This paper investigates the impact of the international equity market integration to the international nonsynchronous trading effects (INTE). The paper finds that the financial market integration would increase INTE, in general, and the impact monotonically decreases over the lag length. However empirical evidence suggests that the increase is asymmetric among developed and emerging markets. Further theoretical investigation reveals that the level of volatility and autocorrelation are positively related to the increase in INTE. The paper concludes that the relatively higher level of volatility and autocorrelation in emerging markets could mitigate the increase in INTE from financial market integration.
This paper investigates the impact of financial equity market integration to international nonsynchronous trading effects (INTE). It should be emphasized that this paper looks into the cross market nonsynchronous trading. It is different from the nonsynchronous trading within one market, which most existing studies investigate. The paper employs a stochastic process to provide a theoretical framework. The contributions are clear: First it provides a theoretical framework to investigate the nonsynchronous trading effect. Second, it theoretically and empirically investigates the effect of equity market integration to INTE. And third, the paper provides price level evidence instead of return level evidence.
The paper finds the theoretical and empirical evidences that more integrated equity markets increases the INTE. However the empirical findings show that the increase is asymmetric between developed and emerging markets. Further investigation reveals that the higher volatility and autocorrelation in emerging equity markets mitigate the increase in INTE incurred by the equity market integration. The paper also finds that the size of the effect of equity market integration to nonsynchronous effects decreases over time.
It is well known that nonsynchronous trading causes auto correlation in equity price and return. There are plenty of literature that addresses problems of nonsynchronicity which can be thought of as the random, or deterministic, arrival of data within the interval of measurement and its consequences on the measurement of beta (See Atchison, Butler and Simonds, [
A careful investigation of existing literature in nonsynchronous trading effects reveals that there are several potential aspects that could be further investigated. (1) The cross autocorrelation is much less explored compare to the autocorrelation within the index due to its complex nature. The latter induces autocorrelation mainly from thinly traded securities that have less liquidity while the first induces autocorrelation due to equity market operation time difference. This paper is interested in the cross market nonsynchronous trading effect and refers it to INTE. There could be two consequences of nonsynchronous trading to international equity market data. (2) Most of the existing literature investigate return series instead of price series. Although positive price autocorrelation means positive return autocorrelation in AR process, the direct effect of nonsynchronous trading is price autocorrelation. Return autocorrelation is a product of the price autocorrelation. This paper tackles the price autocorrelation.
(3) With more integrated world’s financial markets, such nonsynchronicity problem could be investigated in a different angle. Schotman and Zalewska [
The rest of the paper is organized as follows. Section 2 introduces the theoretical model. Section 3 empirically tests the two hypotheses. Section 4 reinvestigates the theoretical model to explain the discrepancy between developed and emerging markets. Section 5 concludes the paper.
In modeling shock emissions among international equity markets, there are some domestic shocks that are irrelevant to international markets. These shocks should be excluded from the analysis; hence the term “shock” in this paper refers to the shocks that have international influence only. For modeling purpose, I take a common practice in literature investigating international equity market volatility spill over (See Baele [
This paper employs a modified version of the model developed by Lo and Wang [
dependence on factors. Assume that the logarithm of the index levels
trends
Assume that
where
are correlated standardized Wiener processes with correlation coefficient
Therefore
According to Hong and Satchell [
where
All results are conditional on time zero information.
Remark 1: The conditional covariance of index price and factor price can be written as
See Appendix A of Lo and Wang [
Remark 2: The conditional
A proof of Remark 2 is in the Appendix. Equation (6) is the amount of autocovariance induced from nonsynchronous trading.
The system of Equations (2) and (3) represents the one factor model that the demeaned log price of index
Proposition 1: The nonsynchronous trading effect,
Proposition 1 is an immediate result of Remark 2, dividing the both sides by the variance.
Proposition 2: The impact of equity market integration to nonsynchronous trading effects can be expressed as
Proposition 2 is a partial derivative of Proposition 1 with respect to kappa. This measure becomes positive as long
Proposition 2 implies that more integrated international financial equity markets increases the level of INTE. Intuitively, this is because more integrated financial markets allow shocks from one market to be transferred more widely to other markets.
Proposition 3: The sensitivity of
Proposition 3 is a partial derivative of Proposition 2 with respect to s. Given
From the theoretical finding of Proposition 2 and 3, two hypotheses could be drawn.
H1: The more integrated financial market equity market increases the INTE.
H2: INTE monotonically decreases with respect to time.
In next section, the two hypotheses are empirically tested.
In the empirical investigation of INTE, daily data is more suitable than weekly or monthly data. Schotman and Zalewska [
Intervalling effect (Dimson [
Two measures are computed according to the theoretical section:
Since quarterly non-overlapping price correlations of most liquid stock indices using daily frequency, the data is less likely to suffer from autocorrelation induced by overlapping data and intervalling effect. The autocorrelation and cross autocorrelation observed in the sample data would be mostly due to INTE between equity markets. We see that Canadian the European stock markets move more closely with the US stock market compare to those of East Asian countries. This conforms to the general expectation. All correlation distributions are negatively skewed and this indicates that relatively high level of correlations are more likely compare to relatively low level of correlations. This is also consistent with the findings of the previous studies.
. List of sample international equity markets. The table below presents the list of 11 sample international equity markets used in this paper. Country is the country that the index belongs to, Index is the name of the index and Notation is the notation for the index that this paper refer to.
Country | Index | Notation |
---|---|---|
United States | S&P 500 Index | S&P 500 |
Canada | S&P/TSX Composite Index | S&P/TSX |
Germany | Deutsche Borse Ag German Stock Index | DAX |
United Kingdom | FTSE 100 Index | FTSE100 |
France | CAC40 Index | CAC40 |
Netherland | Amsterdam Exchange index | AEX |
Japan | NIKKEI Index | NIKKEI |
China | HANGSENG Index | HANGSENG |
Korea | KOSPI Index | KOSPI |
Australia | ASX Index | ASX |
Brazil | IBOVESPA Index | IBOVESPA |
. Descriptive statistics of historical correlations. The table below presents the mean, standard deviation, skewness and kurtosis of the historical quarterly non-overlapping correlations between 10 international indices and S&P 500. The sample used in this analysis consists of 10 international equity indices from Q1 of 1998 to Q1 of 2013.
Index | S&P/TSX | DAX | FTSE100 | CAC40 | AEX |
---|---|---|---|---|---|
Mean | 0.7707 | 0.8109 | 0.8064 | 0.7892 | 0.7089 |
Standard Dev | 0.2416 | 0.2094 | 0.1835 | 0.2379 | 0.3036 |
Skewness | −1.5753 | −2.1834 | −1.7757 | −1.8615 | −1.5586 |
Kurtosis | 4.7250 | 8.0303 | 5.9726 | 5.6683 | 4.7992 |
Index | NIKKEI | HANGSENG | KOSPI | ASX | IBOVESPA |
Mean | 0.5785 | 0.6474 | 0.6240 | 0.6990 | 0.6348 |
Standard Dev | 0.3046 | 0.3153 | 0.2880 | 0.2580 | 0.3821 |
Skewness | −0.8373 | −1.9983 | −1.1917 | −1.7278 | −1.7945 |
Kurtosis | 3.2930 | 7.9827 | 4.1332 | 6.2440 | 6.6518 |
on the results that the nonsynchronous trading problem could be resolved by using weekly data instead of daily
data. The sum of squared auto-betas,
lent to Ljung-Box Q statistic using lagged betas instead of autocorrelations.
In order to test the first hypothesis that “The more integrated financial market equity market increases the INTE”, Following regression for each index is ran.
where
. Regression result of Equation (10). The table below presents the regression result of Equation (10) for the 10 sample international equity indices. R Square is the
Index | S&P/TSX | DAX | FTSE100 | CAC40 | AEX |
---|---|---|---|---|---|
R Square | 0.0997 | 0.1295 | 0.3052 | 0.1671 | 0.1204 |
λ | 244.30** | 157.15*** | 113.15*** | 72.52*** | 0.65*** |
(0.013) | (0.004) | (0.0000) | (0.001) | (0.006) | |
α | −2.92 | −18.40 | −48.55** | −10.93 | 0.08 |
(0.970) | (0.680) | (0.0110) | (0.5320) | (0.6600) | |
Index | NIKKEI | HANGSENG | KOSPI | ASX | IBOVESPA |
R Square | 0.1583 | 0.0368 | 0.0626 | 0.1224 | 0.049 |
λ | 921.72*** | 651.05 | 5.73* | 43.92*** | 5134.66** |
(0.001) | (0.1390) | (0.0520) | (0.006) | (0.086) | |
α | −59.39 | 563.67** | 1.95 | −2.65 | 3638.62 |
(0.7430) | (0.0760) | (0.3290) | (0.817) | (0.1000) |
First we note that all but one constant term,
It is interesting to see that the emerging markets, China, Korea and Brazil, do not have significant
The second hypothesis states that the INTE monotonically decreases with respect to time, hence predicts monotonically decreasing
Note that there are total 61 quarters during the sample period. The result of the formal test is reported in
Observed Mean and SD are the parameters for the number of the observed monotonic relationships. Random Mean and SD are the parameters when the size of
. Monotonically decreasing sample relationships. The table below presents the number of monotonically decreasing sample betas for all 5 lagging periods and its percentage with respect to the total sample period of 61 quarters.
S&P/TSX | DAX | FTSE100 | CAC40 | AEX | |
---|---|---|---|---|---|
. Monotonically decreasing sample relationships. The table below presents the number of monotonically decreasing sample betas for all 5 lagging periods and its percentage with respect to the total sample period of 61 quarters. | 51 | 54 | 54 | 54 | 53 |
Otherwise | 10 | 7 | 7 | 7 | 8 |
% of Monotonic Relationships | 83.61% | 88.52% | 88.52% | 88.52% | 86.89% |
NIKKEI | HANGSENG | KOSPI | ASX | IBOVESPA | |
. Monotonically decreasing sample relationships. The table below presents the number of monotonically decreasing sample betas for all 5 lagging periods and its percentage with respect to the total sample period of 61 quarters. | 40 | 47 | 40 | 49 | 45 |
Otherwise | 21 | 14 | 21 | 12 | 16 |
% of Monotonic Relationships | 65.57% | 77.05% | 65.57% | 80.33% | 73.77% |
. Test of the monotonicity. The table below presents the result of testing the statistical significance of the number of quarters that all five betas showed monotonic relationship. Observed Mean is the actual percentage of monotonic relationships, Observed SD is the standard deviation of the distribution of the actual monotonic relationships, Random Mean is the percentage of monotonic relationships if betas follow random order and Random SD is the standard deviation of the distribution of the monotonic relationships if betas follow random order. The sample used in this analysis consists of 10 international equity indices from Q1 of 1998 to Q1 of 2013.
S&P/TSX | DAX | FTSE100 | CAC40 | AEX | |
---|---|---|---|---|---|
Observed Mean | 83.61% | 88.52% | 88.52% | 88.52% | 86.89% |
Observed SD | 37.33% | 32.14% | 32.14% | 32.14% | 34.04% |
Random Mean | 6.25% | 6.25% | 6.25% | 6.25% | 6.25% |
Random SD | 24.21% | 24.21% | 24.21% | 24.21% | 24.21% |
t-stat | 3.50 | 3.17 | 3.17 | 3.17 | 3.29 |
NIKKEI | HANGSENG | KOSPI | ASX | IBOVESPA | |
Observed Mean | 65.57% | 77.05% | 65.57% | 80.33% | 73.77% |
Observed SD | 47.91% | 42.40% | 47.91% | 40.08% | 44.35% |
Random Mean | 6.25% | 6.25% | 6.25% | 6.25% | 6.25% |
Random SD | 24.21% | 24.21% | 24.21% | 24.21% | 24.21% |
t-stat | 4.23 | 3.84 | 4.23 | 3.69 | 3.98 |
Sections 2 and 3 of this paper investigate the sensitivity of the most recent five day factor price with respect to the current index and factor price correlation. The theoretical findings using a bivariate OU process model predict a positive significant relationship between the degree of equity market integration and INTE in international equity markets. However the empirical evidences show that the more integrated equity market increases INTE in developed economies but it does not have significant effect to emerging economies. In order to resolve the discrepancy, we need to take advantage of using a theoretical model and analyze exactly what induces the sensitivity of INTE with respect to the equity market integrity.
Although they are not statistically significant, we still observe positive estimated parameters. This indicates that noise in the data could hide the relationship to be empirically observed. This could be the reason why the empirical result for developing countries is not consistent with the theoretical outcome. One potential explanation could be that developed markets have parameters that make empirical
Proposition 4: The sensitivity of
We can see that the larger
While financial market integration worsens the problem of nonsynchronous trading among the international equity markets in general, emerging markets suffer less from it compared to developed markets. The paper finds that this is due to emerging markets’ higher level of volatility and higher price autocorrelation compared to those of developed markets. These could be indicators of less information efficient market. Ironically speaking, less efficient equity markets of emerging countries mitigate the increase in INTE from more integrated financial equity markets.
This paper builds a theoretical model, using a bivariate Ornstein Uhlenbeck process, to investigate the impact of the international equity market integration to INTE. The paper finds that the financial market integration would increase the effect, in general, and the impact monotonically decreases over the lag length. However the paper empirically finds that the increase is asymmetric between developed and emerging markets. Further theoretical investigation reveals that the level of volatility and autocorrelation in equity market are positively related to the increase in INTE due to financial market integration. The paper finds that relatively higher level of equity index volatility and autocorrelation in emerging markets could mitigate the increase in INTE from financial market integration.