GEORGIA DOT RESEARCH PROJECT 16-38 FINAL REPORT
BELTLINE BICYCLIST FACILITY PREFERENCES AND EFFECTS ON INCREASING TRIPS
OFFICE OF PERFORMANCE-BASED MANAGEMENT AND RESEARCH 15 KENNEDY DRIVE FOREST PARK, GA 30297-2534 March 2019

1.Report No.: FHWA-GA-19-1638

2. Government Accession 3. Recipient's Catalog No.: No.:

4. Title and Subtitle: BeltLine Bicyclist Facility Preferences and Effects on Increasing Trips
7. Author(s): Dr. Kari Watkins, Dr. Giovanni Circella, Dr. Patricia Mokhtarian, Calvin Clark, and Reid Passmore 9. Performing Organization Name and Address: Georgia Institute of Technology, Civil and Environmental Engineering 12. Sponsoring Agency Name and Address: Georgia Department of Transportation Office of Research 15 Kennedy Drive Forest Park, GA 30297-2534 15. Supplementary Notes:

5. Report Date: March 2019
6. Performing Organization Code 8. Performing Organ. Report No.:
10. Work Unit No. 11. Contract or Grant No.: PI #: 0015304
13. Type of Report and Period Covered: Final, Dec 2016 April 2019;
14. Sponsoring Agency Code:

16. Abstract: The objectives of this study were to investigate perceptions of users and potential users of bicycle infrastructure and to investigate the impact of multi-use paths on bicycle trips. Data were collected through a before-and-after survey in 2017 and 2018 (wave 1 N=1,335, wave 2 N=713) of residents near the Eastside Atlanta BeltLine extension and the Westside Atlanta BeltLine trail, along with residents in neighboring control communities of Grant Park and South Atlanta. Survey data was analyzed using statistical models such as analysis of variance, ordinary least squared regression, and segmented models. The analyses indicate positive perceptions of comfort and safety along with willingness to try biking on facilities with a greater degree of separation from traffic. Although results suggest that those residing near the BeltLine projects perceived a greater degree of neighborhood improvements for both biking and walking, there does not appear to be any statistically significant difference between the BeltLine and control communities in terms of actual changes in frequency of bicycling. The findings of this research suggest that although the BeltLine has had a positive impact on perceptions of the bikeability of the neighborhood it is not enough to spur substantial changes in behavior.

17. Key Words: Bicycle infrastructure; Atlanta BeltLine; Multiuse Trail; Protected Bicycle Lanes; Quasi-experimental Design; Panel Survey

19. Security Classification (of this report): Unclassified

20. Security Classification (of this page): Unclassified

18. Distribution Statement:

21. Number of Pages:
150 pages

22. Price:

ii

GDOT Research Project No. 16-38 Final Report
BELTLINE BICYCLIST FACILITY PREFERENCES AND EFFECTS ON INCREASING TRIPS
By Dr. Kari Watkins, Dr. Giovanni Circella, Dr. Patricia Mokhtarian,
Calvin Clark, and Reid Passmore Georgia Tech Research Corporation
Atlanta, Georgia Contract with
Georgia Department of Transportation Office of Roadway Design/ Office of Design Policy
Research and Development Branch In cooperation with
U.S. Department of Transportation Federal Highway Administration
March 2019
The contents of this report reflect the views of the authors who are responsible for the facts and the accuracy of the data presented herein. The contents do not necessarily reflect the official views or policies of the Georgia Department of Transportation or the Federal Highway Administration. This report does not constitute a standard, specification, or regulation.
iii

Table of Contents
List of Tables............................................................................................................... v List of Figures........................................................................................................... viii Executive Summary..................................................................................................... x Introduction ...............................................................................................................1
Research Approach ............................................................................................................3 First-Wave Survey Description ....................................................................................7
Survey Method ..................................................................................................................7 Survey Design ....................................................................................................................8 Data Cleaning ..................................................................................................................13 Survey Response..............................................................................................................14 First-Wave Survey Statistics ...................................................................................... 15 Summary Statistics Separated by Site ...............................................................................15 Summary Statistics Segmented by Rider Status.................................................................22 First-Wave User Preference Analysis ......................................................................... 29 Infrastructure Images.......................................................................................................29 Image Response Results ...................................................................................................32 User Preference Models...................................................................................................38 First-Wave Survey Conclusions .........................................................................................53 Second-Wave Survey Description .............................................................................. 55 Multi-wave Demographic Statistics ........................................................................... 59 Summary Statistics for Matched Respondents ..................................................................59 Second-Wave Survey Analysis ................................................................................... 65 General Perceptions of Changes in Transportation............................................................65 Recognition and Use of New Bicycle Facilities ................................................................... 77 Second-Wave User Preference Analysis ............................................................................82 Before-and-After Analysis ......................................................................................... 87 Changes in User Preference Analysis.................................................................................87 Changes in Cycling Frequency...........................................................................................91 Conclusions............................................................................................................... 93 Implementation Recommendations .......................................................................... 95 Appendix A: First-Wave Survey ................................................................................. 97 Appendix B: Second-Wave Survey ........................................................................... 115 Appendix C: Complete Demographics...................................................................... 131
iv

List of Tables

Table

Page

1. Survey Responses by Site......................................................................................... 14 2. Survey Responses by Version .................................................................................. 14 3. Survey Respondents' and Study Area Population Household Incomes
(Wave 1) ................................................................................................................... 16 4. Survey Respondents' and Study Area Population Household Sizes (Wave 1)......... 17 5. Survey Respondents' Residence Types (Wave 1) .................................................... 18 6. Survey Respondents' Genders (Wave 1) ................................................................. 18 7. Survey Respondents' Ages (Wave 1) ....................................................................... 19 8. Survey Respondents' Races (Wave 1)...................................................................... 20 9. Survey Respondents' Employment Status (Wave 1) ............................................... 21 10. Number of Vehicles Owned by Survey Respondents (Wave 1)............................... 21 11. Number of Bikes Owned by Survey Respondents (Wave 1).................................... 22 12. Respondents' Stated Bike Confidence Level (Wave 1) ............................................ 22 13. Distribution of Rider Segments by Neighborhood (Wave 1) ................................... 23 14. Survey Respondents' Household Income by Rider Type (Wave 1).......................... 24 15. Survey Respondents' Household Sizes by Rider Type (Wave 1).............................. 24 16. Survey Respondents' Residence Types by Rider Type (Wave 1) ............................. 25 17. Survey Respondents' Gender by Rider Type (Wave 1) ............................................ 25 18. Survey Respondents' Age by Rider Type (Wave 1) .................................................. 26 19. Survey Respondents' Race by Rider Type (Wave 1) ................................................ 26 20. Survey Respondents' Employment Status by Rider Type (Wave 1) ........................ 27 21. Number of Vehicles and Bikes Owned by Survey Respondents by Rider Type
(Wave 1) ................................................................................................................... 28 22. Respondent's Stated Level of Confidence by Rider Type (Wave 1)......................... 28

v

23. Self-Reported Frequency of Use for Multi-use Paths .............................................. 36 24. Self-Reported Frequency of Use for Each Infrastructure Type for Two-lane
Roads without Parking ............................................................................................. 36 25. Self-Reported Frequency of Use for Each Infrastructure Type for Two-lane
Roads with Parking................................................................................................... 37 26. Self-Reported Frequency of Use for Each Infrastructure Type for Four-lane
Roads without Parking ............................................................................................. 37 27. Self-Reported Frequency of Use for Each Infrastructure Type for Four-lane
Roads with Parking................................................................................................... 37 28. Average Ratings for Comfort, Safety, and Willingness to Try for Protected
Bike Lanes and Multi-Use Paths............................................................................... 41 29. Linear Regression for Expressed Comfort, Safety, and Willingness to Try,
Including only Infrastructure Characteristics........................................................... 43 30. Linear Regression for Expressed Comfort, Safety, and Willingness to Try by
Infrastructure and Individual Characteristics .......................................................... 46 31. Linear Regression for Expressed Comfort by Infrastructure and Individual
Characteristics, Segmented by Rider Type .............................................................. 50 32. Linear Regression for Expressed Safety by Infrastructure and Individual
Characteristics, Segmented by Rider Type .............................................................. 51 33. Linear Regression for Expressed Willingness to Try by Infrastructure and
Individual Characteristics, Segmented by Rider Type ............................................. 52 34. Survey Responses for Waves 1 and 2 for each Neighborhood................................ 58 35. Genders of Respondents of both Wave 1 and Wave 2............................................ 59 36. Ages of Respondents of both Wave 1 and Wave 2 ................................................. 60 37. Races of Respondents of both Wave 1 and Wave 2 ................................................ 60 38. Household Incomes of Respondents of both Wave 1 and Wave 2 ......................... 61 39. Household Sizes of Respondents of both Wave 1 and Wave 2 ............................... 61 40. Residence Types of Respondents of both Wave 1 and Wave 2 .............................. 62 41. Employment Status of Respondents of both Wave 1 and Wave 2 ......................... 62 42. Number of Vehicles Owned per Household of Respondents of both Wave 1
and Wave 2 .............................................................................................................. 63
vi

43. Number of Bikes Owned per Household of Respondents of both Wave 1 and Wave 2 ..................................................................................................................... 63
44. Stated Bike Confidence Level of Respondents of both Wave 1 and Wave 2 .......... 64 45. ANOVA Results for Mean Responses for Sidewalk Availability ............................... 75 46. ANOVA Results for Mean Responses for Sidewalk Quality ..................................... 75 47. ANOVA Results for Mean Responses for Bike Safety .............................................. 76 48. ANOVA Results for Mean Responses for Bike Lane/Trail Availability ..................... 76 49. ANOVA Results for Mean Responses for Bike Lane/Trail Quality............................ 77 50. Linear Regression for Expressed Comfort, Safety, and Willingness to Try,
Including Only Infrastructure Characteristics .......................................................... 83 51. Linear Regression for Expressed Comfort, Safety, and Willingness to Try by
Infrastructure and Individual Characteristics .......................................................... 85 52. Linear Regression for Expressed Comfort, Safety, and Willingness to Try,
Including Only Infrastructure Characteristics .......................................................... 89 53. Linear Regression for Expressed Comfort, Safety, and Willingness to Try by
Infrastructure and Individual Characteristics .......................................................... 90 54. Changes in Bike Commuting Frequency from First to Second Wave ...................... 91 55. Changes in Frequency of Other Trips by Bike from First to Second Wave.............. 92 C-1. Wave 1 & 2 Survey Respondents by Gender ......................................................... 131 C-2. Wave 1 & 2 Survey Respondents by Age ............................................................... 131 C-3. Wave 1 & 2 Survey Respondents by Race ............................................................. 132 C-4. Wave 1 & 2 Survey Respondents by Household Income....................................... 133 C-5. Wave 1 & 2 Survey Respondents by Household Size ............................................ 134 C-6. Wave 1 & 2 Survey Respondents by Residence Type ............................................ 134 C-7. Wave 1 & 2 Survey Respondents by Employment Status ..................................... 135 C-8. Wave 1 & 2 Survey Respondents by Number of Vehicles ..................................... 136 C-9. Wave 1 & 2 Survey Respondents by Number of Bikes .......................................... 137 C-10. Wave 1 & 2 Survey Respondents by Cycling Confidence Level ............................. 138
vii

List of Figures

Figure

Page

1. Map of BeltLine Current Segments............................................................................ 2 2. Map of BeltLine Treatment and Control Neighborhoods.......................................... 8 3. Image for Multi-use Paths Used in Survey............................................................... 10 4. Images of Infrastructure Configurations for Different Roadway Layouts Used
in Survey................................................................................................................... 12 5. Combinations of Bicycle Infrastructure Used in Survey Versions 1 and 2............... 30 6. Combinations of Bicycle Infrastructure Used in Survey Versions 3 and 4............... 31 7. Distribution of Comfort Perceptions for Each Image (Wave 1)............................... 33 8. Distribution of Safety Perceptions for Each Image (Wave 1) .................................. 34 9. Distribution of Willingness to Try Perceptions for Each Image (Wave 1) ............... 35 10. Average Expressed Comfort Levels for Each Lane/Parking Configuration by
Bicycle Infrastructure Type ...................................................................................... 39 11. Average Expressed Safety Levels for Each Lane/Parking Configuration by
Bicycle Infrastructure Type ...................................................................................... 39 12. Average Expressed Level of Willingness to Try for Each Lane/Parking
Configuration by Bicycle Infrastructure Type .......................................................... 40 13. Distribution of Responses for Perceived Changes in Traffic Congestion................. 66 14. Distribution of Responses for Perceived Changes in Parking Availability ............... 67 15. Distribution of Responses for Perceived Changes in Availability of Taxi/ Uber/
Lyft ........................................................................................................................... 68 16. Distribution of Responses for Perceived Changes in Public Transit Route
Coverage .................................................................................................................. 69 17. Distribution of Responses for Perceived Changes in Public Transit Frequency ...... 69 18. Distribution of Responses for Perceived Changes in Sidewalk Availability ............. 70 19. Distribution of Responses for Perceived Changes in Sidewalk Quality ................... 71 20. Distribution of Responses for Perceived Changes in Bicycle Safety........................ 72

viii

21. Distribution of Responses for Perceived Changes in Availability of Bicycle Lanes and Trails........................................................................................................ 72
22. Distribution of Responses for Perceived Changes in Quality of Bicycle Lanes and Trails.................................................................................................................. 73
23. Chart of Mean Responses for Pedestrian- and Bicycle-related Questions.............. 74 24. Distribution of Responses for the Question, "Have you seen this added in
your community?" for Each Infrastructure Type and for Each Neighborhood ....... 78 25. Distribution of Responses for the Question, "Have you used it?" for Each
Infrastructure Type (for those who have seen it) and for Each Neighborhood ...... 80 26. Distribution of Responses for the Question, "Do you like it?" for Each
Infrastructure Type (for those who have seen it) and for Each Neighborhood ...... 81
ix

Executive Summary
The BeltLine is a major infrastructure initiative in the city of Atlanta that is ultimately intended to convert a 22-mile ring of a former rail corridor into a multimodal walk/bike/transit corridor around the dense urban center of Atlanta. In spite of the interest in implementing multimodal transportation infrastructure there is little known in terms of the actual impacts of such projects on perceptions and travel behavior. The objective of the project summarized by this report is to use the opening of two of these segments in Fall 2017, the Westside trail (3 miles) and the Eastside trail extension (1.25 miles), to investigate the impact of such multi-use paths on perceptions of bikeability and bicycle trip making for those who reside near these facilities.
The data for this project was collected using a two-wave panel survey deployed in May 2017 (N=1,335) and May 2018 (N=713). Those residing within a half-mile of the two segments were included in the "treatment" group, and nearby neighborhoods (in South Atlanta and Grant Park) with similar land-use and demographic makeup were also included in the study as a "control" group. Thus, the research had a quasi-experimental design with the survey serving as an instrument for a before-and-after-with-controls natural experiment.
The first-wave survey was 12 pages and took approximately 30 minutes to complete. To keep from biasing responses towards those who are more interested in biking, the survey was designed as a general transportation survey including questions regarding attitudes, technology usage, home/work/commute, mode characteristics (for driving, transit, biking, and walking), perceptions of bicycle infrastructure, and sociodemographics. Respondents were also shown several images of hypothetical roadways including variations in bicycle accommodations, the number of vehicular lanes, and the presence of on-street parking, and asked to rate the extent to which cycling on such a road would be comfortable, safe, or something they would try.
x

Responses to questions on the perceptions of cycling images were used to estimate a linear regression model. Respondents showed a significant increase in perceptions for bicycle facility types that provided greater degrees of separation from automobiles, while the presence of on-street parking was also a clear deterrent to perceptions. The presence of an additional lane of automobile traffic was a negative factor for perceptions in some cases, though this variable was not consistently significant in all models. Respondents were also segmented into different rider type groups: potential cyclists (N=648), recreational cyclists (N=330), utilitarian cyclists (N=234), and those who cannot bike (N=97). Segmented regression models reveal that the perceptions of some characteristics may vary amongst rider types. For example, those identified as recreational cyclists had coefficients in models of safety and comfort that were significantly different (and negative) from the rest of the sample for the number of vehicular lanes, indicating that this factor may be strongest among the potential cyclist group.
The second-wave survey was a condensed version (20 minutes) of the first-wave survey that was sent to all those who responded to the first survey. Additional questions were added asking about perceptions and recognition of changes that may have occurred in transportation in the community in the previous year. This survey revealed that respondents near the recently completed BeltLine segments perceived a more positive change in both walkability and bikeability than those in the respected control sites. Those near the Westside trail recognized drastically more positive improvements in bikeability than those in their control site of South Atlanta, though the difference in perceived improvement between those near the Eastside extension and those in the control site of Grant Park were much less pronounced. Despite the apparent differences in perceptions of bikeability attributed to the treatment, there appears to be very little in terms of changes in bicycle trip frequency.
The findings from this research project provide GDOT and other agencies with evidence regarding the impact of multi-use trails. GDOT and other agencies should use this document to give priority to the implementation of protected bicycle facilities. Multi-use paths such as the
xi

BeltLine should also be implemented, particularly in areas that may be lacking in bikeability and walkability. Finally, this project shows the importance of conducting regular before-and-after studies on further infrastructure projects, such as projects like these using the same or a similar survey instrument.
xii

Introduction
Pursuit of cycling as a sustainable transportation alternative is desirable for several reasons. However, accurate and robust data to support decisions on where and how to best develop new cycling infrastructure remain elusive. Data on current bicycling has many gaps, but more importantly, there is almost no data on potential cyclists--who they are, the barriers that inhibit their cycling, and how infrastructure investments may help to overcome these barriers. As a result, planners have little understanding of the latent demand from either current or potential cyclists who do not feel safe due to current infrastructure.
This project is an addition to the National Academies' Transportation Research Board (TRB) National Cooperative Highway Research Program (NCHRP) Project 08-102, Bicyclist Facility Preferences and Effects on Increasing Bicycle Trips, to be published by the study team in 2019. The objective of that study is to understand how both current and potential cyclists respond to different types of cycling infrastructure, thus facilitating a quantification of demand that is both induced and generated through mode and route shifts. In contrast to previous research that has predominantly been conducted in communities where cycling is widely accepted and automobile drivers are conditioned to the presence of cyclists, this study focuses on communities in the southern United States, where cycling for transportation is relatively new and rapidly expanding. Using such communities as illustrative examples of evolving cycling infrastructure, the study team is conducting a comprehensive investigation of personal preferences and attitudes, current behaviors, and propensities to bicycle in response to different types of bicycle infrastructure investments and facility designs.
One major infrastructure initiative in the city of Atlanta is the Atlanta BeltLine, a 22-mile ring around the dense urban center of Atlanta that will convert a former rail corridor into a multimodal walk/bike/transit corridor. Full build-out is anticipated to include 33 miles of trails around the ring
1

and connecting to it. During the study site selection of NCHRP 08-102 in 2015, one 2.2-mile section of the ring trail (the Eastside trail) and two other connecting trails were already open. Two additional sections of the ring trail were expected to open during the timeline of NCHRP 08-102: a 1.25-mile Eastside trail extension and a 3-mile Westside trail, shown in Figure 1.
FIGURE 1 Map of BeltLine Current Segments
2

The NCHRP study began in August 2015, but in the process of identifying locations, the BeltLine project schedules had slipped and the two BeltLine projects no longer fit into the NCHRP project schedule. Therefore, three other study areas were chosen for the NCHRP study: Anniston, Alabama (sharrows and bike lanes), Opelika, Alabama (protected bike lanes) and Chattanooga, Tennessee (protected and buffered bike lanes). This project, therefore, is a supplement to the NCHRP project to allow the study team to deploy the NCHRP project survey in the BeltLine communities.
Research Approach The research team's approach to understanding the relative preference for and relative
effectiveness of various kinds of bicycle facilities among current and potential cyclists is crosssectional and quasi-experimental. Specifically, this project investigates the revealed preferences of existing cyclists and stated preferences of potential cyclists through a panel dataset collected through two waves of an online and paper survey. The first-wave survey was distributed among a sample of current and potential users in the study areas to evaluate their personal attitudes, preferences, and behaviors before the opening of planned bicycle facilities. The second-wave survey included many of the same questions, with additional questions relating to perceptions of new infrastructure changes. To enable the researchers to measure changes rigorously, while avoiding biasing respondents toward exaggerating any changes, both surveys had a similar structure.
This approach provides a rigorous basis for estimating both induced demand as well as demand that results from mode and route shifts. Key dependent variables include the following:
Preference for facility types Likelihood of cycling Revealed amounts of actual cycling
3

The study controls for a number of explanatory variables including the following: Individual sociodemographic characteristics Personal attitudes, personality traits, lifestyles, and preferences Household characteristics and living arrangements Work characteristics and schedule Current travel behavior patterns for both commuting and leisure trips Residential location and land use characteristics Community environment (e.g., extent of bicycle network, community support, population characteristics, geography) Features of bicycle facilities (e.g., on-road bike lanes, off-road bike trails, intersection control) for both existing facilities and future projects
The research is of a quasi-experimental or natural experiment design. The purpose of such a design in this case is to measure perceptions and behaviors in a "treatment" group before and after a treatment is implemented, which in this case is the opening of the Westside trail and the Eastside extension. The measurements at two separate points of time increases the robustness of the research by allowing for the analysis of a change associated with the treatment. The robustness of the analysis is further augmented by the inclusion of control groups that are similar in nature to the treatment groups with the only difference being the lack of the treatment. The combination of these characteristics allows for a difference-in-difference analysis, where the differences between first and second observations in the treatment group can be compared to the differences between the first and second observations in the control group. The design of this study enables the research team to disentangle background changes in attitudes and demographics that may be confounded with the influence of the new infrastructure. This before-and-after-with-control-group approach is considered to be a robust quasi-experimental design that protects against a number of common
4

threats to validity. It will provide strong evidence for the impacts of various infrastructure improvements on cycling behavior.
5

This page is intentionally left blank. 6

First-Wave Survey Description
Survey Method The initial sample of respondents invited to complete the first-wave survey was built with a
stratified random sampling methodology. For the "treatment" neighborhoods, the researchers focused on the residents that live within a radius of 0.5 mile from the location of the coming BeltLine segment. For the "control" neighborhoods, the researchers identified adjacent, similarsized areas comprising contiguous areas matched on key variables, including population and employment density, mean income, household size, race and ethnicity, and presence of student population. These comparisons were done using American Community Survey (ACS) 5-year data and were verified using demographic data purchased with the addresses from the targeted marketing company. The two control neighborhoods identified were in areas near Grant Park (control for the Eastside treatment) and South Atlanta (control for the Westside treatment) as shown in Figure 2.
7

FIGURE 2 Map of BeltLine Treatment and Control Neighborhoods
The intent of the survey was to: (1) identify the composition of the population of current and potential bicycle users, and their characteristics; (2) assess the size of the persuadable market of potential bicycle users; (3) assess preferences for "treatments," e.g., different types of bicycle infrastructure and facilities; and (4) investigate the relationships of several dimensions of interest, including users' personal attitudes and preferences, current lifestyles, land-use patterns, and sociodemographic traits, with current travel behavior and the propensity to engage in bicycle use. Questions were designed to address all of these issues.
Survey Design The survey instrument was 12 pages and took approximately 30 minutes for the respondent to
complete. This allowed a nice balance of a thorough dataset, but limited time commitment from participants. To reduce potential response biases, the content of the survey was purposefully
8

broader than just cycling to ensure that participants remained interested and did not quit the survey if they did not recognize themselves as the "biking type." To the extent practical, the researchers reused questions from previous surveys both to rely on previously tested and vetted questions and to maximize opportunities for cross-study comparisons of results. The resulting survey contained six sections, including:
A. Attitudes B. Technology usage C. Household location D. Daily travel E. Bicycling experience F. Demographics The complete survey instrument is found in Appendix A. Particular attention was given to attitudinal questions regarding car dependence, environmental concerns, exercise, land use, mode preferences, peer influence, time pressure, and multitasking for the survey. To assess bicycle preferences, the research team used Adobe Photoshop to modify an image of a generic low-rise downtown streetscape into 16 images, with all combinations of four bike infrastructure classes (i.e., sharrows, bike lanes, buffered bike lanes, and protected bike lanes); presence or absence of onstreet parking; and two versus four traffic lanes. The background image was intended to be seen as a small-town downtown or central point in a lower-density area of an urban environment to allow it to be familiar to residents from a variety of urban settings. An additional image of a multi-use trail was also used, but due to the nature of this type of infrastructure it was impossible to use the common streetscape. It was impractical to ask each respondent to rate all 17 images, so the researchers prepared four different versions of the survey, using a modified factorial design that gave each respondent six images to evaluate. Each respondent was presented with one image from each of the four types of
9

on-street infrastructure (i.e., sharrows, bike lanes, buffered bike lanes, and protected bike lanes) for the same roadway characteristics, and at least one additional image from among those four types that differed either in whether parking was present or not, or in whether the street was two-lane or four-lane. The sixth image was either another "double" from among the four infrastructure types, or portrayed a multi-use path as shown in Figure 3. These combinations ensured that across the entire sample, specific comparisons of interest could be made. All 17 images were tested in focus groups and some modifications were applied. Figure 4 displays the images used for the 16 on-street infrastructure configurations.
FIGURE 3 Image for Multi-use Paths Used in Survey The survey was pretested with graduate students, the NCHRP panel, and members of the public. Both an online version and a paper version were prepared. All four versions of the final survey are attached to this report in Appendix A. The survey is intended to be generic enough for use across the country for future comparison of results in varying locations (beyond the scope of this project). The survey was deployed in May 2017 and responses were collected throughout that summer. A printed version of the full survey (including a URL for an online version) was mailed to over 17,000 residents of the study area. The research team provided a 1-800 number and email address
10

to field questions or comments from respondents. Each paper survey was entered (coded) twice, and the two datasets were compared to ensure no coding errors were introduced during the dataentry process.
11

12

FIGURE 4 Images of Infrastructure Configurations for Different Roadway Layouts Used in Survey

Data Cleaning A general screening process was utilized during the data collection process and a more in-depth
review for missing data has followed in this phase. Unfinished surveys and those with a low portion of questions answered were removed entirely from the working database. An additional assessment was undertaken on a section-by-section basis, using commonly accepted methods to fill in small amounts of missing data, and excluding cases with an unacceptable amount of missing data. Cases were evaluated for inclusion or imputation on different completion criteria for each section, as follows:
Section A (Attitudes): Cases with more than five missing items were deleted; otherwise, missing items were imputed using expectation maximization.
Section B (Technology Use): Uncleaned. Section C (Household Information): Uncleaned. Section D (Daily Travel): Logical variables were introduced to account for any
discrepancies between employment data and commute pattern data. Section E (Bicycling Experience): For key dependent variables and segmentation
variables, all missing responses were excluded from the respective models. Section F (Sociodemographics): Where available, responses with small amounts of
missing sociodemographic data were supplemented with information from the targeted marketing database. After cleaning, there was data from 1,335 respondents. Each person responded to 6 different images, so there were up to 8,010 possible image responses for each of the 4 questions (i.e., comfort, safety, willingness to try, and frequency), though cases were excluded from their respective models due to item non-response.
13

Survey Response

In total, the researchers received 1,335 responses to the survey: 408 online and 927 on paper. Responses were distributed by site, as shown in Table 1.

Area
Eastside Westside Grant Park South Atlanta Total

TABLE 1 Survey Responses by Site

Households Contacted 4,509 5,035 4,411 3,815 17,770

Responses
433 235 477 190 1335

Response Rate 9.6% 4.7%
10.8% 5.0% 7.5%

Treatment / Control
Treatment Treatment
Control Control

As discussed previously, four different survey versions were used to limit the number of images that any one respondent saw. The four versions were evenly divided among the six sites. As shown in Table 2, the responses were fairly evenly distributed, as well.

TABLE 2 Survey Responses by Version

Version Number
1 2 3 4

Responses
332 339 363 301

Percent of Total
24.9% 25.4% 27.2% 22.5%

14

First-Wave Survey Statistics
Summary Statistics Separated by Site The final section of the survey included several demographics questions to illuminate the
participant's personal and household characteristics and allow comparison to the populations to which the respondents belong. Note that in most cases the most appropriate comparison is 5-year 2014 ACS data at the block group level, but in others the targeted marketing data received from Direct Mail, from which the original addresses were obtained, was used for comparison to the respondents. To control for possible discrepancies between the sample and the population shown in the tables below, models will include sociodemographic variables.
Individual demographics questions were also asked, but the researchers are not able to compare to the populations to which the respondents belong as this data is not readily available at the population level.
A breakdown of household incomes by study site is presented in Table 3. As discussed earlier, individuals in higher income brackets were overrepresented in the combined study area, but the individual study areas show that most of this comes from the Eastside and Grant Park study areas. Each treatment area has a comparable distribution to its respective control area. Note that for the sake of brevity, the percentage of respondents reported in this section only includes those who answered the questions.
15

TABLE 3 Survey Respondents' and Study Area Population Household Incomes (Wave 1)

Household Income
$15,000 or less $15,001 $30,000 $30,001 $50,000 $50,001 $75,000 $75,001 $100,000 $100,001 $125,000 More than $125,000 Prefer Not To Answer

Eastside (N=393) Responses* Population

8 2.0%

15%

15 3.8%

13%

36 9.2%

19%

56 14%

18%

63 16%

14%

59 15%

6.4%

156 40%

16%

33

Grant Park (N=426) Responses* Population

13 3.1%

16%

17 4.0%

11%

31 7.3%

12%

63 15%

16%

68 16%

13%

64 15%

12%

170 40%

20%

38

Household Income

Westside (N=199)
Responses* Population

$15,000 or less

42 21%

31%

$15,001 $30,000

36 18%

24%

$30,001 $50,000

32 16%

22%

$50,001 $75,000

31 16%

12%

$75,001 $100,000

32 16%

8.1%

$100,001 $125,000

10 5.0%

2.6%

More than $125,000

16 8.0% 1.8%

Prefer Not To Answer 22
*Percentage of respondents electing to answer the question.

South Atlanta (N=163)
Responses* Population

27 17%

35%

33 20%

23%

25 15%

14%

26 16%

14%

23 14%

5.4%

8 4.9% 3.3%

21 13%

5.1%

34

Household size by study area is presented in Table 4. Single households were generally underrepresented, with the exception of South Atlanta. Households of two were overrepresented, also with the exception of South Atlanta.

16

TABLE 4 Survey Respondents' and Study Area Population Household Sizes (Wave 1)

Household Size
1 2 3 4 5+

Eastside (N=420) Responses* Population

171 39%

58%

184 42%

30%

40 9.2% 7.4%

22 5.1% 4.7%

3 0.7% 0.7%

Grant Park (N=459) Responses* Population
132 28% 39% 201 42% 36% 50 10% 13% 60 13% 9.0% 16 3.4% 3.5%

Household Size

Westside

South Atlanta

(N=221)

(N=180)

Responses* Population Responses* Population

1

72 31%

38%

83 44% 39%

2

85 36%

29%

50 26% 27%

3

31 13%

14%

27 14% 16%

4

14 6.0% 8.0%

8 4.2% 8.5%

5+

19 8.1% 11%

12 6.3% 9.2%

*Percentage of entire sample size, note that due to non-responses percentages may not add

up to 100%.

Table 5 shows the breakdown of residence types by study site compared to the targeted marketing (TM) data. Note that the TM data reported only "single-family" and "multi-family" dwellings, which correspond loosely to "Detached" and "Duplex," and "Apartment" and "Other," respectively. Other than Eastside, most of the study sites were represented by detached resident types. Respondents along the Eastside extension were much more likely to live in an apartment than respondents in other areas.

17

TABLE 5 Survey Respondents' Residence Types (Wave 1)

Residence Type
Detached Duplex Apartment Other

Eastside

(N=432)

Responses*

TM

179 41%

39%

67 15%

183 42% 61%

3 0.7%

Grant Park

(N=477)

Responses*

TM

352 74%

65%

74 16%

48 10% 35%

3 0.6%

Residence Type

Westside (N=233)

South Atlanta (N=189)

Responses*

TM

Responses*

TM

Detached Duplex

183 78%

74%

125 66%

63%

16 6.8%

9 4.7%

Apartment

31 13% 26%

49 26% 37%

Other

3 1.3%

6 3.2%

*Percentage of entire sample size, note that due to non-responses percentages may not add

up to 100%.

Responses for gender were compared to the population (from the targeted marketing data) for each site as shown in Table 6. In each case, there were more female respondents than male respondents, but this trend was even more prevalent in the Westside.

TABLE 6 Survey Respondents' Genders (Wave 1)

Gender
Female Male

Eastside

(N=424)

Responses

TM

237 55% 53%

187 43% 47%

Grant Park

(N=470)

Responses

TM

263 55% 53%

207 43% 47%

Gender

Westside

(N=222)

Responses

TM

South Atlanta

(N=187)

Responses

TM

Female

153 65% 56% 103 54% 55%

Male

69 29% 44%

84 44% 45%

*Percentage of entire sample size, note that due to non-responses percentages may not add

up to 100%.

18

Age distributions compared to populations of each site (from the ACS population data) are presented in Table 7. Respondents under 35 were severely underrepresented in each study area. Like the combined study area data, older respondents were overrepresented.

TABLE 7 Survey Respondents' Ages (Wave 1)

Age
1834 3549 5064 65+

Eastside (N=428) Responses* Population

149 34%

49%

166 38%

28%

78 18%

16%

35 8.1%

6.7%

Grant Park (N=471) Responses* Population

120 25%

42%

191 40%

33%

112 23%

18%

48 10%

6.9%

Age

Westside (N=222)

South Atlanta (N=187)

Responses* Population Responses* Population

1834 45 19%

30%

33 17%

39%

3549 51 22%

29%

55 29%

30%

5064 72 31%

26%

59 31%

22%

65+

54 23%

16%

39 21%

9.2%

*Percentage of entire sample size, note that due to non-responses percentages may not add

up to 100%.

The racial breakdown of respondents by site is presented in Table 8. The majority of respondents were White or African American, but there were vastly more White respondents in Eastside and Grant Park than in South Atlanta and Westside. Still, African Americans were underrepresented even in South Atlanta and Westside.

19

TABLE 8 Survey Respondents' Races (Wave 1)

Race
White African American
Hispanic Asian
Native American Other

Eastside (N=425) Responses* Population
340 79% 55% 49 11% 37% 10 2.3% 3.9% 23 5.3% NA 1 0.2% NA 9 2.1% 8.6%

Grant Park (N=458) Responses* Population
371 78% 58% 62 13% 36% 18 3.8% 5.7% 8 1.7% NA 4 0.8% NA 8 1.7% 6.3%

Race

Westside

South Atlanta

(N=215)

(N=184)

Responses* Population Responses* Population

White

54 23% 4.7%

59 31%

19%

African American 163 69% 93% 116 61% 71%

Hispanic

6 2.6% 1.7%

3 1.6% 11%

Asian

3 1.3% NA

5 2.6% NA

Native American 7 3.0% NA

3 1.6% NA

Other

5 2.1% 2.7%

7 3.7% 9.8%

*Percentage of entire sample size, note that due to non-responses and respondents possibly

giving more than one answer percentages may not add up to 100%.

The employment status breakdown for each site is presented in Table 9. Eastside and Grant Park showed a larger percentage of respondents that work full-time, while South Atlanta and the Westside BeltLine had more sizable portions of respondents that do not work.

20

TABLE 9 Survey Respondents' Employment Status (Wave 1)

Employment Status
Full time Part time 2+ jobs Homemaker Don't work

Eastside* (N=428)

346

80%

32

7.4%

18

4.2%

11

2.5%

35

8.1%

Grant Park* (N=472)

354

74%

44

9.2%

19

4.0%

14

2.9%

59

12%

Employment Status

Westside* (N=223)

South Atlanta* (N=186)

Full time

102

43%

100

53%

Part time

30

13%

27

14%

2+ jobs

23

9.8%

12

6.3%

Homemaker

11

4.7%

5

2.6%

Don't work

71

30%

53

28%

*Percentage of entire sample size, note that due to non-responses and respondents possibly

giving more than one answer percentages may not add up to 100%.

Vehicle ownership data for each site is presented in Table 10. While South Atlanta and Westside show a sizable portion of respondents that do not own a vehicle, the majority of respondents had at least one car per household when considering all the study sites.

TABLE 10 Number of Vehicles Owned by Survey Respondents (Wave 1)

Vehicles per Household

Eastside* (N=428)

Grant Park* (N=471)

Westside* (N=223)

South Atlanta* (N=183)

0

21 4.8% 26 5.5% 41 17% 37 19%

1

194 45% 149 31% 92 39% 65 34%

2

170 39% 229 48% 63 27% 62 33%

3

28 6.5% 49 10% 18 7.7% 15 7.9%

4

12 2.8% 11 2.3% 6 2.6% 3 1.6%

5+

3 0.7% 7 1.5% 3 1.3% 1 0.5%

*Percentage of entire sample size, note that due to non-responses percentages may not add up to

100%.

21

Bicycle ownership for each site is represented in Table 11. As with vehicles, about 20% more of the respondents did not own any bicycles in South Atlanta and Westside compared to Eastside and Grant Park.

TABLE 11 Number of Bikes Owned by Survey Respondents (Wave 1)

Bikes per Household

Eastside* (N=428)

Grant Park* (N=472)

Westside* (N=220)

South Atlanta* (N=183)

0

105 24% 112 23% 99 42% 99 52%

1

129 30% 95 20% 59 25% 38 20%

2

113 26% 140 29% 38 16% 25 13%

3

36 8.3% 48 10% 10 4.3% 13 6.8%

4

26 6.0% 35 7.3% 11 4.7% 4 2.1%

5+

19 4.4% 42 8.8% 3 1.3% 4 2.1%

*Percentage of entire sample size, note that due to non-responses percentages may not add up to

100%.

The bike confidence levels stated the respondents are tabulated in Table 12. There was a greater percentage of respondents who could not bike in South Atlanta and Westside than Eastside and Grant Park.

TABLE 12 Respondents' Stated Bike Confidence Level (Wave 1)

Bike Confidence

Eastside* (N=430)

Grant Park* (N=473)

Westside* (N=222)

South Atlanta* (N=184)

Can't Bike

13 3.0% 21 4.4% 38 16% 25 13%

Not Very Confident 66 15% 69 14% 44 19% 40 21%

Somewhat Confident 119 27% 137 29% 43 18% 38 20%

Very Confident

232 54% 246 52% 97 41% 81 43%

*Percentage of entire sample size, note that due to non-responses percentages may not add up to 100%.

Summary Statistics Segmented by Rider Status

The same household characteristics were also computed based on segments of different rider status among the combined study group. The four rider statuses are potential rider, recreational,

22

utilitarian, and those that cannot bike. The criteria for inclusion in one of these categories comes from the responses to questions regarding bicycling confidence, cycling distances for recreation/utilitarian purpose, and cycling trip frequency for commute/other purposes. The four segments and their criteria are:
1. Potential cyclist (N=648)--those who report zero miles of cycling per month, but report being able to ride a bike, regardless of confidence level.
2. Recreational cyclist (N=329)--those who bike a non-zero distance per month, but do not bike more than once a month for utilitarian purposes.
3. Utilitarian cyclist (N=235)--those who bike more than once a month for utilitarian purposes and bike at least a mile a week, on average.
4. Cannot bike (N=97)--those who state that they cannot ride a bicycle. The statistics presented do not have a comparison to the population, as there is no readily available population-level data for rider type segmentation. Note that those who did not answer the bike confidence question were not included in the segmentation. The distribution of respondents in these segments is shown in Table 13.

TABLE 13 Distribution of Rider Segments by Neighborhood (Wave 1)

Rider Status

Eastside* (N=430)

Grant Park* (N=473)

Westside* (N=222)

South Atlanta* (N=184)

Potential

183 42% 222 47% 133 57% 110 58%

Recreational

120 28% 139 29% 33 14% 37 19%

Utilitarian

114 26% 91 19% 18 7.7% 12 6.3%

Can't Bike

13 3.0% 21 4.4% 38 16% 25 13%

*Percentage of entire sample size, note that due to non-responses percentages may not add up to 100%.

23

Income for each of these segments is presented in Table 14. Those who stated they cannot bike were drastically overrepresented by those in the lowest income categories. Conversely, recreational and utilitarian cyclists were vastly overrepresented by those in the highest income categories.

TABLE 14 Survey Respondents' Household Income by Rider Type (Wave 1)

Household Income

Potential* (N=630)

Recreational* (N=320)

Utilitarian* (N=232)

Cannot Bike* (N=90)

$15,000 or less $15,001 $30,000 $30,001 $50,000

39 6.0% 9 2.7% 6 2.6% 27 28% 69 11% 9 2.7% 7 3.0% 14 14% 66 10% 24 7.3% 22 9.4% 11 11%

$50,001 $75,000 94 15% 40 12% 31 13% 9 9.3%

$75,001 $100,000 89 14% 52 16% 38 16% 6 6.2%

$100,001 $125,000 60 9.3% 43 13% 36 15% 2 2.1%
More than $125,000 153 24% 122 37% 81 34% 5 5.2%
Prefer not to answer 60 9.3% 21 6.4% 11 4.7% 16 16%
*Percentage of entire sample size, note that due to non-responses percentages may not add up to 100%.

Distributions for household sizes by rider type are presented in Table 15. Single-person households appeared to make up the largest portions of respondents who could not bike. Larger households made up the majority in the other segments.

TABLE 15 Survey Respondents' Household Sizes by Rider Type (Wave 1)

Household Size

Potential* (N=640)

Recreational* (N=326)

Utilitarian* (N=232)

Cannot Bike* (N=90)

1

240 37% 96 29% 71 30% 45 46%

2

241 37% 138 42% 101 43% 31 32%

3

76

12%

44

13%

20 8.5%

6

6.2%

4

42 6.5% 27 8.2% 29 12%

5

5.2%

5+

27 4.2% 13 4.0%

6

2.6%

2

2.1%

*Percentage of entire sample size, note that due to non-responses percentages may not add up to 100%.

24

Residence types for each rider type are presented in Table 16. The residence type was pretty consistent across rider type, but most residences were detached.

TABLE 16 Survey Respondents' Residence Types by Rider Type (Wave 1)

Residence Type

Potential* (N=646)

Recreational* (N=328)

Utilitarian* (N=235)

Cannot Bike* (N=96)

Detached

410 63% 217 66% 137 58% 60 62%

Apt

81 13% 38 12% 37 16% 5 5.2%

Duplex

149 23% 71 22% 60 26% 28 29%

Other

6 0.9% 2 0.6% 1 0.4% 4 4.1%

*Percentage of entire sample size, note that due to non-responses percentages may not add up to 100%.

Responses for gender are reported by rider type in Table 17. Females made up the majority of the cannot bike segment but also the potential and recreational segment. Male riders represented the vast majority of utilitarian riders.

TABLE 17 Survey Respondents' Gender by Rider Type (Wave 1)

Gender

Potential* (N=637)

Recreational* (N=324)

Utilitarian* (N=231)

Cannot Bike* (N=96)

Female

402 62% 181 55% 88 37% 73 75%

Male

235 36% 143 43% 143 61% 23 24%

*Percentage of entire sample size, note that due to non-responses percentages may not add up to 100%.

Respondents' ages for each rider type are presented in Table 18. Not surprisingly, a large part of those who cannot bike are those 65 years old or older. Utilitarian cyclists are likewise more likely to be under 44. The other two rider types were most likely to be 3044.

25

TABLE 18 Survey Respondents' Age by Rider Type (Wave 1)

Age

Potential* (N=639)

Recreational* (N=328)

Utilitarian* (N=234)

Cannot Bike* (N=93)

<30

158 24% 93 28% 93 40% 2 2.1%

30-44

206 32% 147 45% 95 40% 12 12%

45-64

177 27% 67 20% 39 17% 34 35%

65+

98 15% 20 6.1% 7 3.0% 45 46%

*Percentage of entire sample size, note that due to non-responses percentages may not add up to 100%.

Respondents' race by rider type is presented in Table 19. Most of the respondents who cannot bike were African American, while the majority of all the other three rider statuses were White.

TABLE 19 Survey Respondents' Race by Rider Type (Wave 1)

Race

Potential* (N=623)

Recreational* (N=323)

Utilitarian* (N=227)

Cannot Bike* (N=96)

White

369 57% 245 74% 182 77% 24 25%

African American 222 34% 63 19% 25 11% 71 73%

Hispanic

16 2.5% 49 2.7% 9 3.8% 0 0.0%

Native American 6 0.9% 3 0.9% 4 1.7% 1 1.0%

Asian

19 2.9% 10 3.0% 8 3.4% 1 1.0%

Other

15 2.3% 3 0.9% 9 3.0% 2 2.1%

*Percentage of entire sample size, note that due to non-responses and respondents possibly giving more than one answer percentages may not add up to 100%.

Table 20 shows the employment status breakdown for each rider type group. As expected with the overrepresentation of senior adults in the "cannot bike" category, a majority of those in that category do not work. Potential, recreational, and utilitarian cyclists were also much more likely to work full-time.

26

TABLE 20 Survey Respondents' Employment Status by Rider Type (Wave 1)

Employment Status

Potential* (N=640)

Recreational* (N=326)

Utilitarian* (N=235)

Cannot Bike* (N=94)

Full time

425 66% 256 78% 192 82% 24 25%

Part time

73 11% 30 9.1% 20 8.5% 9 9.3%

2+ jobs

29 4.5% 20 6.1% 19 8.1% 2 2.1%

Homemaker

16 2.5% 15 4.6% 6 2.6% 3 3.1%

Don't work

116 18% 26 7.9% 13 5.5% 57 59%

*Percentage of entire sample size, note that due to non-responses and respondents possibly giving more than one answer percentages may not add up to 100%.

Vehicle and bike ownership broken down by rider types are presented in Table 21. Zero-vehicle households made up the majority in the group of respondents who cannot bike. Households with three or more vehicles made up the majority in the potential, recreational, and utilitarian rider groups, indicating that those who cannot bike are less likely to own more than one vehicle. Unsurprisingly, the majority of respondents who cannot bike do not own any bikes. Recreational and utilitarian cyclists were more likely to own more than one bike, but potential cyclists were still about as likely to have a bike as they were to not have one.

27

TABLE 21 Number of Vehicles and Bikes Owned by Survey Respondents by Rider Type (Wave 1)

Vehicles per Household
0 1 2 3 4 5+

Potential* (N=636)
55 8.5% 262 40% 260 40% 38 5.9% 13 2.0% 8 1.2%

Recreational* (N=324)
11 3.3% 105 32% 158 48% 38 12% 10 3.0% 2 0.6%

Utilitarian* (N=232)
14 6.0% 92 39% 88 37% 28 12% 6 2.6% 4 1.7%

Cannot Bike* (N=91)
37 38% 34 35% 13 13% 4 4.1% 3 3.1% 0 0.0%

Bikes per Household

Potential* (N=633)

Recreational* (N=325)

Utilitarian* (N=232)

Cannot Bike* (N=90)

0

304 47% 23 7.0% 3 1.3% 73 75%

1

161 25% 84 26% 64 27% 8 8.2%

2

114 18% 133 40% 62 26% 4 4.1%

3

28 4.3% 39 12% 37 16% 2 2.1%

4

22 3.4% 25 7.6% 23 9.8% 3 3.1%

5+

4 0.6% 21 6.4% 43 18% 0 0.0%

*Percentage of entire sample size, note that due to non-responses percentages may not add up to 100%.

Table 22 shows respondents' stated level of bike confidence, segmented by rider type. By definition, all those who stated they cannot bike are in the category of "cannot bike." Respondents of all confidence levels were present in the potential rider group. There are higher representations of more confident riders in both the recreational and utilitarian categories.

TABLE 22 Respondent's Stated Level of Confidence by Rider Type (Wave 1)

Confidence Level
Can't Bike Not Very Confident Somewhat Confident
Very Confident

Potential (N=648)
0 0.0% 200 31% 211 33% 237 37%

Recreational (N=329)
0 0.0% 18 5.5% 98 30% 213 65%

Utilitarian (N=235)
0 0.0% 1 0.4% 28 12% 206 88%

Cannot Bike (N=97)
97 100% 0 0.0% 0 0.0% 0 0.0%

28

First-Wave User Preference Analysis
Infrastructure Images
The images presented to respondents were created in Adobe Photoshop. One common roadway setting was chosen as a base image to control for urban environment, weather, and other contextual variables. Variations were based on different types of bicycle infrastructure, the presence or absence of on-street parking, and the number of automobile lanes (one versus two in each direction). Each scenario exhibited a moderate amount of automobile traffic that would allow for near free-flow conditions with a reasonable amount of opportunity for auto-to-cyclist interactions. The images were designed such that the background scenery could be related to by urban dwellers as an in-town neighborhood and by rural dwellers as a small town.
Seventeen total images were prepared, as shown in Figure 5 and Figure 6. The infrastructure includes sharrows, bike lanes, buffered bike lanes, and barrier-protected bike lanes (also referred to as separated bike lanes). Two of the protected bike lanes were one-way, while the other two were twoway. An image for a multi-use path was also created, though due to the nature of this type of infrastructure a different road environment had to be used.
For each image, respondents were given the prompt: "Bicycling on a road [trail] like this is...", with the sentence being completed in each of three ways (perceptions): "Comfortable," "Safe," and "Something I'd try." For each perception, they were asked to choose the most appropriate response on a 5-point Likert-type scale (Strongly Disagree, Disagree, Neutral or No opinion, Agree, or Strongly Agree). Respondents were randomly assigned one of four versions of the survey, each of which had a different combination of infrastructure images. Each version had a base road configuration (e.g., two lanes with on-street parking, or four lanes with no parking) for which a sequence of all four on-street infrastructure types were shown. Two other images were also included, from among the other road configurations and/or multi-use trails, so that each respondent was presented with six infrastructure combinations, and several were repeated between surveys.
29

FIGURE 5 Combinations of Bicycle Infrastructure Used in Survey Versions 1 and 2
30

FIGURE 6 Combinations of Bicycle Infrastructure Used in Survey Versions 3 and 4
31

Image Response Results Figure 7, Figure 8, and Figure 9 visually show the distribution of respondents' perceptions of
comfort, safety, and willingness to try, respectively. These figures are grouped so that for each lane combination each row is progressively more separated from traffic. The agreement with each perception markedly increases with each degree of separation from traffic and decreases with the addition of on-street parking. The differences between scenarios varying only by the number of lanes are subtler. More rigorous analysis is necessary to delve into the underlying patterns.
32

2LSH (335) 2LBL (659) 2LBB (662) 2L1C (336) 4LSH (299) 4LBL (298) 4LBB (627) 4L2C (297) 2PSH (327) 2PBL (682) 2PBB (325) 2P2C (325) 4PSH (361) 4PBL (655) 4PBB (352) 4P1C (357)
MU (992)

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Strongly Disagree Disagree Neutral Agree Strongly Agree

*2L=two lanes, 4L=four lanes, 2P=two lanes with parking, 4P=four lanes with parking, SH=sharrow, BL=bike lane, BB=buffered bike lane, 1C=one-way protected cycletrack, 2C=two-way protected cycletrack, and MU=multi-use path.
**Number in parentheses is the number of responses for the associated configuration
FIGURE 7 Distribution of Comfort Perceptions for Each Image (Wave 1)

33

2LSH (335) 2LBL (662) 2LBB (660) 2L1C (335) 4LSH (297) 4LBL (298) 4LBB (627) 4L2C (296) 2PSH (328) 2PBL (684) 2PBB (325) 2P2C (324) 4PSH (360) 4PBL (655) 4PBB (353) 4P1C (359)
MU (992)

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Strongly Disagree Disagree Neutral Agree Strongly Agree

*2L=two lanes, 4L=four lanes, 2P=two lanes with parking, 4P=four lanes with parking, SH=sharrow, BL=bike lane, BB=buffered bike lane, 1C=one-way protected cycletrack, 2C=two-way protected cycletrack, and MU=multi-use path.
**Number in parentheses is the number of responses for the associated configuration
FIGURE 8 Distribution of Safety Perceptions for Each Image (Wave 1)

34

2LSH (333) 2LBL (656) 2LBB (658) 2L1C (332) 4LSH (294) 4LBL (296) 4LBB (624) 4L2C (294) 2PSH (327) 2PBL (680) 2PBB (323) 2P2C (324) 4PSH (358) 4PBL (648) 4PBB (351) 4P1C (355)
MU (985)

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Strongly Disagree Disagree Neutral Agree Strongly Agree

*2L=two lanes, 4L=four lanes, 2P=two lanes with parking, 4P=four lanes with parking, SH=sharrow, BL=bike lane, BB=buffered bike lane, 1C=one-way protected cycletrack, 2C=two-way protected cycletrack, and MU=multi-use path.
**Number in parentheses is the number of responses for the associated configuration
FIGURE 9 Distribution of Willingness to Try Perceptions for Each Image (Wave 1)

Frequency
Multi-use paths were the most frequently used of all infrastructure types. The breakdown of reported frequency of biking on such a path is presented in Table 23. Over half of respondents reported biking on something similar at least sometimes or often. This is likely a representation that members of the general population are more likely to have biked on a multi-use path rather than on

35

an on-street infrastructure, especially in areas anticipating BeltLine extensions. For example, many people will bike for a one-time recreational event, but never develop the habit. This type of ride is much more likely to take place on a multi-use path than on any other type of infrastructure.
TABLE 23 Self-Reported Frequency of Use for Multi-use Paths

Multi-use Path

Responses Never
Sometimes Often
Not Sure

976 29% 36% 33% 1.8%

On-street facilities were biked less frequently. Table 24, Table 25, Table 26, and Table 27 display the reported frequencies for each infrastructure for two-lane roads without parking, two-lane roads with parking, four-lane roads without parking, and four-lane roads with parking, respectively. Roughly half of respondents report never having used bike lanes and roads with sharrows, though two-thirds to four-fifths (and higher) of the relevant samples report never using buffered or protected bike lanes.

TABLE 24
Self-Reported Frequency of Use for Each Infrastructure Type for Two-lane Roads without Parking

Responses Never
Sometimes Often
Not Sure

Sharrow
332 45% 31% 20% 3.6%

Bike Lane
655 38% 36% 22% 3.5%

Buffered Bike Lane
654 60% 23% 7.6% 8.6%

One-way Cycletrack
327 74% 14% 6.4% 5.2%

36

TABLE 25
Self-Reported Frequency of Use for Each Infrastructure Type for Two-lane Roads with Parking

Responses Never
Sometimes Often
Not Sure

Sharrow
329 51% 27% 16% 5.5%

Bike Lane
680 47% 31% 18% 4.4%

Buffered Bike Lane
322 70% 13% 4.3% 12%

Two-way Cycletrack
332 72% 19% 5.4% 3.9%

TABLE 26
Self-Reported Frequency of Use for Each Infrastructure Type for Four-lane Roads without Parking

Sharrow

Responses Never
Sometimes Often
Not Sure

290 47% 28% 21% 3.4%

Bike Lane
288 36% 40% 20% 3.5%

Buffered Bike Lane
607 63% 23% 6.3% 8.6%

Two-way Cycletrack
289 70% 21% 4.8% 4.5%

TABLE 27
Self-Reported Frequency of Use for Each Infrastructure Type for Four-lane Roads with Parking

Responses Never
Sometimes Often
Not Sure

Sharrow Bike Lane

355 53% 29% 16% 2.0%

641 53% 31% 11% 5.0%

Buffered Bike Lane
341 66% 24% 4.7% 5.3%

One-way Cycletrack
350 74% 18% 3.7% 4.3%

37

User Preference Models As previously described, survey respondents were presented with different configurations of
roadway characteristics and infrastructure types, and asked to state their perceived levels of comfort, safety, and willingness to try the presented infrastructure. Responses were converted to numeric values, with Strongly Disagree equal to 1 and Strongly Agree equal to 5. The average ratings for comfort, safety, and willingness to try are presented in Figure 10, Figure 11, and Figure 12, respectively. Each version of the survey focused on the continuum of the four infrastructure types within the same traffic lane and parking lane combination, plus two additional images duplicated from the other survey versions. To avoid the potential framing effects introduced by the insertion of these additional images "out of sequence," only the responses for the four in-sequence images are included in the descriptive analysis presented here (sample sizes of between 266 and 308 responses for each mean); all responses are included in the regression analysis reported below.
The characteristics of the bicycle infrastructure portion of the roadways for the sharrow, bike lane, and buffered bike lane cases were consistent between roadway configurations. However, protected bike lanes had two variations, one-way and two-way, only one of which was presented for a given configuration in order to limit the number of images presented. The broken lines on the graphs show the point in the progression of bicycle infrastructure where barrier-protection is introduced, and two different protected bicycle infrastructure types are portrayed. The two-lane/no parking and four-lane with parking configurations had one-way protected bike lanes (indicated by the dotted line), while the four-lane/no parking and two-lane with parking arrangements had twoway protected bike lanes (indicated by the dash-dot lines). Given the close clustering of the four means for this infrastructure type, the figures indicate that the differences in ratings between protected bike lane scenarios may be unrelated to roadway characteristics.
38

4.50

4.00

3.50

3.00

2.50

2.00 Sharrow

Bike Lane

*Note that axis does not start at 0

Buffered Bike Protected Bike

Lane

Lane

FIGURE 10
Average Expressed Comfort Levels for Each Lane/Parking Configuration by Bicycle Infrastructure Type

4.50

4.00

3.50

3.00

2.50

2.00 Sharrow

Bike Lane

*Note that axis does not start at 0

Buffered Bike Protected Bike

Lane

Lane

FIGURE 11
Average Expressed Safety Levels for Each Lane/Parking Configuration by Bicycle Infrastructure Type

39

4.50

4.00

3.50

3.00

2.50

2.00 Sharrow

Bike Lane

*Note that axis does not start at 0

Buffered Bike Protected Bike

Lane

Lane

FIGURE 12
Average Expressed Level of Willingness to Try for Each Lane/Parking Configuration by Bicycle Infrastructure Type

Ratings for these three different measures tended to follow the same patterns. This indicates that respondents did not make much distinction between the different questions (comfort versus safety versus willingness to try) for each image, which may result, for example, from a lack of experience that would allow one to rate a given infrastructure as safe but not comfortable, or vice versa.
Each of the three measures improved for each increased degree of separation provided by the bicycling infrastructure, indicating a positive benefit associated with separation from moving and parked cars. Each version of the survey began the infrastructure image section with a sharrow configuration, which allows the sharrow infrastructure layouts to serve as a base measurement for each lane configuration. In each version, the sharrow configurations received the lowest ratings, and the existence of any sort of spatial separation was influential in increasing each perception measure. Average ratings for each traditional bike lane scenario were higher than those for sharrows on the same roadway configuration. The difference is more pronounced for bicycle lanes without

40

adjacent curb parking. Buffered bike lanes received higher average ratings than traditional bike lanes, and also saw the same disutility of parking lanes.
As previously mentioned, two different protected bike lane scenarios were tested in the survey. Table 28 shows the average ratings for each of the protected bike lane scenarios along with the multi-use path. The presence of the barrier was effective in overcoming the obstacles created by the inclusion of parking or extra traffic lanes. The differences between perceptions of protected facilities appeared to be more related to whether the facility was one-way or two-way than the configuration of the rest of the roadway. The multi-use path received ratings comparable to those of the one-way and two-way protected bike lanes.
TABLE 28 Average Ratings for Comfort, Safety, and Willingness to Try
for Protected Bike Lanes and Multi-Use Paths

Comfort

One-way Protected

Two-Lane/ Four-Lane No Parking with Parking

3.64

3.92

Two-way Protected

Two-Lane Four-Lane/ with Parking No Parking

4.25

3.42

Multi-Use Path
4.10

Safety

3.59

3.91

4.20

3.30

4.07

Willingness to Try

3.79

3.94

4.28

Note: (1=Strongly Disagree, 5=Strongly Agree)

3.70

3.89

Infrastructure and Roadway Trait Models

While the descriptive analysis of the preceding subsection is useful, it is also desirable to control for a number of covariates whose effects might otherwise be confounded with those of infrastructure type and roadway configuration. Linear regression models were built using the multiple responses by 1,335 respondents for each of the three dependent variables (comfort, safety, and willingness to try), as presented in Table 29. Dummy variables for each infrastructure type, along with the presence of on-street parking and additional lanes of traffic, were included in the

41

models. Although linear regression models have limitations for application to Likert-type data, they can serve as a reliable approximation with four or more ordinal response levels with "little worry."1
An issue resultant from the survey design was the emergence of a framing effect. Six of the seventeen images appeared on more than one version of the survey. One of those six images was the multi-use path, which had consistent scores in every version where it appeared. The other five saw more variance in responses between versions. Specifically, these images attracted different responses when they were out-of-sequence (e.g., the "two-lane/no parking bike lane" image in Version 1 of Figure 5) than when they were in-sequence (the same image in Version 2).
Dummy variables were included in the regression to capture the variation due to the framing effects of the preceding image--specifically, the interaction effects occurring when the bicycle infrastructure type changed at the same time as the removal of parking or extra lanes of traffic. Three such variables were created: Bike Lane (BL)-No Parking, Buffered Bike Lane (BBL)-No Parking, and BL-Two Lanes. The BL-No Parking variable was set to 1 for the second image in Version 1, which added a bike lane and removed parking compared to the preceding image; the BBL-No Parking variable was set to 1 for the two-lane buffered bike lane image in Version 1 along with the four-lane buffered bike lane in Version 4, both of which added a buffer to the bike lane and removed parking compared to the preceding image; and the BL-Two Lanes variable was set to 1 for the second image in Version 3, which introduced a bike lane and removed the additional lanes of traffic compared to the preceding image. A fourth dummy variable was also considered for the two-lane one-way protected bike lane without parking image in Version 2; however, this variable was eventually excluded because it undermined the stability of the model, perhaps due to empirical collinearity issues related to the infrequent appearance of one-way protected bike lanes.
1 Bentler, P.M., and C.-P. Chou (1987). Practical issues in structural modeling. Sociological Methods and Research 16, 78117.
42

TABLE 29 Linear Regression for Expressed Comfort, Safety, and Willingness to Try, Including only Infrastructure Characteristics

43

Variable Constant

Comfort

Coefficient P-value

2.83 ***

<0.001

Safety Coefficient 2.54 ***

P-value <0.001

Willingness to Try

Coefficient P-value

3.25 ***

<0.001

Bicycle Infrastructure Types

Bike Lane (BL) Buffered BL (BBL) One-way Protected Two-way Protected Multi-Use

0.61 *** 1.02 *** 1.67 *** 1.46 *** 1.44 ***

<0.001 <0.001 <0.001 <0.001 <0.001

0.64 *** 1.16 *** 2.01 *** 1.79 *** 1.69 ***

<0.001 <0.001 <0.001 <0.001 <0.001

0.39 *** 0.67 *** 1.23 *** 1.09 *** 1.08 ***

<0.001 <0.001 <0.001 <0.001 <0.001

Roadway Characteristics

Parking Four Lanes

-0.21 *** -0.06 *

<0.001 -0.23 *** 0.022 -0.03

<0.001 -0.23 *** 0.228 -0.13 ***

<0.001 <0.001

Framing Effects

BL-No Parking

0.26 ***

<0.001 0.42 ***

<0.001 0.16 *

BBL-No Parking

0.19 ***

<0.001 0.30 ***

<0.001 0.12 *

BL-Two Lanes

0.18 **

0.004 0.25 ***

<0.001 0.09

# of Responses

7889

7890

7838

R2

0.236

0.312

0.121

Adjusted R2

0.235

0.311

0.120

*Significant at P = 0.050 or better; **Significant at P = 0.010 or better; ***Significant at P < 0.001

0.032 0.040 0.280

As shown in Table 29, the dummy variables for each infrastructure type were significant. The coefficients for each of the on-street infrastructure variables (BL, BBL, and Protected Lanes) were also significantly different from each other, supporting the earlier finding that greater separation of cyclists from cars increases all three measures of effectiveness. The multi-use dummy coefficient was not substantially different from the protected bike lane coefficients; however, it was still included separately in the model because the multi-use images excluded the effects of roadway characteristic variables.
The framing effect terms were significant in each model. These variables show sensitivity to the comparative removal of a perceived negative aspect (i.e., parking or additional travel lane) that is not explained by the variables, indicating the absence of that aspect alone. For example, when an image without parking was presented after an image with parking, it tended to receive a higher rating than if it were preceded by an image that also had no parking.
While the framing variables picked up the influence of multiple simultaneous changes from image to image, the "Parking" and "Four Lanes" variables represented the overall effects of roadway characteristics. The parking variable was significant in all models, indicating that the overall effect of parking was still significant, even after accounting for the strong impact of the removal of parking in the few images affected by framing. Interestingly, the variable for the number of traffic lanes alone was not consistently significant between models. The coefficients are negative in each model, though with a lower magnitude than the parking coefficient, likewise leading to reduced levels of significance. The insignificance in the Safety model is accompanied by a highly significant framing variable, while the highly significant coefficient of Four Lanes in the Willingness to Try model is accompanied by an insignificant framing variable.
Additional Influence of Sociodemographic Traits
Sociodemographic data was also collected using the survey instrument. The influence of covariates such as demographic and other characteristics on the perceptions of interest is
44

additionally informative in its own right. The previous linear models were supplemented with sociodemographic data, as presented in Table 30. As explained previously regarding imputing data, for the few cases where this information was not reported, data obtained from targeted marketing data sources was used as an estimate. Each model was estimated step-wise, with insignificant sociodemographics being dropped from the model, while judgment was used for inclusion of borderline-significant variables that were significant in other models. The best of each model is presented in Table 30. In all three models, age, student-status, and gender were significant with similar signs between models. Older individuals rated scenarios lower in general, as did full-time students and women.
45

TABLE 30 Linear Regression for Expressed Comfort, Safety, and Willingness to Try by Infrastructure and Individual Characteristics

46

Variable

Comfort

Coefficient

P-value

Safety

Coefficient

P-value

Willingness to Try Coefficient P-value

Constant

3.40 ***

<0.001

2.84 ***

<0.001

4.54 ***

<0.001

Bicycle Infrastructure Types

Bike Lane (BL) Buffered BL (BBL) One-way Protected Two-way Protected Multi-Use

0.62 *** 1.03 *** 1.68 *** 1.47 *** 1.46 ***

<0.001 <0.001 <0.001 <0.001 <0.001

0.65 *** 1.18 *** 2.03 *** 1.81 *** 1.73 ***

<0.001 <0.001 <0.001 <0.001 <0.001

0.39 *** 0.66 *** 1.23 *** 1.09 *** 1.11 ***

<0.001 <0.001 <0.001 <0.001 <0.001

Roadway Characteristics

Parking Four Lanes

-0.21 *** -0.04

<0.001 0.110

-0.23 *** -0.01

<0.001 0.62

-0.22 *** -0.07 *

<0.001 0.017

Framing Effects

BL-No Parking BBL-No Parking BL-Two Lanes

0.26 *** 0.19 *** 0.22 ***

<0.001 <0.001 <0.001

0.42 *** 0.31 *** 0.28 ***

<0.001 <0.001 <0.001

0.16 * 0.13 * 0.16 *

0.019 0.014 0.036

Sociodemographics

Age Full-Time Student Driver's License Asian African American Hispanic Other Female Children in Home

-0.01 *** -0.30 ***
-0.16 *
0.17 **
-0.16 ***

<0.001 <0.001
0.017
0.004
<0.001

-0.01 *** -0.19 **
0.17 ** 0.12
0.11
-0.12 ***

<0.001 0.004 0.003 0.072
0.098
<0.001

-0.02 *** -0.29 ***
-0.28 *** -0.36 ***
0.22 ** -0.26 * -0.36 ***
0.08 **

<0.001 <0.001
<0.001 <0.001
0.001 0.010 <0.001 0.008

# of Responses R2 Adjusted R2

7721 0.263 0.261

7703 0.333 0.331

7682 0.237 0.236

*Significant at P = 0.050 or better; **Significant at P = 0.010 or better; ***Significant at P < 0.001

The only race/ethnicity variables that were significant in the comfort model were Asians and Hispanics. The negative coefficient for Asians indicates that this group generally views infrastructure as less comfortable, while Hispanics view infrastructure as more comfortable, all else equal.
The coefficient for holding a driver's license was significant only in the safety model. The positive coefficient for driver's license may indicate that those with a license feel more control over the safety of the roadway in general. The coefficients for Asians and Hispanics were borderline significant in this model.
In addition to the coefficients for Asians and Hispanics, the coefficients for African Americans and Other races were also significant in the model for willingness to try. The remaining two race/ethnicity options not represented are Native American (which had only 2 respondents) and White (which essentially acts as the base). In each sizable ethnic/racial group other than White the coefficient was negative, indicating a general lack of willingness to try cycling for other groups, apart from Hispanics.
Sociodemographic characteristics seemed to play a larger role in the willingness to try model than for the other two perceptions, as seen by the increase in the R2 value from 0.121 (Table 29) to 0.237 (compared to increases of 0.027 and 0.021, respectively, for the other two models). This indicates that individual characteristics have a stronger relationship to potential users' decisions of whether to use a certain type of infrastructure than to their perceptions of whether it is safe or comfortable in general.
Segmented Models: Ridership Status A segmented model was developed to investigate how the influence of the other explanatory
variables differs by rider group. The sample was segmented using the previous criteria for rider statuses of potential rider, recreational, utilitarian, and those that cannot bike:
47

1. Potential cyclist (N=648)--those who report zero miles of cycling per month, but report being able to ride a bike, regardless of confidence level.
2. Recreational cyclist (N=330)--those who bike a non-zero distance per month, but do not bike more than once a month for utilitarian purposes.
3. Utilitarian cyclist (N=234)--those who bike more than once a month for utilitarian purposes and bike at least a mile a week, on average.
4. Cannot bike (N=97)--those who state that they cannot ride a bicycle. The potential cyclist population was used as the base, and incremental-difference coefficients were reported for segments with significant differences from the base group. Not all segments were significantly different from the base in each model. Each segmented model started from the previously reported ordinary least squares models for comfort, safety, and willingness to try, respectively. Dummy variables were introduced for the "recreation," "utilitarian," and "cannot bike" segments, using the "potential cyclists" as the base. The incremental effects for each segment were estimated using interaction terms between the main effect explanatory variables and the segment dummy variables, piecewise removing insignificant variables (constraining them to be 0). Insignificant variables were included in cases with borderline significance, where a main effect was insignificant but an associated interaction effect was significant, and/or in cases where the coefficient is necessary for interpretation of a similar variable, such as for different types of bicycle infrastructure. A segmented model for expressed comfort is presented in Table 31. Those unable to bike had negative coefficients for each infrastructure type, indicating a neutralizing effect on the positive main effects. This presents a problem for models that include those unable to bike, as they serve as a dampening effect on measurements of perceptions of target cyclists. Utilitarian cyclists also had slightly negative coefficients for one-way protected bike lanes and multi-use paths, indicating that recreational and potential cyclists are the segments driving positive perceptions of comfort for these more protected infrastructure types. The negative coefficient of the four lanes variable for
48

recreational cyclists is indicative that this segment may be a driving force for the sometimes significantly negative effect of extra lanes of traffic.
A segmented model for expressed safety is presented in Table 32. Like the previous model, those unable to bike had compensatory negative coefficients for bicycle infrastructure variables. Utilitarian cyclists and recreational cyclists had positive coefficients for two-way protected bike lanes, with recreational cyclists also having a positive coefficient for multi-use path. In addition to recreational cyclists having a negative coefficient for the number of vehicular lanes, utilitarian cyclists also had a similar coefficient in this model. This discrepancy may indicate a difference in perceptions of comfort and safety for utilitarian cyclists in terms of riding with more lanes of traffic.
A segmented model for expressed willingness to try is presented in Table 33. Notably, the only roadway characteristics to be significant in any segmentation were the parking and four lanes variables for those unable to bike. Both were positive, with higher magnitudes than the negative base coefficients, implying that those who cannot bike express a greater willingness to try in the presence of parking and additional traffic lanes, and are even more likely to express it than other groups. The change in sign for these coefficients was unexpected; however, based on the rather large magnitude of the negative constant term for that group, it is important to note that this group is still substantially less willing to try in comparison to the other groups. The coefficients for age are all significant and have similar magnitudes, with only the base being negative. This indicates that age is a deterrent for those in the potential cyclist group, but does not have a significant effect among the recreational, utilitarian, and unable groups.
49

TABLE 31 Linear Regression for Expressed Comfort by Infrastructure and Individual Characteristics, Segmented by Rider Type

50

Variable Constant

Main Effects
3.06

P-value <0.001 ***

Recreation 0.19

P-value <0.001 ***

Incremental Effects

Utilitarian P-value

0.47

0.012 *

Unable 0.20

P-value 0.111

Bicycle Infrastructure Types

Bike Lane Buffered Bike Lane One-way Protected Two-way Protected Multi-use

0.63 <0.001 ***

1.05 <0.001 ***

1.75 <0.001 ***

1.49 <0.001 ***

0.18

1.54 <0.001 ***

0.033 *

-0.20 -0.20

0.010 ** 0.008 **

-0.20 -0.28 -0.48 -0.69 -0.54

0.185 0.077 0.009 ** 0.004 ** 0.003 **

Roadway Characteristics

Parking Four Lanes

-0.21 <0.001 *** -0.002 0.438

-0.14

0.011 *

Framing Effects

BL-No Parking BBL-No Parking BL-Two Lanes

0.26 <0.001 *** 0.20 <0.001 *** 0.23 <0.001 **

Sociodemographics

Age Asian Hispanic Student (full-time)
Note: 7,659 Responses

-0.007 <0.001 ***

-0.17 0.011 *

0.95

<0.001 ***

0.22

0.003 ***

-0.31

0.011 *

-0.20 0.002 ** *Significant at P = 0.050 or better; **Significant at P = 0.010 or better; ***Significant at P < 0.001; R2=0.212; Adj R2=0.210

TABLE 32 Linear Regression for Expressed Safety by Infrastructure and Individual Characteristics, Segmented by Rider Type

51

Variable

Main Effects

P-value

Recreation P-value

Incremental Effects Utilitarian P-value

Unable P-value

Constant

2.53

<0.001 ***

0.10

0.010 *

0.35

<0.001 *** 0.16

0.212

Bicycle Infrastructure Types

Bike Lane (BL)

0.66

<0.001 ***

-0.16

0.305

Buffered BL (BBL)

1.20

<0.001 ***

-0.26

0.095

One-way Protected

2.08

<0.001 ***

-0.65 <0.001 ***

Two-way Protected Multi-use

1.76

<0.001 ***

0.29

1.73

<0.001 ***

0.16

0.001 **

0.18

0.040 *

0.055

-0.59 -0.55

0.013 * 0.003 **

Roadway Characteristics

Parking

-0.23

<0.001 ***

Four Lanes

0.05

0.099

-0.13

0.021 *

-0.13

0.026 *

Framing Effects

BL-No Parking BBL-No Parking BL-Two Lanes

0.43

<0.001 ***

0.31

<0.001 ***

0.28

<0.001 ***

Sociodemographics

Age

-0.007

<0.001 ***

Female

-0.72

0.002 **

Driver's License

0.24

<0.001 ***

Student (full-time) Hispanic Asian
Note: 7,639 Responses

-0.07

0.275

-0.40

0.042 *

0.16

0.034 *

-0.32

0.040 *

-0.13

0.065

0.63

0.002 **

*Significant at P = 0.050 or better; **Significant at P = 0.010 or better; ***Significant at P < 0.001; R2=0.3478; Adj R2=0.3451

TABLE 33
Linear Regression for Expressed Willingness to Try by Infrastructure and Individual Characteristics, Segmented by Rider Type

52

Variable

Main Effects P-value

Recreation P-value

Incremental Effects Utilitarian P-value

Unable P-value

Constant

3.74

<0.001 ***

-0.65

0.004 **

-0.20

0.428

-1.89 <0.001 ***

Bicycle Infrastructure Types

Bike Lane (BL) Buffered BL (BBL) One-way Protected Two-way Protected Multi-use

0.32

<0.001 ***

0.59

<0.001 ***

1.15

<0.001 ***

1.02

<0.001 ***

1.19

<0.001 ***

Roadway Characteristics

Parking Four Lanes

-0.21 -0.05

<0.001 *** 0.152

0.37 <0.001 ***

0.24

0.014 *

Framing Effects

BL-No Parking BBL-No Parking BL-Two Lanes

0.44

<0.001 ***

0.25

<0.001 ***

0.18

0.043 *

Sociodemographics

Age Female African American Education Vehicles per Driver
Note: 7,619 Responses

-0.009

<0.001 ***

0.008

0.009 **

0.009

0.032 *

0.009

0.015 *

-0.19

<0.001 ***

-0.16

<0.001 ***

0.62 <0.001 ***

0.03

0.012 *

0.15

<0.001 ***

-0.48

<0.001 ***

0.50

0.016 *

0.40

0.026 *

*Significant at P = 0.050 or better; **Significant at P = 0.010 or better; ***Significant at P < 0.001; R2=0.327; Adj R2=0.323

First-Wave Survey Conclusions Results from the first-wave survey suggest similar trends between perceived comfort, safety,
and willingness to try infrastructure. Respondents responded more positively to images containing bicycle facilities providing a higher degree of separation from drivers, with protected bike lanes and multi-use paths being the best. Parking was a clear deterrent for all measures of perception/preference, while an increase in the number of automobile lanes did not appear to negatively affect perceptions. Protected bike lanes seemed effective in reducing the negative effects of parking and traffic lanes.
Linear regression models were used to predict stated preferences for perceived comfort, safety, and willingness to try bicycle infrastructure. The estimated coefficients for the bicycle infrastructure variables were significantly positive and significantly different from each other in each model, implying a significant difference between each type of infrastructure type on perceptions. For the pooled sample, the variable for parking was significantly negative, though the variable for the number of lanes of traffic was not significant. Framing effects were also accounted for in the regression models, where images that removed parking or an extra lane of travel (compared to the previous image shown) were given a dummy variable to capture the relative changes in perception from image to image associated with the order in which the images were presented. Each of these variables was significant.
User characteristics were significant in explaining variations in comfort, safety, and willingness to try. The addition of sociodemographic information was more influential in improving explanatory power for the willingness to try model than for the other two dependent variables. Age, gender, and student status were significant in all models, with older individuals, females, and full-time students having a decreased perception of comfort, safety, and willingness to try cycling, all else equal. Other characteristics were only significant in some models.
53

Perceptions were also modeled using segmentations based on rider types, including potential cyclists, recreational cyclists, utilitarian cyclists, and those unable to bike. These models also saw a comparatively larger impact on the willingness to try model than on the other two perceptions. Those who are unable to bike had a number of coefficients that consistently differed from the rest of the sample, indicating the need to exercise caution in including the perceptions of members of this group with the rest of the population. Age was positive and statistically significant in the willingness to try model for all segments except the base, essentially cancelling out the influence of the main effects of age, indicating that the overall effects of age are not substantial for those that either bike currently or are unable to bike anyway.
54

Second-Wave Survey Description
The intent of the second survey was to repeat observations from the first-wave survey and to assess perceptions and recognitions of any changes that may have occurred. Key questions from the first-wave survey were repeated verbatim, with other questions being removed in favor of brevity. The survey was pretested with graduate students and members of the public. Both an online version and a paper version were prepared. The resulting survey (which can be found in Appendix B) was 9 pages, taking approximately 20 minutes to complete, and contained four sections, including:
A. Attitudes B. Daily travel C. Bicycling experience D. Demographics With the ever-changing nature of some transportation systems, the researchers wanted to gauge the general perceptions of changes in transportation in each neighborhood, including for automobiles, transit, walking, and biking. This also helped the research team avoid leading respondents about specific changes, and provide a reasonable basis for comparing perceptions of bicycle infrastructure. A general question on perceptions was included to fulfill this purpose, as presented below:
55

In addition to general perceptions, the research team also wanted to measure recognition of changes in bicycle facilities. The goal was to measure recognition of the addition of any bicycling facility as well as properly identify what facility was added. Recognition in treatment sites would be compared to those of the respective control sites (which have not received bicycling facilities during the study period). From this data, models would be estimated to explain differences in recognition.
Parallel to the questions of recognition, the researchers also asked respondents whether they have used the new bike facilities and if they like them. These responses would also be compared between treatment and control pairs and models would be developed to predict usage of and sentiments toward new facilities. These questions are reproduced below:
56

The survey was deployed in May 2018 and responses were collected throughout the summer. The invitation list for the second-wave survey was composed of all respondents from the first wave. Printed versions of the survey were mailed to all on the list. Additionally, email invitations with a URL to take the survey online were sent to all subjects who had provided an email address. The research team provided a 1-800 number and email address to field questions or comments from respondents. Each paper survey was entered (coded) twice and the two datasets were compared to ensure no coding errors were introduced in the data-entry process.
As is typical for panel surveys, the response rate for the second-wave survey was much higher than the first wave. This is generally the case due to the weeding out effect of the first-wave survey. The total number of responses for each neighborhood (after removing severely incomplete responses) for both waves is presented in Table 34.
57

TABLE 34 Survey Responses for Waves 1 and 2 for each Neighborhood

Area

Households Invited

Eastside*

4,509

Grant Park

4,411

Westside*

5,035

South Atlanta

3,815

Total

17,770

*Indicates treatment location

Initial Responses
433 477 235 190 1335

Initial Rate
9.6% 10.8% 4.7% 5.0% 7.5%

Follow-up Responses
231 265 108 109 713

Follow-up Rate
53% 56% 46% 57% 53%

58

Multi-wave Demographic Statistics

Summary Statistics for Matched Respondents

This section contains a summary of the demographics for those individuals who responded to both waves of the survey. Although second-wave invitations were sent only to those who responded to the first wave, there was no way the researchers could force the same individual or member of the household to respond to each wave. To determine whether a second-wave respondent matched with a first-wave respondent, the research team checked for consistency of gender, age (accounting for the passage of time), and race/ethnicity. Only those that matched on all three criteria were identified as a matched respondent. Out of the 713 responses, 612 were from matched respondents. The summaries reported in this section are for only matched respondents, with summaries from 2017 and 2018, where applicable. For complete summaries, see Appendix C.
Distributions for gender are presented in Table 35. Each site was slightly overrepresented by females, as is typical for mail-out/mail-back surveys.

Gender
Female Male

TABLE 35 Genders of Respondents of both Wave 1 and Wave 2

Eastside (211)
54% 46%

Grant Park (225)
58% 42%

Westside (82)
61% 39%

South Atlanta (94)
52% 48%

The ages (in 2018) of respondents are presented in Table 36. Eastside had a somewhat larger portion of younger respondents than Grant Park, while South Atlanta was younger than Westside.

59

Age
1834 3549 5064 65+

TABLE 36 Ages of Respondents of both Wave 1 and Wave 2

Eastside (211)
30% 43% 18% 9.0%

Grant Park (224)
22% 38% 25% 14%

Westside (81)
10% 28% 28% 33%

South Atlanta (93)
18% 30% 32% 19%

The race and/or ethnicity of respondents is presented in Table 37. Note that respondents were instructed to select all options that apply, so percentages may exceed 100%.

TABLE 37 Races of Respondents of both Wave 1 and Wave 2

Race/Ethnicity
White African American
Hispanic Asian
Native American Other

Eastside (206)
81% 13% 3.4% 6.3% 0.0% 1.0%

Grant Park (222)
83% 13% 2.3% 1.4% 1.8% 1.8%

Westside (78)
22% 72% 5.1% 0.0% 0.0% 3.8%

South Atlanta (93)
38% 59% 1.1% 2.2% 1.1% 3.2%

For the remainder of the demographic statistics presented in this section, values are reported for both 2017 and 2018 responses to show how these characteristics may have changed. Values in parentheses are the number of responses, which may vary due to item non-response in one but not both of the survey years. Household incomes are presented in Table 38. There are minor fluctuations between income groups for each site, but the overall distributions are rather consistent.

60

TABLE 38 Household Incomes of Respondents of both Wave 1 and Wave 2

Household Income
$15,000 or less $15,001 $30,000 $30,001 $50,000 $50,001 $75,000 $75,001 $100,000 $100,001 $125,000 More than $125,000

Eastside

2017 (192) 3.6% 4.2% 6.3% 16% 18% 15% 38%

2018 (196) 2.0% 2.6% 7.1% 14% 15% 18% 41%

Grant Park

2017 (205) 2.4% 4.9% 11% 16% 14% 15% 37%

2018 (200) 2.5% 4.0% 10% 14% 15% 16% 39%

Westside

2017 (78) 21% 22% 13% 12% 21% 5.1% 7.7%

2018 (70) 14% 19% 24% 11% 13% 13% 5.7%

South Atlanta

2017 (82) 13% 20% 13% 15% 18% 3.7% 17%

2018 (76) 13% 13% 12% 20% 17% 5.3% 20%

Household sizes reported for each wave are presented in Table 39. There are also minor fluctuations here, but again, the overall distributions are consistent between years.

TABLE 39 Household Sizes of Respondents of both Wave 1 and Wave 2

Household Size
1 2 3 4 5+

Eastside

2017 (205) 40% 41% 11% 4.9% 2.0%

2018 (205) 43% 41% 10% 4.9% 1.0%

Grant Park

2017 (220) 32% 42% 13% 11% 1.8%

2018 (221) 34% 41% 12% 10% 2.3%

Westside

2017 (78) 41% 37% 5.3% 11% 5.3%

2018 (75) 39% 39% 8.9% 5.1% 7.6%

South Atlanta

2017 (89) 45% 33% 13% 4.7% 4.7%

2018 (85) 44% 30% 15% 4.5% 6.7%

Residence type for each wave is reported in Table 40. Very few discrepancies exist between years, indicating that the overwhelming majority of respondents did not move between survey waves, or if they did, they at least moved to a similar residence type as before.

61

TABLE 40 Residence Types of Respondents of both Wave 1 and Wave 2

Residence Type
Detached Duplex
Apt Other

Eastside

2017 (211) 45% 16% 38% 0.5%

2018 (211) 40% 18% 42% 0.9%

Grant Park

2017 (225) 75% 12% 12% 0.4%

2018 (225) 72% 16% 12% 0.0%

Westside

2017 (81) 76% 6.1% 16% 1.2%

2018 (82) 77% 8.6% 14% 1.2%

South Atlanta

2017 (93) 70% 7.4% 17% 3.2%

2018 (94) 72% 6.5% 19% 2.2%

Employment status in 2017 and 2018 is presented in Table 41. The share of respondents who don't work decreased between waves, indicating that more people gained employment than lost employment or retired. Note that respondents were instructed to select all that apply, so percentages may exceed 100%.

TABLE 41 Employment Status of Respondents of both Wave 1 and Wave 2

Employment Status
Full time Part time 2+ jobs Homemaker Don't work

Eastside

2017 (211) 81% 4.3% 4.7% 3.3% 10%

2018 (211) 81% 7.1% 3.8% 2.4% 7.1%

Grant Park

2017 (225) 66% 12% 5.3% 6.2% 20%

2018 (225) 70% 12% 4.4% 2.2% 15%

Westside

2017 (82) 37% 9.8% 8.5% 2.4% 51%

2018 (82) 35% 16% 9.8% 1.2% 43%

South Atlanta

2017 (94) 51% 13% 4.3% 3.2% 28%

2018 (94) 57% 12% 7.4% 3.2% 27%

The number of vehicles per household is presented in Table 42. Vehicle ownership appears to be relatively stable on the aggregate between the two survey waves.

62

TABLE 42 Number of Vehicles Owned per Household of Respondents of both Wave 1 and Wave 2

Number of Vehicles
0 1 2 3 4 5+

Eastside

2017 (209) 5.8% 41% 46% 5.3% 1.4% 0.5%

2018 (208) 6.7% 45% 39% 6.7% 1.4% 1.0%

Grant Park

2017 (225) 5.8% 33% 49% 9.4% 1.8% 0.9%

2018 (224) 5.8% 33% 49% 8.4% 1.3% 1.3%

Westside

2017 (81) 18% 39% 30% 9.1% 2.6% 1.3%

2018 (77) 16% 44% 26% 8.6% 3.7% 1.2%

South Atlanta

2017 (91) 18% 36% 38% 3.4% 3.4% 1.1%

2018 (87) 18% 35% 34% 8.8% 3.3% 1.1%

The number of bikes per household is presented in Table 43. The consistency of bike ownership between waves indicates that the impact of the year between surveys and the treatment itself do not have a measurable impact on bike ownership, which helps remove access to a bike as a potential causal channel of any changes in perceptions or behavior.

TABLE 43 Number of Bikes Owned per Household of Respondents of both Wave 1 and Wave 2

Number of Bikes
0 1 2 3 4 5+

Eastside

2017 (209) 27% 28% 25% 10% 4.8%

2018 (207) 28% 29% 26% 7.7% 3.8%

5.3% 4.8%

Grant Park

2017 (225) 25% 19% 29% 10% 6.3%

2018 (224) 24% 19% 31% 11% 5.8%

10% 9.3%

Westside

2017 (76) 49% 26% 17% 3.9% 1.3%

2018 (80) 45% 28% 16% 7.5% 2.5%

2.6% 1.3%

South Atlanta

2017 (91) 50% 24% 17% 5.8% 2.3%

2018 (86) 53% 25% 13% 5.5% 2.2%

0.0% 1.1%

The distributions of bike confidence levels are presented in Table 44. Although there is some fluctuation between the confidence levels, the share of respondents who cannot bike is rather consistent between waves.

63

TABLE 44 Stated Bike Confidence Level of Respondents of both Wave 1 and Wave 2

Bike Confidence
Can't Bike Not Very Confident Somewhat Confident
Very Confident

Eastside

2017 (211) 1.9% 14% 33% 51%

2018 (211) 3.3% 14% 26% 57%

Grant Park

2017 (225) 3.6% 15% 33% 48%

2018 (224) 4.0% 15% 33% 48%

Westside

2017 (80)
13%

2018 (75)
14%

18% 21%

26% 24%

43% 41%

South Atlanta

2017 (91) 10% 17% 24% 49%

2018 (93) 10% 29% 16% 45%

64

Second-Wave Survey Analysis
General Perceptions of Changes in Transportation The two new questions that were introduced in the second-wave survey relied on recollection
of recent trends, so the responses to these questions were analyzed for all second-wave respondents regardless of whether they were matched respondents. The first new question included perceptions about general transportation trends. This question was written in a general sense to capture a holistic perspective of how transportation has changed in the previous year. Although the bikeinfrastructure items are the variables of greatest interest, results from the other items are included here for completeness. Note that the sample sizes (before adjusting for item non-response) of each site are Eastside=231, Grant Park=265, Westside=108, and South Atlanta=109.
The two automobile-related items, congestion and parking, are reported in Figure 13 and Figure 14, respectively. The share of respondents expressing changes for the worse for congestion are in the majority, especially in the denser neighborhoods of Eastside and Grant Park. Parking availability in Eastside and Grant Park was also perceived as worsening, while there were many more respondents expressing no change in Westside and South Atlanta.
65

*Indicates treatment location, consistent throughout the rest of this section
FIGURE 13 Distribution of Responses for Perceived Changes in Traffic Congestion
66

FIGURE 14 Distribution of Responses for Perceived Changes in Parking Availability An item for the availability of ride-hailing options was also presented and is summarized in Figure 15. The share of respondents in each site expressing positive changes outweighed those expressing negative changes. The directionality of responses for this item is the reverse from the automobile-focused items, which indicates a lack of "yea-saying," or the tendency of respondents to over-agree on some items.
67

FIGURE 15 Distribution of Responses for Perceived Changes in Availability of Taxi/ Uber/ Lyft The two transit items, route coverage and frequency, are reported in Figure 16 and Figure 17, respectively. For both of these items, the overwhelming majority of respondents perceived no changes, which may be an indication of consistency between transit operations or a lack of attention paid to transit.
68

FIGURE 16 Distribution of Responses for Perceived Changes in Public Transit Route Coverage
FIGURE 17 Distribution of Responses for Perceived Changes in Public Transit Frequency
69

The two pedestrian-related items, sidewalk availability and sidewalk quality, are presented in Figure 18 and Figure 19, respectively. There appears to be a consistent pattern of respondents in the treatment sites perceiving greater improvements in pedestrian infrastructure. Although the purpose of this research project is to investigate the impact of the BeltLine on bike trips, a side benefit is the quantification of the perceived pedestrian improvements linked to the BeltLine.
FIGURE 18 Distribution of Responses for Perceived Changes in Sidewalk Availability
70

FIGURE 19 Distribution of Responses for Perceived Changes in Sidewalk Quality The three bike-related items--safety, bike lane/trail availability, and bike lane/trail quality-- are presented in Figure 20, Figure 21, and Figure 22, respectively. For each of these measures there appears to be little difference in perceptions between Eastside and Grant Park. On the other hand, the differences between Westside and South Atlanta are much more pronounced, especially for perceived improvements in availability and quality of bike lanes/trails. Despite differences among each site, there is a consistent trend of positivity in each site.
71

FIGURE 20 Distribution of Responses for Perceived Changes in Bicycle Safety
FIGURE 21 Distribution of Responses for Perceived Changes in Availability of Bicycle Lanes and Trails
72

FIGURE 22 Distribution of Responses for Perceived Changes in Quality of Bicycle Lanes and Trails
Although the distributions of responses are similar between Eastside and Grant Park, the positivity in both cases is informative. This may be an indication that the two neighboring sites are interconnected enough by bike that the impacts of the BeltLine permeate through both neighborhoods. The smaller share of positive responses than those in Westside may also be an indication that while things have improved, the extension of an existing trail is not as monumental as an entirely new trail.
The inclination of respondents to express perceived improvements in bicycle infrastructure may be a representation of general changes that have occurred for cycling throughout Atlanta over this time period. It may also be a representation of the impact of the visibility of the BeltLine that extends beyond the half-mile buffer used to define neighborhoods in this study.
Although the distributions of responses are important for understanding the general spread of responses, it is also valuable to investigate the mean responses. Responses were coded to numeric values (Much Worse=1 and Much Better =5), and mean values were calculated for each item for
73

each neighborhood. Figure 23 shows a graph of the mean responses for the pedestrian- and bikerelated items for each neighborhood.
5 4 3 2 1

Eastside* Grant Park Westside* South Atlanta
Eastside* Grant Park Westside* South Atlanta
Eastside* Grant Park Westside* South Atlanta
Eastside* Grant Park Westside* South Atlanta
Eastside* Grant Park Westside* South Atlanta

Sidewalk Sidewalk Quality Availability

Bike Safety

Bike Lane/Trail Availability

Bike Lane/Trail Quality

FIGURE 23 Chart of Mean Responses for Pedestrian- and Bicycle-related Questions

Analysis of variance (ANOVA) tests were performed on the mean responses to test for statistical significance of the differences in means between groups. The sample was subdivided between treatment (Eastside and Westside) vs control (Grant Park and South Atlanta) and Westside with control (Westside and South Atlanta) vs Eastside with control (Eastside and Grant Park). An interaction term for the two dummies was also included to test for the difference in the effects of the two treatments.
The ANOVA results for sidewalk availability are presented in Table 45. The significance of the treatment confirms that treatment areas perceived a significantly better change in sidewalk availability.

74

TABLE 45 ANOVA Results for Mean Responses for Sidewalk Availability

Degrees of
Freedom

Sum of Squares

Mean Square

P-value

Treatment

1

7.2

7.18

0.001

***

Westside/South ATL

1

0.8

0.84

0.237

Treatment (Westside)

1

0.0

0.003 0.942

Residuals

704

420

0.60

ANOVA results for sidewalk quality are presented in Table 46. The treatment is also significant for this measure, as is the Westside and South Atlanta, indicating a perceived improvement associated with the treatment, as well as somewhat better improvements overall that were reported in both Eastside and Grant Park.

TABLE 46 ANOVA Results for Mean Responses for Sidewalk Quality

Treatment

Degrees of
Freedom

Sum of Squares

Mean Square

P-value

1

9.4

9.44

0.001

**

Westside/South ATL

1

3.6

3.59

0.043

*

Treatment (Westside) 1

0.3

0.31

0.549

Residuals

703

615

0.88

Results of the ANOVA for bike safety are presented in Table 47. Although there appears to be greater improvement in Westside, the ANOVA results are borderline significant at best, indicating that the data does not strongly point toward a significant difference by site.

75

TABLE 47 ANOVA Results for Mean Responses for Bike Safety

Degrees of
Freedom

Sum of Squares

Mean Square

P-value

Treatment

1

0.5

0.50

0.404

Westside/South ATL

1

0.0

0.01

0.899

Treatment (Westside) 1

2.3

2.32

0.072

.

Residuals

696

498

0.72

Table 48 contains the ANOVA results for bike lane/trail availability. The dummy for each group is significant. The impact of both treatments on average has a significant association with better improvements. Additionally, the general pattern of perceiving improvements is higher on average in Eastside and Grant Park, though the improvements associated with the Westside treatment are even higher.

TABLE 48 ANOVA Results for Mean Responses for Bike Lane/Trail Availability

Treatment

Degrees of
Freedom

Sum of Squares

Mean Square

P-value

1

9.3

9.28 <0.001 ***

Westside/South ATL

1

7.2

7.19

0.002

**

Treatment (Westside) 1

12.4

12.45 <0.001

***

Residuals

700

499

0.71

ANOVA results for bike lanes/trail quality are presented in Table 49. The treatment is significant for this item as well, further confirming the positive impact associated with the BeltLine. Although the Westside and South Atlanta neighborhoods are not significantly different from the Eastside and Grant Park neighborhoods for this measure, the impact of the Westside treatment is significantly different from that of the Eastside treatment.

76

TABLE 49 ANOVA Results for Mean Responses for Bike Lane/Trail Quality

Degrees of
Freedom

Sum of Squares

Mean Square

P-value

Treatment

1

12.6

12.60 <0.001

***

Westside/South ATL

1

1.5

1.47

0.142

Treatment (Westside) 1

14.2

14.19 <0.001

***

Residuals

699

475

0.68

The perceptions of general transportation trends analyzed in this section shed light on some common themes. Automobile measures were perceived as worse, while transit measures were mostly noncommittal. Pedestrian infrastructure was perceived as improving significantly more in the BeltLine treatment locations than in their controls. Bike infrastructure was generally perceived as improving in each site. The improvements associated with the BeltLine treatment areas are significantly greater than those perceived in their controls. However, the impacts of the two treatments were significantly different from each other, indicating that the newly constructed Westside trail may be more influential in triggering perceptions of improvements than the extension of the Eastside trail.

Recognition and Use of New Bicycle Facilities
The second-wave survey also included another new question designed to assess the extent of recognition of new bicycle facilities that had been implemented since the first-wave survey. Each respondent was presented images of each of five bike facility types (sharrows, bike lanes, buffered bike lanes, protected bike lanes, and multi-use paths), though unlike the previous infrastructure images, these were presented without any other roadway characteristics. Respondents were asked if they have seen that type of facility implemented in their community since May 2017, and if they have seen it, if they have used it and if they like it. The response distributions for the Seen question are presented in Figure 24.

77

FIGURE 24 Distribution of Responses for the Question, "Have you seen this added in your community?" for
Each Infrastructure Type and for Each Neighborhood
A sizeable portion of respondents in each site stated that they had seen each facility type in their community. Additions of sharrows, bike lanes, buffered bike lanes, and protected bike lanes may vary between locations. As the purpose of this report is to outline impacts of the BeltLine, the discussion herein focuses only on multi-use paths, with the other facility types serving as a primer to help respondents understand that some facilities they may have seen are different types.
78

Recognition of the path was highest in Eastside, followed by Grant Park, then Westside and South Atlanta. This seems to be counterintuitive based on the findings reported in the previous section, as the Westside treatment area was viewed as having improved more than the Eastside treatment. This seeming disagreement may be an indication that the treatment improved perceptions of cycling, but respondents in the Westside area were not able to properly identify the BeltLine as a multi-use path. Recognition was only marginally higher in Eastside than Grant Park, further strengthening the idea that the two neighborhoods are well-connected, while the differences between Westside and South Atlanta appear to be greater.
The distributions of responses of those who have used each facility type (if they have seen it) are presented in Figure 25. Those in Eastside were most likely to have used a multi-use path, followed by Grant Park, then Westside and South Atlanta. Both treatment areas had higher shares of respondents that have used a multi-use path than their respective controls areas, though the Eastside and Grant Park areas were substantially higher. This may be evidence that the connection of the Eastside extension to the original trail, while not as influential in improving perceptions, is more useful as it connects into a more well-established network. The Westside trail, on the other hand, is a fairly novel facility, so while it may have been successful in improving perceptions, it simply has not had time for a network of use or compatible development to accompany it.
79

FIGURE 25 Distribution of Responses for the Question, "Have you used it?" for Each Infrastructure Type
(for those who have seen it) and for Each Neighborhood
The distributions of responses of those who like each facility type (only for those who have seen it) are presented in Figure 26. While the patterns of usage by site is explainable by treatment, the percent of those who like it is not as apparent. In Eastside and Grant Park the percentages of those who like protected bike lanes is higher than that of multi-use paths, which may be a reflection of the business of the Eastside trail and extension. In Westside and South Atlanta, multi-use paths
80

were more liked than protected bike lanes, which may reflect a slight preference for multi-use paths among those who have not seen as much bike infrastructure.
FIGURE 26 Distribution of Responses for the Question, "Do you like it?" for Each Infrastructure Type
(for those who have seen it) and for Each Neighborhood 81

Second-Wave User Preference Analysis The series of six photoshopped images from the first-wave survey were repeated in the second
wave, with each respondent being assigned the same version for each wave. Models of similar form as the first-wave models were estimated on the second-wave data. Table 50 includes a summary of the linear regressions for comfort, safety, and willingness to try by infrastructure characteristics. These models generally reflect those that were estimated on the first-wave sample (Table 29), though the coefficient for the number of automobile lanes went from significantly negative in the willingness to try model in the first wave to not significant in the second wave. The R2 in each model is also higher in the second wave.
82

TABLE 50 Linear Regression for Expressed Comfort, Safety, and Willingness to Try, Including Only Infrastructure Characteristics

83

Variable Constant

Comfort

Coefficient

P-value

2.72 ***

<0.001

Safety Coefficient
2.57 ***

Pvalue <0.001

Willingness to Try

Coefficient

Pvalue

3.06 ***

<0.001

Bicycle Infrastructure Types

Bike Lane (BL) Buffered BL (BBL) One-way Protected Two-way Protected Multi-Use

0.68 *** 1.11 *** 1.89 *** 1.78 *** 1.62 ***

<0.001 <0.001 <0.001 <0.001 <0.001

0.64 *** 1.12 *** 1.99 *** 1.94 *** 1.68 ***

<0.001 <0.001 <0.001 <0.001 <0.001

0.47 *** 0.72 *** 1.39 *** 1.31 *** 1.28 ***

<0.001 <0.001 <0.001 <0.001 <0.001

Roadway Characteristics

Parking Four Lanes

-0.24 *** -0.02

<0.001 -0.28 *** 0.576 0.02

<0.001 -0.13 ** 0.580 -0.03

0.007 0.478

Framing Effects

BL-No Parking

0.29 ***

<0.001 0.33 ***

<0.001 0.30

BBL-No Parking

0.30 ***

<0.001 0.34 ***

<0.001 0.29

BL-Two Lanes

0.25 **

0.009 0.33 ***

<0.001 0.21

# of Responses

3633

3624

R2

0.301

0.336

Adjusted R2

0.299

0.334

*Significant at P = 0.050 or better; **Significant at P = 0.010 or better; ***Significant at P < 0.001

** *** .
3610 0.149 0.147

0.004 0.001 0.054

Sociodemographic data was added to the previous models, and the resulting models are presented in Table 51. The significance of much of the sociodemographics did not drastically change for the second-wave models, though the smaller sample size makes it less likely to have as many significant variables in each second-wave model. The coefficients for age, females, and African Americans were consistent in the willingness to try models between waves.
84

TABLE 51 Linear Regression for Expressed Comfort, Safety, and Willingness to Try by Infrastructure and Individual Characteristics

Variable Constant

Comfort Coefficient 2.66 ***

P-value <0.001

Safety Coefficient 2.16 ***

P-value <0.001

Willingness to Try

Coefficient

P-value

3.41 ***

<0.001

Bicycle Infrastructure Types

Bike Lane (BL) Buffered BL (BBL) One-way Protected Two-way Protected Multi-Use

0.67 *** 1.13 *** 1.89 *** 1.81 *** 1.58 ***

<0.001 <0.001 <0.001 <0.001 <0.001

0.63 *** 1.13 *** 1.99 *** 1.98 *** 1.65 ***

<0.001 <0.001 <0.001 <0.001 <0.001

0.45 *** 0.71 *** 1.41 *** 1.31 *** 1.29 ***

<0.001 <0.001 <0.001 <0.001 <0.001

Roadway Characteristics

85

Parking Four Lanes Framing Effects BL-No Parking BBL-No Parking BL-Two Lanes

-0.23 *** -0.03
0.28 ** 0.29 *** 0.27 **

<0.001 0.378
0.002 <0.001
0.007

-0.28 *** 0.01
0.33 *** 0.35 *** 0.34 ***

<0.001 0.747
<0.001 <0.001 <0.001

-0.15 ** 0.05
0.29 ** 0.27 ** 0.33 **

0.001 0.281
0.006 0.001 0.002

Sociodemographics

Income Group

0.052 ***

<0.001 0.036 ***

<0.001

0.068

Education Level

-0.038 *

0.020

Driver's License

0.23 *

0.030

0.22

Age

-0.014

Female

-0.30

African American

-0.36

# of Responses

3229

3196

R2

0.311

0.345

Adjusted R2

0.308

0.343

*Significant at P = 0.050 or better; **Significant at P = 0.010 or better; ***Significant at P < 0.001

***
. *** *** *** 3181 0.245 0.241

<0.001
0.061 <0.001 <0.001 <0.001

86

This page is intentionally left blank. 86

Before-and-After Analysis
The purpose of maintaining consistency in much of the survey between wave 1 and wave 2 was to allow for comparisons between the two waves and to quantify change. This section includes an analysis of changes in both perceptions and behavior among matched respondents.
Changes in User Preference Analysis User preference models were estimated, with the wave 2 responses as the dependent
variable and the wave 1 responses and a dummy variable for treatment neighborhoods included as explanatory variables, as shown in Table 52. These models differ from the previously presented regression models as the wave 1 response is expected to explain a large amount of variation in wave 2 responses. For example, a model with a coefficient of 1 for wave 1 response and no other significant variables would indicate that wave 2 responses are equal to wave 1 responses. The relatively low values on the wave 1 responses in the models presented here indicates that even after using similarly constructed measures of preferences as predictors, there is still a large amount of variation built into these constructs. The dummy treatment variable is intended to capture the portion of that variation that is associated with respondents in one of the treatment areas. The lack of significance for the treatment coefficient in the comfort model indicates that there is not enough evidence from the data of any association of residing near a treatment and having a general increase in comfort toward biking. The significantly negative coefficient for the treatment variable in the safety models indicates that those who are near treatments are more likely to rate hypothetical cycling scenarios as less safe, which may be an indication that these respondents have become conditioned to seeing higher quality bike infrastructure and are thus less likely (albeit slightly) to rate other facilities as safe. Conversely, the treatment coefficient for willingness to try is borderline significant, indicating a slight association with those near the BeltLine treatments being more willing to try other facilities in general.
87

The models were re-estimated including sociodemographic cheristics and are shown in Table 53. The addition of sociodemographics in these models was enough to push the treatment coefficient for the willingness to try model from marginally significant (p=0.056) to significant (p=0.021), strengthening the association that, after controlling for demographics, the treatment locations are associated with a slightly higher willingness to try.
88

TABLE 52 Linear Regression for Expressed Comfort, Safety, and Willingness to Try, Including Only Infrastructure Characteristics

89

Variable
Constant Wave 1 Response Treatment

Comfort
Coefficient
1.55 *** 0.42 *** -0.04

Pvalue <0.001 <0.001
0.199

Safety
Coefficient
1.53 *** 0.42 *** -0.08 **

Pvalue <0.001 <0.001
0.008

Willingness to Try

Coefficient

Pvalue

1.25 ***

<0.001

0.55 ***

<0.001

0.06 .

0.056

Bicycle Infrastructure Types

Bike Lane (BL) Buffered BL (BBL) One-way Protected Two-way Protected Multi-Use

0.42 *** 0.71 *** 1.18 *** 1.16 *** 1.00 ***

<0.001 <0.001 <0.001 <0.001 <0.001

0.36 *** 0.63 *** 1.11 *** 1.16 *** 0.95 ***

<0.001 <0.001 <0.001 <0.001 <0.001

0.27 *** 0.39 *** 0.74 *** 0.72 *** 0.66 ***

<0.001 <0.001 <0.001 <0.001 <0.001

Roadway Characteristics

Parking Four Lanes

-0.15 *** 0.003

<0.001 0.926

-0.17 *** 0.04

<0.001 0.216

-0.03 0.01

0.526 0.791

Framing Effects

BL-No Parking

0.20 *

0.015

0.16 .

0.050

0.24

BBL-No Parking

0.20 ***

0.001

0.22 ***

<0.001

0.20

BL-Two Lanes

0.19 *

0.038

0.22

0.011

0.14

# of Responses

3602

3599

R2 Adjusted R2

0.429 0.427

0.464 0.462

*Significant at P = 0.050 or better; **Significant at P = 0.010 or better; ***Significant at P < 0.001

** **
3564 0.423 0.421

0.005 0.004 0.134

TABLE 53 Linear Regression for Expressed Comfort, Safety, and Willingness to Try by Infrastructure and Individual Characteristics

90

Variable

Comfort

Coefficient

P-value

Safety Coefficient P-value

Willingness to Try Coefficient P-value

Constant

1.19 ***

<0.001

1.21 *** <0.001

1.56 *** <0.001

Wave 1 Response

0.43 ***

<0.001

0.43 *** <0.001

0.50 *** <0.001

Treatment

-0.05

0.135 -0.07 *

0.025

0.08 *

0.021

Bicycle Infrastructure Types

Bike Lane (BL)

0.38 ***

<0.001

0.34 *** <0.001

0.25 *** <0.001

Buffered BL (BBL)

0.69 ***

<0.001

0.62 *** <0.001

0.39 *** <0.001

One-way Protected

1.13 ***

<0.001

1.10 *** <0.001

0.79 *** <0.001

Two-way Protected

1.13 ***

<0.001

1.16 *** <0.001

0.75 *** <0.001

Multi-Use

0.94 ***

<0.001

0.92 *** <0.001

0.71 *** <0.001

Roadway Characteristics

Parking

-0.14 ***

<0.001 -0.17 *** <0.001 -0.05

0.225

Four Lanes

-0.01

0.830

0.04

0.234

0.06

0.145

Framing Effects

BL-No Parking

0.18 *

0.036

0.15 .

0.084

0.22 *

0.015

BBL-No Parking

0.19 **

0.002

0.22 ***

0.001

0.20 **

0.006

BL-Two Lanes

0.21 *

0.021

0.25 **

0.006

0.23 *

0.017

Age

0.003 *

0.016 0.003 **

0.006 -0.005 *** <0.001

Income Group

0.041 ***

<0.001 0.035 *** <0.001 -0.14 *** <0.001

Female

-0.14 *** <0.001

African American

0.045 *** <0.001

# of Responses R2 Adjusted R2

3206 0.442 0.440

3214 0.470 0.468

3177 0.447 0.444

*Significant at P = 0.050 or better; **Significant at P = 0.010 or better; ***Significant at P < 0.001

Changes in Cycling Frequency

In each wave of the survey, respondents were asked to report their frequency of making trips using certain modes, both for commute purposes and other purposes. Respondents were divided into groups based on their bike trip frequency in wave 1. Table 54 and Table 55 show crosstabulations for each group within each neighborhood and the number of those in each group who decreased, increased, or did not change in frequency for commute trips and other trips, respectively.

TABLE 54 Changes in Bike Commuting Frequency from First to Second Wave

First Wave Frequency

Eastside

Grant Park

Decreased No change Increased Decreased No change Increased

Never

0

94

12

<1 day a month

4

1

0

13 days a month

5

1

1

12 days a week

5

3

0

34 days a week

7

3

0

5 days a week

0

4

0

Total

21

106

13

0

106

10

2

3

4

1

4

1

1

4

1

4

2

2

0

3

0

8

122

18

First Wave Frequency

Westside

South Atlanta

Decreased No change Increased Decreased No change Increased

Never

0

20

2

<1 day a month

2

2

0

13 days a month

1

1

1

12 days a week

0

0

0

34 days a week

0

0

1

5 days a week

1

0

0

Total

4

23

4

0

42

1

3

0

0

1

1

0

0

0

0

0

0

0

0

0

0

4

43

1

91

TABLE 55 Changes in Frequency of Other Trips by Bike from First to Second Wave

First Wave Frequency
Never <1 day a month 13 days a month 12 days a week 34 days a week 5 days a week
Total

Eastside

Decreased No change

0

71

7

16

10

15

16

10

2

4

4

3

39

119

Increased
17 10 12 2 0 0 41

Grant Park

Decreased No change

0

87

13

15

13

10

7

12

5

2

4

0

42

126

Increased
22 9 5 2 2 0 40

First Wave Frequency

Westside Decreased No change

Never

0

47

<1 day a month

1

3

13 days a month

2

3

12 days a week

1

1

34 days a week

1

0

5 days a week

0

0

Total

5

54

Increased
9 1 1 0 1 0 12

South Atlanta

Decreased No change Increased

0

60

10

5

2

1

3

1

1

0

0

1

2

0

0

0

0

0

10

63

13

As shown in the tables, the vast majority of respondents are not commuting or making other trips by bike in both waves. There is some movement of respondents to begin making commute or other trips by bike, but similar numbers of respondents increased as decreased overall. Standout results are that more respondents increased commute trips in Grant Park than in the Eastside, perhaps indicating that the extension of the BeltLine opened up other neighborhoods than those along the trail. For other trips, the Westside has a noticeable increase and limited decrease in bike trips, while the comparable control (South Atlanta) had a similar increase and decrease.

92

Conclusions
The research presented in this report investigated preferences for bicycle infrastructure and the impact of the BeltLine on travel behavior. Surveys were deployed in two waves in the neighborhoods of the two BeltLine treatments of interest (Eastside Extension and Westside Trail) and their similar control neighborhoods (Grant Park and South Atlanta, respectively). The first wave of the survey was sent out in May 2017, roughly 6 months before the completion of both projects, while the second wave of the survey was sent out in May 2018, roughly 6 months after the opening of both facilities.
Results from the first wave were used to analyze preferences for and perceptions of a variety of bicycle facilities. Images were created in photoshop to identify specific roadway characteristics--namely on-street parking, the number of automobile lanes, and the type of bicycle facility--and presented to respondents. The resulting models indicate a clear preference and positivity toward bicycle facilities that are more separated from vehicles. Parking was also identified as a consistent negative, though protected infrastructure was enough to overcome those negatives. Models segmented by rider type (based on cycling frequency and purpose) indicate that different rider types have different tastes for certain infrastructure characteristic, such as a preference of recreational cyclists for multi-use paths.
Results from the second wave were used to assess perceptions of how transportation in the communities has changed over the previous year. Results indicate that there is a perception in all study areas that private automobile conditions have worsened, while ride-hailing availability has improved and transit conditions have remained roughly the same. There is a perception within the treatment locations that pedestrian infrastructure has improved to a greater extent than within control locations, though the trend was positive in both cases. Perceptions of bicycle facility availability and quality were positive in all locations, with Eastside and Grant Park expressing
93

similar amounts of improvement while Westside expressed a significantly greater amount of perceived improvements than South Atlanta. Perceptions of both pedestrian and bike improvements in each site can be attributable to the BeltLine. The differences between these perceptions between the two neighborhood pairs may be an indication that while the impact of both BeltLine treatments appears to be comparable for pedestrian perceptions, the Eastside area has either already seen the bulk of the improvements that came with the original Eastside BeltLine segment or that the improvements associated with the extension have already begun to spill into Grant Park.
Comparisons in responses were also conducted for those who responded to both waves. Preferences and perceptions (as measured by the hypothetical images) were similar on average between the two waves, though there was a lot of individual variability. Despite this variability, a slight but significant difference was identified in treatment locations of a decrease in perceived safety and an increase in willingness to try, indicating that those near the BeltLine treatments were more likely in general to express a lower level of perceived safety for roadway configurations but a higher level of willingness to try biking on them.
94

Implementation Recommendations
One of the primary purposes of GDOT research is to inform future planning, design, operations, and maintenance practices at the agency. There are several key policy takeaways from the research presented in this report that should be carried forward for implementation of the research.
First and foremost, throughout both waves of the survey, respondents showed a clear preference and positivity toward bicycle facilities that are more separated from vehicles. Parking was also identified as a consistent negative, though protected infrastructure was enough to overcome those negatives. GDOT should focus on implementing protected bicycle infrastructure and multi-use trails to encourage bicycle trip-making behavior.
Second, the implementation of multi-use trails such as the BeltLine have positive impacts on the impression of sidewalk quality and availability as well as bicycle facility quality and availability. Facilities such as the BeltLine are noticed and appreciated by residents. This gives further evidence that multi-use trails should be encouraged and funded.
Finally, through this study, a ready-made survey to assess future sections of the BeltLine and other bicycle infrastructure has been developed. As policy, GDOT should ensure that as infrastructure is constructed, before-and-after surveys such as this one are conducted to better understand preferences and impacts over time.
95

This page is intentionally left blank. 96

Appendix A: First-Wave Survey
97

98

99

100

101

102

103

104

<VERSION 1>

105

106

<VERSION 2>

107

108

<VERSION 3>

109

110

<VERSION 4>

111

112

113

114

Appendix B: Second-Wave Survey
115

116

117

118

119

120

<VERSION 1>

121

122

<VERSION 2>

123

124

<VERSION 3>

125

126

<VERSION 4>

127

128

129

130

Appendix C: Complete Demographics

Gender
Female Male
Gender
Female Male
Age Group
18-34 35-49 50-64 65+
Age Group
18-34 35-49 50-64 65+

TABLE C - 1 Wave 1 & 2 Survey Respondents by Gender

Eastside

Unmatched Wave 1 Wave 2 (429) (230)

Matched (211)

55% 55%

54%

43% 45%

46%

Grant Park

Unmatched Wave 1 Wave 2 (473) (261)

Matched (225)

55% 56%

58%

43% 44%

42%

Westside

Unmatched
Wave 1 Wave 2 (226) (102)

Matched (82)

65% 61%

61%

29% 39%

39%

Southside

Unmatched
Wave 1 Wave 2 (188) (107)

Matched (94)

54% 53%

52%

44% 47%

48%

TABLE C - 2 Wave 1 & 2 Survey Respondents by Age

Eastside

Unmatched Wave 1 Wave 2 (428) (230)

Matched (211)

34% 30%

30%

38% 44%

43%

18% 17%

18%

8.1% 10%

9.0%

Grant Park

Unmatched Wave 1 Wave 2 (471) (261)

Matched (224)

25% 23%

22%

40% 36%

38%

23% 26%

25%

10% 15%

14%

Westside

Unmatched Wave 1 Wave 2 (222) (100)

Matched (81)

19% 9%

10%

22% 28%

28%

31% 30%

28%

23% 33%

33%

Southside

Unmatched Wave 1 Wave 2 (186) (105)

Matched (93)

17% 19%

18%

29% 27%

30%

31% 32%

32%

21% 22%

19%

131

TABLE C - 3 Wave 1 & 2 Survey Respondents by Race

Race
White African American
Hispanic Asian
Native American Other

Eastside

Unmatched
Wave 1 Wave 2 (414) (226)

Matched (206)

77% 79%

81%

11% 12%

13%

0.9% 2.2%

3.4%

4.6% 7.5%

6.3%

0.0% 0.0%

0.0%

1.8% 1.3%

1.0%

Grant Park

Unmatched
Wave 1 Wave 2 (452) (258)

Matched (222)

76% 83%

83%

12% 13%

13%

2.9% 2.7%

2.3%

1.3% 1.2%

1.4%

0.2% 1.9%

1.8%

2.7% 1.9%

1.8%

Race
White African American
Hispanic Asian
Native American Other

Westside

Unmatched Wave 1 Wave 2 (210) (101)

Matched (78)

19% 19%

22%

63% 68%

72%

5.1% 6.9%

5.1%

0.9% 0.0%

0.0%

0.4% 2.0%

0.0%

0.9% 4.0%

3.8%

Southside

Unmatched Wave 1 Wave 2 (176) (106)

Matched (93)

30% 37%

38%

59% 57%

59%

2.1% 2.8%

1.1%

1.1% 2.8%

2.2%

0.0% 1.9%

1.1%

1.1% 2.8%

3.2%

132

TABLE C - 4 Wave 1 & 2 Survey Respondents by Household Income

Household Income
$15,000 or less $15,001 - $30,000 $30,001 - $50,000 $50,001 - $75,000 $75,001 - $100,000 $100,001 - $125,000 More than $125,000

Eastside

Unmatched

Matched

Wave 1 Wave 2 2017 (393) (209) (192)

2018 (196)

2.0% 1.9% 3.6% 2.0%

3.8% 2.9% 4.2% 2.6%

9.2% 6.7% 6.3% 7.1%

14% 12.9% 16% 14%

16% 14.8% 18% 15%

15% 18.2% 15% 18%

40% 42.6% 38% 41%

Grant Park

Unmatched

Matched

Wave 1 Wave 2 2017 2018 (426) (233) (205) (200)

3.1% 3.0% 2.4% 2.5%

4.0% 3.9% 4.9% 4.0%

7.3% 9.0% 11% 10%

15% 12.9% 16% 14%

16% 15.5% 14% 15%

15% 15.5% 15% 16%

40% 40.3% 37% 39%

Household Income
$15,000 or less $15,001 - $30,000 $30,001 - $50,000 $50,001 - $75,000 $75,001 - $100,000 $100,001 - $125,000 More than $125,000

Westside

Unmatched

Matched

Wave 1 Wave 2 2017 (199) (87) (78)

2018 (70)

21% 13.8% 21% 14%

18% 19.5% 22% 19%

16% 23.0% 13% 24%

16% 12.6% 12% 11%

16% 14.9% 21% 13%

5.0% 11.5% 5.1% 13%

8.0% 4.6% 7.7% 5.7%

Southside

Unmatched

Matched

Wave 1 Wave 2 2017 2018

(163) (89)

(82) (76)

17% 14.6% 13% 13%

20% 14.6% 20% 13%

15% 11.2% 13% 12%

16% 18.0% 15% 20%

14% 19.1% 18% 17%

4.9% 5.6% 3.7% 5.3%

13% 16.9% 17% 20%

133

TABLE C - 5 Wave 1 & 2 Survey Respondents by Household Size

Household Size
1 2 3 4 5+

Eastside

Unmatched

Matched

Wave 1 Wave 2 2017 (420) (222) (205)

2018 (205)

39% 38% 40% 43%

42% 44% 41% 41%

9.2% 11.3% 11% 10%

5.1% 5.4% 4.9% 4.9%

0.7% 1.8% 2.0% 1.0%

Grant Park

Unmatched

Matched

Wave 1 Wave 2 2017 2018 (459) (258) (220) (221)

28% 31% 32% 34%

42% 42% 42% 41%

10% 15.1% 13% 12%

13% 10.5% 11% 10%

3.4% 1.6% 1.8% 2.3%

Household Size
1 2 3 4 5+

Westside

Unmatched

Matched

Wave 1 Wave 2 2017 (221) (95) (78)

2018 (75)

31% 40% 41% 39%

36% 38% 37% 39%

13% 5.3% 5.3% 8.9%

6.0% 11.6% 11% 5.1%

8.1% 5.3% 5.3% 7.6%

Southside

Unmatched

Matched

Wave 1 Wave 2 2017 2018

(180) (98)

(89) (85)

44% 45% 45% 44%

26% 33% 33% 30%

14% 12.2% 13% 15%

4.2% 5.1% 4.7% 4.5%

6.3% 5.1% 4.7% 6.7%

TABLE C - 6 Wave 1 & 2 Survey Respondents by Residence Type

Residence Type
Detached Duplex
Apt Other

Eastside

Unmatched

Matched

Wave 1 Wave 2 2017 (432) (230) (211)

2018 (211)

41% 45% 45% 40%

15% 16% 16% 18%

42% 38% 38% 42%

0.7% 0.9% 0.5% 0.9%

Grant Park

Unmatched

Matched

Wave 1 Wave 2 2017 2018 (477) (263) (225) (225)

74% 45% 75% 72%

16% 16% 12% 16%

10% 38% 12% 12%

0.6% 0.9% 0.4% 0.0%

Residence Type
Detached Duplex
Apt Other

Westside

Unmatched

Matched

Wave 1 Wave 2 2017 (233) (103) (81)

2018 (82)

78% 78% 76% 77%

6.8% 5% 6.1% 8.6%

13% 16% 16% 14%

1.3% 1.9% 1.2% 1.2%

Southside

Unmatched

Matched

Wave 1 Wave 2 2017 2018 (189) (105) (93) (94)

66% 68% 70% 72%

4.7%

9%

7.4% 6.5%

26% 19% 17% 19%

3.2% 4.8% 3.2% 2.2%

134

TABLE C - 7 Wave 1 & 2 Survey Respondents by Employment Status

Employment Status
Full time Part time 2+ jobs Homemaker Don't work

Eastside

Unmatched

Matched

Wave 1 Wave 2 2017 (428) (231) (211)

2018 (211)

78% 79% 81% 81%

6.5% 5.2% 4.3% 7.1%

1.4% 4.8% 4.7% 3.8%

2.3% 3.5% 3.3% 2.4%

8.1% 12% 10% 7.1%

Grant Park

Unmatched

Matched

Wave 1 Wave 2 2017 2018 (455) (265) (225) (225)

72% 68% 66% 70%

7.5% 10.9% 12% 12%

1.5% 5.3% 5.3% 4.4%

1.7% 6.4% 6.2% 2.2%

12% 19% 20% 15%

Employment Status
Full time Part time 2+ jobs Homemaker Don't work

Westside

Unmatched

Matched

Wave 1 Wave 2 2017 (211) (108) (82)

2018 (82)

40% 36% 37% 35%

10% 8.3% 9.8% 16%

5.1% 7.4% 8.5% 9.8%

3.8% 2.8% 2.4% 1.2%

30% 51% 51% 43%

Southside

Unmatched

Matched

Wave 1 Wave 2 2017 2018 (177) (109) (94) (94)

48% 50% 51% 57%

12% 15.6% 13% 12%

3.2% 4.6% 4.3% 7.4%

1.6% 2.8% 3.2% 3.2%

28% 28% 28% 27%

135

TABLE C - 8 Wave 1 & 2 Survey Respondents by Number of Vehicles

Number of Vehicles
0 1 2 3 4 5+

Eastside

Unmatched

Matched

Wave 1 Wave 2 2017 (428) (227) (209)

2018 (208)

4.8% 5.7% 5.8% 6.7%

45% 40.1% 41% 45%

39% 47.6% 46% 39%

6.5% 4.8% 5.3% 6.7%

2.8% 1.3% 1.4% 1.4%

0.7% 0.4% 0.5% 1.0%

Grant Park

Unmatched

Matched

Wave 1 Wave 2 2017 2018 (471) (261) (225) (224)

5.5% 6.1% 5.8% 5.8%

31% 32.6% 33% 33%

48% 49.4% 49% 49%

10% 9.2% 9.4% 8.4%

2.3% 1.9% 1.8% 1.3%

1.5% 0.8% 0.9% 1.3%

Number of Vehicles
0 1 2 3 4 5+

Westside

Unmatched

Matched

Wave 1 Wave 2 2017 (223) (99) (81)

2018 (77)

17% 19.2% 18% 16%

39% 40.4% 39% 44%

27% 27.3% 30% 26%

7.7% 10.1% 9.1% 8.6%

2.6% 2.0% 2.6% 3.7%

1.3% 1.0% 1.3% 1.2%

Southside

Unmatched

Matched

Wave 1 Wave 2 2017 2018 (183) (100) (91) (87)

19% 19.0% 18% 18%

34% 35.0% 36% 35%

33% 38.0% 38% 34%

7.9% 4.0% 3.4% 8.8%

1.6% 3.0% 3.4% 3.3%

0.5% 1.0% 1.1% 1.1%

136

TABLE C - 9 Wave 1 & 2 Survey Respondents by Number of Bikes

Number of Bikes
0 1 2 3 4 5+

Eastside

Unmatched

Matched

Wave 1 Wave 2 2017 (428) (227) (209)

2018 (207)

24% 25.6% 27% 28%

30% 29.5% 28% 29%

26% 23.3% 25% 26%

8.3% 10.1% 10% 7.7%

6.0% 6.2% 4.8% 3.8%

4.4% 5.3% 5.3% 4.8%

Grant Park

Unmatched

Matched

Wave 1 Wave 2 2017 2018 (472) (261) (225) (224)

23% 25.3% 25% 24%

20% 21.5% 19% 19%

29% 29.1% 29% 31%

10% 9.2% 10% 11%

7.3% 6.5% 6.3% 5.8%

8.8% 8.4% 10% 9.3%

Number of Bikes
0 1 2 3 4 5+

Westside

Unmatched

Matched

Wave 1 Wave 2 2017 (220) (99) (76)

2018 (80)

42% 50.5% 49% 45%

25% 25.3% 26% 28%

16% 17.2% 17% 16%

4.3% 3.0% 3.9% 7.5%

4.7% 2.0% 1.3% 2.5%

1.3% 2.0% 2.6% 1.3%

Southside

Unmatched

Matched

Wave 1 Wave 2 2017 2018

(183) (99)

(91) (86)

52% 50.5% 50% 53%

20% 23.2% 24% 25%

13% 18.2% 17% 13%

6.8% 5.1% 5.8% 5.5%

2.1% 2.0% 2.3% 2.2%

2.1% 1.0% 0.0% 1.1%

137

TABLE C - 10 Wave 1 & 2 Survey Respondents by Cycling Confidence Level

Confidence Level
Can't Bike Not Very Confident Somewhat Confident
Very Confident

Eastside

Unmatched

Matched

Wave 1 Wave 2 2017 2018 (430) (231) (211) (211)

3.0% 2.2% 1.9% 3.3%

15% 13.0% 14% 14%

27% 33.3% 33% 26%

54% 51.5% 51% 57%

Grant Park

Unmatched

Matched

Wave 1 Wave 2 2017 2018 (473) (262) (225) (224)

4.4% 3.8% 3.6% 4.0%

14% 14.9% 15% 15%

29% 31.3% 33% 33%

52% 50.0% 48% 48%

Confidence Level
Can't Bike Not Very Confident Somewhat Confident
Very Confident

Westside

Unmatched

Matched

Wave 1 Wave 2 2017 (222) (100) (80)

2018 (75)

16% 14.0% 13% 14%

19% 20.0% 18% 21%

18% 26.0% 26% 24%

41% 40.0% 43% 41%

Southside

Unmatched

Matched

Wave 1 Wave 2 2017 2018 (184) (105) (91) (93)

13% 10.5% 10% 10%

21% 18.1% 17% 29%

20% 23.8% 24% 16%

43% 47.6% 49% 45%

138