- September 1, 2015
September 1, 2015
The U.S. Supreme Court's ruling last year in Halliburton Co. v. Erica P. John Fund, Inc. upheld the ability of investors to recover losses that result from public misstatements and misrepresentations by directors and officers of U.S.-listed companies. The ruling, however, also made it possible for defendant\ companies in class action cases to disprove, at the early, class certification stage of a case, plaintiffs' assertions that a company's disclosure of information somehow correcting alleged misrepresentations caused share prices to fall.
The implications of the price impact portion of the ruling became clearer on July 25, 2015, when the federal judge in the original Halliburton case allowed class certification for only one of six alleged corrective disclosures. Nathan Associates has analyzed alleged corrective disclosures in relevant cases filed since the Supreme Court ruling in Halliburton. Dr. Stephen E. Christophe, a finance professor and academic affiliate of Nathan Associates, offers the following commentary on the results of that statistical analysis.
By Stephen E. Christophe, Ph.D.
Now that the federal judge in Halliburton has established the\r\nimportance of proving stock price response for class certification, a statistical\r\nexamination of ?price impact? is in order. \r\n\r\n Under my direction, Nathan has performed\r\nsuch an examination. The firm evaluated the statistical significance of stock price\r\ndeclines for 228 alleged corrective disclosures identified in 124 class action complaints\r\nfiled from June 2014 through June 2015. These represent all identified Rule\r\n10b-5 complaints, excluding those involving ADRs, filed since the Supreme\r\nCourt?s decision in Halliburton.\r\nNathan keeps a Rule 10b-5 depository of Defendant companies? stock price\r\nreaction, statistical information, and potential investor exposure. \r\n\r\n
This article explores the distributions of both statistically significant and statistically non-significant price changes on trading days corresponding to alleged corrective disclosures. Among the discoveries was a range of stock price declines where the likelihood of statistically significant price movements and non-significant price movements was the same. To obtain our results, we applied a single-firm, multivariate regression analysis for each defendant company in the 124 cases.
We limited the analysis to the allegations and class period identified in the first filed class action complaint for each defendant company. The multivariate regression incorporated adjustments for price movements in both the general stock market and an index specific to the company's industry of operation. The control period that was used to estimate the but-for price of the defendant's company stock what the price would have been absent release of company-specific information linked to the alleged fraud consisted of a time interval with a minimum of 100 trading days in each case.
We first established that the likelihood\ of a statistically significant price impact (i.e., within a 95 percent confidence interval) increases when a stock price declines substantially.  A more detailed analysis found:
- When an alleged disclosure was followed by a statistically significant stock price impact, those price declines ranged from 93 percent (-93%) to 4 percent (-4%), with a median decline of approximately 20 percent (-20%).
- When an alleged disclosure was not followed by a statistically significant price change, the steepest decline in a share price was 11 percent (-11%), and this population also included shares that rose, with the price rising 15 percent (+15%) in the instance of greatest\ increase. The median price decline for the non-statistically significant group was only 4 percent (-4%).
- In the overlapping population of corrective disclosures where the stock price declines ranged from 11 percent (-11%) to 4 percent (-4%)?about half exhibited statistical significance at the 95 percent confidence interval.
The graphic below shows the respective histograms and normal density overlays of the percentage price declines associated with those disclosures found to be statistically significant and those found to be non-significant. Again, note the overlapping distribution of cases with statistically significant and non-significant price declines. A t-test was also performed to examine whether the average price change on statistically significant days is equal to that on statistically non-significant days. As expected, that hypothesis is rejected soundly because the test results in a t-statistic of 14.35 (with a corresponding p-value less than .0001). The average decline in price on statistically significant days is substantially greater (more negative) in absolute magnitude.