CHAPTER VI

CONCLUSIONS

The results clearly show that neural networks can successfully be used in predicting airline market activity.  As was apparent through several of the result sets, other factors not included in the independent variables of this research were present.  Perhaps other, more specific, conclusions can be drawn from this study concerning the way markets react to different events.  The events of September 11, 2001, as well as the introduction of low cost competition, had an obvious impact on the markets, but what were not as obvious were the effects of an exit of a low cost carrier from a market as was seen on the ATL - LAX market.  This clearly showed that the increase in capacity brought about by the introduction of a low cost carrier did not return to previous levels upon the exit of that carrier.  Likewise, fares tended to remain low, and the overall effect of the low cost introduction could be felt years later.

    Additional conclusions can be drawn in situations where a competing market has higher control over price than the market in question.  This was clearly seen in the DFW - ORD market where DFW - MDW fares had a higher impact on the passenger boardings between DFW and ORD than its own fares did.  This was echoed between JFK and MCO where passenger boardings were more sensitive to fare changes on the LGA - MCO market.  Each situation was due to low cost presence in their respective markets, and the JFK - MCO market continues to see stiff competition with capacity steadily increasing and fares remaining low ($117.06 average Q3, 2005).

   The neural network approach to forecasting market performance is promising; however, the results are extremely dependent upon the factors input into the network.  Even with high correlation between fares and passengers, as was seen on the ATL - LAX market, this provides no clear conclusions due to the fact that fares were constantly being adjusted during the quarter, which could not be differentiated by the neural network.