Stock and Industry Return Characteristics Around Price Shocks

ABSTRACT

The author investigates positive and negative price shocks in individual securities and the degree to which they affect related firms in the same industry. This price contagion effect is significant with initial price shocks leading to substantial long-term abnormal returns across firms in the same industry over time. Price shocks also have predictive value regarding future earnings and revenues for the firm in question and its industry overall. Positive (negative) price shocks that are continued over time are associated with higher (lower) Sharpe ratios suggesting that abnormal returns are not simply a form of compensation for greater expected future volatility.

KEYWORDS:

• Price shock
• Factor model
• Abnormal returns
• Industry contagion

Acknowledgments

The author thanks Sila Alan, Walter Hlawitschka, and Ying Zhang for helpful comments.

Notes

1. Results throughout article are qualitatively similar when defining a shock at the individual firm level as a move that is greater than 2 standard deviations above normal for a given company. This definition is not used primarily because practical application of such a rule is more difficult than using a simple percentage based move in this case 10%.

2. Following the literature, reversal in this case refers to abnormal returns in the opposite direction of the initial shock, whereas continued refers to abnormal returns in the same direction as the initial shock.

3. A type I error in this context is one in which investors react to an event that ultimately proves meaningless.

4. As Compustat data are quarterly, the same value decile data is used for each month in a given quarter. The other factors are calculated from CRSP and hence vary monthly. All results are robust to using a linear interpolation of the value decile to create unique monthly values based on the Compustat data.

5. Results are robust to using varying sample sizes and using 0 as the value for certain variables (e.g., in place of earnings or book values) when those variables are missing.

6. Results are robust to using shorter periods for defining shocks, however, using a longer period allows for the type of information delays found in past literature, as well as slower information communication that characterizes the earlier portion of the sample period. Using the longer-period results in more price shocks, which should bias against finding results to the extent that these are anything other than legitimate shocks.

7. Factors are reported as deciles here with 10 being the largest decile and 1 being the smallest, but similar correlations hold when using calculated values for each factor.

8. Again results are qualitatively similar when using the stronger price shock variables, and I omitted those variables here for readability.

9. Hence the table actually shows the results of 48 (6*8) different regressions with only the B1 coefficient reported for the sake of economy of space. Full regression results for any of the 48 regressions are available from the author by request.

10. Results are qualitatively similar when using the 20% Price Drop Dummy.

11. While the 10% Drop binary variable does not distinguish between industries where a single stock experiences a drop versus multiple stocks seeing a drop, as a robustness check I used a second binary variable that accounted for multiple stocks in a single industry experiencing severe drops and found qualitatively similar results.

12. Similar results hold using alternative time frames for calculating the look-ahead Sharpe ratio including the 6-month, 2 year, 3-year, and 4-year time horizons.

13. Results are similar when using 20% Drop and Rise Dummy variables.