Improving ADA Complementary Paratransit Demand Estimation (2007)

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

Research Stemming from the Regression Analysis The regression analysis results in some surprising findings about how community and service characteristics affect demand for ADA paratransit demand. Each of these findings suggests pos- sible further research as discussed below. In addition, some refinements and extensions of the regression analysis are discussed. Age and Disability The regression analysis found no significant effect from the percentage of people in older age groups or with various types of disability. In the case of disability, the lack of any observ- able effect most likely indicates that the Census measures do not correspond well to ADA paratransit eligibility. In the case of age, the result is somewhat surprising, since older peo- ple typically account for a high percentage of paratransit riders. However, younger people with disabilities are often very frequent riders and generate a disproportionate share of rid- ership. For example, in a planning project for the Whatcom Transportation Authority in Bellingham, Washington, Nelson\Nygaard determined that 60% of riders were age 65 and older, but people under age 65 made 58% of trips. If it is true, as found in the regression, that ridership is proportional to total population rather than population in older age groups, then paratransit ridership may grow far less dramatically than expected. A relationship based on total population is also much easier to use for predicting ridership than one based on older population, since projections of total population are more available than projections by age category. The lack of relationship between age and ADA paratransit ridership could also be taken as an indication that many older people require demand responsive services other than ADA paratransit. In recent work for the National Cooperative Highway Research Program, Bai- ley and others analyzed National Transit Database and Census data and found a strong pos- itive connection between growth in overall demand responsive transit ridership and growth in the 75 to 84 and 85 and older age groups. Their analysis did not distinguish ADA para- transit ridership from other demand responsive ridership such as general public dial-a-ride service.41 Additional regression analysis with a larger sample could help to test these findings. Disaggregate analysis could serve the same function. 54 C H A P T E R 7 Research Agenda 41 ICF Consulting, NCHRP Web-Only Document 86: Estimating the Impacts of the Aging Population on Transit Ridership, National Cooperative Highway Research Program, TRB, January 2006.

Poverty Rate and Incomes The model shows a very strong connection between higher poverty rates and reduced demand for ADA paratransit. This suggests research to explore the following: • How and why does the poverty rate in a community depress ADA paratransit demand? • Is the effect really as strong as suggested by the regression results? • Is demand reduced mainly due to limited incomes of individual travelers or due to commu- nity characteristics related to widespread poverty? In general, demand for goods increases with income. In the case of public transportation, it might be thought that increasing income would go with more availability of other modes and there- fore decreasing demand. Some research has in fact found a negative elasticity of transit ridership with respect to income.42 Two hypotheses for the observed impact of poverty rate (corresponding to lower incomes) on paratransit demand area as follows: • Individual people with disabilities might in fact use paratransit more as their incomes decrease, and the observed effect reflects mainly differences in communities. • Lower incomes might correspond with higher paratransit ridership up to a point, but at the very low income levels associated with poverty status (which is disproportionately common among people with disabilities), total travel demand is so depressed that this overwhelms the effect of mode availability. Both of these effects could be at work. Research that tests these findings and elaborates how they work would be valuable. This research would need to look at the choices of individual con- sumers. (An additional consideration is discussed in the later section about population growth.) Transit Service Availability The regression results suggest that communities with more transit service have more paratransit demand. This contrasts with an expectation that more transit service would lead to lower paratransit demand, since transit would be a more viable alternative than in cities with less transit service. Clearly adding transit service does not increase paratransit usage. A possible explanation for the observed effect was suggested in the model development chapter, namely that the effect is a result of less dependence on private automobiles for travel in cities with more transit service. The observed effect might be explained if transit riders who can no longer ride transit are more likely to use paratransit than drivers who can no longer drive. In other words, if a significant fraction of people are used to travel by public transportation, then they would cre- ate a lot of demand for paratransit when they can no longer use conventional service. However, if nearly everyone is accustomed to drive for all of their trips, and if they drive until they can no longer do so, then they would create much less demand for paratransit, since they are unlikely to consider transit or paratransit as a realistic alternative when they can no longer drive. The effect would be intensified if, on average, people lose the ability to drive later than they lose the ability to ride transit. These speculations suggest fundamental research on travel needs and preferences of people with disabilities and older people. Questions would include the following: • How do people make choices between driving, getting rides, taking transit, and using para- transit in response to becoming disabled or in response to age-related limitations? Research Agenda 55 42 McLeod Jr., M. S.; Flannelly, K. J.; Flannelly, L.; Behnke, R. W., “Multivariate Time-Series Model of Transit Ridership Based on Historical, Aggregate Data: The Past, Present, and Future of Honolulu,” Transportation Research Record 1297, Transportation Research Board, National Research Council, Washington, DC, 1991, pp. 76–84.

• How are these choices influenced by incomes, family situation, and availability of each mode, especially transit and paratransit service? As in the case of research about poverty rate and incomes, this research would need to look at the choices of individual consumers. Cross-Sectional Effects and Changes Within One Paratransit System The model was estimated by comparing ridership across various systems. The model is most useful as an aid for comparing paratransit systems. Effects within one paratransit system (“lon- gitudinal effects”) might be different or take a long time to occur. For example, the analysis pro- duced a cross-sectional price elasticity of −0.77 for paratransit demand. This result suggests that paratransit trip making is much more sensitive to fares than is general transit ridership. In para- transit systems without capacity constraints, this might be expected given the general low income of people with disabilities and the relatively high fares that characterize many paratransit sys- tems. In the first interim report for this project, evidence from the literature was presented that the estimated fare elasticity at individual paratransit systems is between −0.2 to −0.8. The litera- ture review also found some evidence of elasticity over −1.0 when fare levels are high. Another possibility is that the estimated cross-sectional price elasticity of −0.77 corresponds to long-term effects but not necessarily short-term effects. Research on response to transit fares has shown that long-term elasticities are much larger than short-term elasticities. The on-line TDM Encyclopedia of the Victoria Transport Policy Institute quotes the following results for transit fare elasticities from research by the British Transport Research Laboratory:43 Buses: Short-run −0.4 Medium run −0.56 Long run 1.0 Metro rail: Short run −0.3 Long run −0.6 The results of the regression analysis are reasonably consistent with these long-term elasticities. Similar considerations would apply to other factors in the model, especially the on-time window. Further research about differences between cross-sectional and longitudinal effects and between long-term and short-term effects would help practitioners applying model results. Disaggregate analysis might provide some evidence on these questions. A relatively simple analysis would apply cross-sectional analysis to fixed-route transit ridership to see what difference there is in the impact of fares measured this way or as a short-term response to fare changes in one system. Additional Issues about Fares It is possible that differences in cost of living or incomes between service areas would affect responses to fares. People with lower incomes would be expected to see a $2.00 fare as a stronger disincentive to travel than people with higher income. This might explain some of the impact of poverty rate seen in the regression analysis. It could also be true that the response to a given 56 Improving ADA Complementary Paratransit Demand Estimation 43 The Demand for Public Transit: A Practical Guide, Transport Research Laboratory, Report TRL 593, 2004, quoted at http://www.vtpi.org/tdm/tdm11.htm.

percentage change in fare would be different depending on incomes. A preliminary attempt to create a “poverty-adjusted fare” produced no significant change in the estimated fare elasticity. Conceivably using area-wide median income to adjust fares would work better. However, this would be a difficult analysis, and probably beyond the abilities of most paratransit system staff who might want to apply the resulting model. Another issue is whether the response to fares changes is different at relatively low fares com- pared with relatively high fares. The regression model assumes that a given percentage fare change always produces the same percentage demand change. In other words, a change from $1.00 to $1.25 produces the same percentage ridership drop as a change from $2.00 to $2.50. Analysis of the model results indicates that the data fit this assumption reasonably well. (That is, residuals show no pattern when plotted against system base fare.) However, it may be that the sample is too small to detect differences, or that the assumption is true in the long run but not in the short run. Analysis with a larger sample of systems might help answer this question. A related issue is the impact of fare discounts and zone charges. Discounts for passes or tick- ets and extra charges for zones were not included in the analysis because of concerns about reliability of the fare revenue data provided by some representative systems. If better fare data can be obtained, it should improve the analysis since some systems provide substantial dis- counts for passes and others have significant zone charges. It would also allow model users to estimate the impact of passes or zone charges. Some systems may have difficulty providing the necessary information, whether for researchers or even for their own use. Allocating pass sales revenue is often difficult. Also, the National Transit Database does not require that systems separate fare revenue for ADA paratransit from fare revenue for other demand-responsive service. Population Growth An odd feature of the regression model is that it predicts paratransit demand as of about 2005 as a function of population and poverty status in 2000. Since the model indicates that demand increases in proportion to population (as long as other factors are constant), this mismatch should not affect the overall conclusion that any given percentage growth in population should translate to the same percentage growth in demand. It is likely that population growth between 2000 and 2005 has been greater in some of the representative systems than others. If there is a correlation between differential population growth rates and the other factors, then the estimated impacts of other factors would be biased. For example, it is possible that areas with high poverty rates have grown less than areas with low poverty rates. If this is true, then the actual impact of the poverty rate alone may be somewhat less than estimated in the model. However, the impact of these differences over a period of only 5 years would probably account for a small part of the observed effect. If it were possible to deter- mine these differential growth rates, then it is possible that the regression analysis could be slightly more accurate. Analysis with a Larger Sample Experiments with omitting various representative systems from the regression give us confi- dence that the model results are valid and should apply to other systems. However, it would be even better if additional systems could be added to the analysis. Once the model results are pub- lished, it is possible that more systems would volunteer to be included in a similar effort in the future. If some candidate variables can be eliminated as a result of the research to date, then the effort to collect the data could be reduced. Research Agenda 57

Long-Term Trends The chapter about long-term trends noted that several subjects appear as important but hard- to-predict components of future ADA paratransit demands. They include the following: • To what extent will future drivers convert to transit and paratransit riders if they eventually experience disabilities that compromise their abilities to drive? This is closely connected to the issues discussed regarding the regression analysis results concerning availability of transit ser- vice. Further exploration would require large surveys including people who use paratransit and people who do not, despite presence of significant disabilities. • Will advances in automotive technologies significantly extend driving abilities beyond what is possible today? Exploration of this topic might be a reasonable topic for a research project. • How will public transit agencies react to the increasing decentralization of homes and com- mercial activities? Will transit agencies follow these new developments or continue to focus on central city services? • If human service program funding declines, will it decrease or increase ADA paratransit demands? • To what extent will potential advances in medical technologies and treatment protocols influ- ence ADA paratransit demands? The trends discussed in the chapter about long-term trends also point to many issues that go beyond the limited topic of ADA complementary paratransit. As often noted, ADA paratransit is not intended as a comprehensive solution to the transportation needs of people with disabil- ities. Fundamental research into the travel needs and choices of people with disabilities could help to understand needs in a broader framework than is attempted in this project. Disaggregate Analysis A disaggregate analysis of ADA paratransit demand could permit integrating ADA paratran- sit into the demand models used for regional transportation planning. The analysis could also help elucidate fundamental issues about travel behavior by people with disabilities. Chapter 6 concludes that appropriate data to estimate disaggregate models of paratransit demand do not yet exist but could be obtained. The chapter describes two needed types of data: (a) A diary-based survey of at least 500 individuals who are registered to use paratransit, based in one or more metro areas where exemplary paratransit service exists and where other data sources such as coded zonal land use data and road and transit networks are available. (b) A few supplementary questions in a large regional household survey, asking each person about disabilities that prevent use of specific modes, as well as ADA registration, if applica- ble. This does not need to be in the same region as survey (a) above. The first of these data sets could probably be obtained for roughly $50,000 to $100,000 within the scope of a reasearch project. Alternatively it may be cost effective to piggy-back onto one or two regional household travel surveys that are going on anyway. This would be a less expensive way of collecting the data because the questionnaire will already be designed. Even if it is necessary to add questions specific to ADA paratransit, there would be minimal extra cost. The cost per completed survey would be less than for the rest of a regional household sur- vey because the names and numbers of the eligible people are already known, and they will probably be a population that is easier to contact and more willing to participate than most households. 58 Improving ADA Complementary Paratransit Demand Estimation

The second of these data sets, as indicated, can only be obtained at reasonable expense through a regional household travel survey. Note that ADA-eligible people probably compose only about 2% of the general population.44 One member of the study team is currently assisting in the design of the 2006–2007 Chicago area household travel survey, with a large sample size, and it may be possible to include a ques- tion or two about ADA eligibility and certification in that survey. Besides Chicago, two regional surveys that will happen next year are Washington, DC, and New York City/New Jersey. All of these present possibilities for obtaining travel diaries from paratransit riders. Of these surveys, Chicago may present the best opportunity for paratransit research in the short run. Another team member is currently working on paratransit issues in Chicago and pre- viously participated in an FTA compliance review. Paratransit service in Chicago has improved greatly in recent years, although many riders would still question whether it is good enough for use in this research. Also, service is undergoing major restructuring. New York is a second pos- sibility, since it is one of our representative systems. An issue with New York would be whether it is too untypical to be considered transferable to other places. Paratransit service in the Wash- ington, DC, area has been the subject of extensive controversy, which would make it unsuitable for this research. As estimated in Chapter 6, once the necessary data are available, model estimation and subsequent coding of the model application to run in a regional model framework would likely require contracting a modeling consulting firm for a period of 9 to 18 months, for a typical budget in the range of $100,000 to $200,000. A very different style of disaggregate research would use a multi-city or national travel survey such as the National Household Travel Survey (NHTS). With information about each person and average characteristics of paratransit and other modes in each city, it may be possible to esti- mate a model that explains how many trips each person makes by each mode of travel as a func- tion of personal, household, and modal variables. This type of analysis was done in a recent NCHRP project about the future of transit ridership by older people.45 Unfortunately, the 2001 NHTS did not include questions that would indicate ADA paratransit eligibility, although it did record use of transit agency paratransit by respondents. If a future NHTS were to include the appropriate questions, it would provide a basis for useful disaggregate analysis of how people with disabilities choose between ADA paratransit and other modes. Summary of Potential Research The preceding discussion suggests three possible areas for further research: disaggregate mod- eling, additional regression analysis, and cross-section research on transit fares (see Exhibit 7-1). At least the last two could be done as an immediate follow-up for roughly $150,000 to $300,000, if there is interest. For the first it needs to be determined how soon there will be an opportunity to cooperate with a regional household travel survey. An order-of-magnitude estimate for the disaggregate mod- eling project would be $200,000 to $400,000, including data collection, modeling, and project management. These cost estimates would need to be refined following discussion with the panel. Research Agenda 59 44 Thatcher, R H; Gaffney J E; ADA Paratransit Handbook, U.S. DOT., September 1991. 45 ICF Consulting, NCHRP Web-Only Document 86: Estimating the Impacts of the Aging Population on Transit Ridership, (Project 20-65[4]): Contractor’s Final Report, National Cooperative Highway Research Program, Transportation Research Board.

60 Improving ADA Complementary Paratransit Demand Estimation Research Topics Addressed When to Do Disaggregate analysis of travel choices by people with disabilities Travel needs and choices of people with disabilities in general. Effect of incomes, mode availability, and prior travel experience. Likely impacts of future income, driving, and settlement trends on paratransit usage. Depends on opportunities to cooperate with a regional household travel survey or inclusion of necessary questions in a future National Household Travel Survey. Further regression with larger sample, population adjustments, additional measures of income, fare data Cost-of-living and income effects on fares. Impact of discounts and zone charges. Discussion of whether fare elasticity is constant or not. Connections between population growth and other factors. Possible immediate follow-up project or following the 2010 Census. Cross-sectional analysis of transit ridership Connections between cross- section, short-term, and long- run elasticities. Possible immediate follow-up project. Exhibit 7-1. Opportunities for additional research.