Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 26
26 factors can account for only a small fraction of the total vari- not seem to have high riding levels (55), possibly because they ations that are observed. Other, seemingly unobservable atti- were designed into new roads, that is, built in anticipation of tudinal and possibly historical factors seem to dwarf the effect riding rather than in response to it. of the factors that planners and policymakers can control. Seemingly, the only way around these problems would be The report contains findings for residents of Minneapolis to study the same geographic area over a period of time as facil- and St. Paul. The team is not convinced that these findings are ities change. The relevant comparison is not comparing people applicable to other locations. Twin Cities' residents may dif- living at location A to different people living at location B, but fer from those of other places with respect to lifestyles and rather comparing the people at A with themselves as the provi- preferences for bike use. Even within the Twin Cities area, the sion of facilities changes over time. This would be an expen- same regressions yielded different results depending on the sive prospect using surveys; development of a low cost method geographical scope of the study area, that is, including sub- of counting bikes over a large number of different streets and urbs in the analysis changed the results. The options, avail- bike facilities would be of great value for this purpose. ability, and manner in which bicycle facilities are valued likely differ substantially between populations, especially urban versus suburban (54). A SKETCH PLANNING METHODOLOGY Because the impact of the unobservable variables are large relative to the variables of interest, it is likely that what is being This section outlines a simple sketch planning method observed, both in this model and in others, is the effect of atti- to estimate the number of daily bicyclists in an area using tudinal variables acting on policy variables through spurious readily available data. This method could be used for gen- correlations. What seems to have been observed in this model eral political purposes, justifying expenditures by reference are geographic spikes in bicycling. These spikes happened to to the number of bicyclists and the benefits that they receive be positively correlated with some facility measures and neg- from cycling. It could be used to estimate demand on new atively with others, but that in a causal sense had little or noth- facilities by assuming that it will be some fraction of the ing to do with any of them. While other work of this type in the total amount of riding in the surrounding area. Finally, this literature has typically not had to deal with quite such a wide method could be used indirectly to estimate changes to the range of bicycling levels, it seems probable that these types of amount of cycling resulting from facility improvements, correlations might be driving their results to a large extent, too, assuming that changes will be some (probably small) fraction given the typically low explanatory power of these models. of the existing total. A third issue is that because the level of bicycling is so low, the range of sampling error can be many times larger than the sample mean for any realistic sample size. The effect is that the Discussion regression is trying to predict (measured) variable values that The basic assumption that motivates the model described could be off by a factor of five or more from their true values. here is the idea that a large portion of total bicycling is done A sample of 1,000 people would yield 9 cyclists on a given day at the national average level; the 95% confidence interval for by a small fraction of cyclists who ride frequently, and that this sample ranges from 3 to 15 cyclists. This is a very big dif- many of these frequent riders are bike commuters who will ference in relative terms. Seemingly very high or low levels be observed as such in the census commute to work data. The could easily just be sampling aberrations. Yet these inaccurate assumption is that the basic riding frequency table described measurements could strongly influence the estimated parame- in the previous section will hold more or less across different ter values. Obviously this will be a problem with any model of areas, so that an area with a lot of commuter cyclists will also bicycling behavior, but it seems likely to be compounded by have a lot of total cycling, and an area with few commuters the need in traditional models to incorporate a large number of will have little total riding. In other words, commuting by explanatory variables. bike, while it is a small fraction of the total bicycling in a Finally, there is always the problem that even a positive given area, can still be used as a "leading indicator" of what correlation between riding and facilities could be causation might be happening with other types of cycling. in the other direction--that is, the large number of cyclists The team used three different geographical divisions to creates the political climate to build the facilities in the first study the relationship among commute shares, total daily place. Given limited funds, agencies are unlikely to spend bicycling rates, and total weekly bicycling rates. These divi- them on striping bike lanes unless there is a problem such as sions were metropolitan statistical areas, states, and zones crashes, or bikes and motor vehicle conflicts. Thus, retrofit- of 10,000 to 30,000 people within the Minneapolis-St. Paul ted lanes are probably a response to existing cycling to a area. The results of these analyses are described in more large extent, and thus will naturally be associated with high detail in Appendix A. The primary results are summarized in levels of cycling after they are built. Indeed, the other main the following subsections, along with some key facts from "result" of the Twin Cities model was that on-street bike lanes the demand measurement analysis. were very strongly and positively associated with increased On any given day, roughly 1% of the adults in the United riding. By contrast, lanes in some newer cities in California do States ride a bicycle. Over large geographic areas such as met-
OCR for page 27
27 ropolitan areas or states, this number could range roughly This is a baseline level of demand. Then the findings for the between about 0.3% and 2.5%. Over smaller areas such as spe- higher levels of bicycling for residents in the immediate cific parts of metropolitan areas, the range could go as high as vicinity of a facility are applied to this baseline to estimate 15%. The possible range can be reduced somewhat for a given the rates that would exist after the facility is built. This area by considering the bicycle commute to work share. The process is described in more detail in Chapter 4. best fit for a regression relating percentage of adults who bicy- It should be noted that this process is estimating the daily cle in a given day to the census bike commute to work share average number of adult bicyclists from the area around the varies considerably across different sized geographic units. facility, based on observed relationships from around the However, based on this analysis, it appears that bounds can United States. It is not estimating the number of people who be placed on the range of values that are likely to be observed. will actually use the facility itself. A given facility will likely The observed lower bound for the number of daily adult bicy- be used by many people who do not live near it, and some local clists is equal to the commute share (even though in this case residents may ride but not on the facility. For purposes of esti- there are more total bicyclists than commuters, since they mating benefits, one wants to know the number of cyclists in are calculated from different denominators). A "most likely" general (on or off the facility), and how the presence of a value would be 0.4% plus 1.2 times the commute share; this facility might impact that number, based on empirical obser- was the best fit at the MSA level, and also describes the vation. Estimating total users, including those from outside United States as a whole. An upper bound would be about the immediate area, would have required a level of data that 0.6% plus 3 times the commute share; this is slightly higher is not available and details of local geography that would be than the slope observed at the neighborhood level. hard to account for in a general way. Two important points are worthy of note. First, the range Once the demand estimate has been reduced to a range of is large in relative terms; bicycling days per adult are 8 to 10 possible values based on readily available data and a set of times larger in the high-bicycling cities and states compared tested relationships between different measures of bicycling with the lower level places. The difference between neigh- demand, the user can apply his or her own judgment and local borhoods within a city can be even larger. These variations knowledge to choose a most likely point within the range that seem far larger than can be reasonably explained by differ- has been determined. It could be that circumstances are so ences in formal policies and facilities, especially given that unusual in a given situation that the user will even want to some low-cycling areas have similar circumstances to other choose a point outside the recommended range. high-cycling areas. It seems that local attitudes and perhaps The kinds of factors that the user might want to consider in history play a substantial role in the perception of bicycling this step could be things like design details of the facility, spe- as an appealing or normal thing for an adult to do. Thus, these cial local land uses that could affect bicycling demand, how guidelines leave considerable scope for the planner to apply the facility might fit into a larger system, and so on. These local knowledge and judgment to modify the estimated range kinds of factors would not be included in the primary demand of demand levels estimated. range estimate for two reasons. First, there is no compelling The second important point is that while the range is very evidence in terms of how these factors affect demand or that large in relative terms, it is very small in absolute terms. An the effect is sufficiently predictable and reliable that it can be estimate of total usage or total benefits in a given local context included in a model that can be applied across a variety of is unlikely to be off by very much in absolute terms because locations. For example, universities tend to be associated with the numbers overall are so small. Because costs tend to be rel- higher levels of bicycling in general, but the size of this effect atively small as well, even somewhat inaccurately predicted seems to be highly variable depending on the specific situa- demand is unlikely to lead to poor decisions on major invest- tion. Second, the number of possible local details would be so ments. The demand model outlined should be accessible to large as to be unmanageable, especially when there is no decision-makers and provide a known range of outcomes.iii basis for attaching specific numbers to most of them anyway. Although examples can be listed to prompt the user to think of additional local factors, putting together a comprehensive Description of Process list would be nearly impossible. The demand estimation method, in summary, provides a The process for estimating the increase in demand due to range of possible demand levels for a given situation, based a facility is based on two steps derived from the team's on a simple method derived from high quality, nationally research described in this chapter and in the Appendix A consistent data and from well-tested relationships among and B. The first step uses the bicycling commute share in the various measures of bicycling demand. The user would then area around the proposed facility to generate low, medium, be able to choose a most likely estimate based on this range and high estimates of the total number of current adult bicy- by applying local knowledge and judgment regarding other clists in the area (i.e., the average number on a given day). more qualitative factors.