Empirical results Directory UMM :Data Elmu:jurnal:M:Multinational Financial Management:Vol11.Issue1.2001:

Table 2 Daily average abnormal returns AARs surrounding the publication of rumors a Event days t-value AARs − 0.99 − 10 − 0.17 − 9 − 0.03 − 0.02 − 8 − 0.26 − 1.29 − 7 0.21 1.49 − 6 − 0.05 − 0.08 − 5 − 0.20 − 0.21 − 4 3.29 0.52 − 3 0.29 2.14 − 2 4.89 0.76 − 1 0.78 3.60 – – + 1 0.07 0.36 + 2 − 2.65 − 0.43 + 3 − 0.04 − 0.06 + 4 − 1.12 − 0.37 + 5 − 0.04 − 0.44 + 6 − 1.44 − 0.21 + 7 0.21 1.31 + 8 − 0.01 − 0.04 + 9 − 0.10 − 0.09 + 10 − 1.42 − 0.42 a This table presents the abnormal returns surrounding the publication date t = 0 in the Ekonomik Trend. Abnormal return is calculated as the difference between the actual and expected return. Expected return is generated from the market model parameters. The ISE Composite index used as a proxy for market. The t-statistics tests the null hypothesis that the average abnormal returns are equal to zero. Statistically significant at 1. Statistically significant at 5. The event study methodology is employed to analyze the effects of rumorsgos- sips on stock prices as surveyed by Brown and Warner 1985. The analysis period extends from day − 30 to + 30 relative to publication date t = 0. 5

4. Empirical results

To determine how stock prices are influenced by rumors, I pose three period analyses. The first examines the effect of rumors around the publication date. The 5 Standard event methodology is used to measure the stock price reaction for N firms in each of the group of firms. The standard event study methodology is not spelled out here. second looks for unusual activity prior to publication of rumors. Finally, I analyze price movement after the publication date. The empirical results are reported on Tables 2 – 4. The daily average abnormal returns AARs for all rumors are calculated over − 30 and + 30 period relative to the event day 0. Only AARs for − 10 and + 10 period are reported on Table 2. The results indicate that the sample experiences statistically significant positive abnormal returns prior to publication of rumors. The AARs are 0.52, 0.29, 0.76 and 0.78 for the days − 4, − 3, − 2 and − 1, respectively. These results are statistically significant at 1 level. The AARs following the publication day are mostly negative and statistically insignificant. For example, AARs are 0.07 on day + 1, − 0.43 on day + 2, − 0.04 on day + 3, and − 0.37 on day + 4. Only AARs on day + 2 is statistically significant at 5 level. The average cumulative abnormal returns CARs for all rumors are reported on Table 3. In the top part of table, the CARs around the publication of rumors are reported. The results indicate that firms experience positive abnormal returns just before and after publication date. For example, during − 1, + 1, − 2, + 2, and Table 3 Cumulative abnormal returns CARs surrounding the publication of rumors a t-value CARs Windows Combined periods 0.85 2.80 −1, +1 3.09 1.19 −2, +2 3.42 1.35 −5, +5 1.89 0.44 −10, +10 − 0.75 −20, +20 − 0.28 − 1.22 −30, +30 − 0.49 Prior to publication date 2.16 −5, −1 6.13 1.82 4.06 −10, −1 1.48 −20, −1 2.60 After publication date − 0.81 +1, +5 − 1.31 − 1.38 − 1.39 +1, +10 − 1.78 +1, +20 − 1.54 a This table presents the cumulative abnormal returns in combined, pre, and post publication period relative to publication date t = 0 in the Ekonomik Trend. Abnormal return is calculated as the difference between the actual and expected return. Expected return is generated from the international market model parameters. The t-statistics tests the null hypothesis that the cumulative abnormal returns are equal to zero. Statistically significant at 1. Statistically significant at 5. Statistically significant at 10. H . Kiymaz J . of Multi . Fin . Manag . 11 2001 105 – 115 111 Table 4 Daily average abnormal returns AARs based on the content of rumors a Undervalued stocks Earnings expectations Unclassified rumors Rumors without content Days Purchases by foreign investor Salesexport expectations n = 68 n = 22 n = 23 n = 6 n = 108 n = 128 t-value AARs AARs t-value AARs t-value AARs t-value t-value AARs t-value AARs 0.15 − 0.62 − 0.56 − 0.15 − 0.35 0.81 0.36 − 10 0.30 0.30 0.72 − 1.55 − 0.52 − 0.40 − 1.15 − 1.94 0.10 0.27 0.24 0.82 0.46 0.66 − 0.79 1.13 − 9 1.17 − 0.37 − 0.11 − 0.16 0.11 − 0.15 − 0.32 − 0.36 − 8 − 0.58 − 1.49 0.04 − 0.31 − 0.30 0.43 0.38 1.05 0.58 1.37 − 0.09 0.23 0.04 − 7 − 0.02 0.13 0.45 0.02 − 0.34 0.87 1.13 − 0.11 0.29 − 0.07 − 6 0.21 − 0.29 − 0.78 0.44 0.40 0.08 − 0.11 0.75 1.74 − 0.33 − 0.04 − 1.03 − 0.14 − 1.93 − 5 0.50 0.57 0.19 0.13 0.75 0.65 1.13 0.09 − 0.41 0.83 2.94 − 0.02 − 0.19 2.44 − 0.51 − 0.47 − 4 0.71 0.97 1.97 0.30 1.46 0.47 0.32 1.35 0.21 − 3 − 2.27 − 2.58 0.51 0.21 0.67 0.35 − 0.55 − 1.38 1.96 6.32 − 0.30 0.11 2.38 0.15 0.15 − 2 0.22 − 0.12 0.37 1.41 2.17 2.18 5.88 0.37 0.89 − 0.89 − 0.46 − 0.56 − 1 – – – – – – – – – – – – 0.81 − 1.46 − 2.40 0.20 0.60 0.21 0.04 + 1 0.15 0.84 − 1.10 − 1.05 0.40 − 0.28 − 0.09 − 0.51 − 0.35 − 1.04 − 0.87 − 0.11 − 1.60 + 2 − 0.35 − 0.68 − 0.79 − 2.34 0.24 0.10 0.13 − 0.24 − 0.49 − 0.56 − 1.07 0.40 0.56 − 0.98 − 1.40 + 3 − 1.16 0.63 − 0.60 − 0.99 − 0.21 − 0.09 − 0.20 0.24 0.42 − 0.62 + 4 − 1.29 − 1.74 − 2.07 0.30 − 0.07 0.50 1.04 1.65 − 0.13 − 0.12 − 0.44 − 0.80 − 0.22 1.58 − 1.36 + 5 0.02 − 0.39 − 0.95 − 0.20 − 0.12 − 0.34 − 0.74 − 0.21 − 0.68 + 6 − 0.74 − 0.08 0.06 0.24 − 0.04 − 0.10 0.08 0.79 0.64 0.12 1.37 0.00 0.21 + 7 0.16 0.41 − 0.53 0.21 − 0.93 − 0.40 − 0.74 − 0.08 − 0.06 − 0.11 − 0.17 1.05 − 0.57 − 0.52 + 8 − 1.05 1.07 2.01 − 0.46 + 9 − 1.17 0.21 − 0.07 − 0.06 1.16 − 0.22 − 1.75 − 0.80 0.68 0.39 0.68 − 0.34 − 0.26 − 1.33 0.31 − 0.35 0.76 0.81 − 1.15 + 10 − 2.08 a This table presents the average abnormal returns based on the content of rumors, surrounding the publication date t = 0 in the Ekonomik Trend. Abnormal return is calculated as the difference between the actual and expected return. Expected return is generated from the market model parameters. The ISE Composite index used as a proxy for market. The t-statistics tests the null hypothesis that the average abnormal returns are equal to zero. Event day 0 represents the day of publication of rumorsgossips and corresponds to Sundays. Statistically significant at 1. Statistically significant at 5. Statistically significant at 10. − 5, + 5 periods, CARs are 0.85, 1.19 and 1.35, respectively. All of them are statistically significant at 1 level. In longer time periods firms experience negative insignificant abnormal returns. For example, during − 20, + 20 and − 30, + 30 periods, CARs are − 0.28 and − 1.22, respectively. Based on these results, one would conclude that stock market rumors provide valuable information to in- vestors. Trading based on rumors would provide statistically significant abnormal returns. I further analyze the behavior of stock prices in the period prior and after the publication of rumors. In the middle part of Table 3, the CARs in the pre-publica- tion period for several windows are reported. The CARs in the pre-publication periods are positive and statistically significant. For example; during − 20, − 1, − 10, − 1 and − 5, − 1 periods, the CARs are 1.48, 1.82 and 2.16, respec- tively. While the results of − 20, − 1 period is statistically significant at 5, others are statistically significant at 1 level. The significant positive stock price reaction in the pre-publications days may be explained by two interpretations. The first one is related to the possible use of information by those who initially posses it. These may include insiders, who may use information for their trading, and stock analysts, who may supply information to their clients for trading. The second interpretation can be attributed to the nature of the HOTS column itself. Typically, the stocks mentioned in HOTS are those that recently have been performing well. This may be one of the reasons why almost all HOTS rumorsgossips are favorable. Finally, in the bottom part of Table 3, CARs in the after-publication period are analyzed. The CARs for + 1, + 5, + 1, + 10 and + 1, + 20 windows are − 0.81, − 1.38 and − 1.78, respectively. None of these results is statistically significant. The negative gains in the after publication period would support the view that trading based on the rumors would not benefit to investors and informa- tion does not have any value. These results are contrary to those of recent US studies, which reports a positive price reaction after the publication date. The results of this study seem to suggest the possible dissemination of information prior to the publication date. To further analyze the differences in stock price reaction with respect to the contents of rumors, I classify gossips into six sub-groups. The results of AARs are reported on Table 4. When the pre-publication period is examined, earning expecta- tions rumors, purchases by foreign in6estor rumors, and unclassified rumors groups have positive significant abnormal returns. For example; earning expectations rumors group experiences abnormal returns of 0.65 and 0.67 on days − 4 and − 2. Both findings are statistically significant at 5 level. Unclassified rumors group shows the highest abnormal returns in pre- publication period. The AARs are 2.18, 1.96 and 0.83 for the days − 1, − 2, and − 4, respectively. All of them are statistically significant at 1 level. The rest of the groups do not show any statistically significant specific patterns. In the period following the publication of rumors, the AARs are mostly negative and some of them are statistically signifi- cant. For example; earning expectations group has a return of − 0.62 on day + 4, which is statistically significant at 5 level. Similarly, purchases by foreign investors group experiences a return of − 1.46 on the day + 1. Table 5 reports for CARs for three periods based on the subjects of rumors. In pre-publication period positive CARs are observed for earning expectations ’ rumors, purchases by foreign in6estors rumors, and unclassified rumors groups. For example; earning expectations ’ rumors group experience CARs of 1.54 during − 5, − 1 windows, purchases by foreign in6estors rumors and unclassified rumors groups have CARs of 2.48 and 4.94 in the same time period. While the first two results are statistically significant at 5 level, the last one is highly significant at 1 level. In the after publication period, all CARs are negative but none of them is statistically significant. For combined periods, only CARs for other topics are statistically significant. For example; CARs are 2.37, 3.99, 3.89 in the windows − 1, + 1, − 2, + 2 and − 5, + 5 respectively. The analysis of content of rumors reveals that there are differences in abnormal returns with respect to the content of stock market rumors. Clearly rumors related to earning expectations and purchases by foreign in6estors have greatest impact on stock prices, while others have statistically insignificant effects. Overall, the empirical results indicate the existence of positive statistically signifi- cant abnormal returns in the pre-publication period of rumors. Such findings would refute the strong form of market efficiency, and are in line with the existing literature. The findings pertaining to the post-publication of rumors, on the other hand, show that there are statistically insignificant negative abnormal returns. This suggests that investment strategies based on the published rumors would not generate any wealth gains to investors, implying that information provided by column does not have any value at all. The results seem to suggest the possible dissemination of information prior to publication, and are contrary to those of US studies.

5. Summary and conclusions