168 / 322
期貨未平倉量的資訊內涵及其交易活動之研究
168
three regressions of Model (4), none of the coefficient of spot volatility is statistically
significant.
9
Evidence suggests that hedging demand stimulated by spot volatility is reflected
only in the expected open interest. Our results are consistent with Chen et al. (1995), Chang
et al. (2000) and Pan et al. (2003) that open interest can represent the hedging demand for
futures contracts. All
TTM
variables are negatively correlated to open interest, indicating that
the sum of open interest of the two nearby contracts has cyclical pattern.
In Table 5, we perform regressions that examine whether open interest increases with
unexpected
spot volatility, the implications (4) of Hypothesis 2. Note that in regression
Models (5) and (6), the explanatory variables for spot volatility include both expected and
unexpected
volatility, partitioned using ARIMA model as the description in the data section.
In regressions of Model (5), the coefficients of the expected and unexpected volatility
are both positive with statistical significance, indicating that the expected open interest rise
with the both expected unexpected increased in volatility. This result is consistent with
Chang et al. (2000), who study the S&P 500 index futures using similar model and report
significant influence from both volatility components on open interest. The magnitude of
effect from unexpected volatility, however, is smaller than those from the expected volatility.
This suggests a less influential role of unexpected volatility than expected volatility to the
open interest. This result differs from that of Chang et al. (2000), who report large coefficient
for the unexpected than for the expected volatility. Since large volatility shock is almost
always accompanied by substantial swing in price, hedgers may not instantaneously adjust
hedge positions in response to volatility shock, during which hedging is extremely
expensive. Instead, hedgers either wait for the end of the transient volatility shock or adjust
positions according to the previous (expected) level of volatility. Because volatility tend to
be clustering in time and can be forecasted with acceptable precision (Engle, 2004), the
expected volatility provides important guidance to hedging strategies. This could be the
reason why we observe stronger linkage between open interest and expected volatility but a
lesser degree with unexpected volatility. Nevertheless, the effect of unexpected volatility in
Model (5) remains significant holding constant the expected volatility, days to expiration,
and weekly patterns.
9 It is somehow surprising to find that volatility is not significant linked the unexpected open interest. We
speculate that, because volatility tends to cluster over time, hedgers can forecast volatility to certain
extent and thus adjust their hedge positions ahead of the occurrence of volatility spikes. This would make
the volatility more correlative to expected open interest than unexpected open interest. A detailed analysis
in the timing of hedge may be warrant for future research.