Edmund J. Zolnik

Context in Human Geography: A Multilevel Approach to Study Human–Environment Interactions∗

One of the shortcomings of quantitative methods in human geography is an overreliance on models that can only account for the average outcomes of spatial events. Although methodological advancements in the form of multilevel models now allow human geographers to account for average outcomes as well as deviations about average outcomes of spatial events, it is not without question that such models can fully account for the cultural, political, and social contexts that affect human behavior. In this article I argue for the adoption of a multilevel approach to the study of human behavior to address some of the criticisms of quantitative methods in human geography. Using a transportation geography research agenda as an example, I discuss how a multilevel approach can complement qualitative methods. Such mixed research methods offer human geographers the opportunity to identify the appropriate role for quantitative methods and explore how context affects behavior.

The Effect of Sprawl on Private-Vehicle Commuting Outcomes

Amongst the hypothesized beneficial impacts of sprawl on urban quality of life is a decrease in commuting times. Unfortunately, empirical evidence to substantiate the commuting time benefits of sprawl is scant. To address this void in the urban planning literature, a multilevel approach is adopted to study how sprawl affects the duration and length of private-vehicle commutes. Using microdata from the 2001 National Household Transportation Survey for individuals and microdata on various measures of sprawl for metropolitan areas, multilevel models of commuting times and distances are estimated for a subsample of private-vehicle commuters. Results somewhat support the hypothesis that sprawl decreases commuting times for private-vehicle commuters—employment decentralization is associated with shorter commuting times. However, this finding is based on the statistical significance of one of the sprawl measures; not on its practical significance in explaining commuting times which is, in fact, negligible. Coupling this with the finding that the multilevel model only explained about 5% of the variance in private-vehicle commuting times suggests that there is considerable room available for improving the model specification.

The effects of sprawl on private-vehicle commuting distances and times

Sprawl is a well-known consequence of the suburbanization of the United States population. However, only recently have scholars shown interest in the effects of sprawl on transportation outcomes. For example, little research on how sprawl affects private-vehicle commuting distances and times is evident in the literature. To explore further the relationship between sprawl and private-vehicle commuting distances and times, a multilevel approach is adopted. Using microdata for households and macrodata for Metropolitan Statistical Areas, multilevel models of the effects of four dimensions of sprawl on commuting distances and times are estimated for private-vehicle commuters. Results show that shorter-distance, private-vehicle commutes are associated with lower density, more centeredness, and better accessibility, while shorter-time, private-vehicle commutes are associated with more centeredness. Overall, results point toward centralizing employment as the urban planning policy most likely to reduce private-vehicle commuting times, if only modestly.

Poisson Multilevel Methodology of Bicycle Levels of Service for Road Networks

Methodologies for bicycle levels of service have been developed for urban and suburban road segments, as well as for rural road segments. Today, however, the utilitarian (commuter) cyclist requires access to urban, suburban, and rural environs to travel safely between home and work. To complement bicycle level-of-service methodologies that incorporate mental stressors in road segments, this study developed a bicycle level-of-service methodology that incorporates physical stressors in road networks with the greatest influence on the frequency and the severity of cycling injuries from bicycle–motor vehicle collisions. To do so, the methodology integrates output from two different statistical models. To model the frequency of bicycle–motor vehicle collisions on all road segments (n = 65) in a regional road network, a Poisson model was used. To model the severity of all bicycle–motor vehicle collisions (n = 25) on the road segments in a regional road network, a multilevel model was used. To complete the Poisson multilevel bicycle level-of-service methodology, the outputs from the Poisson model of the frequency of bicycle–motor vehicle collisions and from the multilevel model of the severity of bicycle-motor vehicle collisions were used as the inputs in a geographic information system. Through analysis of the study area road network, samples of utilitarian cycling routes that minimize distance alone and that maximize safety alone were selected to quantify the benefits and the costs to utilitarian cyclists in terms of time and safety for both selections.

The North American City Revisited: Urban Quality of Life in Canada and the United States

An international comparative urban analysis requires an effort to reconcile not only the variables, but also the units of observation from each country. In a multivariate analysis, failure with regard to the former jeopardizes the validity of the results while failure with regard to the latter risks a priori bias in the interpretation of the results. This comparative, multivariate analysis of urban quality-of-life variables from Canada and the United States calls into question the distinctiveness of the Canadian City in the North American City debate because of such potential bias. As a point of departure from past research, not only is an attempt made to harmonize the set of urban quality-of-life variables with regard to definition, year, and geography, but also to control for the definitional, geographic, and hierarchical differences between the Canadian and U.S. urban systems. Such efforts reveal the Canadian City is indeed distinct from the U.S. City, but also distinct from the North American City, which is a regional entity encompassing Canada and the United States.

Growth Management and Resident Satisfaction with Local Public Services

The lack of timely survey data on resident satisfaction with local public services makes it difficult to verify the legitimacy of resident concerns with the effects of rapid population growth. This study attempts to verify the legitimacy of such concerns by analyzing survey data on resident service satisfaction in a rapidly growing county in Northern Virginia. Using an empirical model that simultaneously incorporates individual- and community-level predictors of resident service satisfaction seems to support residents concerns; resident service satisfaction is significantly lower in the area of the county (the eastern part) where most new development has occurred. This finding suggests that residents living in comparatively low density communities in the fastest-growing area of the county are not very satisfied with services. One explanation is that these services are not yet in place so they compare unfavorably with the services residents experienced in the communities from which they came.

Eliminating Bus Stops: Evaluating Changes in Operations, Emissions and Coverage

Bus systems in the United States are unattractive to many potential riders because of their lack of efficiency, especially with regard to travel time. One of the reasons services are not more efficient has to do with the spacing of bus stops. After using a nearest facility algorithm with an 800 m walking distance threshold to identify eligible bus stops in the current bus system in the city of Fairfax, Virginia, the impacts of their elimination on operations, emissions, and coverage are estimated. Results indicate that eliminating some bus stops (about 40% of current stops) could improve travel times and reduce operating costs by the same percentage (23%). In addition, bus-related emissions such as carbon monoxide (CO, 34%), volatile organic compounds (VOC, 18%), and nitrogen oxide (NOx, 10%) could all be substantially lower. Surprisingly, the loss in coverage due to eliminating stops would not be large (10% of the total population of the city of Fairfax).

The Geographic Distribution of U.S. Unemployment by Gender

A study of the geographic distribution of U.S. unemployment takes up the challenge to explore regional variation in unemployment by gender. To do so, the differential effects of local and regional unemployment on the likelihood that individual women and men are unemployed and differences in the underlying patterns of unemployment by gender are explored. Using a unique tabulation of micro data for labor market areas, two series of multilevel models for women and men are fit. The multilevel model results suggest that regional variation in unemployment is more evident for men than for women and, contrary to the empirical evidence, the underlying pattern of unemployment for men is more clustered than for women. The article concludes by discussing the policy implications of the finding that most of the variation in labor market outcomes for both women and men is attributable to individual-level rather than labor market area-level characteristics.

Multilevel Models of Commute Times for Men and Women

The commuting time discrepancy between men and women is known as the commuting time gender gap. Empirical evidence for the gender gap seems to be conclusive. However, recent research on commuting times in San Francisco (Gossen and Purvis 2005) and Philadelphia (Weinberger 2007) suggests that the gender gap is less ubiquitous than previously thought. To test whether or not the attenuation of the gender gap is idiosyncratic to single-city analyses of commuting times, national data is used to specify three statistical models of private-vehicle commuting times for men-only, women-only, and pooled men–women subsamples from the 2001 National Household Travel Survey (NHTS). The first goal of this chapter is to ascertain what personal characteristics of men and women and what locational characteristics of cities have the greatest affect on private-vehicle commuting times. The second goal of this chapter is to ascertain how much of the variation in commuting times for men and women originates within cities and how much originates between cities.

The role of postdoctoral fellows in technology transfer: evidence from the National Capital Region of the USA

Surveys of postdoctoral fellows in the National Capital Region of the USA examined their role in transferring technologies from national and university research laboratories to the regional economy. Results suggest that postdoctoral fellows identified themselves as having a vital role in technology transfer. However, they exhibited very little interest in self-employment after completing their fellowships. Of those few who were actively preparing for self-employment, initial wealth was a distinguishing factor. Housing affordability was a financial constraint for many study participants, but, overall, they expressed satisfaction with the economic and cultural diversity of the area and identified each as an advantage of settling there permanently. Taken together, educational courses and programs which make postdoctoral fellows aware of entrepreneurial career tracks and mentor them early on in their transition from apprentice to entrepreneur are the most appropriate policy interventions suggested by these results.