Airport Aviation Activity Forecasting (2007)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

7 USES OF FORECASTS Airport aviation activity forecasts may be used for many dif- ferent purposes. Typically, the forecasts are not final objec- tives in and of themselves. An essential ingredient to prepar- ing forecasts is to understand the purpose for which they will be used. In economic terms, the forecasts of activity are usu- ally meant to reflect the demand for aviation services. It is the demand for services that drives the forecasts and therefore helps airport planners provide the appropriate supply in terms of the infrastructure needed to meet the demand. Seen in this light, it is important to keep in mind that observed airport avi- ation activity is driven not just by demand, but by the inter- action between demand for and supply of aviation services. Forecasts are essential demand-side tools that planners and decision makers use to make supply-side assessments and judgments regarding: • Long-Term Airport Planning and Capacity Needs – Airside facilities expansion—runways and taxiways, air cargo facilities. – Landside facilities expansion—terminals, concourses, parking, airport access. • Short-Term Operational Planning – Airport personnel requirements. – FAA tower staffing requirements. – Identification of seasonal, daily, or hourly peaking effects. – Identification of aircraft and passenger travel time and delays. • Financial Planning – Bond issues and use of public funds. – Annual budgeting. – Airport Capital Improvement Plans (ACIP)— planning tool for identifying capital needs (required to receive AIP funding). Various measures of aviation activity tie directly or indi- rectly to the revenues and costs associated with operating an airport. For example, aircraft operations lead to landing fee revenues, fuel sales, and fixed-base operator (FBO) sales, and also drive costs associated with airside personnel, hangar facil- ities, etc. Passenger enplanements are tied to the revenues and costs associated with terminal amenities, parking facilities, etc. In addition to the larger goals identified previously, cer- tain forecast methods may be required (or at least be more suitable) for specific needs, such as identifying complex time dependencies in highly trended or seasonal data or forecast- ing high-fidelity estimates of aircraft and passenger travel times and delays after the overall activity levels have already been forecast. In any case, the different objectives identified here are not mutually exclusive. Clearly the financial plan- ning function affects both short-term and long-term opera- tional planning. Prudent budgeting practices for short-term operational needs typically require budgeted expenditures to exceed actual expenditures so that the organization does not run out of money or is unable to hire adequate personnel dur- ing any given budget cycle. To be “conservative,” this means that aviation forecasts that are produced to support short- term budgeting may typically err on the high side to support conservative budgeting requirements. In contrast, long-term forecasts are often used to support bond offerings or expenditures of public funds. Because of the long lead times needed to plan, design, approve, and build physical capacity at airports, it can be very costly if the fore- casts cause overestimates of the magnitude of the expansion needed. In such an environment, the prudent conservative approach is to not overestimate long-term capacity needs. In addition, debt service coverage may require conservative fore- casts to ensure the ability to repay the debt. Therefore, the sup- porting forecasts may tend to err on the low side. On the other hand, long-term forecasts are also used by airports and regional and state agencies primarily as a way to assess potential future aviation infrastructure needs. In this environment, many studies will focus primarily on uncon- strained demand; that is, activity levels that would be pro- jected to occur in the absence of any bottlenecks or capacity constraints. At airports that are currently congested or that may reach their capacity in the foreseeable future, such fore- casts are often used to assess how much additional capacity would be needed to meet demand. This is very different from constrained forecasts that explicitly account for the effect of existing or likely future capacity constraints. In many cases, a single analysis may entail both types of forecasts—the unconstrained forecast is used to assess additional capacity needs, and the constrained forecast is used to assess how much activity may be curtailed if additional capacity is not forthcoming. In all cases, the forecaster must be clear in iden- tifying to what extent existing or future bottlenecks and capacity constraints will be considered as factors affecting projected activity. CHAPTER TWO INFORMATION AND DATA COLLECTION

These distinctions between short-term/long-term and constrained/unconstrained demand can lead to stark differ- ences in the associated activity forecasts that are produced. However, such differences do not necessarily imply that one is correct and another is incorrect (or that one is more correct than another). If one keeps in mind the different purposes for which forecasts are used, the variations and differences between them can be understood and explained to stake- holders and other interested parties. Aside from evaluating point estimates of future activity, it is also important to consider the uncertainty associated with such estimates; a very important but often neglected area of aviation forecasting. A more thorough discussion of evaluating forecast accuracy and uncertainty is provided in chapter four. METRICS Aircraft operations and passenger enplanement counts are the primary elements used to measure activity at commercial air- ports. Although passenger operations and enplanements are obviously highly correlated with each other, there may well be situations where their projected growth rates differ from one another because of changes in average aircraft size (owing to the introduction of different aircraft types) and/or load factors. Growth in aircraft operations is more likely to directly impact the need for airside facilities such as runways, taxiways, and gates, whereas growth in passenger enplanements more directly affects the need for landside facilities such as terminal space, baggage claim areas, and parking. For passenger service, both aircraft operations and enplane- ments are related to other common measures of activity such as available seat-miles (ASMs) and revenue passenger-miles (RPMs). ASMs are a measure of airline capacity and are equal to the number of seats available multiplied by the number of 8 miles flown. For a given number of operations, ASMs can be computed if one knows the average seat size of the aircraft flown and the average stage length (distance) of the routes. RPMs are a measure of airline traffic handled and are equal to the number of seats sold to passengers multiplied by the num- ber of miles flown. For a given number of ASMs, RPMs can be computed if one knows the average load factor (percentage of seats sold) for the operations in question. Figure 1 shows the relationships among the various com- mon aviation metrics that are relevant for commercial passen- ger transportation. There may be little correlation, however, between commercial passenger traffic and other metrics that may be important, such as nonpassenger (cargo) or noncom- mercial traffic (general aviation/military) at a given airport. The count of based aircraft at an airport is another common forecast metric, and is particularly important at general avia- tion airports where it can drive the need for facilities such as hangars, fueling, etc. In addition, determination of potential changes in aircraft types can drive assessments of needed run- way improvements or runway length at smaller airports. Even at small airports, however, it is important to focus on opera- tional activity—there are many examples of general aviation airports with relatively high numbers of operations even though there are relatively few aircraft based at the airport. Cargo activity is an important revenue source at certain air- ports with large cargo processing activities (e.g., Memphis International Airport), as well as at commercial airports with significant international operations. For such airports, the relevant metrics will typically include cargo operations and freight-tonnage, and it may be useful to distinguish between all-cargo flights and cargo that is carried as belly freight on passenger aircraft. Again, it is important for the forecaster to properly structure the forecast to focus on the most relevant activity measures for the purpose at hand. ASMs (available seat miles) RPMs (revenue passenger miles) Operations g A Passenge AS s RP s Operations Load Factor Load Factor Average Aircraft Seats Average Stage Length Avg. Stage Length, Avg. Aircraft Seats verage Stage Length, Average Aircraft Seats Load Factor Passenger Enplanements (revenue passengers boarding an aircraft) FIGURE 1 Relationships among common passenger aviation activity metrics.

9 Appropriately defined peak periods are an important aspect of facility planning and, depending on the situation, the appropriate measure of peak activity may refer to seasonal, monthly, daily, and/or time-of-day demands. For example, seasonal peaks may be important at vacation destinations; air- ports that have substantial international traffic may have sub- stantial variations in daily demand, whereas airports serving as connecting hubs for large carriers will likely be subject to large hourly peaks. Many airport forecasts identify the peak month for activity and then compute activity measures for the average day in that month. These peak month-average day metrics (sometimes also called the “design day”) can be derived by dividing peak month operations or enplanements by the number of days in the month. An upward adjustment can be made if activity is heavier on weekdays than on weekends. Then, given the design day estimates, peak-hour metrics can be computed by applying a distribution by time of day. It is important to rec- ognize that with this sort of approach, design day and peak- hour activity levels actually may be exceeded at certain times. Nevertheless, they may represent reasonable standards for planning future facility needs. AVIATION DATA SOURCES One of the most important requirements in preparing a useful and realistic forecast is to obtain accurate historical data for whatever metrics are to be projected. A common technique is to develop baseline forecasts for passenger demand, and then translate these into aircraft operations by applying estimates of aircraft size and load factors based on current or projected fleets. With this type of approach, it is important to recognize the necessity to obtain historical data not only for the primary metrics of interest (passengers and operations), but also for these ancillary factors. In addition to the data generated by the airport itself, U.S.DOT and FAA collect a large amount of aviation activity information from a number of different sources as described here. Access to some of the data sets is restricted to authorized users; for others, the data can also be obtained from private third parties who provide data cleaning and checking, as well as formatting and reporting options for end-users. The applicability and accuracy of publicly available data can vary considerably from case to case. Airport authorities often supplement such data by fine-tuning some of their own airport-specific data that are kept in-house and by conduct- ing surveys of local airport users. OPSNET (FAA) OPSNET is the official source of historical National Airspace System (NAS) traffic operations and delays. Daily airport data, collected since January 1990, are available for: • Airport operations (takeoffs and landings) at FAA-funded towers, classified by itinerant and local. • Instrument operations—primary, secondary, and over- flights. Instrument operations are those flown under an Instrument Flight Rules (IFR) flight plan or special Visual Flight Rules (VFR) procedures, or an operation where a terminal control facility controls IFR separation between aircraft. Primary operations refer to departures or arrivals at the airport where the approach control facil- ity is located. Secondary operations refer to departures or arrivals at other nearby airports that are secondary to the primary airport. Overflights refer to a terminal IFR flight that originates outside the control facility’s area and passes through without landing. • Approach operations—approaches made to an airport by an aircraft with an IFR flight plan usually because of low visibility owing to severe weather. These data are classi- fied into the four FAA-standard user categories—Air Carrier, Air Taxi, General Aviation, and Military. These data are primarily recorded by tower operators; in some cases, the data collection is automated, in others, manual entries are made in logbooks. As with any manual data entry scheme, there may be data and classification reliability issues, although the data do undergo internal cleaning and cross- checking functions before being released. OPSNET data are summarized in FAA’s Air Traffic Activ- ity Data System (ATADS), which can be accessed directly from the FAA website (www.faa.gov). It is important to note that the tower data included in OPSNET and ATADS only track activity at FAA and FAA-contracted towers (currently approximately 510 facilities nationwide). Although these facilities make up virtually all of the commercial activity in the United States, it should be noted that there are more than 3,300 facilities listed in the National Plan of Integrated Airport Sys- tems that identifies all facilities eligible to receive grants under the AIP. Enhanced Traffic Management System (FAA) The FAA’s Enhanced Traffic Management System (ETMS) data system is designed to track every flight that enters the U.S. en route system. The en route system is made up of Air Route Traffic Control Centers (ARTCCs) that are responsible for controlling aircraft flying under IFR at high altitudes. Each ARTCC is responsible for a defined airspace, and typically accepts traffic from and passes traffic to another ARTCC or to a Terminal Radar Approach Control (TRACON) facility. TRACONs are normally located near large airports and pro- vide departure and approach control services for aircraft at less than 10,000 ft and within approximately 30–50 nautical miles of an airport. The ETMS system collects and stores data for individual flights, and includes information on the date, time, user identity (operator name and flight number or registration N-number), and latitude and longitude of where the flight entered and exited a given ARTCC.

In principle, ETMS data can be assembled to track the date and time of individual flights to and from a given airport. This can be particularly useful in identifying peak operations by time of day or day of week. However, the sheer volume of ETMS data (currently approximately 50,000 flights per day) can make it difficult to handle if one is interested in longer periods of time. ETMS covers only those flights that interact with the en route system. With relatively few exceptions, local flights that fly entirely under VFR (unless flying in controlled airspaces) or that fly only under the guidance of airport towers will never be seen by the en route system and will not be accounted for in the ETMS data; in practice, this includes many local gen- eral aviation flights. Rough estimates of this coverage gap can be deduced by comparing ETMS flight counts with ATADS operation counts at airports (although such an approach does not account for those flights that may enter the en route sys- tem while traveling to or from nontowered airports not cov- ered by ATADS). In principle, each ETMS flight should account for one operation at takeoff and one operation at land- ing. Based on FY2005 data, ETMS coverage for flights tak- ing off or landing at large airports (those having at least a 1% share of U.S. passenger enplanements) is well over 90%. For middle tier airports not reaching the 1% threshold, but having more than 100,000 annual enplanements, ETMS coverage is on the order of 55% to 60%. At low activity airports (fewer than 100,000 enplanements), ETMS covers only about 15% of activity. Enhanced Traffic Management System Counts (FAA) The Enhanced Traffic Management System Counts (ETMSC) data set combines the raw ETMS data with OPSNET opera- tions data to provide flight counts by hour and to track aircraft equipment by city pair. Daily OPSNET operations data are distributed to 15-min intervals based on the distribution of ETMS records. The system also identifies individual aircraft for fractional ownership, so that users can query the number of hours and when a particular aircraft is in use. ETMSC data are updated daily and can be accessed directly from the FAA website. 5010 Forms—Airport Master Record (FAA) Every airport submits an Airport Master Record (Form 5010) to FAA, which includes counts of based aircraft and (for non- towered airports) annual aircraft operations. Internet access to the latest available Form 5010 data can be obtained from com- mercial organizations. It is important to point out that all oper- ation counts submitted on Form 5010 are essentially self- reported by airport managers or sponsors. This is often the only source of historical operations data for nontowered air- ports, and many states have undertaken efforts to improve data collection for such airports. In addition, FAA is currently 10 conducting a validation study for based aircraft by issuing directives to airport managers to count and list their aircraft by tail number. Form 41 Schedule T-100 (U.S.DOT) The T-100 data set contains aggregated monthly statistics on segment activity between airports for U.S. carriers. It includes data on departures (both scheduled and performed), available capacity and seats, number of passengers transported, and tons of freight and mail transported. Similar information for inter- national segments (one point outside the United States) is collected from both U.S. and foreign carriers. In addition to the segment-based data, there is also a “market” version of the T-100, but it provides largely the same information as the “segment” version, because in most cases it does not accu- rately track true origin-destination (O-D) traffic. In particular, connecting passengers who change planes at an airport are treated as two separate passengers traveling first from point A to point B and then from point B to point C, rather than as a single passenger traveling from point A to point C. One of the primary uses of the T-100 data set is to derive average load factors by carrier, airport, and/or city-pair seg- ment. It should be noted that certain smaller commuter carri- ers and nonscheduled carriers are not required to submit T-100 data. Because of the aggregation to monthly totals, T-100 data cannot be used to assess daily or time-of-day peaks in demand. Ten Percent Ticket Sample (U.S.DOT) U.S.DOT collects a 10% sample of ticket coupons sold by major U.S. carriers; it includes the full itinerary (excluding intermediate stops on through flights) and the total dollar amount paid by each passenger. The sample is drawn from all tickets (both paper and e-tickets) issued by major U.S. carriers and includes Internet sales from carriers’ websites or third- party travel websites. The DB1B database developed from the sample is issued quarterly and contains total counts of the number of passengers during the quarter who traveled on a specified itinerary at a specified total fare (including taxes). The data set includes “trip break” indicators to facilitate dis- aggregating each full itinerary into one or more O-D trips; in addition, a “dollar credibility indicator” is provided to help identify tickets that are outside credible limits (based on cents- per-mile criteria). Fare class indicators are also provided (first class vs. business vs. coach, restricted vs. unrestricted); how- ever, these are typically of limited value because of variations in individual carriers’ criteria for assigning fare classes and because of upgrades often provided to passengers that are not recorded on the original ticket. The DB1B data are probably the best source of information on O-D traffic for U.S. city pairs. As with the T-100 data, how- ever, because of the aggregation across time, they cannot be used to assess daily or time-of-day peaks in demand. In addi-

11 tion, certain commuter airlines do not submit ticket data for the sample; for those commuter carriers who have a ticketing rela- tionship with one or more of the major carriers, however, this is often not a major concern because most of the commuter carrier’s passengers will be connecting to one of the major’s mainline flights at a hub airport, and so their ticket (only a por- tion of which is with the commuter carrier) will be captured in the DB1B data set. Finally, the DB1B data do not provide a complete picture of O-D demand in international markets because foreign carriers do not participate in the sample. Published Airline Schedules The Official Airline Guide (OAG) and Innovata are commer- cial entities that assemble worldwide airline schedules for pub- lication. The schedules can be used to obtain information on commercial fleet assignment and scheduled activity by time of day and season. The data are available for up to six months in advance; therefore, it can sometimes be used to discover potential activity changes planned by airlines for the near term. However, the accuracy of future schedules typically dimin- ishes significantly beyond about three months; some airlines do not even submit schedules that far in advance, and many that do provide estimates based only on current or past sea- sonal scheduling. An important issue in using such data is that airlines and their commuter or codeshare partners may separately submit schedules that both contain the same flights offered for ser- vice, which can lead to double-counting. Since early 1998, OAG has required submitting carriers to include a descrip- tive identifier indicating the actual carrier for each flight. For those users who obtain the raw OAG data, it is still their responsibility to do the necessary crosschecks to eliminate double-counting where appropriate. Third-party providers of scheduled airline data typically provide this data cleaning as part of their service. Terminal Area Forecast (FAA) The TAF produced by FAA’s Office of Aviation Policy and Plans is the official forecast of aviation activity for all National Plan of Integrated Airport Systems airports—this currently includes approximately 3,300 facilities. The fore- casts are utilized for budgeting and planning purposes by the FAA, and can be accessed directly from the FAA website. The forecasts include annual projections for 20 years on a government fiscal year basis for enplanements (broken out by air carrier and commuter), itinerant aircraft operations (broken out by air carrier, air taxi/commuter, general avia- tion, and military), local operations (general aviation and military), instrument operations, and based aircraft counts. Historical data are available back to 1976. Airport master records are used as the initial source of information for based aircraft at all airports, and for aircraft operations at nontow- ered or contract-towered airports. OPSNET is the initial information source on operations for FAA-towered airports. The initial data may be supplemented or revised through information provided in airport master plans, state aviation activity surveys, or other supplemental sources. Although the TAF can provide a basis of comparison for airports preparing their own forecasts, it is important to under- stand the limitations and uses of the TAF projections. First, they are primarily unconstrained demand forecasts—in other words, they are prepared without reference to existing or potential future airport capacity constraints. In this regard, their primary purpose is to help FAA project potential staffing workloads, budgeting, and overall NAS plan development. They are also used for establishment criteria purposes (e.g., to identify whether small nontowered airports may need a tower in the future). It is also important to understand that the TAF for large commercial airports are fundamentally based on estimates of O-D demand and regional demographics; connecting traffic is forecast separately for those airports where it represents a significant share of total passenger traffic. Furthermore, for many smaller general aviation airports, the TAF projections are often simple flat-lined trends based on current activity. Finally, because the TAF projections are annual totals, they cannot be used to assess seasonal, daily, or time-of-day peaks in demand. For all of these reasons, users must use caution when evaluating TAF and assessing how useful they may be as a point of reference in specific cases. For other purposes, the FAA’s Airports Office has investi- gated the issue of capacity constraints and issued a study in 2004 on likely future capacity constraints at U.S. airports, Capacity Needs in the National Airspace System (2004). This analysis, often referred to as the FACT study (Future Airport Capacity Task), identified six airports where additional capac- ity was needed as of 2003, 15 additional airports and 7 metro- politan areas needing capacity increases by 2013, and 18 more airports and 8 metropolitan areas needing capacity enhance- ments by 2020. An update to the FACT study is scheduled to be released in 2007. Other FAA Data Sources FAA annually publishes FAA Aerospace Forecasts, a 12-year forecast of national aviation activity. It includes aggregate forecasts of passenger enplanements, RPMs, fleets, and hours flown for large air carriers and regional and commuter carri- ers; cargo revenue-ton-miles and fleets for large air carriers; fleets, hours flown, and pilot counts for general aviation; and operations for FAA and contract towers by user category. FAA also publishes the General Aviation and Air Taxi Activity and Avionics Survey, which includes current and historical data on general aviation and air taxi aircraft counts and hours by usage and aircraft type; some of the data are broken out by region and state.

Although neither of these sources contain any airport- specific data elements, the aggregate measures can be useful to airport planners who are employing market share fore- casting methods where local activity is calculated as a share of some larger aggregate forecast. DATA COLLECTION AND PREPARATION Regardless of which forecasting method is used, there are a number of standard principles that planners should follow in preparing their data for analysis. First, it is usually best to use all relevant historical data; forecasting from a small number of data points is less likely to be successful. In some cases, however, it may be advisable to ignore older data; for exam- ple, if there is some important discontinuity, such as an industry deregulation that makes the early data irrelevant. Armstrong (1985) found that the longer the forecast horizon, the greater the need for more historical data to obtain accu- rate estimates. Second, it is important to clean the historical data by check- ing for data measurement errors, missing data, outliers, and, if necessary, seasonality. As noted by Armstrong (1985), even small measurement errors can cause large forecast uncertainty; he reported a numerical example from Alonso (1968), where a 1% error in measuring current population could directly cause a two-period ahead forecast of the change in popula- tion to have a prediction interval of ±37%. This was not a real-world assessment using actual data, but rather a labora- tory experiment where the true population parameters were assumed to be known. This was done to isolate the effects of the measurement error. One way to guard against input errors is to collect data from more than one source, if possible. For example, one could check U.S.DOT measures of scheduled commercial passenger operations (from T-100) against OAG schedules. Flagging any significant differences would allow the user to manually inves- tigate the cause. Missing data are a fairly common occurrence; with time series data of the sort typically used in aviation forecasting, particular data points for a certain time period may simply be unavailable. In such a case, those observations with complete data would constitute a usable data set, but it might be possi- ble to obtain useful information from the incomplete obser- vations. If one is employing econometric forecasting tech- niques, there are a few statistical adjustments that can be tried to extract such information (Greene 1993). Undetected outliers can also have an important impact on forecast accuracy. Many statistical software programs can automatically check for potential outliers by calculating means and standard deviations, and then flagging those observations that lie outside predetermined limits; graphical displays that show potential outliers can also be useful. 12 Once an outlier has been identified, one must investigate the reason. If there is some unusual identifiable historical event that is likely to have caused the outlier (e.g., the 9-11 terrorist attacks had an enormous impact on aviation activity across the United States), then one can account for this in structuring the specific model that will be used in the fore- casting process. In general, external historical events such as wars, strikes, boycotts, weather or other environmental dis- asters, policy changes, etc., can have significant impacts on many forms of economic activity. If a cause for a particular outlier cannot be found (or if one is doing a time series analysis that does not allow for exter- nal factors in the model specification), it is often advisable to simply exclude the observation from the analysis. Alterna- tively, a number of ad hoc procedures to adjust identified out- liers have been suggested in the literature. One is to replace the outlier with the overall mean or median of the series. For positively or negatively trended data, it may be better to do the replacement with an average of the immediately adjacent observations. Seasonality can also have significant impacts in studies where the time intervals of interest are less than a year. As noted earlier, seasonal peaks may be important at vacation des- tinations, whereas airports that have significant international traffic may have substantial variations in daily demand, and airports serving as connecting hubs for large network carriers may experience large hourly peaks. Nevertheless, in practice, most airport activity analyses develop models and make fore- casts of annual activity and then translate these estimates to peak values based on current observed relationships between annual and peak activity. Such an approach is valid if one can reasonably expect the annual/peak relationship to be stable over time. In cases where it is desirable to explicitly forecast daily, monthly, or quarterly activity, there are a number of ways to make seasonal adjustments. In econometric models, one can use dummy variables to essentially distinguish each relevant time period. Alternatively, one can attempt to “de-seasonalize” the observations themselves. Many software programs are available to do this; perhaps the best known and most com- monly used is the Census Bureau’s X-12 program (see Find- ley et al. 1998), which can be used to weight observations and adjust for seasonality, trend, and outliers. Finally, when using explanatory variables in the analysis, an observation with an unusual value of an explanatory vari- able can have a significant impact on the estimates produced by the statistical estimator. Such “leverage points” may be worthy of further investigation; indeed, any observations that have a strong influence on the statistical estimates should be identified. There are a number of ways to identify and test for influential observations (see Belsey et al. 1980 and White and McDonald 1980), and many statistical software packages can automate the suggested procedures.

13 DATA ISSUES AT NONTOWERED AIRPORTS Estimating flight activity at nontowered airports (or at towers with limited operating hours) can be difficult. TRB is currently conducting a separate synthesis study (ACRP S10-01, Count- ing Aircraft Operations at Small and Non-Towered Airports) on this issue. One method that has been used in the past is to identify a relationship between operations and some other independent factor, such as fuel sale records, based aircraft counts, or activity at a nearby towered airport. This often leads to inaccurate estimates, because the activity relationship with the independent factor may not be stable over time. Another option is to interview FBOs at the airport who may be able to provide accurate information about activity levels. In addition, FAA has published Model for Estimating General Aviation Operations at Non-Towered Airports (2001), a document describing a statistical model to estimate operations at non- towered airports based on data from other towered and non- towered airports. Other more direct methods include visual observation or the use of one or more types of automatic counters. In most cases, it is considered too expensive to collect a true census of information over long periods; an alternative that is typically employed is to use sample counts to estimate activity. If sam- pling is used, it is important to develop a valid sample design to ensure that the sampled operations are representative of activity throughout the year. For example, general aviation air- craft operations typically vary based on weather, day of the week, and season. A common plan is to sample for 14 con- secutive days, four times a year, once in each quarter (see Ford and Shirack 1984 and 1985 for further discussion of sampling techniques and counting instruments.) Visual observation relies on human observers actually being present at the airport to count operations, a potentially very expensive way to collect data. Often it is most feasible to do visual counts only during daylight hours, which can lead to inaccuracies unless it is known that most operations occur only during the day. There are a variety of counting instruments available for survey use. These include pneumatic tubes, inductance loops, and acoustical counters. A pneumatic tube placed on a runway registers a count as an aircraft rolls over it. The counter may not be 100% accurate owing to mechanical error, placement of the tube, recording of nonaircraft movements on the runway, etc. If placed on a taxiway instead of a runway, it will actually record ground movements to and from the runway, and so will not record touch-and-go operations or missed approaches. In addition, it cannot distinguish the type of operation (takeoffs vs. landings, local vs. itinerant). The inductance loop counter is another alternative. Unlike a pneumatic tube, which is portable, an inductance loop is a wire that is installed in the pavement of the runway. Opera- tions are counted as aircraft pass over the loop or fly over it within a few feet of the surface. Similar to pneumatic tubes, inductance loop counters will not record missed approaches, will miss most touch-and-go operations, and cannot distin- guish takeoffs from landings or local vs. itinerant operations. Acoustical counters are probably the most popular form of counting operations at small nontowered airports. These devices essentially use microphones placed at strategic points near the runway to record the sound of departing aircraft. Trained personnel and, more recently, software programs, can listen to the noise signatures and accurately identify depar- tures, single versus multi-engine aircraft, and touch-and-go operations, while correctly ignoring nondeparture sounds. DRIVERS OF AIRPORT AVIATION ACTIVITY Many airport forecasts use econometric methods that utilize explanatory variables—these are measures of factors thought to explain changes in the demand and/or supply of aviation activities. There are many potential factors that may affect both the supply of and demand for aviation activities. Again, it is important to keep in mind the purpose for making avia- tion forecasts; if it is to provide guidance for long-term capacity needs, then the forecast should focus on the demand for services. On the other hand, it may be appropriate in some circumstances to explicitly account for existing or future sup- ply-side constraints that may limit activity. In this case, there may be a wide variety of additional factors that may (or should) affect the forecasting process. The factors affecting aviation activity can be broadly categorized into the follow- ing areas: • Macroeconomic and demographic factors such as the level of and growth in the economy, population, incomes, etc.; • Airline market factors, including fares, flight frequency, and schedules; • Air transport production costs and technology; • Regulatory factors; • Infrastructure constraints and improvements; and • Substitutes for air travel. Air travel is fundamentally a derived demand. In the case of business travel, it represents an input to productive activity; in the case of leisure travel, it is part of the consumption of a broader activity (e.g., taking a vacation or visiting friends or relatives). In both cases, air travel demand derives from the desire or need to be at a certain location for a certain purpose, and perhaps at a certain time. Leisure travelers may have more flexibility in their travel plans and so may be more willing than business travelers to trade off certain attributes of travel (e.g., time spent en route) for others (e.g., lower fares). It is also important to note that some explanatory factors may primarily affect the demand for air travel as measured by enplanements, whereas others may primarily affect aircraft

operations (takeoffs and landings). In this context, it is impor- tant to keep in mind that the demand for air travel is likely to respond to traditional economic variables such as price and income, whereas the number of aircraft operations depends on how carriers choose to serve that demand (by means of sched- ules, fares, and amenities) in the market. Macroeconomic and Demographic Factors For commercial airports that are directly connected to the rest of the commercial air transportation network, demand is likely to depend on broad macroeconomic factors that tie closely to business cycles. Most commercial airport activity forecasts include factors such as real gross domestic product (GDP) and real income, measured at the local or regional level, as primary drivers of demand. Where more specific geographic data are available, corresponding local or regional measures of GDP and income can be used. Some studies rely on estimates of total real GDP and/or income for the relevant region, whereas others use per capita measures combined with estimates of population growth. Other possible demand drivers include employment levels or unemployment rates; measures of con- sumer confidence, which is often seen as a leading indicator of future economic activity; and shares of income accounted for by high-income households. Many of these metrics are highly correlated with each other; therefore, analysts often use only a small number of them when constructing econometric models to project future aviation activities. Regardless of which specific macroeconomic or demo- graphic factors are employed, one potential issue that must be addressed is how to define the appropriate catchment area for the airport in question. When the airport in question is the only one providing commercial service in its geographic area, the catchment area probably coincides well with the metro- politan area for which standard macroeconomic measures are produced by local, regional, and/or national entities. How- ever, in multi-airport regions one must also consider factors that may influence leakage of traffic from one airport to another and how passengers select one airport over another— such considerations are discussed in the following section. Airline Market Factors The price of air travel has an important explanatory impact on demand. As prices decline, traffic will increase, holding all else constant. Real fares (adjusted for inflation) or yields (price per mile) are the conventional measures for prices in the air travel industry. However, it is important to recognize that ticket prices are only one of a number of attributes that passengers may consider when deciding on how much air transportation to consume. The “full price of travel” is a stan- dard concept in air travel demand studies; in essence, the idea is that passengers may also care about schedule convenience, the en route time in traveling to their final airport destination point, connecting time on the ground at intermediate airports, 14 and ground access and egress times to and from the O-D air- ports. In studies of the full price of travel, the time spent in some or all of these activities is measured and valued, and then added to the fare paid by the consumer to arrive at a “full price” for the trip. The full price of travel approach is com- mon in individual choice modeling of air travel demand, whereas more aggregate demand models tend to use the more traditional measures of (money) price and income. Since the 9-11 terrorist attacks, the increased access time required to pass through security lines before departure has essentially led to an increase in the full price of travel. How- ever, aside from this change, for many airport forecasting pur- poses it is often the case that the ancillary factors affecting the full price of travel can be safely ignored, because they are not likely to change significantly over the forecast horizon; the airport is fixed in location and large scheduling changes are not expected. However, in the case of forecasting for a new airport or an airport that faces direct competition from other airports in the same geographic region, passenger demand will be influenced by comparisons of flight frequency and schedule convenience, travel times, and other amenities at each airport. In such cases, comparative measures of the full price of travel between existing airports and the new airport (rather than just fares or yields) may be more relevant for pro- jecting demand from specific catchment areas. There are many other airline market factors that may affect future aviation activity at an airport and that are often consid- ered in well-prepared forecasts. These include: • Low-cost carriers or other new entrants—Assumptions about if and when such carriers may offer or expand ser- vice at an airport can have a significant impact on pro- jected future activity. • Regional jets—The impact of regional jets on the airline industry was quite significant starting in the late 1990s. Initially they were used primarily to connect low- volume markets to carrier hubs that were too far away for service by turboprop aircraft. More recently it has been recognized that although passengers prefer jet ser- vice to prop service, the high costs of small regional jets (per available seat-mile) has limited their usefulness. Many carriers are now seeking to reduce these costs through redeployment and re-bidding of contracts with their regional partners. • Changes in service from competing airports—For those airports that compete regionally with other airports for traffic, any projected changes at those airports (such as increased fare competition, congestion, or service by low-cost carriers) can have important effects on activity at the airport in question. • Industry consolidation—The airline industry has con- solidated substantially since the 1980s when large num- bers of new carriers entered the industry after deregula- tion took hold in the late 1970s. Mergers can have a significant impact on activity at a given airport (e.g., the

15 large decline in operations at Lambert–St. Louis Inter- national Airport following American Airline’s takeover of TWA). • Taxes and fees—There are several excise taxes and fees assessed by the federal government or airport operators that affect fares faced by passengers and operating costs faced by carriers. Currently these include: – Federal taxes to support NAS  Passenger ticket tax—7.5% (applies to domestic travel).  Passenger flight segment tax—$3.40 per enplane- ment (applies to domestic travel; certain rural air- ports are exempt).  International arrival/departure tax—$15.10 per arrival and departure.  Alaska/Hawaii international arrival/departure tax— $7.30 per arrival and departure.  Cargo waybill tax—6.25% (applies to domestic freight).  Commercial jet fuel tax—4.3 cents per gallon.  Non-commercial jet fuel tax—21.8 cents per gallon.  Non-commercial gasoline tax—19.3 cents per gallon. – Federal fees to support Homeland Security  September 11 fee—$2.50 per enplanement (cer- tain small airports exempt).  Aviation Security Infrastructure Fee—carrier- specific fee.  Animal and Plant Health Inspection Service pas- senger fee—$5.00 per international passenger arrival.  Animal and Plant Health Inspection Service aircraft fee—$70.50 per international aircraft arrival.  U.S. Customs and Border Protection user fee— $5.00 per international passenger arrival.  Immigration user fee—$7.00 per international pas- senger arrival. – Local passenger facility charges  Passenger facility charge—up to $4.50 per enplane- ment at eligible U.S. airports. A study by Yamanaka et al. (2006) found that the average effective tax rate on domestic airline passenger travel in the United States is approximately 16%. This does not include taxes on international services or taxes and fees assessed directly on carriers. Air Transport Production Costs and Technology Even in the simplest airline market, production costs directly affect the amount of services that airlines are willing to sup- ply. Two of the most important cost factors in the airline industry are fuel and labor. Many long-range aviation forecast studies consider the impacts of potential changes in fuel prices, although many larger carriers try to limit the impact of large swings in prices by hedging a portion of their fuel pur- chases. Any potential changes in labor costs brought about by scope clause changes and overall cost reductions (either nego- tiated between management and their unions or imposed as a result of the many bankruptcy filings that have characterized the domestic industry) must also be factored in. An increasingly important cost faced by commercial carri- ers is the set of landing and usage fees charged by airports for use of their facilities. For many airports, such fees are set to directly recover the costs of operating the airport, which may include large current expenditures for, say, capacity expan- sion. This can lead to wide variations in the fees charged; for example, the landing fee at Atlanta Hartsfield Airport is $0.46 per 1,000 lb, whereas the fee at New York’s LaGuardia Airport is $6.35 per 1,000 lb. These differences primarily reflect variations in the cost of operation. At Atlanta in 2005, airport operating expenses averaged approximately $2.20 per commercial enplanement, whereas the corresponding rate at LaGuardia was more than $16 per enplanement. Such large differences can affect carrier decisions about where to offer service. However, direct negotiations between airports and air- lines often result in bilateral agreements where an airport may reduce or waive certain usage fees in exchange for service commitments by a carrier. All of these potential factors can have important influences on current or projected airport activ- ity levels. Advances in aircraft technology also can affect airport activity. As the major manufacturers design new equipment with lower net operating costs (through increases in fuel efficiency, increased cargo capacity, extended range, and advances in engine technology), airport forecasts must take into account the introduction of these new aircraft and their potential impacts on both operation and passenger counts. For example, three manufacturers have made significant invest- ments in a new category of aircraft called Very Light Jets (VLJ), and more than 3,000 orders have already been placed. These jets, with a maximum takeoff weight of fewer than 10,000 lb and designed for single-pilot operation, will be able to operate from short runways. Some industry experts project that these aircraft will see widespread use in point-to-point air taxi service. New aircraft can also have impacts on commer- cial passenger demand through improvements in passenger amenities. Commercial airlines have also made advances in the sophis- tication of their yield management programs, allowing them to more efficiently fill available seats on their flights. This is evi- denced by the dramatic rise in domestic system load factors across the industry over the past decade (from 67.4% in 1996 to 77.3% in 2005—see Aerospace Forecasts . . . 1999 and FAA Aerospace Forecasts . . . 2006). This technical ability to fill more seats can have important implications for both oper- ation and enplanement forecasts, although there is a practical upper limit to how much higher average load factors can go in the future.

Regulatory Factors Aircraft operations at some airports are affected by regulatory constraints, including environmental rules regarding noise and emissions, time-of-day restrictions, and, in a few cases, direct quotas on the number of operations allowed. Under Federal Aviation Regulation Part 150, FAA has established specific metrics regarding noise exposure at air- ports. Although FAA does not directly impose specific noise limit levels at airports, capacity expansions that use federal money must follow guidelines regarding changes in noise lev- els and attempt to mitigate increases. In addition, aircraft are classified into one of four noise categories—from Stage 1 (loudest) to Stage 4 (quietest). The Stage 4 noise rule adopted in 2005 requires that all new designs for jet aircraft and large transport aircraft submitted on or after January 1, 2006, meet Stage 4 limits. More recently, concerns about emissions by aircraft on air quality and climate change have been raised that may potentially affect airport operations. Internationally, the International Civil Aviation Organization has promulgated increasingly stringent standards for emissions during take- off and landing. Domestically, local authorities and envi- ronmental groups have responded to regulations under the Clean Air Act seeking to reduce emissions of nitrogen oxide (NOx) during takeoff and landing. NOx emissions during cruise conditions are also a growing concern. According to the NASA website, “proposed research and technology objectives are to reduce NOx emissions by a factor of three within 10 years and by a factor of five within 25 years.” In the past, some airports have attempted to impose noise-related or other restrictions on aircraft operations. Since the passage of the Airport Noise and Capacity Act of 1990, however, such restrictions cannot be imposed without rigorous study and approval from FAA. For many years, FAA-sanctioned limits on operations have existed at four large commercial airports—Ronald Reagan Washington National in Washington, D.C.; LaGuardia International and JFK International in New York; and O’Hare International in Chicago. Since the late 1960s, operations at these airports had been limited by means of the High Density Rule, which established slot controls (the right to take off or land) at each facility. In 2000, the U.S. Congress passed the so-called “Air 21” legislation, which mandated that slot controls be eliminated at Chicago’s O’Hare, at JFK, and at LaGuardia; controls are to remain in effect at Washington’s Reagan National. Since that time, FAA has sought voluntary carrier agreements to limit operations at O’Hare, and is seeking to implement new rules at LaGuardia to prevent airline over- scheduling and large increases in congestion delays at these facilities. 16 Infrastructure Constraints and Improvements Except in the case where one is preparing an unconstrained demand forecast, physical capacity constraints can have impor- tant consequences for forecasting future airport activity and the bottlenecks that can occur in a variety of ways. Runway capacity is often a limiting factor; this can occur owing to the number, length, or orientation of runways and taxiways, and weather, which often plays a central role in determining the number of hourly takeoffs or landings that an airport can accommodate. Gate capacity can be another limiting factor. Although many airports have “common use” facilities, others enter into agreements whereby specific airlines can control the use of particular gates; thus, they can limit access by their competitors, although such agreement may put limits on this practice by imposing certain minimum usage requirements. (In addition, airports receiving AIP funds must be in compli- ance with grant assurances that include requirements regard- ing competition.) At airports where capacity limits come into play, there may be a natural tendency for airlines to collectively overschedule the airport. This is because each extra flight added early or in the middle of the day is likely to impose congestion costs on many flights scheduled to depart or arrive later that day; how- ever, the carrier scheduling the extra flight will only take account of the impact it has on its own later flights. This can lead to a higher number of total flights than are optimal from a social welfare standpoint; a study by Brueckner (2002) sup- ports this argument. Capacity limits may also be reached on the landside as the number of passengers approaches the capacity of terminal or parking facilities. Increasing congestion at terminal curbsides and security checkpoints since the 9-11 terrorist attacks repre- sents new capacity constraints that may affect future activity levels. When activity at an airport begins to approach capacity increasing congestion results, this in turn increases the costs to both airlines and their passengers. The level of congestion and delay tends to increase as the level of operational activity continues to rise. The relationship between aircraft operations and delay is often captured in a “delay curve,” as shown in Figure 2. Typically, such a curve is used to assess the mone- tary value of the delay (to passengers and/or airlines), and these values then can be fed into estimates of the full price of travel for passengers and the cost of production for airlines. FAA’s Advisory Circular on Airport Capacity and Delay (1983) provides guidance on computing airport capacity and delay by means of an Annual Service Volume methodology, which accounts for variations in runway use and configura- tion, aircraft mix, weather conditions, etc. Infrastructure improvements can also have important effects on activity forecasts. With expanded capacity, con- gestion levels may decrease, and this may induce an increase

17 in activity that would otherwise not occur, subsequently resulting in an increase in congestion levels. The effects of induced demand on airport activity and congestion levels are an important part of forecasting that should be accounted for in appropriate situations. Substitutes for Air Travel Traditionally, air travel has been thought to be subject to competition from competing modes only on shorter-haul routes, where travel by automobile, train, or, in some cases, bus may be a practical alternative. Mode choice studies by Morrison and Winston (1985) and others indicated that sig- nificant substitution may take place on shorter-haul routes. However, recent advances in communications technology suggest that teleconferencing may become a viable way to conduct business, which may therefore affect the demand for business travel, regardless of the length of haul. Some recent airport forecasting studies have attempted to account for this by explicitly directly reducing activity enplanements and/or operations projections after the fact. The 9-11 terrorist attacks have also had an impact on air travel substitution, particularly in short-haul markets as travel- ers consider total travel time and the “hassle factors” associated with airport security procedures. A report by the International Air Transport Association (“The Air Transport Industry . . .” 2006) indicates that U.S. passenger enplanements in July 2006 were still 12% below the levels of July 2001. Delay Airport Capacity Operations FIGURE 2 Relationship between airport capacity and delay.