Dynamic Relations between Order Imbalance, Volatility and Return of Top Losers

1
Dynamic Relations between Order Imbalance,
Volatility and Return of Top Losers

• Authors Yong-Chern Su, Han-Ching Huang, Po-Hsin
  Kuo and Peiwen Chen (National Taiwan University)
• Discussant Yin-Feng Gau (National Central
  University)
• NTU Finance Conference 2008

2
Summary

• 61 daily loser stocks, TAQ database,
  12/1/2005-12/31/2005
• Use GARCH(1,1) to study the relation b/tw
  volatility and order imbalance
• Use the regression of return on lagged order
  imbalance to study the predictability of returns
• Use VAR of return and order imbalance to study
  the causal relation
• Trading strategy based on order imbalance buy if
  order imbalance is positive sell if order
  imbalance is negative

3
Comments

• Where is the story? better if focused on a
  few important themes
• Contemporaneous relations b/tw
• return (Rt) and order imbalance (OIt) eq (1)
  Chordia and Subrahmanyam (2004)
• volatility and order imbalance eq (2) Chan and
  Fong (2000)
• Lagged relations b/tw
• Rt and OIt, OIt-2, , OIt-4 eq(3) Chordia and
  Subrahmanyam (2004)
• Rt and OIt-1, OIt-2, , OIt-5 eq(4) Chordia
  and Subrahmanyam (2004)
• causal relation (Granger, 1969 Chen and Wu,
  1999) b/tw return and order imbalance
• trading strategy based on the order imbalance

4
Questions about Data

• Data frequency daily or tick-by-tick?
• P.3 the order imbalance is defined as daily net
  share volume .
• How long dose period t in models span?
• If the estimation of models are based on
  intraday data, how fine the data are?
• If using the intraday data, the intraday
  seasonality in volatility should be checked and
  adjusted
• Is there intraday seasonality in OIB?
• Why use the sample of top loser? Motivation?

5
Questions about Empirical Results

• Is the same model (e.g. GARCH(1,1)) fitted to all
  61 stocks studied
• should take care of the seasonality in intrady
  volatility, leverage effect, day-of-the-week
  effect
• Table 2 shows for 40 of sample returns,
  coefficient of OIt is positive and significant in
  eq(2), whereas for 40 of sample, the coefficient
  of OIt is negative and significant in eq(2) ? How
  to judge the effect of OIt on volatility
• As shown in Panel A of Table 6, three of four
  trading strategies earn negative returns, and one
  strategy obtains 1.06 return. ? not substantial
  or significant enough for a profit, because the
  transaction costs and taxes are not considered
  yet