.8403 (1.62) 0.5140 (0.33) 20.6436*** (24.78) 1.0132*** (7.49) 0.0948 (0.18) 0.9097 (0.75)

1.20

Brazil

0.0025 (0.90) 0.0171 (0.69) 0.0284** (2.35) 0.0007 (1.32) 0.0112 (0.83) 0.0012 (0.28)

1.95*

0.7510

F-statistic

Adjusted R-square

1.54

0.5509

2.89***

0.0573

0.96

0.0351

2.72***

0.5492

2.77***

0.4762

2.06**

0.2141

Panel B: Summary results for Ret t 5g0 1o i 51212gi Ret t 2 i 1o j 51212dj Vol t 2 j Country

Number of observations

g0

Argentina

98

Brazil

63

Chile

98

Colombia

98

Mexico

98

Venezuela

75

0.2625 (0.74) 0.0676 (1.02) 0.0872 (1.03) 0.0266 (0.87) 0.0192 (0.56) 0.4081* (1.94)

g1 21.0389*** (25.46) 0.3327* (1.78) 20.0358 (20.07) 0.8626 (0.67) 0.7611 (1.27) 20.7209*** (23.14)

g2

d1

21.0449*** (23.35) 20.3013 (20.57) 20.2753 (21.20) 20.5015 (20.33) 20.4688 (20.88) 20.4602 (21.60)

2.1752 (1.02) 1.8969 (0.61) 3.3069 (0.98) 3.45 (1.45) 0.5148 (0.45) 1.2546 (0.81)

d2 1.3183 (0.60) 24.5633 (20.98) 2.5891 (0.68) 20.8191 (20.16) 0.8016 (0.59) 21.0780 (20.73)

a

This table provides summary results for a vector autoregression (VAR) analysis of the relation between price changes (returns) and volume for six Latin American stock markets over the period January 1986 through April 1995. Turnover, the percentage of market capitalization traded in a given period, is used as a measure of volume. Return is calculated as the natural logarithm of the price relative using the IFC Global total return index in U.S. dollars. Only the parameters for the first two lags are reported here. Panel A presents the results for the regressions testing price changes (returns) Granger causing volume, while Panel B presents the results for the regressions testing volume Granger causing price changes (returns). The data series have been corrected for first-order autocorrelation before running the tests in either panel. Also shown for each regression are the F-statistic and the adjusted R-square. t-statistics are in parentheses. * Statistical significance at the 10% level. ** Statistical significance at the 5% level. *** Statistical significance at the 1% level.

currency returns. Those results are not materially different than the results obtained using U.S. dollar returns and are therefore omitted.

5. Conclusions In this paper we investigate the relation between stock prices and trading volume for six Latin American emerging stock markets. Using monthly index

data, we regress volume on price changes and find strong evidence of a positive correlation between volume and returns. In addition, we present evidence in general for these markets that volume leads returns, but that returns do not lead volume. In short, this set of emerging markets with different institutions and information flows than the developed markets, do not present similar stock price– volume lead–lag relation to the preponderance of studies employing U.S. data. The implication of

K. Saatcioglu, L.T. Starks / International Journal of Forecasting 14 (1998) 215 – 225

these results is that differences in institutions and information flows in this set of emerging markets are important enough to affect the valuation process of equity securities and warrant further analysis.

Acknowledgements The authors would like to thank the IFC for availability of the Emerging Markets Database and participants in seminars at the University of Texas at Austin and at the 1997 Business Association of Latin American Studies meetings.

References Aggarwal, R., Leal, R., Hernandez, L., 1993. The aftermarket performance of initial public offerings in Latin America. Financial Management 22 (1), 27–53. Barry, C., Lockwood, L., 1995. New directions in research on emerging capital markets. Financial Markets, Institutions and Instruments 4, 15–36. Bhagat, S., Bhatia, S., 1996. Trading Volume and Price Variability: Evidence on Lead–Lag Relations from Granger-Causality Tests. Working paper, University of Colorado at Boulder. Blume, L., Easley, D., O’Hara, M., 1994. Market statistics and technical analysis: the role of volume. Journal of Finance 49 (1), 153–181. Chan, W.S., Tse, Y.K., 1993. Price–volume relation in stocks: a multiple time series analysis on the Singapore market. Asia Pacific Journal of Management 10 (1), 39–56. Copeland, T.E., 1976. A model of asset trading under the assumption of sequential information arrival. Journal of Finance 31, 1149–1168. Crouch, R.L., 1970a. A nonlinear test of the random walk hypothesis. American Economic Review 60 (1), 199–202. Crouch, R.L., 1970b. The volume of transactions and price changes on the New York Stock Exchange. Financial Analysts Journal 26 (4), 104–109. Divecha, A.B., Drach, J., Stefek, D., 1992. Emerging markets: a quantitative perspective. Journal of Portfolio Management 19 (1), 41–50. Epps, T.W., 1975. Security price changes and transaction volumes: theory and evidence. American Economic Review 65 (4), 586–597. Epps, T.W., 1977. Security price changes and transaction volumes: some additional evidence. Journal of Financial and Quantitative Analysis 12 (1), 141–146. Epps, T.W., Epps, M.L., 1976. The stochastic dependence of security price changes and transaction volumes: implications for the mixture-of-distributions hypothesis. Econometrica 44 (2), 305–321. Errunza, V.R., 1994. Emerging markets: some new concepts. Journal of Portfolio Management 20, 82–87.

225

Gallant, A.R., Rossi, P.E., Tauchen, G., 1992. Stock prices and volume. Review of Financial Studies 5 (2), 199–242. Godfrey, M.D., Granger, C.W.J., Morgenstern, O., 1964. The random-walk hypothesis of stock market behavior. Kyklos 17 (1), 1–30. Granger, C.W.J., Morgenstern, O., 1963. Spectral analysis of New York stock market prices. Kyklos 16 (1), 1–27. Harvey, C.R., 1995. Predictable risk and returns in emerging markets. Review of Financial Studies 8 (3), 773–816. Hiemstra, C., Jones, J.D., 1995. Testing for linear and nonlinear Granger causality in the stock price–volume relation. Journal of Finance 49, 1639–1664. International Finance Corporation, 1994. The IFC Indexes: Methodology, Definitions, and Practices. World Bank, Washington, DC. International Finance Corporation, 1995. Emerging Stock Markets Factbook. World Bank, Washington, DC. Jennings, R., Starks, L., Fellingham, J., 1981. An equilibrium model of asset trading with sequential information arrival. Journal of Finance 36, 143–161. Karpoff, J.M., 1987. The relation between price changes and trading volume: a survey. Journal of Financial and Quantitative Analysis 22 (1), 109–126. Kennedy, P., 1993. A Guide to Econometrics, 3rd ed. MIT Press, Cambridge MA. Rogalski, R.J., 1978. The dependence of prices and volume. Review of Economics and Statistics 60 (2), 268–274. Smirlock, M., Starks, L.T., 1985. A further examination of stock price changes and transactions volume. Journal of Financial Research 8 (3), 217–225. Smirlock, M., Starks, L.T., 1988. An empirical analysis of the stock price–volume relationship. Journal of Banking and Finance 12 (1), 31–41. Tiao, G.C., Box, G.E.P., 1981. Modeling multiple time series with applications. Journal of the American Statistical Association 76 (376), 802–816. Tse, Y.K., 1991. Price and volume in the Tokyo stock exchange. In: Ziemba, W.T., Bailey, W., Hamao, Y. (Eds.), Japanese Financial Market Research. North Holland, Amsterdam, pp. 91–119. Biographies: Laura T. STARKS is the Charles E. and Sarah M. Seay Reagents Chair in Finance at the Graduate School of business, University of Texas. Her recent research focuses on the evaluation, compensation and performance of portfolio managers, the role of institutional investors in markets and corporate governance and on the differences in analysts and markets across country borders. Her publications have appeared in the Journal of Finance, Journal of Financial Economics, Review of Financial Studies, Management Science, Journal of Business, and other leading finance and accounting journals. Kemal SAATCIOGLU is a Ph.D. candidate in Finance at the graduate School of Business, University of Texas at Austin. He expects to receive his Ph.D. in August 1999. He holds a B.S. in Mechanical Engineering from Bogazici University in Istanbul, an M.S. in Mechanical Engineering from Carnegie Mellon University and an M.B.A. from Texas Tech University. His research interests include securities markets and corporate finance.