Addressing Uncertainty about Future Airport Activity Levels in Airport Decision Making (2012)

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.

A Framework and Methodology for Addressing Uncertainty about Future Airport Activity Levels in Airport Decision Making P a r t I I

43 Part II provides guidance on the application of a systems analysis framework and a series of related methodologies for addressing uncertainty in airport decision making. The framework and related methodologies have been devel- oped from the material summarized in Part I and refined through application to a number of case studies (described in Part III). The systems analysis framework and the related methodologies are designed to assist airport decision mak- ers with: • Identifying and characterizing risks (threats or opportuni- ties), including their plausibility and magnitude; • Assessing the impact of these threats and opportunities (i.e., determining what can happen, to whom, and when); and • Developing response strategies to avoid or lessen the impact of threats or to foster the realization of opportunities. The systems analysis framework is designed to be general enough to accommodate a variety of airports and projects and to be scalable in order to match the methodology with the resources and needs of each airport. The framework allows planners to consider a broad range of events and risks and helps them anticipate possible changes that may follow. It is not intended to replace the master planning process or any other planning or decision-making model. Instead, the framework augments the master plan with methodologies that allow airport planners to analyze risk and uncertainty and incorporate relevant mitigation measures into the plan- ning process. 6.1 Overview of the Framework As illustrated in Figure 11, the systems analysis framework is composed of five key steps: 1. Identify and quantify risk and uncertainty. Using a combination of data-based and judgment-based method- ologies, identify and attempt to quantify risks and uncer- tainties facing the airport. The ultimate output from this step is a risk register (detailed in Section 7.4), which sum- marizes the risks and can feed information into other steps of the process. 2. Assess cumulative impacts. This step involves analysis and modeling to assess the impact of the identified risks occur- ring in various combinations and the implications for air- port traffic development. 3. Identify risk response strategies. Based on the output from Steps 1 and 2, identify risk response strategies that will help avoid or mitigate negative risks and exploit or enhance positive risks. 4. Evaluate risk response strategies. Undertake qualita- tive and quantitative evaluation of the risk response strategies identified in Step 3 to demonstrate value for money and effectiveness. This may result in revisions to the risk response strategies. 5. Risk tracking and evaluation. This final step is slightly different from the others because it represents an ongoing process of review and revision. It involves tracking risks and traffic over time and flagging potential issues, taking action prescribed in the risk response strategies if poten- tial risks do materialize, and making revisions to the risk register and risk response strategies. Each of these steps is described in detail in Chapters 7 to 11. 6.2 Tailoring the Framework The guidebook provides different tracks that the method- ology can follow, with each track having different data, analy- sis, and resource requirements and generating output with differing levels of detail and depth. The four tracks suggested are summarized in Table 2. The selection of the track is at the discretion of the user, although it may be useful to base the selection on the size of C h a p t e r 6 Introduction

Figure 11. Systems analysis methodology overview. Table 2. Methodology tracks. Step Track A Track B Track C Track D Mostly Qualitative Some Quantification Quantitative, with LimitedStakeholder Involvement Quantitative, with Peer Review and Structured Elicitation 1. Identify and quantify risk and uncertainty Development of the risk register based largely on the guidebook combined with qualitative analysis, visual aids, and informal elicitation within the airport. Development of the risk register based largely on the guidebook combined with qualitative analysis, visual aids, and formal elicitation (e.g., Delphi) within the airport. Development of the risk register based on quantitative analysis, where possible, combined with formal elicitation (e.g., Delphi) within the airport and with key stakeholders. Development of the risk register based on quantitative analysis, where possible, combined with formal elicitation (e.g., Delphi and structured workshops) with airport management/planners, subject matter experts, and a wide range of stakeholder groups. 2. Assess cumulative impacts Based on basic scenario analysis and qualitative approaches. Based on basic scenario analysis and other simple modeling approaches. Use of more advanced modeling procedures such as Monte Carlo simulation. Use of more advanced modeling procedures such as structure and logic diagrams and Monte Carlo simulation. 3. Identify risk response strategies Based largely on the information provided in the guidebook with informal elicitation within the airport. Based on the guidebook and research on examples and best practice at other airports with informal elicitation within the airport. Based on research of examples and best practice at other airports and informal elicitation within the airport and with key stakeholders. Based on research of examples and best practice at other airports and formal elicitation within the airport and with stakeholders. 4. Evaluate risk response strategies Largely qualitative and basic quantitative assessment. Largely qualitative and basic quantitative assessment. Quantitative analysis such as expected net present value. Quantitative analysis such as expected net present value. 5. Risk tracking and evaluation Tracking of traffic against forecasts and trigger points and annual review of risk register. Tracking of traffic against forecasts and trigger points and annual review of risk register. The risk register is updated continuously (possibly using a database system) whenever new pieces of information come in. Full periodic reviews of the risk register. Major risks may be assigned to specific airport staff (risk managers) for tracking and updates. The risk register is updated continuously (possibly using a database system) whenever new pieces of information come in. Full periodic reviews of the risk register.

45 airport and size of planning project being contemplated, as illustrated in Figure 12. Figure 12 is for guidance only—issues around the budget and time available for such analysis will also be important. It may be that a large airport only has time for a quali- tative (Track A) approach; this can still provide insight for the decision maker. Equally, a small airport may find value in pursuing a more quantitative approach (Track C or D). Similarly, it is not necessary to stick to one track through- out the process. For example, an airport could undertake Track D (highly quantitative) for Steps 1 to 4 but select a more modest approach to the risk tracking and evaluation (Step 5) if it is unable to commit significant ongoing resources to this component. Figure 12. Selection of the system analysis methodology track (A to D).

46 7.1 Categories of Airport Activity Risk and Uncertainty Airport risks and uncertainty include both threats and oppor- tunities and can be grouped within the following categories: • Macroeconomic: Events in the general economy that can have implications for air traf- fic, such as a national reces- sion, demographic changes (e.g., aging population), or more localized events such as the loss of a major employer. • Market: Events affecting the supply of, and/or demand for, aviation services in the airport catchment area. For example, entry of a new car- rier, loss of an incumbent carrier, airline mergers, and emergence of a new airport in the region. • Regulatory/policy: Changes in regulations and rules governing the activities of airlines and/or airports. This can also include new environmental regulations on noise or emissions or the introduction of cap-and-trade policies. • Technology: Innovations that may influence the supply of and demand for airport services, such as new aircraft mod- els that reduce the cost of air travel and open up opportu- nities for new routes. • Social/cultural: Changes in the attitude of society and business toward the use and value of air travel (e.g., use of Internet technologies to conduct meetings rather than face-to-face meetings requiring air travel). • Shock events: Unpredictable, infrequent events with poten- tially significant impacts (wars, terrorist attacks, geopolitical instability, etc.). • Statistical or model: Forecasts of airport activity are based on analytical models. Such models can be mis-specified (i.e., they do not correctly represent the underlying relationships) or are subject to estimation error. Also, the historical rela- tionships captured in the model may not continue into the future due to structural changes in the market. These categories are related to risk and uncertainty associ- ated with air traffic activity. However, airports face other risks outside of this: obtaining funding (e.g., cuts to state or city budgets), opposition from local communities, and changes to local land use regulations. These risks are not covered in this guidebook, although it is feasible that these risks can also be addressed by the systems analysis methodology. Airport activity uncertainty may further be characterized in terms of the geographic scale and/or reach of their expected impacts, including: • Airport-specific impacts: A single airport would be affected by the event; • Local or regional impacts: A group of airports located in relatively close proximity would be affected (e.g., all five London airports); • State or national impacts: All airports within a state or country would be affected; or • Global impacts. Table 3 provides common examples of these categories of risk and uncertainty. The list is not exhaustive but may pro- vide a starting point for Step 1. 7.2 Approach and Tools for Identifying and Quantifying Risk and Uncertainty In order to identify and quantify the risks and uncertainty facing an airport, the following questions need to be answered for each possible risk factor: C h a p t e r 7 Step 1: Identify and Quantify Risk and Uncertainty

47 Table 3. Categories of airport activity risks and uncertainties, with examples. Increased security requirements for air cargo could result in some cargo shipments being transported by other modes. New/revisions of airport taxes and passenger duties National Ecological departure tax in Germany in 2011. Nighttime restrictions or bans Local Restrictions on night operations can harm an airport’s ability to attract and sustain air services. This is particularly the case for air cargo since shippers require 24-hour operations. Aviation cap and trade or new carbon taxes Global/national Entry of aviation into the EU Emission Trading Scheme (ETS) in 2012. Category Risk/Uncertainty Scale of Impact Comments and Examples Macro- economic Economic shifts/general business cycle Global/national Global recession in 2008/09; U.S. recession in 2001. Fuel price volatility Global Fuel price peak in 2008; Gulf War 1991. International monetary crisis altering trade and exchange rates Global Asia 1997. Closure of a major local business State/local Market Exit/collapse of a major carrier Local TWA at St Louis in 2001; DHL shutting down domestic shipping operations at Wilmington International Airport in 2008. Entry of a new carrier stimulating traffic Local Southwest Airlines at multiple airports; Allegiant Air at Bellingham. Airlines merger Global/national Continental Airlines merging with United Airlines to form the world's largest carrier. Increased competition from surrounding airports for air passengers Regional/local Oakland vs. San Francisco airport. Often associated with an LCC starting operations at a nearby airport. Competition from airports for air cargo Regional/local Shippers have considerable flexibility to change cargo routings and will do so for relatively small cost or efficiency improvements. Increased GA or military activity Local May affect commercial operations. Connecting traffic operations at an airport can be transferred to other airports or simply downsized by the carrier. Local Examples of airport losing significant connecting traffic: Baltimore, Pittsburgh, St. Louis. Changes in aircraft size Local Carrier decides to switch from mainline to regional services (e.g., American Airlines at St. Louis in 2003). Modal competition Regional/local Development of high-speed rail on certain corridors; competition from truck, rail, and marine for certain types of cargo movements. Changes in seating capacity, aircraft utilization, and load factor Global Regulatory/ Policy Liberalization of certain air markets Global/national EU–U.S. Open Skies Agreement. Privatization of airlines National British Airways privatized in 1987. New security requirements Local Various new requirements since 9/11. (continued on next page)

48 • What is the particular risk/uncertainty? • What is the probability or likelihood of that risk/uncertainty occurring? • What will be the impact of the risk/uncertainty factor if it were to occur, both in the short- and long-term? This information will be obtained from brainstorming and elicitation techniques, as well as analysis of historical data and other quantitative methods. A number of iterations of the process may be required in order to obtain all of the relevant information. The steps are as follows: 1. The team leading the risk project develops an initial list of risks and uncertainties. The initial list can be developed using information within this guidebook, as well as from analysis of historical events and the current business and economic environment. 2. Formal and informal elicitation exercises are undertaken with airport management and other stakeholders, using the initial list of risks and uncertainties to develop more infor- mation, including basic estimates of probability and impact. 3. The probability and impact information is refined using quantitative analysis and other evidence (e.g., review of similar events or information from literature reviews) to produce a draft risk register (explained in more detail in Section 7.4). 4. Additional elicitation exercises to review, confirm, or revise the risk register. 5. Finalization of the risk register. 7.2.1 Analysis of Historical Data In some cases, historical data can be used to determine the likelihood and probable impact of recurring events. In this context, recurring events refer to those events that have occurred at least once in the past and may occur again. They may be recurring within the aviation industry but not neces- sarily at a specific airport. This approach, however, has some limitations. First—and this is a critique applicable to most forecasting techniques used in airport planning—the past is not necessarily a good indicator of what will happen in the future. Second, isolating the impact of a specific event on a measure of airport activity may not be straightforward due to data limitations (data may be limited in quantity, quality, or both) or inadequate statis- tical skills. Third, historical data is only marginally relevant to Technology New aircraft design reducing operating costs Global The Boeing 747 significantly reduced the cost of long-haul air travel. Reductions in belly space Global The amount of belly space on some aircraft designs is limited, reducing the scope for the carriage of cargo. Social/ Cultural Using of internet technology Global Firms and individuals may be willing to replace face-to-face meetings with Internet-based approaches (e.g., WebEx video conferencing). Changing attitudes to climate change Global Concerns regarding greenhouse gas emissions from air transport may lead to curtailing of air travel. For example, some companies now publish their carbon footprint; public pressure to reduce their footprint may result in less air travel by employees. Shock Events War Regional/national First Gulf War. Pandemic Global/regional Swine flu; SARS. Terrorist attack, hijackings, and geopolitical instability Global/national 9/11, London liquid explosives 2006, Northwest flight Amsterdam–Detroit 2009. Natural disaster Regional/local Hurricane Katrina 2005; Indian Ocean tsunami 2004; Icelandic volcanic ash 2010. Statistical/ Model Mis-specification or errors Local The model does not correctly represent the underlying relationships. Structural changes Local Future changes in the underlying relationships (e.g., the relationship between economic growth and traffic growth). Category Risk/Uncertainty Scale of Impact Comments and Examples Table 3. (Continued).

49 assess the likelihood of rare events or to quantify threats and opportunities whose frequency and/or impacts vary with no apparent pattern over time. Planners choosing to evaluate the probable impact of future events based on historical data can: • Use evidence and priors published in the literature. • Use an existing airport activity forecasting model, cali- brated specifically for the airport being reviewed, and con- duct sensitivity analysis and/or scenario testing. • Perform statistical analysis of airport activity data, whereby the impact of a past event—or a series of past events—is estimated while accounting for the influence of other fac- tors (i.e., holding everything else constant). An example of this approach is the quantification of the impact that terrorist attacks would have on air traffic in the United States using historical data from the 9/11 attack. The result of that historical event was an immediate reduction in passenger travel in the U.S. domestic and international mar- kets. Total passenger enplanements in the United States during the month of September 2001 decreased by 45.3% compared to the previous month. Additionally, historical data indicates that it took the industry 33 months (until June 2004) to return to the same level of activity as before the attacks (Bureau of Transportation Statistics, 2011). These numbers may be used as a basis for quantifying the probable impacts of similar events (i.e., major terrorist attack) although they would have to be adjusted—using judgment—to reflect changes in the market since the 9/11 attack. Since historical responses to specific events occurred under specific conditions that may not apply today, quantifying risk impacts using historical evidence may lead to misleading con- clusions. As an alternative, a range of probable impacts may be estimated for each event drawing on relevant historical data. In this approach, data on recorded occurrences similar to the one under examination can be analyzed throughout history, focus- ing on the type of impacts that each event had on relevant out- puts, their frequency of occurrence, and any other data that can help differentiate this event from others. Based on this analysis (which can be as statistically sophisticated as the data allows), a range of possible impacts may be created for each event. 7.2.2 Elicitation Techniques Eliciting information from airport management and other stakeholders will be a key element of the risk and uncertainty identification and quantification process. Depending on the track selected (or the resources available), this information may simply be obtained from within the airport organization (e.g., Track A or B). However, the elicitation can be extended to include external subject matter experts (from academia or consulting), colleagues from other airports, airlines and other customers, government officials, and representatives of other stakeholders (e.g., air navigation, community, and business groups), as may be the case with Track C and D. Drawing from a wider group can lead to the identification of a greater number of risks but will also create challenges in terms of managing the process and achieving a degree of consensus. One solution is to conduct specialty workshops focusing on specific risk catego- ries. For example, a separate session on technology risks may be facilitated to identify risks in that category. These discussions can be held in the context of formal workshops and involve a variety of elicitation and group aggregation techniques. Table 4 provides an overview of the methods available to airport planners to elicit probability and/or measures of impact and to summarize (or aggregate) elicited opinions. Technique Description Pros Cons Delphi Refinement of experts’ opinions by providing feedback through a series of surveys, without open interactions. Consensus may be reached relatively quickly. No direct interactions between experts. Statistical groups One-time survey of experts’ opinions, without interactions. Experts cannot influence each other. Consensus may not be reached. Nominal groups Refinement of experts’ opinions by a series of survey-based sessions, with interactions. Consensus may be reached relatively quickly. Discussions may be time- consuming; some experts may be influenced by others. Unstructured interacting groups One-time survey of experts’ opinions, with interactions, possibly in a workshop setting. Consensus may be reached through discussions. Discussions may be time- consuming; some experts may be influenced by others. Table 4. Overview of elicitation and group aggregation techniques.

50 Subjective assessments of preferences, probabilities, or impacts are typically best obtained via choices (e.g., “choose A or B?”) rather than open-ended opinions (e.g., “I like A”). To assist the participants in determining risk probabilities and impacts, visual aids should be used such as the qualitative risk assessment matrix (heat diagram) shown in Figure 13. The participant selects the cell that represents the likelihood/impact combination for a particular risk/uncertainty factor. (This can be done individually or as a group exercise.) Risks marked in the red (or sometimes yellow) areas are defined as “hot” [i.e., they have the potential for significant harm (or benefit)]. Another approach (which can be combined with the heat diagram) is to ask the participants to provide a probability for an event. In doing so, there should be clarity about what that probability represents. Is it the probability that the event will occur at some point in the next 20 years, or is it the prob- ability that the event will occur in any given year? The differ- ence between the two is quite large—the first is equivalent to rolling a die once while the latter is equivalent to rolling a die 20 times. Furthermore, in some cases the probability may vary over time. For example, the probability of new aircraft technology being developed may be very low over the next 5 years, but higher further into the future (10+ years). Thought also needs to be given to what is being affected. Some risks will affect total passenger volumes, while others will affect only specific sectors (e.g., international, connect- ing). In addition, some risks may affect only air cargo, peak hour operations, general aviation, or aircraft movements but not passengers (e.g., fleet changes). Having obtained input on a range of risk factors, the infor- mation gathered can be represented in a simplified form, as illustrated in Figure 14. The summary plot diagram can effec- Li ke lih oo d Very High High Moderate Low ´ Very Low Very Low Low Moderate High Very High Impact on Activity Figure 13. Illustrative example of qualitative risk assessment matrix (heat diagram). Economic recession Fuel price spikes New FAA taxes Terrorist attack Loss/failure of Carrier X Entry of new carrier (e.g., LCC) Pandemic Open Skies Liberalization High Speed Rail Competition Major tourism event Increased security requirements New aircraft technology Economic boom 5% 10% 15% 20% 25% 30% 35% -5 -4 -3 -2 -1 0 1 2 3 4 5 Pr ob ab ilt y Impact Opportunity >< Threat Macroeconomic Market Regulatory/Policy Technology Social/Cultural Key: Shock Event Use of Internet for meetings Figure 14. Illustrative example of a summary plot of identified risks and uncertainties.

51 tively provide feedback to the participants and help identify critical uncertainties (those with high probabilities and/or high impacts). 7.3 Advanced Approaches to Quantifying Probabilities and Impacts Section 7.2 discussed the probability and impacts of risks and uncertainties as fairly straightforward point values or ranges. This may be sufficient for Track A or B analysis or where time and resources are limited. However, enhance- ments can be made to the analysis, which are discussed in the following. 7.3.1 Direct and Indirect Impacts An important consideration in the quantification of impacts is the distinction between direct and indirect impacts. In a direct impact, the occurrence of an event directly affects the activity of the airport being analyzed. Examples of events that create direct impacts are the destruction of airport infrastruc- ture by a hurricane or the de-hubbing/downsizing of an air- line at a specific airport, which directly affects the number of passengers that use that particular facility. On the other hand, an indirect impact is when the occurrence of an event indi- rectly affects the activity at the airport (usually through a well- established transmission mechanism). Examples of indirect impacts include a global economic recession or an increase in jet fuel prices. A recession will reduce employment, consumer confidence, and disposable income, ultimately weakening the demand for air travel. Likewise, increases in jet fuel costs can feed through into higher ticket prices, which dampen demand (or result in air service cutbacks by carriers). It may be neces- sary to undertake additional analysis to understand the impact of certain variables on traffic levels. For example, ACRP Report 48: Impact of Jet Fuel Price Uncertainty on Airport Plan- ning and Development contains parameters on the sensitivity of traffic levels to changes in fuel prices, which can be used to estimate the impact of fuel price increases or decreases (Spitz and Berardino, 2011). 7.3.2 Probability Distributions Rather than expressing the impact of an event (if it occurs) as a single figure such as the percentage or absolute change in traffic, the impact can be characterized by a probability distribution. This distribution represents a range of possible values, along with an estimate of how likely these different outcomes may be. This can be done to address uncertainty about the outcome or to reflect the range of outcomes that have occurred in the past. Determination of the distributions can involve elicitation methods, analysis of historical data, or a combination of the two. A large number of distributions are available to char- acterize potential outcomes. However, for elicitation and review purposes, it is common to consider only a small set of representative distributions, as illustrated in Figure 15. Selection of an appropriate distribution, however, should always be guided by the characteristics of the event being considered. It is recommended that probability distributions be defined with a limited number of data points (e.g., most likely value, 10th percentile, 90th percentile) rather than technical parameters such as the mean or variance. In the context of nominal or interacting groups, the extremes of a probability distribution (e.g., the minimum and maximum impact values) should be elicited first to avoid anchoring on a single, most likely value. In some cases it may be possible to estimate distributions from historical data. For example, data could be collected on quarterly or annual GDP growth rates and a distribution fit- ted that approximates the distribution of data. Distribution fitting can be done with many statistical packages or with spe- cific risk analysis software. (The latter software is described in Section 8.2.2.) An example of distribution fitting is shown in Figure 16. The histogram bars are the observed historical distribution of GDP growth rates, and the black line is the statistical distri- bution fitted to the histogram. 7.3.3 Duration of the Event Different events will have different durations. Some may be short lived, such as a pandemic or hurricane that may be expected to cause traffic to decline for, possibly, 6 months before it recovers to pre-event levels. Other events may be more long lived, such as the 9/11 attacks from which traffic took several years to recover. And there may be some events that result in a structural change in traffic from which there is no full recovery (e.g., the loss of a major carrier). Event duration can also be specified as a probability distri- bution, in the same way as the event impact. 7.3.4 Correlations and Dependencies All risks and uncertainties are not necessary independent of each other. The occurrence of one event may increase or decrease the probability of another event occurring. For example, high fuel prices can increase the likelihood of an economic recession, or the entry of an aggressive LCC can increase the likelihood of an incumbent carrier exiting. Fur- thermore, there can also be dependencies over time, such

52 Density Function Description Uniform: a distribution where all values within a range of potential outcomes have the same probability. For example, a uniform distribution should be used to characterize the impacts of a threat that may lead to a 10% reduction in airport activity, a 20% reduction, or any value in between, with the same probability. The uniform distribution is fully specified with a minimum value and a maximum value. Discrete: a distribution where each potential outcome is represented by a single value and a corresponding probability. A discrete distribution is defined by a list of possible discrete values and corresponding probabilities, where the sum of all probabilities is equal to 1. Normal: a distribution that is often used as a first approximation to describe random variables that tend to cluster symmetrically around a single mean. The normal (or Gaussian) distribution uses the mean (location of the peak) and the variance (the measure of the width of the distribution) as input parameters and can be used to represent risks made up of the sum of a large number of random variables. Generalized triangular: a distribution that uses the median, lower percentile (such as 10%), and upper percentile (such as 90%) as input parameters. Based on these parameters, a triangular distribution is fitted to the data, and the absolute minimum and maximum are calculated as a function of the distribution. This distribution is often used for event risks, where there is equal probability of an input parameter being lower or higher than the median. PERT (program evaluation and review technique ): a special form of the beta distribution. The beta distribution allows for a skew to the data, either upward or downward, and therefore can be used to represent risks where, for example, the upper extreme is further from the median than the lower extreme. The PERT distribution uses the median, minimum (or lower percentile, such as 10%), and maximum (or upper percentile, such as 90%) as input parameters. Derived from material by Palisade Corporation, @RISK for MS Excel. Figure 15. Examples of probability distributions used in the quantification of uncertainty.

53 that the probability of an event depends on whether it has occurred before. More formally, this can be expressed as: • Dependencies across risks: The occurrence of Event k may increase or decrease the probability of Event j occurring. • Dependencies over time: The occurrence of Event k in Year t may increase or reduce the probability of Event k occurring in Year t + s, where s is the interval considered in the assessment of probabilities. Such dependencies can be captured by specifying corre- lation coefficients between variations. However, in practice this can be difficult to communicate in the elicitation process and can greatly increase the complexity of the analysis. 7.4 Developing a Risk Register It is recommended that the information from the identifica- tion and quantification of risks and uncertainties be captured in a risk register. This register forms the basis for much of the work in the subsequent steps and the ongoing tracking of risks. The risk register may include several fields, grouped within two broad categories, as follows: Risk Identification • Risk ID code; • Risk name and brief description; • Risk status: active, dormant, or retired; • Risk category; and • Date the risk was first identified. Risk Evaluation • Probability of occurrence; • Description of the impact; • Metric or metrics being affected (e.g., number of aircraft operations, passengers); • Magnitude of impact, defined as a single value or a prob- ability distribution; • Duration of impacts; and • Recovery—expected extent of recovery. Which fields are included is at the discretion of the user, and additional fields can be added to provide supplemen- tary information. An example of a risk register is provided in Table 5. 0 10 20 30 40 50 60 70 -1.5% -0.5% 0.5% 1.5% 2.5% 3.5% 4.5% 5.5% 6.5% 7.0% Fr eq ue nc y GDP Growth Rate (Mid-Point) Histogram of GDP Growth Data Fitted Distribution Figure 16. Example of distribution fitting (using illustrative data).

Risk Identification Risk Evaluation Risk ID Risk Category Status Threat or Opportunity Probability/ Likelihood Description of Impact Impact On Magnitude of Impacts (on Traffic) Low Medium High ExpectedDuration Expected Recovery E1 Macro- economic Rapid increase in fuel prices 10% Rising fuel prices result in increased operating costs, which may either be passed onto consumers in higher fares, (lowering demand) or result in carriers cutting back services (or a combination of the two). Aircraft ops, passengers Generally short-term Full E2 Macro- economic Economic slowdown/ recession 10% Economic recession leads to declining passenger volumes and service reductions by airlines. Aircraft ops, passengers Short to medium-term Full R1 Regulatory/ policy New security measures for cargo 20% Increased security measures for cargo reduces bellyhold cargo activity. Cargo volumes Long-term Partial M1 Market Loss or failure of major carrier 30% The exit of Airline X due to economic conditions or other factors. Aircraft ops, passengers Long-term Limited S1 Shock event Swine flu pandemic 5% Swine flu outbreak centered in the local area resulting in passengers avoiding the airport. Aircraft ops, passengers Short-term Full … … … … … … … … … … … … … … … … … … … … … … Table 5. Example risk register. (All estimates are for illustrative purposes only.)

55 It may also be beneficial to add fields to incorporate infor- mation on the risk response strategies in Step 3 and the risk tracking and evaluation in Step 5. To facilitate this, the risk register can be developed as a spreadsheet or database system, which would offer ease of updating and tracking. An example of such a database is provided in Figure 17. This software Source: HDR Inc., Risk Management System for the SR 520 Bridge Replacement and HOV Program, on behalf of Washington State Department of Transportation (2011). Figure 17. Screen shot from an example of a risk management and tracking database. encompasses all aspects of the risk management process, including heat diagrams and the risk register, into an easy- to-use database that can be controlled by the management team. Clearly, this requires a greater investment of time and resources but may be suitable for Track C or D applications of the methodology.

56 Step 2 involves integrating the risks identified in Step 1 into a structural model of uncertainty. The purpose of this model is to evaluate the com- bined effect of multiple risks on airport activity and help define and assess alternative courses of action (response strategies). The first undertaking con- sists of developing an analysis of the risks identified in the previous step, paying atten- tion to the way the relation- ships between events, variables, and outcomes will be modeled, as well as to the transmission mechanisms between them. The goal is to create a model that captures—with as much preci- sion as possible—the impacts uncertain events will have on relevant indicators of airport activity. Once a model for quantifying the impacts of uncertainty is in place, the next activity consists of quantifying cumulative impacts of uncertain events on airport activity. To do this, it is necessary to define the different risk scenarios that will be analyzed as well as the characteristics of each one. Tools such as scenario analysis and Monte Carlo simulation are com- monly used at this stage. Finally, an effective assessment of cumulative risk impacts requires that the outcome of this process be expressed in terms that allow airport planners an easy identification of risk response strategies. 8.1 Developing a Model The term “model” is interpreted fairly broadly in this guidebook—it can range from a simple trend model based on assumed growth rates to a complex multivariate model of the airport. It is anticipated that most airports will fall into one of two camps: 1. Airport planners have access to a calibrated activity forecast- ing model (e.g., multivariate regression model of demand, simulation model), which can be used for uncertainty analy- sis and scenario testing. 2. Airport planners do not have a forecasting model, instead relying on outside forecasts (e.g., the FAA TAFs). In the first case, the existing model can be used as the basis for the assessment of cumulative impacts. For example, the model may contain parameters related to economic activity, which can be used to assess the impact of macroeconomic risk factors. The model has the benefit that it already con- tains information on the transmission mechanisms by which chance events and other sources of uncertainty affect relevant variables and outcomes. Nevertheless, modifications may be necessary to allow for risk factors not addressed in the model. For example, the model may not contain any parameters specifically related to shock events (pandemics, terrorism attacks, etc.). In the case where there is no access to a forecasting model but the airport does have an outside forecast, the cumula- tive impacts can still be assessed by considering the likely deviation from the forecast. For example, the loss of a car- rier may cause traffic to drop below the forecast level, and then some or all of the lost traffic may gradually be recov- ered (as other carriers enter the market). This is illustrated in Figure 18. Whether the airport has access to a forecasting model or not, there are various tools and techniques available to enhance existing models or develop new models to better assess the cumulative impact of uncertainty. These techniques can be used to map out how the uncertainty events may occur, their implications for activity levels, and the interactions between events. These techniques are: C h a p t e r 8 Step 2: Assess Cumulative Impacts

57 • Structure and logic diagrams, • Decision trees, • Influence diagrams, • Program flowcharts, • Stock and flow diagrams (system dynamics), and • Reference class forecasting. All six of these techniques are described in Appendix F. However, the two most relevant and accessible techniques— structure and logic diagrams and reference class forecasting— are described in the following. (Decision trees are also discussed in Step 4 as a tool to evaluate risk response strategies.) 8.1.1 Structure and Logic Diagrams A structure and logic (S&L) diagram is a graphical rep- resentation of a model where each box is a variable (input, intermediate output, output), and links between boxes are operations (add, multiply, divide, etc.). S&L diagrams reflect cause-and-effect relationships among economic, finan- cial, demographic, policy, and political factors. Figure 19 is an example of an S&L diagram for estimating aircraft movements. 8.1.2 Reference Class Forecasting Section 4.2.1 introduced the concept of reference class forecasting. The basic idea is that a forecast is evaluated or even developed by referencing against actual outcomes from similar airports. It is recommended that this approach be incorporated into the forecasting and uncertainty analysis process where practical. This can be done in a fairly unstruc- tured way by comparing forecasts and cumulative event impacts against similar airports or events in the past. For example, in the latter case, the impact of a carrier exiting can be compared against previous examples. Undoubtedly, there will be differences in the circumstances of the airports and various factors that may result in different outcomes to pre- vious events, but this approach can still provide useful guid- ance regarding future traffic development. 8.2 Analyzing the Cumulative Impact of Risks Two general approaches are recommended for this analysis: • Scenario analysis, and • Monte Carlo simulations. The scenario analysis presented here is a less techni- cally demanding approach that is suitable for Tracks A and B, while Monte Carlo is more technically demanding but provides a richer output and may be suitable for Tracks C and D. Again, the approach selected is at the discretion of the user. Time En pl an ed P as se ng er s Original Forecast Traffic impact of carrier exit and partial recovery Figure 18. Analyzing the impact of air carrier exit on forecast traffic.

58 Source: Hickling Corporation (1990, p. 51). Figure 19. Structure and logic diagram for estimating aircraft movements.

59 8.2.1 Scenario Analysis Scenario analysis is a process of analyzing the impact of possible future events by considering alternative outcomes. In this case, the scenarios examine the impact of the occur- rence of a series of uncertain events that have a defined impact on relevant variables and result in a specific outcome. A num- ber of separate scenarios may be developed and played out to assess the impact of different sets of events occurring together. Since scenario analysis consists of skipping forward to the outcome of a series of events, it is important to keep in mind that the outcomes are, strictly speaking, expected outcomes with an implicit probability of occurrence. Selection of the events to be considered in the scenarios can be based on the heat diagrams and summary plots described in Step 1. The events considered in combination will be those flagged in Step 1 as having high probabilities and/or high impacts and acting in the same direction (i.e., to either increase or lower traffic). For example, based on the risk register infor- mation, a scenario may be developed that considers a combi- nation of the occurrence of the following upside events: • Entry of a highly stimulative LCC; • New aircraft technology, which lowers operating costs, leads to trans-Atlantic service; and • A new manufacturing plant opens locally, generating a specified number of passenger trips per year and a certain tonnage of air cargo. Each event can be further specified based on reasonable assumptions and analysis. For example, the traffic impact of the LCC can be estimated assuming a given frequency, air- craft size, and load factor, and then further refined to allow for the new service diverting traffic from existing services (e.g., only two-thirds of the traffic carried by the LCC is incremental). Overlapping impacts of the event can also be addressed. For example, some of the traffic generated by the new manufacturing plant will be carried on the LCC entrant, and so the overall gain has to be netted out. Further, the sce- nario can be specified in terms of developments over time (e.g., the entrance of the LCC is assumed to occur in the first 5 years of the forecast period, while the new aircraft technol- ogy is assumed to not occur for another 10 years). The traffic outcomes (whether in terms of passengers, cargo, operations, or even peak hour passengers) can be generated using the airport’s existing forecasting model or considering deviations from an existing forecast, as described in Section 8.1. Multiple scenarios can be devel- oped to address upside and downside risks and impacts to specific traffic segments. Clearly, this approach has a lot in common with the high/ low forecasts described in Chapter 2, which are commonly applied to air traffic forecasts. However, there are some cru- cial differences or enhancements: • The scenarios are developed from a comprehensive risk register and thus provide a more considered means of eval- uating a wide range of significant risk factors. • The scenarios (or at least some of them) are designed to produce extreme results in order to demonstrate the wide scope of potential outcomes and to test the robustness of the airport system and its plans. • The scenarios are a critical input into the planning process rather than an often-ignored adjunct to the forecasts. The scenario approach provides an accessible means of evaluating the overall risk profile facing an airport, although it has some shortcomings. Most notably, it provides little information on probabilities (the information generated relates largely to outcomes) and has limited ability to address interactions between events and developments over time. An example of the scenario approach is described in the Bellingham International Airport case study in Part III. 8.2.2 Monte Carlo Simulation Monte Carlo simulation was described in Section 4.2.2. In essence, Monte Carlo simulation involves running the fore- cast model multiple times (generally thousands of times), each time with the inputs (and in many cases, the parame- ters) being randomly generated based on the probability dis- tribution assumed for each input (or parameter). Under this approach, each forecast produced is associated with a prob- ability of occurrence based on the individual probabilities of occurrence associated with the variables within the model. The probabilities associated with each outcome allow more quantitative analysis to be undertaken, providing airport planners with a richer set of information. As in the case of scenario analysis, a successful probabilis- tic risk assessment requires a robust structural model and a detailed characterization of risks. The difference, however, lies in the fact that under this approach, every possible com- bination of risks can be modeled and quantified, putting a higher burden on the assumptions made about the interac- tions between variables and their estimated magnitudes. Figure 20 illustrates a simple example of the use of the Monte Carlo simulation approach to analyze the impact of risk and uncertainty on future traffic operations at an airport. The total number of aircraft operations at an airport—identified by F— is modeled as a function of airport taxes, average load factor, and average aircraft capacity. However, it is also assumed that these three variables, along with the price elasticity of demand for air travelers (i.e., the elasticity of demand for that specific airport by passengers with respect to airport taxes),

60 have uncertain behavior with well-defined probability den- sity functions. Through the use of Monte Carlo simulations, several point estimates for the total number of aircraft oper- ations are calculated based on individual draws from each probability distribution function associated with each. At the end of the simulation, all point estimates for aircraft opera- tions obtained through this process can be used to construct a probability distribution function for this output. As Figure 20 illustrates, Monte Carlo simulation can handle both direct and indirect impacts (discussed in Sec- tion 7.3.1), both of which can be modeled with considerable complexity. An example of a direct impact is a major carrier exiting an airport. A number of characteristics of this particu- lar event can be modeled and randomized: • Probability of exit: Specified as a probability of exit in any given year (e.g., 5% probability). With the probability expressed in this way, not only is the occurrence of the event randomized (i.e., whether it occurs), but also the timing of the event. In some iterations the event will occur in the first year of the forecast, in other iterations it will occur at the end of the forecast period, and so forth. • Impact of event: This can be specified as a percentage or absolute decline in traffic immediately after exit of the car- rier. The size of the decline can be specified as a probabil- ity distribution [e.g., triangular distribution with a median value of 25% (loss of traffic), a 10% percentile of 20%, and a 90% percentile of 35% (see Section 7.3.2 for information on probability distributions)]. The values are based on the carrier’s share of traffic, with a range used to reflect uncer- tainty about the carrier’s future size. A more advanced approach would be to include a risk factor reflecting the carrier’s growth at the airport and link the carrier exit vari- able to the growth variable. • Extent of recovery: The proportion of lost traffic that returns due to capacity in-fill by other carriers. This can be a fixed number or probability distribution (e.g., a uniform distribution with range of 50% to 100% recovery of traf- fic). This range could reflect the level of recovery found in previous examples of carrier exit. • Time to recover: The time taken to reach the full extent of recovery. Again this variable can be specified as a prob- ability distribution (e.g., a uniform distribution ranging between 2 and 4 years). Indirect impacts involve variables not directly related to traffic. In Figure 20, the impact of airport taxes is modeled as an indirect impact: the tax increase (which is randomized) Figure 20. An illustration of the use of Monte Carlo simulation techniques to account for multiple sources of uncertainty.

61 output distribution). After all, the model output itself indi- cates a very low probability of such an outcome. However, many of the unexpected events that have occurred at airports described in Chapter 3 would likely have been assessed as low probability before the event occurred. In fact, the case study of Baltimore/Washington International Thurgood Marshall Airport in Part III placed a probability of only 0.5% on the event that did occur (loss of international traffic). Obviously, basing the airport planning entirely on such extreme outcomes is not desirable. However, there may be would result in increased ticket prices and, through a fare elasticity (which can also be randomized), result in a decline in traffic. Monte Carlo simulation can also be used to address con- cerns about statistical or model error. For example, the impact of a changing relationship between traffic growth and GDP growth can be explored by randomizing the GDP parameter within the model. The Monte Carlo method can be very powerful—large numbers of uncertainties can be considered simultaneously, each of which can have different, randomized characteristics. Interactions or correlations between the variables can also be modelled, as well as different timings of events. Given the complexity and the need for repeated random sampling of inputs or variables, Monte Carlo is performed by a computer. There are a number of software products that can be used to conduct Monte Carlo simulation, as discussed in the Monte Carlo Software text box. The output from the Monte Carlo simulations can be pre- sented in a number of ways. Figure 21 shows forecast traffic for a given year. The histogram provides information on the probability density—or probability of occurrence—of each bin (i.e., a small interval for the number of passengers). The S-curve shows the cumulative probability (probability of not exceeding, along the right axis) associated with each bin. For example, the chart indicates that, when all risks and sources of uncertainty are being considered simultaneously, there is an 80% probability that the number of enplanements at the airport will be 2.2 million or less in year 10 (or a 20% prob- ability that traffic will exceed 2.2 million in year 10). So-called tornado diagrams can also be derived from the Monte Carlo output and help identify critical risk factors (i.e., those input variables that contribute most to the dis- persion of forecast traffic), as illustrated in Figure 22. In this example, variation in economic growth is found to cause the traffic forecast to vary from -1.8 million to +3.5 passengers relative to the expected or most likely forecast. Figure 23 shows a time-series plot of the mostly likely or base forecast along with the prediction interval produced from Monte Carlo. The darkest gray range shows the 25th to 75th percentile range—50% of the forecasts produced in the Monte Carlo were within that range. The outer band shows the 5th to 95th percentile range—90% of all forecasts gener- ated in the Monte Carlo simulation were within that range. (In other words, based on the model developed, there is 90% probability that future traffic will lie within this range.) 8.3 Examining Extreme Outcomes With the Monte Carlo analysis in particular, there may be a temptation to ignore or pay little attention to the extreme outcomes produced by the analysis (i.e., the far tails of the Monte Carlo Software It is possible to conduct Monte Carlo simulation using a standard spreadsheet package such as Microsoft Excel. For example, Excel contains a number of statistical functions that can be used to model probability distributions and can gener- ate pseudo-random numbers. (Most computers cannot generate genuinely random numbers but instead produce approximations known as pseudo-random numbers). Visual Basic macros may be required in order to produce large itera- tions of the model and to collect the output data, depending on the complexity of the model. There are a number of software products on the market that offer Monte Carlo functionality in combination with Microsoft Excel, which may be particularly useful for airport forecasting. The user can set up a forecasting model in Excel in the usual manner, and then use the add-on fea- tures of these software packages to specify prob- ability distribution, run multiple iterations, and collect, analyze, and visualize the output data. Some have additional functionality enabling distribution fitting and model optimization. In addition, there are a wide variety of stand- alone packages that can run Monte Carlo simu- lations. Some of these combine Monte Carlo with decision analysis techniques (e.g., deci- sion trees). There are also specialist packages designed for engineering, project planning, or scientific research. An Internet search using search terms such as “Monte Carlo software” and “risk analysis soft- ware” will identify software options currently available on the market.

-3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Average Aircraft Size New Security Measures Population Growth New Carrier Entry Fuel Prices Economic Growth Expected Traffic Change (Millions of Passengers) Figure 22. Illustrative tornado diagram for key risks affecting airport traffic. 0% 20% 40% 60% 80% 100% 0% 2% 4% 6% 8% 10% 12% 14% 1, 00 0 1, 10 0 1, 20 0 1, 30 0 1, 40 0 1, 50 0 1, 60 0 1, 70 0 1, 80 0 1, 90 0 2, 00 0 2, 10 0 2, 20 0 2, 30 0 2, 40 0 2, 50 0 2, 60 0 2, 70 0 2, 80 0 2, 90 0 3, 00 0 3, 10 0 Cu m ul at iv e Pr ob ab ili ty Pr ob ab ili ty Passenger Enplanements (Thousands) Probability Density (Left Hand Scale) Cumulative Probability (Right Hand Scale) Figure 21. Illustrative probability density and cumulative probability output from the Monte Carlo simulations.

63 As such, it is recommended that the output from the analy- sis draw attention to low-probability, high-impact outcomes. For example, the time-series plots from the Monte Carlo simu- lation could contain information on low-probability forecasts, as illustrated in Figure 23, which highlights the boundary of forecasts with a 1% probability using a dotted line. value in examining these extreme outcomes and consider- ing whether the airport plan is robust to such extremes (or can be made robust). While these outcomes are not true black swans (which would lie outside the model), they are indicative of potential gray swans that could be avoided or mitigated. 0 100 200 300 400 500 600 700 Year 0 Year 5 Year 10 Year 15 Year 20 Pa ss en ge r E np la ne m en ts (T ho us an ds ) Original Forecast 25th / 75th Percentile Range 10th / 90th Percentile Range 5th / 95th Percentile Range 1% Boundary Figure 23. Illustrative prediction intervals from the Monte Carlo simulation.

64 Having identified and quan- tified the risks and uncertain- ties in air traffic activity and assessed their cumulative impacts, the next step in the methodology is to identify risk response strategies. The following sections define risk response strategies and set out a number of approaches through which they can be developed. 9.1 Overview of Risk Response Strategies The risk and uncertainties facing airports present both threats and opportunities. As set out in Table 6, there are four broad categories of response to these threats and opportunities. This guidebook focuses on response strategies in the areas of planning, management, and business development. Finan- cial approaches to risk mitigation, such as insurance and other financial instruments, are outside the scope of this guide- book, although there is certainly value to these approaches. Other strategies, such as public–private partnerships (PPPs) or privatization, are generally not available to U.S. airports at this time. As such, most of the strategies discussed for Step 3 fall into the avoid/exploit and mitigate/enhance categories. 9.2 Specific Risk Response Strategies in Airport Planning Based on practices at other airports, research by academ- ics, and the experience of the project team, a number of risk response strategies are put forward in Table 7. Further details on these strategies can be found in Part I and Appendix E. This list is not exhaustive but is designed to provide a starting point for identifying strategies. Airports may find or devise other strategies that are suited to their situation and their risk register. Table 7 identifies the broad risk category that each strategy primarily addresses: • Macroeconomic, • Market, • Regulatory/policy, • Technology, • Social/cultural, and • Shock events. As can be seen, it is often the case that the strategies address a broad range of risks. One key finding derived from this research is that many risk strategies were applicable regard- less of the risk profile or even the circumstances of the airport (e.g., airport size, number of carriers). For example, applying a modular design mitigates a wide range of risks (e.g., eco- nomic development, air carrier exit, changes in technology). In a few cases, there may be specific strategies to address spe- cific risks, but in general there are a number of key strategies that can be applied to a wide set of circumstances. 9.3 Developing Ideas for Risk Response Strategies The general approach to developing a set of response strat- egies corresponding to a predefined risk profile is similar to that of risk identification. Given the nature of the risks being analyzed, there is no stand-alone method or tool that can offer the correct set of strategies. Furthermore, given the diversity of airport activity risks, the set of recommended strategies should be flexible and scalable enough to be imple- mented by airports of different sizes and locations. There are two primary approaches that may aid the risk response identification process: C h a p t e r 9 Step 3: Identify Risk Response Strategies

65 • Evidence based, and • Judgment based. The evidence-based approach relies on reviewing the most current aviation practices and risk-based demand forecasts. By reviewing practices at other airports, planners can under- stand how to develop and implement response strategies. This approach can also be used to assess the pros and cons of various strategies and areas for improvement based on past performance. Chapter 5 provides an overview of the current best practices being used to address risk and uncertainty at airports. (Additional details are available in Appendix E.) Decision makers may also find it useful to seek out additional examples of airports with similar characteristics to their own. The judgment-based approach is based on elicitation from stakeholders and subject matter experts. This can be achieved using Delphi, nominal group, or other elicitation techniques described in Step 1 (see Section 7.2.2). For example, a workshop can be held after the risks and uncertainties have been identi- fied and quantified. The purpose of the workshop is to elicit recommendations and consensus on response strategies that are feasible and likely to align with the airport’s overall stra- tegic plans. Workshop participants can engage in developing response strategies using the same aggregation techniques as those identified for risk quantification. An additional advantage of this approach is that the risk profile can be further analyzed and refined during the workshop. A similar approach is scenario planning (described in Sec- tion 4.3), where participants are presented with various fore- cast outcomes from Step 2 (e.g., very high growth, very low growth, exit of the home carrier) and asked to devise response strategies to address these outcomes. This approach provides a realistic and plausible future scenario (or set of scenarios) upon which the response strategies can be based rather than an abstract list of risk factors. A possible shortcoming, how- ever, is that the scenarios being analyzed (and for which a response strategy has been formulated) may be different from actual future conditions. Therefore, the participants should be encouraged to adopt a real options approach (i.e., selecting risk response strategies that provide the maximum amount of flexibility for the airport). The response strate- gies should avoid committing to long-term courses of action since this creates inflexibilities that are costly to correct in case changes need to be made in the near to mid future. Threats Opportunities Avoid Action is taken to eliminate the impact of a risk. Some threats can be avoided entirely by changing operations or eliminating practices deemed risky. This will often incur a cost. Eliminating risky practices may disappoint stakeholders or degrade the overall business case. Exploit Make a proactive decision to take action and show that an opportunity is realized. Transfer The impact of the risk is transferred to another party, willing and better able to handle the risk (such as an insurance company or investors in a futures market). This typically involves payment of a fee (e.g., outsourcing to a skilled expert) or a premium (e.g., insurance). Share Assign ownership of the opportunity to a third party who is best able to capture the benefit for the operation. Examples include forming risk-sharing partnerships, teams, or joint ventures, which can be established with the express purpose of managing opportunities. Mitigate Action is taken to lessen the expected impact of a risk. Mitigation generally requires positive actions and can have a resource cost. These actions should be considered new practices and controlled like any other airport operations. They may affect the airport operating budget but are often preferable to a do-nothing approach (see discussion on evaluation in the next section). Enhance Take action to increase the probability and/or impact of the opportunity for the benefit of the operation. Seek to facilitate or strengthen the cause of the opportunity and proactively target and reinforce the conditions under which it may occur. Accept No action is taken. After trying to avoid, transfer, or mitigate the threats, the operation will be left with residual risks—threats that cannot be reduced further. In active acceptance, airport management may set up a contingency reserve fund to account for the residual expected value of the remaining risks. A passive form of acceptance simply acknowledges the risk and moves forward with existing practices without reserves, which may seem sensible for risks with small expected values. Accept Take no action when a response may be too costly to be effective or when the risk is uncontrollable and no practical action may be taken to specifically address it. Table 6. General risk response strategies to threats and opportunities.

66 Strategy Risk Types Addressed Comments Land banking: reserving or purchasing land for future development Macroeconomic, market Mitigates or hedges against the upside risk associated with strong traffic growth. The airport authority (or government) has the flexibility to expand and take advantage of high traffic growth, but it is not committed to do so. Reservation of terminal space: similar to land banking, this involves setting aside space within the terminal for future use (e.g., for security processes) Macroeconomic, market, regulatory/policy Mitigates risks from high traffic growth or changes in traffic mix (e.g., from domestic to international or O/D to transfer). Also mitigates risks from changes in government policy or regulation (or the overall security environment). The space can be designed in such a way that it remains productive in the short-term (e.g., using it for retail that can be removed quickly). Trigger points/thresholds: next stage of development goes ahead only if predetermined traffic levels are reached Macroeconomic, market, regulatory/policy, technology, social/cultural, shock events Addresses both upside and downside risks and can be applied to specific traffic categories. For example: - Total passengers, - Domestic or international passengers, - Total aircraft operations, and - Large aircraft operations (triggering a runway extension). Since many construction projects have long lead times (due to planning, construction, etc.), the trigger should be specified to allow for this lag. For example, the trigger to expand the terminal facilities may be when passenger volumes reach 90% of existing capacity, allowing time for the additional facilities to be built before the terminal reaches full capacity. This approach is applicable to not only capital developments. For example, a downside trigger could be determined for certain traffic markets, so if traffic falls below that level, additional air service development work would be undertaken. The trigger does not necessarily have to be traffic based. For example, information from airport marketing may trigger actions or capital improvements to accommodate new air service. Modular or incremental development: building in stages as traffic develops Macroeconomic, market, technology, social/cultural Avoids airports committing to large capacity expansion when it is uncertain whether and how the traffic will develop. At the same time, they can respond to strong growth by adding additional modules. Provides the flexibility to delay or accelerate expansion as traffic develops. Also mitigates risks from traffic mix changes—facilities designed to serve one traffic type can also be designed for incremental development. This option is closely linked to the trigger point concept described previously. Common-use facilities/equipment: such as CUTE, CUSS, common gates, lounges, and terminal space Macroeconomic, market, regulatory/policy, technology Mitigates risks from changes in the mix of traffic and carriers operating at the airport. Also has the benefit of reducing the overall space requirements of the terminal. Table 7. Potential risk response strategies.

67 Strategy Risk Types Addressed Comments Linear terminal design and centralized processing facilities Macroeconomic, market, regulatory/policy, technology, social/cultural, shock events Allows the greatest flexibility for airport expansion since it is the most easily expandable in different directions (especially in combination with modular design). It also allows flexibility in the face of changing traffic mix (e.g., O/D vs. connecting). Swing gates or spaces: can be converted from domestic to international traffic (or between types of international traffic) on a day-to-day basis Macroeconomic, market, technology, shock events Mitigates risks from changes in traffic mix. Allows a very fine level of control since it can allow adjustment to changes in traffic during the day as well as long-term developments. Can also reduce overall space requirements since domestic traffic and international traffic often peak at different times of the day. Non-load-bearing (or glass) walls: as with swing gates, terminal space can be converted from one use to another Macroeconomic, market, regulatory/policy, technology, social/cultural, shock events Avoids the airport being locked in to a narrow traffic development path, allowing greater flexibility to manage changes in traffic mix. Unlike swing gates, this is less short-term in nature (not day-to- day). Also allows a broader flexibility in the overall function of the terminal (e.g., converting space from domestic to international, from retail to security). Use of inexpensive, temporary buildings Macroeconomic, market Allows the airport to service one type of traffic (e.g., LCCs) while keeping options open to serve other types (e.g., full service or transfer). Example: Amsterdam Schiphol’s LCC pier. Self-propelled people movers (e.g., buses) rather than fixed transit systems Macroeconomic, market, technology Mitigates risks from changes in traffic growth and traffic mix. The service is easier to expand, contract, and redirect. Tug-and-cart baggage systems Macroeconomic, market, regulatory/policy, social/cultural, technology Provides much greater flexibility to make changes to the operation of the baggage system than does a fixed system. Mitigates risks from upside and downside traffic growth, traffic mix changes, and checked-baggage trends (e.g., less baggage due to checked-baggage charges). Stakeholder consultation Macroeconomic, market, regulatory/policy, social/cultural Helps ensure that stakeholders understand the airport’s plans and enables the airport to respond to concerns (e.g., an airline concerned that the airport is becoming too crowded). Also allows identification of additional risks (including lack of support from certain stakeholder groups). Air service development Market A diversification/hedging strategy to increase the range of carriers and routes operating at the airport, reducing exposure to particular carriers or markets. Development of non- aeronautical revenues and ancillary activities Macroeconomic, market, technology, social/cultural, shock events Revenue diversification (discussed in Section 5.3) can also be an effective risk mitigation strategy. Airports can engage directly (or partner with third parties) in non-aeronautical activities to diversify their sources of income. By relying less on aircraft operations and passenger enplanements, airports can reduce the systemic revenue uncertainty associated with the air travel industry. However, diversification can expose the airport to greater risks from other sectors of the economy. Table 7. (Continued).

68 9.3.1 Augmenting the Risk Register The risk response strategies can be incorporated in the risk register, thus providing a more complete living docu- ment for the management and tracking of risk and uncer- tainty (which will be beneficial in Step 5). In the case of Track A or B, this could be a matter of adding columns to the risk register, which identifies which risk strategies are expected to address which risk factors. Track C or D may involve a more advanced (and more resource-intensive) approach, such as setting up a database system for track- ing risks and mapping them to risk response strategies, as illustrated in Figure 24. The approach chosen to generate response strategies may depend on the methodology track selected: • Track A: evidence-based approach, drawing on the exam- ples in this guidebook; • Track B: evidence-based approach, using research on additional examples and best practice; • Track C: evidence-based approach, combined with infor- mal elicitation from airport management; and • Track D: evidence-based approach, combined with formal elicitation methods involving airport management and other stakeholders. Field Name Example of Content Risk ID M1 Status Active, Dormant, Retired Risk Type Market Date Identified 01-01-2011 Risk Name Loss of major carrier Description Carrier X removes the majority of its operations from the airport either through financial failure or change in strategy. … … Risk response strategies Linked to the following files: RR1; RR8 Field Name Example of Content Risk Response ID RR1 Risk Strategy Air service development Description Air service development program to attract additional carrier to the airport Current Status Targeting airlines Y and Z … … Field Name Example of Content Risk Response ID RR8 Risk Strategy Modular terminal design Description Modular design to allow halting, slowdown, or acceleration of terminal development Current Status Phase 2 triggered in April 2012 … … Figure 24. Illustrative database design for an augmented risk register (Track C or D).

69 The risk response strategies are designed to reduce the like- lihood or impact of potential threats and capitalize on pos- sible opportunities. Inevita- bly, the choice of a strategy to respond to a particular risk is difficult—in particular, because its effectiveness cannot be fully understood until the risk actu- ally occurs. A probabilistic eval- uation of the economic and/or financial value of risk response strategies can be conducted to assist in the selection. The evaluation serves a number of purposes: identify the highest value risk response strategy, demonstrate robustness over a wide range of outcomes, and determine value for money. The last point is particularly important. In some cases (but not always), the risk response strategy may result in additional costs—it is necessary to determine whether the benefits are likely to outweigh these costs when judging the merits of any particular response strategy. Consider the following simplified example. A small regional airport is within driving distance of a larger airport and is faced with one of its carriers considering consolidating its operations at the larger airport (i.e., exiting the regional airport). Information from the regional airport shows that the loss of the carrier would reduce its annual net revenue by $70 million. The management team of the regional air- port estimates that there is a 30% probability that the airline will move to the larger airport, and is evaluating different response strategies. One of the options consists of an aggres- sive lobbying campaign and incentive program to attempt to keep the carrier at the smaller airport. This strategy would cost approximately $10 million and is expected to reduce the probability of the airline moving to the larger airport by 20 percentage points (i.e., from 30% to 10%). If the regional airport decides not to take any action, its expected net rev- enue loss is $21 million ($70 million × 30%). However, if the airport decides to implement the campaign, its net revenue is expected to decrease by only $17 million ($70 million × 10% minus $10 million for the cost of the campaign). By implementing the campaign, the regional airport reduces its expected loss of net revenue by $4 million. This is a sim- plified example, where costs and benefits are estimated for only 1 year. To be more accurate, expected revenues and costs would have to be calculated for multiple years and expressed in present discounted value terms. Sensitivity tests could also be conducted to determine the extent to which the results are driven by certain key assumptions (e.g., the probability assumed). 10.1 Overview of the Assessment Approach In order to perform a detailed evaluation of the risk responses, it is necessary to perform an appraisal of a response strategy under different circumstances, ranging from the traditional evaluation using the best estimate for the effectiveness of the response strategy to situations where the effectiveness is given extreme values (i.e., conducting a stress test). The evaluation process relies on the com- parison between the degree of usefulness of a risk response strategy—in terms of its ability to alter the probability of occurrence of a risk or its impact on airport activity—and its implementation cost. In more advanced analysis, the ben- efits of implementing a risk reduction strategy can be mon- etized and compared to the monetary cost associated with the implementation of the strategy. In some cases, having conducted the initial analysis in Step 4, it may be necessary to loop back to Step 3 to identify additional risk response strategies (or modify selected strategies). This process is summarized in Figure 25. C h a p t e r 1 0 Step 4: Evaluate Risk Response Strategies

70 The approaches for evaluating the risk response strategies can be broadly categorized as follows: • Largely qualitative: relying on judgment, expert opinion, and some basic quantitative approaches; and • Principally quantitative: using output from Step 2 as a means to conduct analytical assessment. In practice, elements from both categories may be used in the process, although it is likely that Tracks A and B will draw mainly from the first category while Tracks C and D will draw more from the latter category. 10.2 Largely Qualitative Approaches to Evaluation The largely qualitative approach involves assessing the risk response strategy (or strategies) against a number of traffic scenarios and evaluating them based on judgment, historical examples, and simple quantification. This approach is most appropriate in combination with the scenario analysis in Step 2. An example is the approach taken to determine the need for a second airport in Sydney, Australia, described in Sec- tion 5.2. The Australian government faced uncertainty as to whether a second airport would be required for the city. The analysis focused on whether land should be reserved that would allow a second airport to be built. This decision was considered under three different traffic growth scenarios, as summarized in Table 8. This evaluation was based on reasoned judgment and did not require complex analysis. It showed that acquiring a site generally provided the best outcome across the scenarios, and as a result, the govern- ment of Australia did acquire a site for the second airport (de Neufville and Odoni, 2003). This approach can be applied to a single response strategy or can be used to consider a number of strategies in combination. (An example of the latter is provided in the Bellingham International Airport case study in Part III.) An expanded version of this approach uses decision trees. Figure 26 shows the Sydney example as a decision tree. The nodes of the tree represent decision points or event outcomes. The advantage of using decision trees is that they can handle Alternatives AlternativesAlternatives Alternatives Risk Response Strategy Net Present Value > 0 Cost of Implementation ($) Change in Probability of Occurrence (%) Legend Decision Input Alternatives Change in Risk Impact (enplanement) Monetization of Impacts ($ per enplanement) Abandon or Redefine Prioritize & Implement “Best” Strategy No Yes Output Figure 25. Flowchart for the assessment of risk response strategies.

71 more dimensions than the tabular approach (e.g., combi- nations of response strategies). However, a complex system could involve large numbers of decision points and events, resulting in a large and potentially unmanageable decision tree. (Computer software is available to make this process more manageable.) Another example of the use of decision trees is provided in Figure 27. In this example, a subset of the relationship between market conditions and airport decision making is illustrated through a series of expected impacts and potential strategies. In this case, a predefined trigger point for passenger enplane- ment loss due to an airline filing for bankruptcy (greater than 5% traffic loss) establishes the need for a mitigation strategy in flexible airport planning (e.g., explore multifunctional air- port development options). If the situation escalates to the loss of the carrier, the 10% accumulated loss in enplanement triggers another mitigation strategy (e.g., promote airport infrastructure to attract investment and other airlines) that is expected to reduce the net expected loss to 5%. Decision tree analysis can be used to provide more quanti- tative output by applying values and probabilities to each out- come. For example, in the case of Sydney, probabilities could be assigned to each traffic outcome (e.g., low: 20%, medium: 60%, high: 20%), and some measure of value applied to each outcome or end node (e.g., cost, revenues, profit). From that, the expected value (E) of each decision can be assessed as the sum of the probability-weighted values: E Value of Decision k Probability of event i Value of event i i N 1 ∑( ) = × = Where k is one of the decisions available (e.g., either acquire a site or do not acquire) and N is the number of outcomes (three in the case of Sydney). Assessed on this basis, the opti- mal decision (or decisions) would be the one that maximizes project value or minimizes project loss. The use of expected value is discussed further in the next section. 10.3 Principally Quantitative Approaches to Evaluation The Monte Carlo simulations described in Chapter 8 pro- vide a rich source of information to be used in the quantitative evaluation of risk response strategies. Typically, thousands of forecasts are generated by the Monte Carlo simulation, each with a probability attached. This allows for a variety of risk- based analytical procedures, including expected values as described in Section 10.2. To calculate expected value, a mea- sure of value must be selected and calculated that represents Figure 26. Decision tree for second Sydney airport. Alternative Low Traffic Growth Medium Traffic Growth High Traffic Growth Acquire the site OK result. Site not needed, but government owns valuable land it can sell. OK result. Site may or may not be needed. Can wait and see. Good result. Site is needed and available. Do not acquire the site Good result. Site not needed. No money or effort expended. Questionable result. Site may be needed and, with growth of city, is more difficult to acquire. Poor result. Site needed and not available. Source: de Neufville and Odoni, Airport Systems: Planning, Design, and Management, copyright the McGraw-Hill Companies, Inc., 2003 Table 8. Decision analysis for a second airport serving Sydney.

72 the desired outcome of the risk response strategies. If the airport planner is looking for risk response strategies that minimize capital spending, the value could be capital costs. Another measure used is net present value (NPV). NPV is a means of producing a single monetary value for an option based on the future cash flow stream (both incom- ing and outgoing, hence net). Future cash flows are converted to a present value using a discount rate, which reflects the time value of money—money today has a greater value than money in the future. This is not due to inflation (NPV gener- ally uses real values) but rather the opportunity cost associ- ated with the project (money invested in the project could have earned returns elsewhere) and its risk profile (money in the future is less certain). The NPV is calculated using the following formula: NPV C r t t t N = +( )=∑ 10 Where r is defined as the discount rate, N is the number of time periods, and Ct is net cash flow in each period. Calcu- lating the NPV of each option allows for a simple ranking of different options. (Favorable options have a higher NPV compared to less favorable options.) It is possible to calculate an NPV associated with each fore- cast generated by the Monte Carlo simulation, reflecting the revenues and costs (including capital costs) resulting from that traffic outcome. Such information may be generated as part of the planning process, although it may need adjustment to account for the various traffic outcomes. Alternatively, to keep the analysis simpler, NPV can be estimated for blocks of fore- casts rather than each specific outcome (e.g., NPV for 5–10 million enplanements, NPV for 10–15 million enplanements). The expected net present value (ENPV) can then be cal- culated as the sum of the product of the NPVs and their probabilities (de Neufville, Scholtes, and Wang, 2006; de Neufville and Scholtes, 2011). As with NPV, ENPV can be used to rank risk response options to aid decision mak- ing. As well as a single ENPV, plots can be made to assess the performance of the risk response strategies over a range of traffic outcomes, as illustrated in Figure 28. In this example (based on fictional data), two options are being examined for an airport terminal: (1) a single-stage design where capac- ity is built in one stage based on a single, baseline forecast; and (2) a flexible design with staged, modular development and options for expansion. The chart shows the NPV of each design against the cumulative probability from the Monte Carlo analysis. (In this case, the cumulative probability is reflective of forecast traffic levels. The bottom of the y axis equates to low traffic levels, while the top reflects high traffic levels). The chart highlights a number of characteristics of the two designs: • The flexible design reduces the probability and size of negative NPV outcomes [e.g., the bottom (left tail) of the distribution for the flexible design extends as far as -$20 mil- lion compared with -$40 million for the single-stage design]. • At the middle of the distribution (around the 50% prob- ability, where traffic develops in line with the baseline fore- cast), the single-stage design has the higher NPV. • The flexible design allows the airport to take advantage of upside opportunities. The NPVs are considerably higher for the flexible design at the top (right tail) of the distribution. • As a result of the reduced downside NPVs and higher upside NPVs associated with the flexible plan, this plan has a higher ENPV than the single-stage plan. These curves are sometimes referred to as value-at-risk-or- gain (VARG) curves (de Neufville, de Weck, and Lin, 2008), Figure 27. Illustration of a decision tree related to changes in market conditions.

73 reflecting a similar concept to value at risk (VAR), which is used widely in the financial industry. Other measures can be used besides NPV, including: • Internal rate of return (IRR): The discount rate that makes the NPV of all cash flows (both positive and negative) from a particular response strategy equal to zero. • First year rate of return (FYRR): The rate of return observed during the first year of implementation of a response strat- egy. This measure is used in determining the optimal timing Figure 28. Illustrative NPV value curves for airport terminal design options. 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% -40 -30 -20 -10 0 10 20 30 40 50 60 70 Cu m ul at iv e Pr ob ab ili ty Net Present Value ($ Millions) Single Stage Design Flexible Design of the implementation of a response strategy. If the FYRR for a specific strategy is smaller than the discount rate, con- sideration must be given to postponing its implementation for another year (or more). • Cost–benefit analysis (CBA): As with NPV, future benefits and costs are discounted. Unlike NPV, CBA can also con- sider noncash factors (such as noise and emission impacts, local community impacts). However, this approach requires considerable additional data and analysis and is often controversial.

74 The first four steps are part of a single exercise to identify and address the risks and uncertain- ties facing the airport. Step 5, however, is an ongoing process of review, revision, and engage- ment. The goal of the risk track- ing and evaluation in Step 5 is to continually assess the risk environment facing the airport, flag new or changing risks, and take action where necessary. The ultimate aim of the risk tracking and evaluation is to foster a high level of risk awareness and responsiveness within the organi- zation. It is recommended to use the risk register as a basis for tracking and evaluating risks in conjunction with various decision-support tools (described in the following) and trig- ger points to assess alternative courses of action. 11.1 Tools to Assist Tracking and Evaluation Various tools and techniques are recommended to aid in the tracking and evaluation of risk and uncertainty. Their selection will depend on the resources and time available. 11.1.1 Tracking Trigger Points The trigger points are established in Step 3. These iden- tify traffic levels (in terms of total passengers, aircraft opera- tions, international passengers, etc.) or other measures (e.g., inquiries from airlines) that would trigger certain actions or developments. Where a trigger point has been met or exceeded, the first task is to evaluate traffic levels to deter- mine whether this is reasonably permanent and likely to be sustained in the future. For example, some trigger points will have long lead times built into them to allow for the planning and construction necessary to build new or augmented facili- ties. It may be prudent to ascertain whether the outlook for short-term traffic growth necessitates accelerating, slowing, or postponing construction (e.g., due to economic conditions). Any such evaluation would involve discussion with relevant airport management knowledgeable about the cause of this traffic growth (or decline), such as marketing and operations staff. Clearly, the extent of the review will depend on the action being contemplated—a major capital project, such as a runway extension, will involve consider- ably more analysis than, say, expansion of the air service development program. Once there is a reasonable consensus that the trigger point has been met, the action or capital development specified can be initiated. As noted previously, downside (or defen- sive) trigger points can also be established that lead to project slowdowns or pauses. 11.1.2 Establish a Risk Management Team Risk tracking is an active endeavor. It requires monitor- ing the risk profile of the airport through regular updates of the risk register and response strategies. Large airports in particular may choose to establish a risk management team, composed of airport personnel from different departments, whose activities have a direct connection to the identi- fied risk factors (e.g., operations for infrastructure risks, marketing for airline-related risk). Large airports may be interested in this approach not only because they have the resources to establish such a team but also because they have greater system complexity, making it more difficult to identify and track risks. The purpose of such a team is to bring together various parts of the organization—marketing, planning, finance, operations, security, and so forth—with each providing a unique perspective on the risk and uncer- tainties facing the airport. C h a p t e r 1 1 Step 5: Risk Tracking and Evaluation

75 The risk management team may assign a risk owner to par- ticular risks who is responsible for tracking and recording any developments related to these risks and the related risk response strategies. The selection of a risk owner for a specific risk depends on a number of factors: • Impact of the specific risk on the risk owner’s activity: risk owners whose activity is threatened by the risk will pay closer attention to it; • Degree of control for implementing the response strat- egy: risk owners should have a role in any avoid/mitigate/ exploit actions related to the risk they own; and • Internal organizational structure of the airport: risk own- ers should have direct access to upper-management staff to discuss the implementation of response strategies. Each risk owner will be responsible for the following: • Tracking their assigned risk(s) and creating awareness within the organization when that risk changes or manifests; • Helping identify and develop any necessary response strategies; • Maintaining documentation of the risks and any action taken to address these risks, including the effectiveness of the actions; and • Considering the timing of any response, how it may affect other responses and/or risks, and overall measures of air- port activity. 11.1.3 Periodic Updates Periodic (e.g., quarterly) update memos can serve as a com- munication tool to summarize the key risks, mitigation status, and changes to the risk profile. This memo can be developed by the risk management team, if one has been established, or by an assigned member of staff. The memo should incor- porate a brief description of the current performance of the airport, any changes to the risk or traffic outlook, and a sum- mary of what has changed since the previous analysis was completed. The update can be provided in tabular form, as illustrated in Figure 29. 11.1.4 Annual Review Approximately once a year, a review should be undertaken to step back and re-evaluate the risk register and the risk response strategies. The review should consider the following issues: • Have any of the risk factors changed in terms of magni- tude or likelihood? • Are there any additional risk factors that need to be added or any that can be removed? • Based on this review, is there a need to revisit the traffic scenarios or re-evaluate possible traffic outcomes? • Based on the previous bullet points, is there a need to adjust or update any of the airport’s plans? The format of the review is flexible—it could consist of a desktop exercise by the risk management team or it could incorporate a workshop with members of the airport man- agement. Scenario planning or gaming exercises could also be undertaken to test and revise the risk response strategies. The purpose of this annual review is not to rewrite the air- port’s plans every year, but to allow the airport to respond to evolving situations and events and to maintain the focus on risk robustness within the airport. 11.1.5 Benchmarking During the process of risk tracking, information can also be sought from other airports nationally and around the world. Events and activities at other airports may provide indicators of risks that could spread to the decision maker’s own airport. In addition, the responses of these airports may provide infor- mation on which actions to take and which to avoid. Risk Area RiskID Description Expected Impact Status/Comments High Medium Low Null (Retired) Regulation/ policy R1 New airport taxes  No government support for new taxes Macro- economic E2 Rapid increase in fuel prices  Awaiting latest long-term EIA forecast Market M11 New carrier entry  Final negotiations with carrier X for new service Note: EIA = Energy Information Administration. Figure 29. Example update memo.

76 11.2 Updating the Risk Register The risk register provides the foundation for much of the risk tracking and evaluation—it contains informa- tion on the risks facing the airport and can also contain information on the risk response strategies. At the same time, the risk register should be updated as new informa- tion becomes available. It can also be used to track the risk response strategies—their implementation, progress, and degree of success—which can provide information to draw on in the future. As mentioned previously, the risk register will likely have to be developed in a computer data- base, or some other software program, to provide the needed functionality. 11.1.6 Information Collection and Management All of the risk tracking and evaluation tools listed in the pre- vious sections highlight the importance of data collection and management in order to monitor ongoing performance. Air- port decision makers should seek to understand the causes of fluctuations in standard measures of activity such as passenger traffic volumes, aircraft operations, and air cargo volumes. Passenger survey data should also be considered as a means of identifying areas of passenger dissatisfaction (or high satisfac- tion) that may influence demand levels and of detecting changes in passenger behavior and characteristics (e.g., aging passenger profile, which may have implications for airport facilities).