A Model to Forecast Aircraft Operations at General Aviation Airports

Transcription

1 Journal of Advanced Transportation, Vol. 31, No. 3, pp A Model to Forecast Aircraft Operations at General Aviation Airports Atef Ghobrial Introduction Forecasting the demand for aviation activities is an important task in economic planning. Traditionally, forecasting the demand for general aviation airports has been done using trend and ratio analyses. Both techniques ignore the changes in many external factors that influence aviation activities. This paper develops an econometric model that relates the number of aircraft operations at a general aviation airport to some socioeconomic characteristics and supply variables. The model was estimated using data for 82 airports in the state of Georgia. The results of estimating the model suggest that the demand is inelastic with respect to total employment. The results also show that presence of aviation related services such as avionics, charter flights, rental, repair and crop dusting are important factors in determining aviation activities. The demand seems to be responsive to the location of the county as a tourist/recreational destination. Air transportation is a vital component in the economic development of communities. Many economic activities (e.g. manufacturing, financial services, agriculture, and tourism) are dependent on a reliable air transport system. If the system serves these industries efficiently, then the overall economic activity will be improved. Benefits derived from the development of airports accrue to both users and local community. Initially, new jobs are created by construction activities at the airport and later by the increase in airport improvement activities. In the case of general aviation airports, Jobs are also created by fixed base operators (FBO) as more services are needed. Many less tangible benefits can also accrue to the community. An airport can provide access to the community s recreation facilities from other parts of the region. It can provide a site for airshows and a base for low cost vacation charter flights. The airport is also valuable in times of emergency to evacuate Atef Ghobrial is a Professor in the Transportation Studies Program, School of Policy Studies, Georgia State University, Atlanta, GA, U.S.A. Received Augunt 1994; Acccptal August 1996.

2 3 12 Atef Ghobrial seriously ill patients and to evacuate the area in case of natural emergencies such as hurricanes and floodings. Aerial fire fighting and airborne search and rescue can be of vital help to the community during times of trouble. Development of general aviation requires large infrastructure in the form of airport and aircraft systems and depends on a technology with a rather long lead time for development and implementation. Consequently, forecasting the demand for aviation activities has become an important task in economic planning. Many economic impact studies of general aviation airports have also been developed to demonstrate the future benefits of these airports to the community. Forecasting the future demand for general aviation airports is an important ingredient in these studies. In addition, due to the limited resources of the FAA to provide construction/- improvement grants to general aviation airports, the FAA uses the forecasts of activities at these airports as input into its decisionmaking process. Forecasting the demand for general aviation airports has traditionally been done using trend analysis by extrapolating the historical trend in aviation demand, and/or ratio analysis by extrapolating the ratio between demand and population or demand and employment ratio. Both techniques assume a linear relationship either over time (trend analysis) or between aviation activities and population or employment (ratio analysis). These techniques have generally been used in developing master plans for general aviation airports. Examples include Peatchtree- Dekalb airport in Georgia, Lakefront airport in Louisiana, and the proposed Madison county airport in Florida. The trend analysis and ratio analysis techniques are discussed in Horonjeff and McKelvey (1994). There are some disadvantages of applying these techniques: The trend analysis technique assumes that the growth rate in demand in the past will continue to be the same in the future. This technique does not explicitly examine the underlying relationship between the projected activity descriptor and the many variables that affect its change. These variables include creation of new industries in the region and improved levels of service at the airport as a result of extending runways, providing avionics and repair services at the airport, and adding a control tower. The trend analysis technique is useful for short- and medium term forecasting (i.e. one to three years) unless significant external changes are foreseen. Similar to trend analysis, ratio analysis technique ignores the impact of changes in external variables on potential demand for general

3 A Model to Forecmt Aircraft Operatio ns aviation airports. One cannot use this technique to assess the combined effects of changes in population and employment on future demand. For example, two communities with the same population but with different employment levels should produce different aviation activities. Forecasting methodologies for airport operations varies from professional judgment to econometric modeling. Earlier studies in the area of forecasting airport activities using applied econometrics focused primarily on air carrier operations. The FAA (1975) developed a multiple regression model to project the national level of annual air carrier enplanements from the period 1978 to The explanatory variables in the model are the number of civilians employed, annual purchase of automobiles, private investment in air transportation plants and equipment, and price of air transportation relative to other transport modes. An econometric intercity travel demand model which has been used in several transportation planning studies, including the Northeast Corridor Project, was adapted for use in the Michigan State Airport System Plan (1975). The independent variables include population, travel cost and time, and frequency of service. Ghobrial and Kanafani (1995) developed a hubbing model that estimates airport traffic in the context of a network. The model takes into consideration presence of competing airports in the region and congestion delay at the hub. The FAA (1989) forecasts the total general aviation fleet with a model that includes the following variables: sales of business aircraft, sales of personal aircraft, aircraft price index, implicit Gross National Product Deflator, rate of interest, income and a measure of business activities. In an attempt to improve demand forecasting for general aviation airports, Ghobrial and Ramdass (1993) developed an econometric model using some explanatory variables. Their model was estimated using data for 20 airports in Florida. This paper extends the work of Ghobrial and Ramdass by enhancing the structure of the model and using a large sample of general aviation airports to estimate the model. The paper develops an econometric model that relates the demand for general aviation airports to some socioeconomic characteristics and supply variables. The model can prove useful in projecting the needed improvements in general aviation airports, and in assisting city, county, and state officials in planning and allocating resources for constructing and upgrading aviation facilities.

4 3 14 Atef Ghobrial The Model We used a multiple regression model to forecast aviation activities at general aviation airports. The dependent variable in the model is the number of annual aircraft operations at a given general aviation airport. Aircraft operations include local and itinerant operations. The former is defined as operations in the local traffic pattern or in the local practice area within a 20 mile radius of the control tower, or executing simulated instrument approaches. Itinerant operations include all aircraft arrivals and departures other than local operations (e.g.; enroute flights). Ideally, one would develop two separate models for local and itinerant aircraft operations. Because a breakdown of aircraft operations by type of operations was unavailable, the dependent variable in the model includes both local and itinerant operations. The independent variables in the model include a set of descriptors of the nature and level of the socioeconomic activities in the county where the airport is located, and another set of supply variables that affect the levels of service at an airport. An investigation of the literature on transportation demand analysis suggests that population and employment are representative of the socioeconomic characteristics of the region. Since the characteristics of a given airport are well known, it is possible to specify supply variables with greater detail. This permits a thorough treatment of the effects of levels of service characteristics on demand. The supply variables in the model include runway length, presence of an air traffic control tower, and presence of aviation services at the airport (e.g.; avionics, repair, charter flights, and rental services). We also included a location-specific variable for airports in tourist/recreational areas or those providing access to urban areas. The location-specific variable measures the relative attractiveness of these airports for aviation demand. Based upon the above discussion, we present a demand model that takes on the following form: D =ap BE 'exp(cyrw +4CT +pal+qav +wcf +XRN +dag +6RP)e (1) where: D = the number of annual aircraft operations at a given general aviation airport. P = population of the county where the airport exists. E = employment (work force) in the county.

5 A Model to Forecast Aircrafl Operations RW = a dummy variable which takes on the value one if the airport has a runway longer than 4OOO feet, and zero otherwise. A 4OOO foot runway is considered the threshold for the levels of service at a general aviation facility according to the types of aircraft that can be accommodated by that facility. Runways longer than 4OOO feet can accommodate heavier types of aircraft. CT = a dummy variable for the presence of an air traffic control tower, which is set to one if the airport has a control tower, and is set to zero otherwise. AL = a dummy variable which takes on the value one if the airport is located in a tourist/recreational area or provides access to urban areas, and zero otherwise. This variable captures the effect of airport location (i.e., urban/tourist versus rural). AV = a dummy variable for presence of avionics service at the airport, which takes the value one if these services exist, and the value zero otherwise. CF = a dummy variable for presence of charter flights services at the airport, which takes the value one if these services exist, and the value zero otherwise. RN = a dummy variable for presence of aircraft rental services at the airport, which is set to one if these services exist, and is set to zero otherwise. AG = a dummy variable for presence of crop dusting and other related services at the airport which is set to one if these services exist, and is set to zero otherwise. RP = a dummy variable for presence of aircraft repair services at the airport which takes the value one if these services exist, and zero otherwise. a, 8, 8, y, p, p, 7, a, X, 4 and 6 are the coefficients to be estimated and E is the error term of estimation. From the above definitions of explanatory variables, one would expect the signs of coefficients 8, 8, y, p, p, 7, o, A, 4 and 6 to be positive. Empirical Results To demonstrate the application of Equation (1) to forecasting activities at general aviation airports, we used the state of Georgia as a case study. The Georgia Department of Transportation has recently completed a study to develop a new statewide airport Plan. It provided the documentation of the needs of airports and airport related facilities to

6 3 16 Atef Ghobrial meet the current and future air transportation in the state. The plan included air carriedair cargo airports, general aviation system, and economic impact study. Data for 82 general aviation airports in 1991 were obtained from information compiled by Georgia Department of Transportation. For each airport, data included the annual number of aircraft operations, description of the airport (i.e. number and length of runways and type of pavement), and description of the aviation-related services available. Figure 1 depicts a distribution of airports in the sample according to their annual number of operations; 62 airports have annual operations of less than 10,OOO and only 5 airports have more than 50,oooO operations. A qualitative analysis of the variables was first conducted to detect possible correlations between them. As in most transportation demand studies, a strong correlation was found between population and employment variables (coefficient of correlation was 0.686). Strong correlation was also found between the dummy variable for presence of e 50 a s z Figure 1. A distribution of the airports in the sample according to annual aircraft operations

7 A Model to Forecast Aircrafl Operations air traffic control tower and the airport location-specific variable (coefficient of correlation was 0.563). This correlation stems from the need for air traffic control facilities to handle the high volume of aircraft operations at airports located in tourist/recreational areas and/or those providing access to urban areas. Strong correlations between the above variables may lead to some serious multicollinearity problems in estimating the demand function. To overcome the possibility of multicollinearity problems, four derivatives of the model in Equation 1 were estimated using different combinations of variables. Model I included all the variables as shown in Equation 1. The population variable was dropped from specifications of models 11, I11 and IV. The population and airport-location variables were excluded from Model 111, whereas the population and presence of air traffic control variables were excluded from model IV. Models I through IV were transformed to log-linear form and were estimated following the ordinary least square procedure (OLS). The results of estimating the models along with t-statistics are depicted in Table 1. The estimated value of the coefficient of determination (R2) in the models varies from to R2 measures the proportion of the variation in the dependent variable that is explained by the regression equation. The value of R2 is not unreasonable for a cross-sectional study of this type. Because most of the data points in the sample are clustered in a relatively small range, the variation in the systematic component of the dependent variable is small, resulting in relatively low values for R2, Hanushek and Jackson (1977). The significance level of the explanatory variables is high, suggesting that the selected explanatory variables are good ones. For the purpose of illustration, Figure 2 depicts a comparison between actual and estimated airport operations using the results of calibrating model 111. Because 88 percent of airports in the sample had less than 30,000 annual operations, only those airports were shown in Figure 2 for the purpose of graph clarity. Forecasting using the model seems to be more accurate for airports that have less than 10,OOO operations. This is again confirmed by the relatively low R2 since most of the data points are clustered in a relatively small range as shown in Figure 1. From Table 1, one can see that the signs of the estimated coefficients in all models agree with their a priori signs. Due to high correlation between the population and employment variables, they had weak statistical significance. Likewise, due to high correlation between the dummy variable for presence of an air traffic control tower (CT) and the airport location-specific variable (AL), both variables were statistically

8 Table 1. Results of estimating the number of aircraft operations at general aviation airports ~odei r I Model U Model III Model IV ble Coefficient t- Statistics Coefficient t- Statistics Coefficient t- Statistics Coefficient Stat ant y Population (thousands) y Employment (work force) y variable for presence of an air control tower y variable for presence of ics services y variable for presence of r flights services y variable for presence of aft repair services y variable for presence of ft rental services y variable for presence of crop g and other related services y variable for runways longer Ooo fcct ion-specific dummy variable for t/rccrutiodurbaoarcas oud I

9 A Model to Forecast Operations , UJ s al 0" FJ 2 a U CI al 0.- E w" /, 0,/" Actual Aircraft Operations Figure 2. A comparison between estimated and actual aircraft operations insignificant. To overcome the problem of multicollinearity between these variables, models 111 and IV were estimated by including the dummy variable for the control tower (CT) in model 111, and the airport location-specific variable (AL) in model IV. We will focus the analysis on models 111 and IV. The estimated coefficients in both models appear to agree with their u priori signs. The variables in model 111 are statistically significant at the 0.05 level, except for the dummy variable for crop dusting which is significant at the 0.1 level. The dummy variable for presence of repair services is statistically insignificant in both

10 320 Atef Ghobrial models 111 and IV. The variables in model IV are statistically significant at the 0.05 level except for the rental service variable which is significant at the 0.1 level. From Table 1, one can see that the signs of the estimated coefficients in all models agree with their Q priori signs. Due to high correlation between the population and employment variables, they had weak statistical significance. Likewise, due to high correlation between the dummy variable for presence of an air traffic control tower (CT) and the airport location-specific variable (AL), both variables were statistically insignificant. To overcome the problem of multicollinearity between these variables, models 111 and IV were estimated by including the dummy variable for the control tower (CT) in model 111, and the airport location-specific variable (AL) in model IV. We will focus the analysis on models 111 and IV. The estimated coefficients in both models appear to agree with their Q priori signs. The variables in model 111 are statistically significant at the 0.05 level, except for the dummy variable for crop dusting which is significant at the 0.1 level. The dummy variable for presence of repair services is statistically insignificant in both models 111 and IV. The variables in model IV are statistically significant at the 0.05 level except for the rental service variable which is significant at the 0.1 level. Because Equation 1 was transformed to log-linear form, the coefficient of the employment variable is interpreted as elasticity of demand. Elasticity is a measure of the relative change in demand due to a relative change in a given explanatory variable. The estimated coefficient of the employment variable in models 111 and IV is roughly This may seem contrary to the common belief that demand is elastic with respect to income. Note that the employment variable in our model measures the work force in a given county which is not necessarily an indicator of the level of income. For example, while employment in a tourist/- recreational area could be high, the average income level is generally low relative to other more technical or industrial jobs. In addition, because the dependent variable includes both local and itinerant operations, the employment elasticity may not be a real indicator of how local operations are influenced by changes in the work force. The employment coefficient in our model suggests that the number of aircraft operations at a given airport is inelastic with respect to employment; a ten percent increase in the work force will likely result in 3.2 percent in aviation activities at the airport. Other dummy variables in the model are specified in exponential forms; meaning that one can compare airport operations with or without

11 A Model to Forecast Aircraft Operatio ns the existence of individual variables. Model 111 shows that the estimated coefficient for presence of an air traffic control tower to be 1.26; meaning that, other things being equal, the number of aircraft operations will increase by 253 percent at airports equipped with a control tower. This increase is significant and reflects the importance of air traffic control in handling certain types of aircraft, operations during bad weather, and ground control at busy general aviation airports such as DeKalb Peachtree airport in DeKalb county. Table 1 shows that the presence of avionics services at airports will likely translate into 119 (model 111) to 203 (model IV) percent increase in aircraft operations at these airports, other things being equal. This is also a significant increase and reflects the attractiveness of these services to the aviation community. The dummy variable for presence of charter flights indicates that, other things being equal, airports providing that service will likely enjoy an increase in traffic by 103 to 113 percent. Again, this is a significant increase in airport demand that presents potential opportunities for airport managers to increase the volume of traffic at their airports. The coefficient for presence of repair services at airports suggest that, other things being equal, the demand will increase by 24 to 38 percent if such services exist. Likewise, the rental services variable indicates an increase of 36 to 48 percent in traffic for airports providing these services, other things being equal. Presence of crop dusting and other related services results in 37 to 44 percent increase in aircraft operations, other things being equal. Runway length seems to be another significant factor in determining the demand for general aviation airports. Table 2 shows that airports with a runway longer than 4O00 feet are likely to encounter 57 to 61 percent more aircraft operations than airports with shorter runways. This range seems to agree with the findings of Ghobrial and Ramdass (1993) which showed that airports with long runways tend to attract about 52 percent more aircraft operations than other airports, other things are equal. This, of course, reflects the capability of the airport to accommodate relatively larger aircraft and to operate in relatively high temperatures and wind speed. The location of airports in tourist and recreational counties or those providing access to urban areas seems also to be significant. These airports are likely to encounter 114 percent increase in aviation demand, other things being equal. Again, this estimate seems to agree with the findings of Ghobrial and Ramdass (1993) who found that general aviation airports in tourist locations in

12 322 Atef Ghobrial Florida tend to attract 135 percent more traffic than other airports, other things being equal. To illustrate the use of the forecasting model in this study, we give a simple example. Assume that a new industry will be created in a given county which currently has a 3,000 feet grass strip. Assume also that the total employment will increase by 10 percent over a period of five years as a result of the new industry. Given the elasticity of demand with respect to employment (0.32), the number of aircraft operations will likely increase by 3.2 percent in five years. Now assume that the runway will be paved and extended to 4,500 feet. The number of aircraft operations will now increase by about 65 percent assuming that the coefficient of runway length in the model to be 0.46 (the estimated coefficient is in model 111 and in model IV). Summary, Conclusions and Limitations of the Study: This papers presented a model to forecast aviation activities at general aviation airports. Unlike commonly known techniques in forecasting demand for general aviation airports such trend and ratio analyses, this papers incorporates some socioeconomic characteristics and supply variables in the model. One can use the model to forecast activities at airports and to assess the effect of improving levels of service on demand. The results showed that the demand for general aviation is sensitive to the presence of air traffic tower and to runway length. The presence of such aviation-related services as avionics, charter flights and rentals, aircraft repair, and crop dusting would increase aircraft activities at general aviation airports. Despite the meaningful results of estimating the model, it is only a step towards modeling aviation activities at general aviation airports. The estimation results showed high statistical significance of the variables in the models but relatively low coefficient of determination (R2); meaning that the variables included in the model are well specified and their contributions to airport demand are well determined. More variables are, however, needed in order to fully explain the changes in aircraft operations at general aviation airports. Sensitivity of the model specifications and correlations between some variables in the model are also drawbacks to this particular of analysis. A more robust specification of the model together with more explanatory variables would certainly yield more useful results. It is also suggested that, depending upon data availability, separate models be developed for local and itinerant aircraft operations.

13 A Model to Forecast Aircraji Operations References Ashford, N. and Wright, P Airport Engineering, John Wiley & Sons, 3rd edition New York. FAA Aviation Forecasts, Fiscal Years , Washington D.C., Federal Aviation Administration, FAA Aviation Forecast Fiscal Years Washington D.C., Federal Aviation Administration, Ghobrial, A. and Kanafani A Future of Airline Hubbed Networks: Some Policy Implications. Journal of Transportation Engineering. Vol. 121, No. 2: Ghobrial A. and Ramdass A The Demand for Aviation Activities at General Aviation Airports: An Empirical Study. The Journal of Aviation/Aerospace Education and Research, Vol. 3, No. 3: Hanushek E. and Jackson J Statistical Methods for Social Scientists, Academic Press, New York. Horonjeff, R. and McLelvey F Planning and Design of Airports, McGraw Hill. Michigan State Airport System Plan, Technical Report, Michigan Aeronautics Commission, Department of State Highways and Transportation, Lansing, Michigan, 1975.