DOES UNCERTAINTY ABOUT A CLOSED-END FUND’S HOLDINGS LEAD TO A DISCOUNT?
4.2. SAMPLE SELECTION AND DATA MANIPULATION
70 4.2.1 Construction of the Rumors Database
Unlike the existing literature discussed earlier, we use our access to a unique hand- collected database of rumors whose construction is now described.26 Since a rumor is not an announcement, its wording can be ambiguous and unclear. Thus, a list of identifiers (keywords) compiled from a selection of M&A rumors reported in S&P Takeover Talk was used to search Factiva, Pro-quest (only for publications/newswires not covered by Factiva), S&P Takeover Talk, Capital IQ, Zephyr, SDC and a variety of newswires. Once a M&A rumor identifier was found in an article, the article was stored and catalogued. Since rumors have a tendency to ricochet throughout other news outlets, a search to identify the first time the rumor was published in the 90 days prior was done. Only reports for the same M&A rumor with the earliest date were retained. To ensure that the retained rumor date was indeed the initial rumor, a continuous search using an additional 90-day window was undertaken.
[Please place table 4.1 about here]
Our initial sample contains 2,250 rumor events that have no publicly reported rumors in the preceding ninety days. The decomposition of this sample by rumor type and the number of rumors that materialize in actual announcements using SDC Thomson Reuters for different time frames are reported in table 4.1. For this sample of M&A rumors, we find that 1,762 targets have actual subsequent M&A announcements and that only 361 of these announcements occur within one year after the rumor started about a possible M&A. We also report the number of acquisitions during the 70 days following the rumor to be consistent with Gao and Oler (2012) who categorize a rumor with no announcement after 70 trading days as misinformation. Although our sample is 14 times larger than theirs, we have approximately the same percentage (12%) of rumored firms announcing a M&A during this 70-day window.
The rumor types obtained from the manual search through the news releases about the rumors provide the existence or not of a source of the rumor. We hypothesize that the credibility of the rumor depends on the source that issued it. We categorize our rumors into more and less credible rumors according to the probability of an M&A announcement happening within 70 days or one year after the rumor date. More-credible rumors are those emanating from reliable sources, including: (i) firms that indicate their interest in looking for
71 a potential acquirer; (ii) firms that confirm the existence of talks about a possible takeover; (iii) rumors disclosed by insiders of the target; (iv) target firms hire financial advisors that generally indicate their interests in pursuing “strategic alternatives”; and (v) rumors attributable to analysts following the stock. Less-credible rumors are not attributable to actual sources, but inferred from stock price movements, changes in options trading, or chatter about possible synergies.
4.2.2 Measuring the Credibility of a Rumor
We verify our conjecture about the relationship between the source and its credibility by calculating the likelihood that the rumor leads to an M&A announcement. We calculate the probability of a M&A announcement within 70 days for the total sample, more- and less- credible rumors and their subcategories. We test the following null hypothesis: H0: Pr(Announcement| More credible)= Pr(Announcement| Less credible). Our test statistic is:
̂ ̂ √ ̂ ̂ where ̂ is the probability of a actual M&A being completed or just an announcement for the more-credible rumored sample, ̂ is the probability of a M&A or an announcement for the less-credible rumored sample, ̂ is the estimate of the common proportion under the null hypothesis calculated as
̂ ⁄ and are the number of rumors in the more- and
less-credible rumor samples respectively. We calculate the probabilities and their significance for M&A announcements that occur within 70 days and one year after the initial rumors. Since M&A announcements do not always result in actual M&As, we also calculate the probabilities of actual M&A within a time frame of 18 months.
[Please place table 4.2 about here]
Table 4.2 reports the probabilities of an announcement or M&A for all, more credible and less credible subcategories. The likelihood of a M&A for a more-credible rumor (15.04%) is approximately double that for a less-credible rumor (8.8%) within 70 days of the initial rumor release date (same period as in Gao and Older, 2012). The corresponding probabilities are 19.97%and 11.74% for one-year post-rumor. These differences are statistically significant at 1%. We obtain statistically significant differences for the various subcategories of the more- credible rumors and their likelihoods. For announcements within 70 days of the rumor date, the announcement likelihoods with the greatest to least likelihoods are for rumors confirmed by targets (24.56%), rumors indicating a financial advisor is hired by the target (19.85%),
72 rumors attributed to an insider (19.09%), rumors initiated by the target itself (16.79%), and finally rumors initiated by an analyst (12.13%). All of these probabilities are statistically different at the 1% level, and statistically different than those associated with less-credible rumors within 70 days at 1%, except for the last most-credible rumor category (analyst is the rumor source), which is significant at the 5% level.
We also find consistent results using a one-year window as in Pound and Zeckhauser (1990) to calculate the probability of a subsequent M&A announcement. The probability of an M&A announcement for more- versus less-credible rumors is approximately double (19.97% vs. 11.74%). Although all the respective percentages are higher, the ranking of the credibility of the various categories of more-credible rumors does not change with this longer window. The announcement likelihoods with the greatest to least likelihoods are for more- credible rumors confirmed by targets (28.07%), rumors indicating a financial advisor is hired by the target (26.59%), rumors attributed to an insider (22.51%), rumors initiated by the target itself (16.79%), and finally rumors initiated by an analyst (21.88%). These probabilities are statistically larger at the 1% level than that for the less-credible rumors within the one year.