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.
1 Research Results The research team examined research on air passenger demand at the airport and system level, including research conducted by academic researchers, government agencies, and industry practitioners. Past studies generally relied on aggregate socioeconomic variables to explain air passenger demand patterns. Air passenger demand analyses conducted for airports rarely reported the details of the models used and the estimated coefficients and estimation statistics, although this may reflect the fact that the analyses were conducted as inputs to larger airport studies, such as master planning projects. The research examined recent trends in the demographic and socioeconomic characteris- tics of U.S. households, especially related to age and household income in order to provide a profile of the characteristics in the population at large. Then, surveys of consumer expen- diture and numerous air passenger and household travel surveys by airports and others were analyzed to illuminate how the characteristics of air passengers and their households were related to their propensity to travel by air. It was found that the propensity to travel by air increased progressively with household income and varied with the age of survey respondents, with the greatest use of air travel by those aged 45 to 54. The earlier analysis of broader socioeconomic trends indicated that these two characteristicsâhousehold income and the age of heads of householdsâare positively correlated for working age households. The distributions of other passenger characteristics, such as gender and travel purpose, also were analyzed and presented in Chapter 3. A case study approach was used to explore how inclusion of disaggregated socioeconomic data might improve the performance of models of air passenger demand. Econometric models of annual airport passenger enplanements were estimated for seven individual air- ports and one multi-airport system. The case study analysis compared the results from models of airport passenger enplanements that included a variable that reflected the changes in the distribution of regional household incomes (along with other aggregate indepen- dent variables commonly used in air passenger demand models) with baseline models that used only the aggregate socioeconomic variables. The eight case studies compared the per- formance of relatively simple models of annual origin and destination (O&D) passenger enplanements. A more detailed analysis was conducted for the Baltimore-Washington regional airport system. This more detailed modeling explored the effectiveness of more complex models of air passenger demand that include disaggregated socioeconomic vari- ables and reported in greater detail in Chapter 4. The case study models were estimated on annual data for the period from 1990 to 2010 and used to generate out-of-sample forecasts for the period 2011 to 2015, using both fore- cast and actual data for the independent variables. These forecast scenarios compared the S U M M A R Y Using Disaggregated Socioeconomic Data in Air Passenger Demand Studies
2 Using Disaggregated Socioeconomic Data in Air Passenger Demand Studies predicted passenger enplanements to the actual enplanements. In general, the inclusion of the household income distribution variable in the models resulted in a small improvement in the accuracy of the predicted enplanements, although for some case study airports neither the baseline nor the alternative simple models provided accurate forecasts. Were there benefits to incorporating the disaggregated socioeconomic data variable that reflected information about regional income distributions into the case study regressions? Based on the analysis of broad socioeconomic trends and of the air passenger and other surveys, it is clear that the frequency with which household members travel by air varies considerably with household characteristics, so it seems reasonable that including disaggre- gated data in air passenger demand models will improve their predictive ability. However, the case study model performance results are mixed, at least for the simple model formulations: ⢠Model goodness of fit: Introducing the disaggregated socioeconomic variable to the model raised the adjusted R-squared consistently but modestly. ⢠Significance of model parameter estimates: In the baseline case study models, which relied only on aggregate socioeconomic variables, the coefficient estimates were usually statis- tically significant to a reasonable degree. Introducing the disaggregated socioeconomic variable, which tended to be highly correlated with one or more of the aggregate socio- economic variables, sometimes resulted in none of the socioeconomic variables having statistically significant coefficient estimates, and sometimes only the estimated coefficient of the aggregate or the disaggregated socioeconomic variable was significant. Successfully incorporating disaggregated socioeconomic variables may require finding ways to overcome strong correlation with other socioeconomic variables. ⢠Out of sample forecasting accuracy: Here there were also mixed results. For some case study airports, the disaggregate model performed better and sometimes the aggregate model performed better. Although the case study analysis of relatively simple econometric models did not indicate any large benefits from incorporating disaggregated socioeconomic variables, the modest cost of doing so suggests that further research and experimentation incorporating these variables is worthwhile. This conclusion is strengthened by the results of the more detailed analysis using more complex models for the Baltimore-Washington airport system, which explored a different way to measure household income distribution. The more detailed models resulted in significantly different implied elasticity of demand with respect to aver- age household income when the household income distribution variable was included in the models. These results have important implications for the use of similar models to pre- pare air passenger forecasts, since average household income and income distribution can change in different ways in the future. Given that the change in model performance represents modest improvements from incorporating disaggregated socioeconomic variables into simple models, the costs of doing so are important to consider. If the demographic and socioeconomic data to be used in air passenger modeling are acquired from a commercial data provider, the regional income and population distributions are often included in the data provided. Because of this, there may be no additional cost to acquire the disaggregated data, and only modest costs to add some additional model runs to those involving more traditional model specifications. In any case, the cost of air passenger demand modeling is trivial compared to the magnitude of the investment decisions based on the resulting forecasts of demand, so the real question is whether including disaggregated socioeconomic data improves the model performance. In addition to examining the use of disaggregated socioeconomic data in the type of econometric modeling commonly used in analyses of air passenger demand, the research
Summary 3 explored new sources of disaggregated data about air passenger activity as well as alternative approaches to incorporating disaggregated socioeconomic data into air passenger demand analysis. These results are presented in in Chapter 5. First, there are new types of disaggregated data that result from the capture of de-identified individual travel and spending patterns through the analysis of cell phone and financial transaction data. These data are of increasing interest to transportation analysts, marketers and others because they provide insights into individual travel behavior and consumer pur- chasing patterns and decisions. This analysis found that while it proved possible to identify specific air trips from a sample of representative financial transaction data and some aspects of those trips, such as the destination and the number of household members making each trip, other aspects, such as the airports used for each trip, proved more difficult or impos- sible to determine. Therefore, although this type of disaggregated data shows some promise for use in air passenger demand studies, further research will be necessary to develop appro- priate techniques to identify more details about each air trip and the individual household characteristics. The research identified different ways of including disaggregated socioeconomic data in conventional econometric models and using the observed differences in propensities to travel by air that can be developed from the analysis of air passenger and other survey data. However, the resources available in the current project did not allow the relative effective- ness of these alternative approaches to be explored in more detail through their application in case study analyses, and this work is left for future research. Research Opportunities The current research project has generated a large amount of relevant information from a detailed analysis of air passenger, household travel, and other surveys, the full analysis of which will require much further work. It has also demonstrated that incorporating these effects in air travel demand models not only can improve the predictive ability of those models, but is essential for them to have good predictive ability, given the underlying changes in the socioeconomic composition of society. However, much work remains to develop robust models of air passenger demand that can be applied in a wide range of settings. This will require a sustained research effort over many years, in such areas as ⢠Air passenger and household travel survey design and implementation to collect data on air traveler socioeconomic characteristics and annual air travel use. ⢠Developing new modeling approaches and model specifications to successfully incorpo- rate disaggregated socioeconomic variables in models of air passenger demand, including applying the modeling approach used for the more detailed analysis of the Baltimore- Washington region to other regions. ⢠Continued analysis of air passenger and household travel survey data to resolve apparent inconsistencies in findings from different surveys and enhance the understanding of how air travel demand varies with household socioeconomic characteristics. These avenues for future research are discussed in Chapter 6.