We propose a Pedestrian Safety Index for crosswalks along main roads in Mexico to reduce fatalities and improve crossing conditions. The pedestrian crosswalk safety index intends to be a useful tool for stakeholders, allowing them to assess crosswalk quality and identify issues to be improved. In order to identify and weight the index criterions, about 53 state of the art documents were reviewed; a non- probabilistic survey was conducted (N=1000); in situ observations were carried out and a multi-criteria analysis method was implemented with a sample of 503 signalized crosswalks in Mexico City. The final index has 19 criterions arranged in 5 macro-criterions related to the infrastructure, design and operation conditions of the crosswalk. One of the main findings is that both land use mix and crossing distance are the most important variables in predicting traffic incidents, both being positively correlated and statistically significant. The second one is that roughly 90% of the surveyed crosswalks in Mexico City do not meet the minimum design criteria to ensure a secure and comfortable crossing.
Rationale & main goals
Traffic incidents are considered as negative externalities derived from car-oriented transit policies. In Mexico, traffic fatalities are the first death cause among children between 5 and 14 years old and the second one between 15 to 29. According to the Ministry of Public Security (Secretaría de Seguridad Pública SSP), within Mexico City 61% of those fatalities happen over main streets.
The main goal of this proposal is to produce a Pedestrian Crosswalk Safety Index (PCSI), in order to reduce fatalities and to improve crosswalk design conditions.
The Pedestrian Crosswalk Safety Index will be developed for signalized pedestrian crossings over main streets. Even though internationally there are different indexes of the same characteristics, Mexico is lacking a validated methodology. Moreover, a standardized method within the new programs in Mexico City regarding “Safety Crosswalks” — developed by two different Mexico City institutions (Agencia de Gestión Urbana (AGU) and the Autoridad del Espacio Público (AEP)) — is missing. In 2014, the recently approved Mobility Law for the Mexico City Metropolitan Area (MCMA) stablishes a ‘Zero’ vision which aims to have zero fatalities related to traffic incidents. In this regard, we see this zeitgeist as an invaluable opportunity to introduce a PCSI that can be used as a tool for the assessment and identification of the most conflictive crosswalks while, at the same time, serve as a guide to improve crossing conditions.
There are two paths to develop PCSI’s: on the one hand, methods derived from direct observation, expert knowledge, and statistical regressions (i.e. Cheng et al., 2014; Bian et al., 2013; Carter et al., 2006); on the other hand, indexes derived from multi-criteria analysis (i.e. Basile et al., 2010; Montella et al., 2010). Our team developed a proposal for a PCSI based on Basile et al. (2010), following a five-step model: i) identification, hierarchization and criterion selection; ii) criteria rating (range of values) and criteria weighting through AHP (Analytic Hierarchic Process (Saaty 1980)); iii) Testing the instrument in field; iv) PCSI calculation; v) ‘validation’ of the PCSI against 2010-2016 fatalities geodatabase (from the SSP).
(i) For criteria selection, we performed an extensive literature review, from which 94 criterions were selected. Later on, those criterions were compared against others derived from previous field work in which 6 crosswalks were filmed and analyzed in order to identify the most problematic variables. From that analysis, an Expert Panel workshop was designed in which the experts selected a final list containing 19 criterions arranged within five macro criterions: Accesibility, Visibility, Design, Horizontal Signaling, Level of Signaling (Traffic Lights).
(ii) For criteria rating (range of values) and weighting, an AHP online method was used (Saaty 1980; 2008, Malczewski, 1999). It allows to weight each criteria at different levels in order to acquire a final value which will become the PCSI. For weighting each criteria the Semantic Scale of Saaty (2008), which scores variables from 1 to 9, was used. The panel was comprised of ten experts (among them were officials of the Mobility Ministery, the SSP and the AGU). All of them weighted each criterion with the aid of an online AHP tool (http://bpmsg.com/ahp-online-calculator/).
(iii) The resulting instrument was tested on the field. A stratified statistical sample of 503 signalized crosswalks were surveyed at day and night. For that purpose a mobile app was developed and installed in digital tablets to a) help the survey teams complete their task in less time and b) gather the collected data into a remote server in real time. The PCSI was computed automatically (iv).
(iv). Next, data was analyzed a) quantitatively and b) qualitatively. For (a), three approaches were used: Zero inflated models (regressions); Fast and Frugal Decision Trees; and a SpaceSytnax (Depthmap) approach (see complete report for further detail). For (b), micro criterions were analyzed and graphed in order to understand the importance of each variable accounting for a better crossing.
Results and discussion
Quantitatively the most important variables for predicting traffic incidents are the land use mix (the more mixed, the more traffic incidents) and the crossing distance (at longer crossing distances there are more chances of being hit by a motorized vehicle). The land use mix variable can be operating as a proxy for pedestrian flow density, Although the Fast and Frugal Decision Trees model was not as statistically robust as the Zero inflated model (it predicts events slightly better than random), both variables were consistent. Regarding the Space Syntax analysis, it should be highlighted that the most connected elements within a given urban network, are linked to more chance of traffic incidents. But probably, the most important conclusion from this research is that although AHP can be useful for assessing qualitative elements, it is not the best way to develop an index intended to be used as a predictive indicator, primarily due to the subjective nature of the method.
On the other hand, the AHP is a very good tool for assessing pedestrian crosswalks. Surprisingly, crossing distance, the variable with the highest weight assigned by the experts panel, is the same variable that better explains the traffic events within the statistical models.
From this qualitative analysis, it can be stated that 95% of pedestrian crossings are not suitable for people with visual impairments; 81% has fixed obstacles over their trajectory; 74% are too long to be deemed safe; 77% of the traffic lightshave too short times for pedestrian to cross at a fixed speed; and in 61% of the surveyed cross walks, pedestrians have to wait too much time to cross.