Last Updated 27 Jul 2020

# Airline Demand Forecast

**3328**(13 pages)

**665**

STIMATION OF AIR TRAVEL DEMAND IN TURKEY ENAR TUNC, Orhan sIvrIkaya* Okan UNIVERSITY Title: ESTIMATION OF AIR TRAVEL DEMAND IN TURKEY Orhan Sivrikaya*(Candidate Phd. ), OKAN UNIVERSITY Tel: 0-532-4265392 Fax: [0-212-4652299] Email: osivrikaya@live. com Enar Tunc, Professor of Industrial Engineering, OKAN UNIVERSITY Keywords * Domestic Air Transportation, City Pair, Origin and Destination, Demand, Forecast, Gravity Model, Multivariate Regression and Detour Factor. Total Page: 11 Abstract

Accuracy in estimating airline market demand is a key element while an airline is planning its short term or long term business plan regardless of its status quo being an incumbent or startup company. Turkish domestic market of air travel industry has been dramatically grown in recent years especially after the deregulation commencing on the renewal of air transportation policy in 2003. However there is not any relevant scientific research in the literature to analyze the determining factors on air travel demand of domestic city pairs in Turkey.

A multivariate regression model is generated in order to fit the air travel demand in number of passengers carried. The model is based on aggregate individual market which consists of on-line city pairs. The model is found significantly representative within the experimental data out of the years 2008 and 2009 including the origin and destination pairs for 40 on-line cities. Then, the model is tested by using 2010 figures in order to compare prediction values with actual figures. Accuracy level is found to be encouraging for potential new airports or potential new routes to be evaluated by using the model estimates. . Introduction The deregulation of air transportation market in Turkey in 2003 has started revolutionary changes in the airline industry. New government having the target to increase the portion of air travel out of all modes of local transportation attempted to encourage more airline companies to enter the market and enable them to offer more attractive prices by tax cutting specific to the airline sector. Price oriented competition has worked very well to generate significant airline passenger traffic.

Haven’t found the relevant content?
Hire a subject expert to help you with **Airline Demand Forecast**

Low Cost Carriers have contributed to exercise a sustainable two digit growth by stealing passenger traffic from bus transportation as a result of shortening the gap between relative prices. Turkish Airlines as a legacy carrier has responded to structural changes in the market by applying dynamic pricing policy and growth strategy to benefit from economies of scale resulting in increase in productivity. Big changes in airline passenger traffic in Turkey create a challenge to testify any claiming model built to estimate air travel demand. Macroeconomic or demographic changes do not seem to be responsible for whole boost in air travel demand.

Competition doubled or tripled available seat capacity on some routes so that it was required a different strategy to generate additional demand to achieve in satisfactory load factor which is a key performance indicator for airline profitability. Airline traffic is most of the time considered as a significant indicator for the performance of the nation’s entire industry since it is highly correlated with the number of business events and interactions with other industries simultaneously. So, it implies that changes in economies may influence airlines traffic indirectly.

However, airline specific parameters like ticket price and degrees of competition are also supposed to be main driver for passenger demand besides the macroeconomic factors. The sustainable success of any organization or company is closely related with how well management or decision makers are able to foresee the future and develop appropriate strategies. The objective of this study is to examine the demand size for air transport in Turkey and show its implications for air transport planning. 2. BACKGROUND It has been seen throughout the results of the previous research in the iterature that one of the most important issue to develop a predictive model is to choose the right combination of the variables which represent the determining factors involved in the model. These variables are categorized by two subgroups (Carson et al. 2010): 1. Geo-economics Factors: which consist of geographical characteristics, economical activities, social factor etc. 2. Service Related Factors: which are related to airline dependent factors. The other prominent aspect of model generation is the level of forecast which can be classified by two groups as well; 1.

Microscopic Model: Airport specific or city pair specific data is involved such that it refers the total number of incoming and outgoing passengers per particular airport or per city pair. 2. Macroscopic Model: Region or country specific data is involved such that it refers to aggregated number of passengers in a region or country regardless of origin or destination city. Aggregate Individual Market (AIM) forecast outperforms the aggregate approach since the forecasting power gained by exploiting heterogeneous information across markets dominates the forecasting power lost due to estimation of many coefficients (Carson et al. 2010).

Local area information appears to be more relevant in determining local O&D travel than of national information such as gross domestic product (Bhadra 2003). ------------------------------------------------- 3. OVERVIEW OF THE determ? nants for air passenger demand ? n turkey Turkey is spread over a wide geographical area and road ways are not adequately constructed for all direction. Hence, air transportation is supposed to have more shares out of total statistics in domestic transportation covering all possible city pairs. While the gap between relative prices is being shortened, more and more people find it affordable to fly.

This study is aiming to find out the determining factors which are concerned to turn potential demand into air travel passengers. The proposed model is not only to explain actual traffic results but also to estimate potential traffic between cross cities which are not connected directly or to evaluate off-line cities to build new airport. Population, gross domestic product per capita and employment rate are considered as the leading macroeconomic dynamics behind air travel demand as depicted in the Table 1. Average fare has a stimulating effect on airline demand as Brons et al. 2002) pointed out that ticket price is an elastic driver for airline demand generation. There are also specific indicators for a particular city pair traffic representing interactivity between the concerning cities such as distance and number of migrants from each other. The number of bus registered in a city is indicating the volume of bus transportation which is considered to be negatively related with air travel demand. Since number of carriers as a degree of competition contributes to market expansion, it is also embedded in the model expecting a positive relation with air travel demand.

Table 1 Commonality in Types of Variables Variable’s Name| Percentage of Occurrence*| GDP| 50. 0 %| GDP per Capita| 35. 7 %| Unemployment Rate| 14. 3 %| Fuel Price| 7. 1 %| Number of Employees| 7. 1 %| Population| 42. 8 %| Average Fare| 57. 1 %| CPI| 14. 3 %| Trade per Capita| 14. 3 %| Exchange Rate| 14. 3 %| Service Frequency| 28. 6 %| Distance| 42. 8 %| Expenditures| 7. 1 %| * The percentages are calculated out of a sample of 14 different relevant articles. Most of the itineraries between city pairs are not directly connected that means air passengers travel with connecting flights via one or more transfer points.

If there is no direct service the dummy variable transit gets 1 and 0 otherwise. Naturally, passengers would not prefer to fly with connecting flights so it is anticipated to be negatively affecting air travel demand. 4. ECONOMETRIC ESTIMATION data, Methodology and results Data availability is main issue when data coverage is decided. Experimental model is based on the data of the two years 2008 and 2009 since all explanatory variables are available within the specified period. There are 40 on-line destinations in domestic network in Turkey.

This number of destinations can theoretically generate 1560 different origin and destinations (O&D’s) on which direct or connecting flights are possible. However experimental sample does not cover data for all possible on-line O&D’s because some city pairs which are at close distance are not meaningful to fly with connecting flights or the concerning flights are not connected each other. There are 231 city pairs which are served with direct flights, whereas the remaining city pairs are found to be flown by connecting flights via an appropriate domestic hub.

Under the assumption of approximately the same number of O&D’s for each year, data size will be duplicated for the two year’s period. Airport statistics for all scheduled carriers are used in the experimental model as a source of the dependent variable. Transfer traffic is removed from the statistics for each city pair, since the proposed model is to estimate pure O&D passenger by using data specific to the corresponding city pairs. Average prices for each city pair are estimated by using airlines’ web site. Road distance between the cities is taken from the web site of the General Directorate Highways of Turkey.

Population of the cities, GDP per capita of the cities, the number of migrants between the cities, the number of bus registered in the city’s account and labour rate of the cities are obtained from the Bureau of Statistics in Turkey. Weighted average of the corresponding city’s population is used, while GDP per capita and the labour rate are being converted to O&D level. A variety of different models exist for passenger volume estimation. Since no single model guarantees accuracy, airlines in fact compare forecasts from several different models.

Within this set of forecasting methods, the most demand models used are of the simple gravity type formulation. (S. C. Wirasinghe et al. 1998). The gravity model for the estimation of domestic passenger volume between city-pairs is examined in this study. By excluding unavailable service-related or market specific input variables, and using cross-sectional calibration data, the model is particularly applicable to city-pairs where no air service exists, historical data is unavailable, or factors describing the current service level of air transportation are not available.

Average price for city-pairs with no air service is estimated by fall back mechanism that it uses the average price which is normalized by distance of the cities having similar market structure. All other explanatory variables are not service related factors and available for the city-pairs with no air service. The gravity model takes the form: D=?. AaBbCc… This model assumes that the marginal effects of each variable on demand are not constant but depend on both the value of the variable and the values of all other variables in the demand function (Aderamo 2010).

In other words, the explanatory variables affect demand in multiplicative manner. Partial derivation of any independent variable proves aforementioned relationship. However, this model can be made suitable for multiple regressions by applying logarithmic transformation. Logarithmic form of the gravity model takes the form: LogD=? 0 + ? 1LogA + ? 2LogB + ? 3LogC +… where ? 0=Log? It is obvious that interdependency is resolved in this form so that multiple regression model can be applied. The proposed multiple regression model is generated by using SAS Jmp 9 tool.

Table 2 shows the matrix of correlation between the independent variables. The results show that some of the variables are interrelated. For example, Log_Migrant has a correlation coefficient of 0. 8661 and 0. 8150 with Log_Pop* and Log_Bus* respectively. Where both Log_Migrant* and Log_Pop* are calculated by taking the product of population of origin and destination cities. However, omitting any of these two variables would substantially reduce the model fit. As the goal is to obtain a reliable estimation of the passenger volume, all interrelated variables were included (Grosche et al. 007). Furthermore, it has been said that if the sole purpose of regression analysis is prediction or forecasting, then multicollinearity is not a serious problem because the higher R2, the better prediction(R. C. Geary, 1963). In order to verify stepwise regression fit of the model, stepwise process by backward direction and minimum AICc selection is used. When all independent variables as depicted in Table 2 are entered, the smallest AICc value 2665. 913 is found. Adjusted R2 as shown in the Table 3 is 0. 823991 which is fairly good.

In the Table 4, adjusted R2’s are compared including the relevant articles in the reference list. This comparison table shows that the studied model efficiency is relatively successful. As shown in the table 5, the F test also shows that the regression is significant since F statistic of 497. 2411 is obviously higher than the critical value of 2. 32 at 0. 01 level of significance. In the table 6, parameter estimates are depicted. As seen in the table, all independent variables are significant at 0. 01 level of two tail significance considering their t-statistics.

Since the coefficients of the regression model represent elasticities of the corresponding variables, how change of any variable affects demand estimation can be determined. The price elasticity of passenger demand is approximately -1. 1 which implies that airline passenger demand in Turkey is elastic. This finding is compliant with the fact that after low cost carriers entered into the market by lowering ticket prices, market size has been tramendously enlarged. Domestic passenger traffic grows higher than the decreasing rate of ticket price.

Both GDP per capita and ticket price seem to have elastic impact on passenger demand estimation. Air transportation and bus transportation seem to be competing each other because of their negative relation. When air service is provided by connecting flight which means transit traffic, air transport demand is decreasing. This result is not surprising because people do prefer to fly directly. Another result is that the number of airlines participating in each O&D market tends to have a positive impact on the number of passengers traveled between O&D pairs, perhaps representing the ffects of choice more than anything else. Lastly, distance and the number of migrants are found positively related with air transport demand as expected. Table 4 Model Efficiency Benchmark| Research Name| Level of Forecast| Author| Year| Independent Variables| Observation| Adjusted R Square| Demand For Air Transport In Nigeria| Aggregate| Adekunle J. Aderamo| 2010| Index of AgricultureIndex of ElectricityGDP| 23| 0. 923| Air Travel Domestic Demand Model in Bangladesh| Aggregate| Md. Jobair Bin Alam Dewan Masud KArim| 1998| PopulationGDPDistance| 31| 0. 8| An Econometric Analysis of Air Travel Demand in Saudi Arabia| Aggregate| Seraj Y. Abed Abdullah O. Ba-FailSajjad M. Jasimuddin| 2001| PopulationTotal Expenditures| 25| 0. 959| Regression Model for Passenger Demand: A case study of Cairo Airport| Aggregate| Dr. Khaled A. Abbas| 2003| Population GDPForeign Tourist| 88| 0. 82| Demand for Airravel In USA| O&D| Dipasis Bhadra| 2003| Density, Interaction, Distance, Marketshare, Fare| 2424| 0. 57| An Aggregate Demand Model in Hub-and-Spoke| Aggregate| Wenbin WeiMark Hansen| 2006| Frequency, Number of Spokes, Fare, Distance, Capacity, Traffic Type| 897| 0. 92| Gravity Model for Airline Passenger Volume Estimation| City-pairs| Tobias GroscheFranz RothlaufArmin Heinzl| 2007| DistancePopulationCatchment Area| 956| 0. 761| The number of migrants indicates the relationship between city-pairs hence it positively affects on point to point air traffic demand. When distance is greater, air transport demand increases due to the fact that people get higher utility comparing to the alternative modes of transportation. In the figure 1, model fit of the experimental data is shown in scatter diagram. There are total 955 observations within experimental data.

A test data is obtained from 2010 actual results which consists of 562 observations. The model predicts 2010 figures with a Mape (Mean Absolute Percentage Error) value 14. 1 %. Actual data of 2010 is refined by excluding the O&D’s having less than 104 yearly passengers flow and detour factors smaller than 3. Logic of this filtering is to choose meaningful connections out of the all itineraries. Although the model is performing significantly well with a relatively high Rsquare value, small discrepancy in prediction value may result in larger inaccuracy in passenger demand estimate because of logarithmic aspect of the regression. . CONCLUSION This study demonstrated that the proposed econometric estimation and using micro data based on local area information can result in substantial insights to O&D travel. The demand model reveals all the quantitative relationships among the used variables, which is helpful for airlines to understand the consequence of change of their decision variables or adjustment of their routing structures, and also useful for the related authority to quantify the benefits of airport capacity expansion and to take into account while airport building plan is being evaluated.

It would be advantageous to extend the time period covered by the analysis. This would enable to examine possible differences in elasticity amongst city-pairs. Extending the data back in time would also provide observations of airfares progress. The model efficiency may be improved for even more reliable estimation, if more independent variables indicating bilateral relations between city-pairs are embedded in the model such as the number of call between city-pairs or credit card statistics of domestic visitors. References S. C. Wirasinghe and A. S. Kumarage, An Aggregate Demand Model for Intercity Passenger Travel in Sri Lanka.

Transportation 25: 77-98, 1998. R. C. Geary, Some Results about Relations between Stochastic Variables: A Discussion Document, Review of International Statistical Institute, Vol. 31, pp. 163-181, 1963. Richard T. Carson, Tolga Cenesizoglu and Roger Parker. Aggregate Demand for USA Commercial Air Travel. Department of Economics, University of California. 2010. Elton Fernandes and Ricardo Rodrigues Pacheco. Air Transportation Analysis: Passenger Demand in Brazil. Aerlines Magazine e-zine edition, issue 33. Adakunle J. Aderamo. Demand for Air Transport in Nigeria. Journal of Economics, 1 (1): 23-31 (2010).

Md. Jobair Bin Alem and Dewan Masud Karim. Air Travel Demand Model For Domestic Air Transportation in Bangladesh. Journal of Civil Engineering The Institution of Engineers, Bangladesh Vol. CE 26, No. 1, 1998. Seraj Y. Abed, Abdullah O. Ba-Fail and Sajjad M. Jasimuddin. An Econometric Analysis of International Air Travel Demand in Saudi Arabia. Journal of Air Transport Management 7 (2001) 143-148. Abdullah O. Ba-Fail and Seraj Y. Abed. The Determinants of Domestic Air Travel Demand in the Kingdom of Saudi Arabia. Journal of Air Transportation World Wide Vol. 5, No. 2 - 2000. Abdullah Omer Ba-Fail.

Applying Data Mining Techniques To Forecast Number of Airline Passengers In Saudi Arabia (Domestic and International Travels). Aviation Institute, University of Nabraska at Omaha, 2004. Senanu Ashiabor, Hojong Baik and Antonio Trani. Logit Models For Forecasting Nationwide Intercity Travel Demand in the USA. Journal of the Transportation Research Board, No. 2007, Washington, D. C. , pp. 1-12 2007. J. Scott Armstrong and Michael C. Grohman. A Comparative Study of Methods for Long-Range Market Forecasting. Management Science, 19 (2), pp. 211-221 1972. Khaled A. Abbas. Conceptual and Regression Models for Passenger Demand Prediction.

Aerlines Magazine e-zine edition, issue 26 2003. Dipasis Bhadra. Demand for Air Travel in the United States: Bottom-up Econometric Estimation and Implications for Forecasts by Origin and Destination Pairs. Journal of Air Transportation Vol. 8, No. 2 – 2003. Radosav B. Jovanovic. Generalised Intrinsic Characteristics as a Forecasting Tool: A dynamic perpective. Second International Conference on Research in Air Transportation – Belgrade, June 24–28 2006. Martijn Brons, Eric Pels, Peter Nijkamp and Piet Rietveld. Price Elasticities of Demand for Passenger Air Travel. Journal of Air Transport Management 8 pp. 65-175 2002. Dail Umamil Asri and Yoriyasu Sugie. Simultaneous Demand Model for Passenger Travel. Proceedings of the Eastern Asia Society for Transportation Studies, Vol. 4, October, 2003. Joyce Dargay and Mark Hanly. The Determinants of the Demand for International Air Travel to and from UK. ESRC Transport Studies Unit, Centre for Transport Studies, University College London, November 2001. Catherine Zhukovskaya. Use of the Generalized Linear Model in Forecasting the Air Passengers’ Conveyances from EU Countries. Computer Modelling and Technologies, Vol. 11, No. 1, pp. 62–72, 2007. Wenbin Wei and Mark Hansen.

An Aggregate Demand Model for Air Passenger Traffic in the Hub and Spoke Network. Transportation Research Part A 40 pp. 841–851, 2006. Matthew G. Karlaftis. Demand Forecasting in Regional Airports. Straer 7 pp. 100-111, Tr. 312, 2008. Tobias Grosche, Franz Rothlauf and Armin Heinzl. Gravity Models for Airline Passenger Volume Estimation. Journal of Air Transport Management 13 pp. 175-183, 2007. Chaug-Ing Hsu and Su-Miao Liu. Predicting City-Pair Air Passenger Traffic Using Grey Topological Forecasting Model. Journal of the Eastern Asia Society for Transportation Studies, Vol. 5, October, 2003.

Haven’t found the relevant content?
Hire a subject expert to help you with **Airline Demand Forecast**

### Cite this page

Airline Demand Forecast. (2018, Jul 31). Retrieved from https://phdessay.com/airline-demand-forecast/