The following paper will discuss the event study that was completed to identify any abnormal returns associated with Apple Computers stock price during the period in which they released a new product into the market.  Then it will discuss whether the market predicted the future success of the product with an abnormal return.  The paper will outline the research process and will briefly discuss how the event study was completed.  Then the conclusions and results will be presented.

                This project used what is called standard event study methodology.  Event studies are generally associated with economics, finance and accounting events.  Event study involves selecting a certain type of event and then studying the security price change in response to the selected event. (Cowan)  Basic financial theory says that we should assume that security prices reflect all relevant information.  Therefore, only unanticipated events should cause a fluctuation in the securities' price.(Fama 1970)  The event for this project was the release of a new product by Apple Computer Corporation. 

            The first step in the project was to ascertain the major products of Apple, and this information was abstracted from several sources.  The biggest contributor was a website based on Apple products located at the url: http://www.theapplemuseum.com/index.html.  The list of products included everything from desktop computers to printers and software produced by Apple Computers.  After the list was composed, introduction dates needed to be assigned to each product.  This was the phase of research that took the most time and was extremely toilsome.  Once again the apple museum website was referenced.  The main problem was that often products were assigned a relative period in which they were introduced and not a specific date.  For this project we needed the exact date of introduction.  Lexis-Nexis was used in an attempt to find specific articles relating to individual products.  This practice yielded poor results and was soon discontinued.  The Wall Street Journal index was referenced next.  This index contains daily excerpts on each news article associated with a company.  The index was sorted by major companies in alphabetical order.  Almost all of the products used for this project, with a date of pre 1997, were listed in this index.  This was a great tool for verifying dates from other sources and finding additional data.  The products that were released in the post 1997 period were cross referenced with the press releases maintained in a database found on the homepage of Apple Computers.  The judgement was made early on, with the guidance of the Professor Piwowar, to use the date that pertained to the actual product being present in the news media.  This meant that if news was leaked before Apple made a press release, then we would use the date that the information was leaked.  The theory is that the market reflects all present and available information, so it was decided that we would use the date news was released or leaked.  At the same time caution was used in order to find the date that the actual product’s dimensions and specifiications were released.  Many articles prior to the release date often made statements that inluded information about a new line or new product coming out, but they did not offer much description.  The dates presented in this case usually were correlated with an event like a trade show or an Apple Computer convention. 

            The next step was to define each product as a success or a failure.   Various articles and websites were used to find this information.  The success or failure question created another dilema.  What should be the criteria for defining a product as a success or a failure?  One could argue that a product is successful if it is innovative and performs well.   The problem with this is that many products may be innovative and functional, but they don’t sell well. “Lisa was Apple's first project that featured Apple's GUI (graphical user interface). At that time the Lisa was the only computer with a GUI on the personal computer market. The Lisa OS was very innovative featuring pull-down menus and it actually looked pretty much similar to the Mac OS. Unfortunately Lisa sold poorly due to its incredibly high price.”(theapplemuseum)   The decision was then made to categorize products as a success if they had profitable and successful sales numbers. 

            This was a point in the research when many products were dropped from the research pool.  Many times it was difficult and probably impossible to find related news articles that stated if a certain product was a success or failure.  At earlier stages of the project several other products were removed because of insignificance or lack of a release date.    

           The next step in the process was to use the Evantus software.  This software was developed by Arnold Cowan a professor at Iowa State University. “Eventus performs state-of –the art event study estimation and testing using the CRSP stock files or other stock return data and provides fast event-oriented data retrieval from the CRSP stock files.”(Cowan)  In order to run the Evantus software, we had to separate the products that were failures from the products that were successes.  Then the dates were formatted in a yyyy/mm/dd style and products with duplicate dates were removed.  The sample size was cut from forty-five failures to twenty-six , and the successes were reduced from ninety- five to thirty-nine.  The dates were now the main input for the program.  Next, a permnode had to be assigned to each date.  A permnode is a unique identifier for a company over time.  Each company has one permnode, no matter how many times it changes its name or it is involved in merger related activity.  For this product one number was assigned for each of the dates because all the dates were associated with apple computer.

            Eventus Software takes an event date then calculates returns for a period of plus and minus two hundred fifty days for the given company.  It then calculates the return for the market for the same period.  The return for the market is found in order to net out the economic and other related factors that may have contributed to the return for the company.  After taking out the market effect on the return of the company we are left with a firm specific return.  After the returns are calculated, Evantus uses SAS, which is a statistical program used to find if there is statistical significance in the returns for a specific event.  Our event was of course the release of products by Apple, and figures 1,2, and 3 show the results.  A period of five business days was used as the parameter for the results.  The successes showed a mean abnormal return of .36% on the day of the event and a positive .04% two days later.  When looking at the day two days ahead of time we see negative .14%.  There was no statistical significance for these numbers at neither the one, five, or ten percent levels.  The next thing that the results showed was the number of positive returns compared to negative returns on each day during the five-day period.  This was also inconclusive when looking at the generalized z.   When looking at the raw number we see almost as many positive returns as we do negative returns.   The next table show the cumulative abnormal return, which is the running total of abnormal returns for the period.   The graph which is figure 1, is based on abnormal returns and seems to show some positive activity when looking at the returns for the event period, but the statistical data showed no real significance.

            Next we looked at the failures. The event day showed a negative .4% on the day of the release. The other dates ranged from negative .47% mean return two days before the event to negative .27% two days after the event.  None of the other days held any statistical significance either.  Next, the number of positive returns to negative returns was looked at for the sample.  This showed statistical significance for the event day where we saw six positive returns compare to seventeen negative returns.  The level of significance was at the five- percent level.  Finally, we looked at the cumulated abnormal returns, which calculates the cumulative effect of the period around the event date.  This showed the same significance for day 0 as the first data did.  These results however, also showed significance for day one at the ten percent level when looking at the number of positive returns compared to negative returns. 

            There are several analysis that one might try to make on the reported data.  First, we might say that the market really does not adjust the price of apple when they release new products.  Second, that when Apple releases new products and they will ultimately be failures, the market price of Apple Computers falls.  At the same time when an Apple Computer releases a good product the stock price is unaffected.  The stock price may be unaffected because the stock market expects Apple to release successful products over time and there for does not reward them.   If this second analysis is correct, then we should be able to predict if a product will ultimately be a failure by evaluating the return for Apple on the date of a the product release.  This would be supported by the trend we see in the graph presented in figure 1 and also by the statistical significance for the number of positive returns compared to negative returns on the event date for the group of failures. 

            The second analysis is an interesting theory but lacks support for several reasons.  The data sample was originally reduced to twenty-six dates for failures and thirty-nine dates for successes.  The numbers were then reduced even more when we used the Eventus software.  One date was rejected because it was from before Apple went public, and several other dates were rejected because the CRSP database does not contain data after the year 2000.  This meant that for failures our sample size was really two small because it contained only twenty-three samples, when the minimum should have been thirty.  Another relevant factor that may have created a disturbance in the project was that some of the failures were released on the same dates as the successes.  If the market truly does reward Apple for good products and punish their market value for bad ones then the effects may have netted out.  At the same time if the theory that their stock price only declines when the product is a failure and does not react to a successful product, then the results would be unaffected by the use of a date in both samples. 

            Even though the results of this event study were inconclusive they do provide a starting point for future research.  As new products are released they can be added to this sample or be included into a new sample.  At the same time an individual could go back and try to find more information on products that were not included in this project.  If one could find significant statistical data that there is correlation between Apples stock price and the ultimate success of their products, we could use this to predict future successes and failures.  At the same time if one could predict if a product will be successful before the release date then that person would be able to predict the movement of Apples stock price on the release date. 

References

Barens Index.  1980-1997. New York N.Y. Dow Jones & Co.

http://www.itp.net/reviews/hardware/99182384277042.htm

http://www.apple.com

http://www.theapplemuseum.com/index.html

Fama. E. F. (1970) "Efficient Capital Market: A Review of Theory and Empirical,"

Journal of Accountancy, 25, pp. 383-417

Malone, Michael.  Infinite Loop: How the World’s Most Insanely Great Computer

Company Went Insane.  1999 N.Y., New York 

Nairn, Geoffrey.  “Challenge Is to Create a Convincing Long-term Strategy: Apple’s           

Ft Information Technology Review; Section: Survey Pg. 6

Smith, Dawn “A Market's Identity Crisis” Brandweek. September 1991 Section: Super

 Brands

Wall Street Journal Index.  1998 – 1999 New York, N.Y.  Dow Jones & Co.