Critical Analysis and Gap in Research

By combining the information above we can make the following inferences:

1. Stock markets adjust to direct company-related news almost instantaneously due to the existence of data-mining programs

2. Because of the data-mining programs arbitrage opportunities arising from news disappear

3. Theoretically, an event in an oligopolistic market can be beneficial for one company, but bad for another

4. Data-mining programs, however, are not sophisticated enough to predict competitors’ stock price reaction to a news article

5. Some events are empirically tested to be similar to Situation X, such as cost reduction

However, the literature does not cover the influence of news on competitors’ stock prices. The articles I mentioned in the “Competitors’ Effect Literature” cover only quantitative data, such as found in financial statements, earnings forecasts and repurchases.

Moreover, text-mining software does not exploit this kind of news. They are, however, able to identify situation Y news, as the article is shared by all the companies within the industry.

Therefore, the combination of news effect on competitors and data mining is the scope of research of this dissertation. It will attempt to test the existence of arbitrage profits from such events.

Methodology

Sample selection

Technological companies that are constituents of S&P Technology & Hardware Equipment (LSP50THE​) are selected for the sample. The sample period is chosen from 25th November, 2011 to 26th November, 2013. Companies that delisted during the period are removed. Also, firms with market capitalisation of less than 1 billion USD are deleted in order to have liquid markets. Finally, companies with competitors which do not fit the above conditions are removed as well.

The competitors for the companies are chosen from Yahoo Finance competitors section. If some of the competitors did not fit the criteria above (i.e. were not listed in the SP technology index) they were not included in the sample. Some companies appeared to have one competitor, while the rest had two.

Filter		# of firms
S&P Technology Index			58
Delisted			11
Illiquid			7
No competitors			11
Final Sample			29
Table 1 – Sample Filters

The technological sector is chosen for the research because oligopolistic market structure prevails in this industry, Pindyck and Rubinfeld (2013). Delisted companies are removed from the sample, as their returns are zero for all non-trading days, which would imply no reaction to competitors’ news due to research design (discussed later). It also removes the ability to test for arbitrage opportunities. Illiquid companies are also removed from the sample as their trades are rare. Moreover, the difference in market capitalisation cannot qualify them to be competitors, as suggested by the Yahoo Finance. Finally, companies, whose competitors fail to fit these filters, are also removed for the same reasons.

Research Design

Ratios of response to significant news analysis

Unlike the previous studies, news pieces in general are difficult to be judged. One can think of an article as a good event, while others may do the opposite. Therefore, I allow the market to decide whether the news was good or not. Good news attracts stock buyers and the share price goes up. If the stock return was positive on the release date, then we can say that the news was generally accepted positively by the investors. On the other hand, if the stock return is negative we can say that, overall, investors sold shares, which implies bad news.

We should mention that not all days have news pieces that are related to the sampled companies. Some portion of volatility is provided by non-company related factors, such as political stability in a different country. Therefore, the paper should be mainly concerned about the effect of significant news’ effect on competitors’ stock prices, and therefore not all days should be included. To identify only significant news we can use a similar technique – if the volatility of the returns was abnormal at time t, then there was a significant news release at time t.

Another problem is economy-wide or industry-wide news. Changes in exchange rates or monetary policy affect all the companies and we do not want these to count for significant company-related news. If a major news drove all the stocks up (or down) by a considerable margin we have to control for these types of events.

For example, Agilent Technologies has 503 observations of daily returns – . To control for the major economic news I subtract the Standard and Poor 500 ( ) index returns from the returns of Agilent Technologies. I am left with which stands for company-related returns (adjusted Agilent Technologies returns at time t). Then the square of the adjusted returns of A is calculated – to directly compare positive and negative volatility. After that outliers in the squared adjusted returns are found in order to identify significant company-related news articles. The outliers are located by using the formula

.

3Q represents a value that is larger than 75% of observations and IQR represents the inter-quartile range. If the inequality holds true, then we can say that at time t there was a piece of news that significantly affected the returns.

Now that we identified important news articles we need to check their effect on the stock returns of the competitors. We conduct a similar procedure with the competitors’ stock returns. For Agilent Technologies, the competitors are Teradyne and Thermo Fisher Scientific. In case of Teradyne, we calculate the adjusted Teradyne returns at time t: and then also find the outliers, which are such if

is true.

Then, if the events of significant returns coincide with each other we can say that one company is affect by the news article of its competitor, or that there was an specific industry-wide, but not economy-wide news, that affected both companies. We shall call these events co-reactions. In case when both companies’ adjusted share prices move in the same direction we can say that situation Y happened and both companies benefited or suffered. If the stocks moved in the opposite directions situation X happened.

Then, the number of situation X co-reactions is calculated and compared to the number of observations. A binomial test is conducted to see if the number is statistically significant from zero. If yes we can conclude that important news for Agilent Technologies can lead to a sharp opposite change in the share price of its competitor, Teradyne. All the steps are repeated for Thermo Fisher Scientific.

The test is conducted across all other companies. However, due to results being mainly insignificant, a less strict test is conducted. Competitors within the sample seem to show a high degree of correlation in the returns and therefore their returns are rarely opposite. To relax the test we identify significant adjusted returns for Agilent Technologies and check whether they coincide with generally opposite adjusted returns for Teradyne. This time not only significant co-reactions are identified, but any opposite returns as well. The number of such events is calculated and divided by the overall number of significant returns for Agilent Technologies. The ratio is then tested if it is statistically different from the general ratio of opposite returns. If yes, we can conclude that important news for Agilent Technologies has a negative effect on the stock price of its rival, Teradyne. Again, the relaxed test is run for other firms.