A股B股的区别英文论文
以下，就是有道富投资为大家翻译的A股B股的区别英文论文
A. Data and Preliminary Statistics
The trading processes for A- (local) and B- (foreign) shares on the Shanghai
Stock Exchange (SHSE) and the Shenzhen Stock Exchange (SZSE) are similar.
Both exchanges run order-driven, automated markets. Neither exchange has
designated market makers. Traders can only submit limit orders, which arrive
at an electronic consolidated open limit order book (COLOB). An incoming order
is automatically matched against the best standing limit order in the COLOB,
according to the price-time priority principle. If it cannot be matched, then it
is added to the COLOB. There is no block trading system that allows liquidity
traders to trade large volumes in an upstairs market. Off-exchange trading
and insider trading are both forbidden, but this is not tightly monitored. The
minimum trade size is 100 shares for local shares in both markets, but 1,000
shares for Shanghai foreign shares and 100 shares for Shenzhen foreign shares.
Our data consist of all time-stamped trades and quotes from January 2000
to November 2001 for all stocks traded on the SHSE and the SZSE.7 We apply
a number of filters to our data. First, we limit the sample to firms that traded
A- and B-shares throughout the sample period, thereby reducing the number
of firms from over 1,000 to 84. Second, we exclude eight firms, because only
for a few days we find nonzero volume in the B-share market. Third, for the
remaining 76 firms, we remove days for which, for exogenous reasons, there
was no or very limited trading in either the A- or the B-share market.8 Fourth,
we remove stale quotes, which are easily recognized through zero depth. Fifth,
the first and last 15 minutes of each trading session are removed from the
sample.
Table I provides trading statistics for the A- and B-shares from January to
December 2000, a period during which both markets were fully segmented.
The table presents the cross-sectional mean, standard deviation, minimum,
and maximum based on the 76 sample stocks. For comparison, the trade and
quote data for the B-shares (in Hong Kong dollars for SZSE and in U.S. dollars
for the SHSE) are converted to yuan using the (fixed) official exchange rate
for the sample period. The average trade price in this period is 14.29 yuan
($1.73) for A-shares and 3.10 yuan ($0.37) for B-shares. This corresponds to an
average B-share discount of 72%, which is in line with previous evidence (see,
e.g., Bailey et al. (1999)).
In terms of volume, the B-share market is about half the size of the A-share
market—an average of 854,000 versus 1,684,000 shares per day. The A-share
market is more actively traded, as expected. Furthermore, we find that while
the trading frequency is higher in the A-share market, the transaction size
is lower. This highlights the importance of controlling for trade size in our
spread decomposition analysis. Although the average quoted spread is 0.027
yuan for A-shares and 0.035 yuan for B-shares, this difference is not significant
at conventional confidence levels. The effective spreads in the two markets are
both equal to 0.035 yuan.
B. Preliminary Analysis of Information Asymmetry Measures
In this subsection, we estimate the price impact coefficients (PI), adverse
selection components (AS), and the probability of informed trading (PIN)
measures, and we study whether these information asymmetry measures agree,
cross-sectionally, as to which securities exhibit the highest information asymmetry.
We then relate these measures to the foreign share discount.
In preparing the data for the estimation of PI and AS, we follow Glosten and
Harris (1988) and truncate the trade size to 100,000 shares to avoid giving too
much weight to large trades. The median truncation frequency is 0.28% and
1.55% for the A- and B-share market, respectively. This is most likely the result
of more institutional investors who trade in larger sizes, in the B-share market.
We also truncate the trade size to 200,000 and 400,000 shares and find that
the results are generally similar.
Panel A of Table II contains estimates of the price impact coefficients for the
76 A- and B-shares. The mean estimates of γ and φ are 9.66 × 10−7 yuan per
share and 8.58 × 10−3 yuan for the A-share market, and 2.61 × 10−7 and 6.35
× 10−3 for the B-share market. Therefore, although A-share volume is higher
than B-share volume (see Table I), A-share depth is actually lower than B-share
depth. This result is interesting as it shows that one should not equate liquidity
with trading volume. Thus, consistent with our model in Section II, even without
trading volume consideration, the information asymmetry between A- and
B-share markets could cause their market depths to be different. This motivates
why in subsequent regression analyses we test the extent to which our information
asymmetry measures explain the B-share discount after controlling for
trading activity.
Panel B of Table II contains cross-sectional statistics on the fixed and variable
adverse selection component (AS) of the spread and on gross profit for the 76
A- and B-shares. The AS component coefficients, z0 and z1, are significant, and
carry the expected sign for all of the 76 securities in the A-share market and
for almost all (66 and 61, respectively) of the securities in the B-share market.
This evidence is consistent with previous literature in that the cost of adverse
selection is a significant component of the spread that increases with the size
of the transaction. The fixed gross profit coefficients, c0, are significant, and
carry the expected sign for all of the securities in the A-share market and for 67
of the securities in the B-share market.
The variable gross profit coefficients, c1, are significant for 55 of the securities in the A-share market and for only 4of the securities in the B-share market. The last column in Panel B reports the
AS component for a median-sized trade. Consistent with the PI estimate, we
find that the average AS component is larger in the A-share market than in the
B-share market, with values of 56.4 × 10−4 and 46.8 × 10−4 yuan, respectively.
A problem in estimating PI and the AS component for the two markets is
that there is a lower trade frequency for B-shares as compared to A-shares. In
addition to estimating the AS component based on trade-by-trade data, we also
estimate based on fixed-length intervals. For instance, in the case of 15 minutes,
price change from t – 1 to t is measured as the price change over the 15-minute
interval, V(t) is defined as the absolute value of net order flow in the 15-minute
interval t, and Q(t) is defined as 1 or –1 depending on whether the net order
flow is positive or negative in the 15-minute interval t. In general, we find
that using alternative estimates of PI and the AS component in our subsequent
regression analyses produces qualitatively similar results.
Panel C of Table II presents cross-sectional statistics on the parameter estimates of the PIN model. Again, we find considerable evidence for privately
informed traders in the A-share market, as the average arrival rate of these
types of traders, μ, is 0.38, which is of the same level of magnitude as the arrival
rates of uninformed buyers (0.44) and sellers (0.51). The arrival rate of
informed traders in the B-share market is lower at 0.11, and again is of the
same level of magnitude as the arrival rates of uninformed buyers (0.06) and
sellers (0.07) in this market. The probability of an information event on a specific
day, α, is higher in the A-share market than in the B-share market, with
rates of 0.36 versus 0.31, respectively. This is consistent with the existence of
more information in the A-share market. However, the average level of the PIN
measure is higher in the B-share market, as this market exhibits a relatively
low number of uninformed trades. All parameters are significant for the majority
of the securities except for the parameter δ, which is the probability of the
news being bad news.
We compare our measures of information asymmetry by verifying whether
they agree cross-sectionally as to which securities exhibit the most information
asymmetry. Figure 4 presents scatterplots of the AS component of the spread
(for a median-sized trade) against PI and the PIN for both the A- and B-share
markets. These plots suggest a stronger relationship between the measures
in the A-share market, with correlations of 89% and 59%, respectively. The
relationship in the B-share market is much weaker.
Finally, scatterplot analysis reveals that the B-share discount appears to be
explained by the proposed information asymmetry measures. Figure 5 plots
foreign share discounts against the differentials of the information asymmetry
measures in the A-share market relative to the B-share market, as measured
by PI, the AS component, or the PIN. For all measures, we find that stocks
with relatively higher information asymmetry appear to command higher Bshare
discounts. The correlations between the three information asymmetry
measures and the discounts are 66% for PI, 67% for AS, and 28% for PIN, and
are statistically significant at either the 1% or 5% level.
D. Changes After the B-Share Market Was Opened Up to Local Investors
In March 2001, regulators opened the B-share market to domestic investors.
We use this regulatory change event to further test our information asymmetry
hypothesis in two ways. First, we expect B-share discounts to shrink or vanish
and, more important, our information asymmetry measures to increase for the
B-share market after this event, because better-informed domestic investors are
now allowed to participate in this market. Second, we repeat our cross-sectional
regressions for the new sample period as a robustness test.We analyze the same
76 stocks for the sample period of April to November 2001.
We find that, consistent with our hypothesis, the discount levels decrease
from an average of 72% to 43%, and the level of informed trading in the Bshare
market increases compared with the model estimates for the preevent
period. The main reason for this gradual, rather than sudden, decline in discounts
is the lack of foreign currency among domestic investors. Panel A in
Table V presents pre- and postentry estimates of PI, and shows that PI almost
doubles (+81%) for the B-share market from 2.6 preentry to 4.7 postentry. This
increase is larger than the 21% increase in the A-share market and is therefore
consistent with the arrival of better-informed domestic investors in the
B-share market. Panel B confirms these results based on the AS component of
the spread. For both the fixed (z0) and the variable (z1) AS component, we find
considerable increases in the B-share market of 52% and 101%, respectively,
after domestic investors were allowed to enter into this market.
We find that
the AS component for median-sized trades increases by 44% from 46.8 cents to
67.4 cents. This increase is again larger than the increase in the A-share market
and thus consistent with the PI findings. Panel C of Table V reveals that
the arrival rate of informed investors in the B-share market increases by 164%.
This rate actually decreases by 29% in the A-share market. Further evidence
of increased information in the B-share market is that the probability of an
information event increases from 0.31 to 0.42, whereas in the A-share market,
the probability increases only from 0.36 to 0.38. Nevertheless, the level of the
PIN measure does not change much in either market, because the changes in the arrival rate of uninformed traders roughly match those of informed
traders.
The regressions of the foreign share discount on the information asymmetry
measures and controls in the postevent period are consistent with earlier findings.
Table VI presents the results for PI, AS, and PIN including and excluding
the control variables. In all of the regressions, the information asymmetry measures
are significantly positive, with the PI and AS component measures explaining
61% and 71% of the variation in foreign share discounts, respectively.
E. Discussion and Interpretation of the Results
Overall, our results provide strong evidence that the information asymmetry
measures, especially the adverse selection component of the bid-ask spread, are
far more important than any of the control variables in explaining the crosssectional
variation in B-share discounts. Our interpretation is that the information
asymmetry measures reflect the extent of private information available
to domestic investors. When there is a higher degree of information asymmetry,
as measured by a higher price impact coefficient or adverse selection component
in the A-share market than in the B-share market, domestic investors are
more willing to pay a higher price than foreign investors, resulting in B-share
discounts. However, several issues are worthy of discussion.
The first issue concerns the type of private information to which we refer. It
is widely believed that Chinese investors trade on rumors rather than fundamentals.
Furthermore, share manipulation is widespread, pushing prices away
from the intrinsic value for a relatively long time period. Mei et al. (2003) argue
that A-share prices are a reflection of speculative bubbles. Thus, it is not clear
whether our information asymmetry measures reflect fundamental news. Although
we agree with this view of the speculative behavior of investors, we do
not think that it should disqualify our price impact coefficient or adverse selection
component from being effective measures of information asymmetry.