Methodology
Introduction
Following the completion of the pilot study in Autumn 2019 and the project restart in summer 2022, the fieldwork methodology was revised to maximise the efficient collection of representative, high-quality data.
This section describes the survey design, fieldwork and data collection approach, and approach to data analysis.
Survey Design
The four-part survey was divided as follows:
Part A – Local Trunk Road Focus
Respondents were first asked to indicate where on a map of their local area most problems occurred. For the section they indicated, they were asked about perceived vehicle speeds, time issues occur, road crossing conditions and walking and cycling conditions along the study area.
Part B – Barriers to Active Travel
In Part B, respondents were asked to indicate the degree to which a range of factors hindered their ability to walk or cycle in a) the study area and b) the wider area (20-minute walk / a mile). The following factors were chosen to be scored as Never, Rarely, Often or Always impacting their ability to walk or cycle:
- Speed of traffic
- Amount of traffic
- Lack of crossing points
- Signalised crossings not allowing adequate time to cross.
- Poor road lighting
- Poor footways
- Noise pollution
- Air pollution
- Fear for personal safety
- Lack of safe cycling infrastructure, hindering cycling
Part C – Resident Views
Part C was a free text question that allowed respondents to provide feedback and give their opinion on what improvements they would like to see in their local area to facilitate travel of any type around their local area.
Part D – About You
Part D asked the respondent general questions about themselves / their household to capture demographic information.
A copy of the survey is included in Appendix B.
FieldWork Approach
Surveys were conducted between September 2022 – November 2022, with two WSP team members visiting each of the settlements to distribute the surveys.
A combined method of in-person/freepost return was chosen for survey distribution, with surveyors introducing themselves and explaining the survey at 1:2 – 1:5 residences depending on population density.
The respondents were left to answer in their own time and could either return the survey to the surveyor later that day or by freepost return. At houses with no response or which were not targeted for in-person discussion, surveys were posted through the letterbox for freepost return.
Sampling
The following settlements were selected as the study areas, with surveys issued to all households living within 50m of the trunk road:
- Rosyth
- Methven
- Freuchie
- Keith
- Fort William
- Inverness
- Skelmorlie
- Lochearnhead
- Drumnadrochit
- Helmsdale
- Newtonmore
- Kincardine
- Mauchline
These settlements were selected based on the combined professional experience of WSP and Transport Scotland as reflective of a variety of trunk road residential contexts.
Response Rate
Across the 13 settlements that were included in the study, there were 589 responses, a response rate of 26%.
Quantitative Data Analysis
This section summarises the approach to quantitative data analysis of the surveys; the results of this analysis are presented in Section 4.
Determination of the primary Dependent Variable
A comparison between perceived and ATC-measured traffic speeds at the location which participants marked on the Part A map was used to address the primary research question, which focused on residents’ perception of speed.
The key dependent variable for the central research question was termed the ‘relative speed estimate’, calculated as:
Measured 85th percentile traffic speed – Estimated traffic speed =
Relative Speed Estimate
As estimated traffic speed was recorded in bands (e.g., 30-35 mph), the maximum speed of the band was used for the above calculation.
Descriptive Statistics
The sample was analysed along the demographic measures described in Table 3-1.
Demographic Characteristic | Response Options |
---|---|
Gender | Female, Male, Other, Prefer Not to Say |
Age | In bands from 18-24 to 85+ |
Number of years living in residence | Free response |
Household description | Retired, Immediate Family, Multi-Occupancy (related), Multi-Occupancy (not related), Single-Occupancy, Couple, Prefer Not to Say |
Mobility-Limiting Disability | Yes / No |
Household vehicle ownership | None, One, Two or More |
Frequency of contact with neighbours | Three or more times a week, Once or twice a week, Once or twice a month, Less often or never |
Pet ownership | Yes / No, free response to indicate type and number |
Due to the geographically-targeted nature of the analysis, the sample was not assessed or weighted against population-wide demographics.
Demographic Analysis
The analysis team used two sample t-tests for unequal variances, conducted in Excel, to compare the mean relative speed estimates between the following demographic groups:
- Male and female respondents
- Disabled and non-disabled respondents
- Respondents with and without a vehicle in the household
- Respondents with and without children in the household
- Respondents with and without a dog and/or cat in the household
Additionally, a linear regression model was used to assess the relationship between age and relative speed estimate.
Coding of Independent Variables
Table 3-2 lists the variables included in the regression models and shows how categorical variables were coded for inclusion in linear and logistic regression modelling:
Demographic Characteristic | Data Source | Coding |
---|---|---|
Gender | Survey | Female (2), Male (1), Other (3), Prefer Not to Say (0) |
Age | Survey | In bands from 18-24 to 85+ - youngest age in band used for regression |
Number of years living in residence | Survey | Numerical data (years) |
Household description | Survey | Retired, Immediate Family, Multi-Occupancy (related), Multi-Occupancy (not related), Single-Occupancy, Couple, Prefer Not to Say |
Mobility-Limiting Disability | Survey | Yes (1) / No (0) |
Household vehicle ownership | Survey | None (0), One (1), Two or More (2) |
Frequency of contact with neighbours | Survey | Three or more times a week (3), Once or twice a week (2), Once or twice a month (1), Less often or never (0) |
Pet ownership | Survey | Dog and / or Cat: Yes (1) / No (0) |
Width of Road | Google Maps | Estimated width in metres |
85th Percentile Speed | ATC | Speed in miles per hour |
Volume of Traffic (10k's) | ATC | 7-day daily average count / 10,000 |
Land Use | Google Maps | Commercial (3), Mixed Use (2), Residential (1), Unoccupied (0) |
Barrier between footway and traffic | Google Maps | Most of carriageway section (as indicated on map) (2), Some of carriageway section (1), None (0) |
Presence of bus stop(s) | Google Maps | One or more bus stop located in carriageway section indicated on the map: Yes (1), No (0) |
Pedestrian crossing in road section? | Google Maps | Pedestrian Island (1), Zebra (2), Controlled (3), None (0) |
Footway width | Google Maps | Estimated width in metres |
School near the section? | Google Maps | School within 100m of road section: yes (1) / no (0) |
Footway | Google Maps | Both sides of road section (2), One side (1), None |
Lanes of Traffic | Google Maps | Number of lanes (Note that this variable was removed from analysis as all considered road sections had two lanes.) |
Model Construction
Three models were prepared using the XLMiner Analysis ToolPak in Excel:
- Multiple linear regression model: relationship between independent variables and relative speed estimate
- Multiple logistic regression model 1: relationship between independent variables and probability of 5+mph relative speed estimate
- Multiple logistic regression model 2: relationship between independent variables and 10+mph relative speed estimate
Model Fitting
The model was then fitted using non-automated backward elimination. The initial model was built including all available predictor (independent) variables. One by one, the variables with the highest p-value (level of statistical significance) were eliminated from the model. The process was repeated iteratively, with the analysis team removing variables until a model was obtained with all variables below the designated alpha value (i.e., p < 0.10). This manual approach allowed the model to be tailored to the specific problem and make informed decisions based on the data and our expertise.
Qualitative Data Analysis
For the ‘Your Views’ section of the survey (Part C), the project team manually coded each response into key categories and summarised the feedback in the Settlement-Specific Datasheets (Appendix A) with additional emphasis placed on more frequently occurring feedback.