By Mary Sue O’Melia and David Brandauer
Why is it so difficult to get an accurate number of passengers in a given transit system? Have you ever sought the total number of passengers and received more than one answer? Such responses do not instill confidence. This month, we delve into passenger statistics – sources and uses of data from manual counts, automated fare collection systems (AFC) and automated passenger counters (APC).
Random sampling and manual counts
A number of transit agencies continue to use a random sampling process whereby data on passenger boardings by trip are manually collected, tabulated and statistically weighted to report monthly and annual passengers. The total number of one-way trips or farebox revenues may be used to expand passenger sample data to an annual total.
Manual counts are generally most accurate on a trip-level and may in fact be used to validate information from driver manifests as well as AFC and APC systems.
The downsides of this method of collecting data are that it is labor intensive, accuracy decreases with the length of shift and, depending on the random sample size, may only be valid at the mode/service type level on an annual basis. While this level of reporting is good for NTD reporting, it is not much use to transit planners and managers who are responsible for agency performance.
Automated fare collection system passenger data
AFC provides detailed information about fare payment and passengers. Information is available by date, time, route/trip, fare type and transaction type. Transactions may include items other than passenger boardings (e.g., wheelchairs, bicycles).
Fare systems typically require driver interaction to correctly record non-cash transactions (e.g., student passes). Consistent driver participation and passenger fare evasion are two of the challenges in getting accurate passenger boarding information from fare systems. Monitoring of “Unclassified Revenues” is useful in identifying driver training and implementation issues.
AFC pre-sets and the use of smart cards have helped to reduce the need for driver interaction. This in turn increases reliability of passenger data, assuming that automated equipment is maintained. Transit agencies typically place fare collection equipment in locations that are clearly visible to drivers. In a well-maintained AFC system, missing data due to equipment failure is not really an issue and most data sets from AFC and smart cards are complete. AFC equipment, however, does not help with the passenger mile reporting required for NTD annual reporting.
Automated passenger counter (APC)
Automated passenger counters provide the basis for collecting not only passenger boardings, but passenger mile data as well. APCs can work on service where the driver is not responsible for fare collection (e.g., free shuttles and trolleys, articulated buses with rear-door entry, rail) and where the passenger loads are such that the driver cannot accurately interact with fare collection equipment on all transactions.
APC data is highly valued by planners as it provides detailed data on boardings and alightings. APC data may be the official agency source for reporting passengers, especially if the entire fleet is equipped and the agency has received FTA approval for the initial benchmarking and long-term maintenance processes. Information by route, stop and time are all available if the equipment is calibrated, maintained, and fully functional. In-service equipment failure is not readily apparent so a method for identifying faulty equipment and of estimating missing passenger data due to equipment failure is required.
If APCs are only available on a portion of the fleet, then the agency must implement a sampling plan that collects information to meet NTD reporting requirements as well as agency internal reporting needs. The potential for bias in the plan, plan implementation by operations, and weighting of sample can impact reported passenger counts.
One thing is clear: manual counts and APC data should be different at the trip level because of end-of-line activity (e.g., drivers getting on and off at layover locations). This causes APC passenger counts to be higher than manual counts. If passenger loads are high, APC passenger counts may be understated. This is why the benchmarking and annual maintenance sample calibration factors are so important.
Conclusion
Agencies may compare the annual or monthly passenger figures from random sampling to AFC and/or APC data. These figures are never the same. The variances may be inconsistent and quite large. While passenger data from individual trips may be compared, monthly and annual comparison of passenger data from manual counts to AFC and APC data will always be different because of expansion factors in random sampling and APC data sets. Driver error, fare evasion and equipment failure may impact manual passenger counts compared to AFC passenger counts.
The FTA recommends that data from a minimum of 100 trips collected with manual counts be compared to APC data on an annual basis as part of the ongoing process to ensure accurate data and equipment calibration. We believe that using a limited sample manual process to validate automated systems is a good practice. Simple manual count random sampling, however, does not provide the level of information needed for service planning and revenue control. What is important is to select one source to be the official passenger figure for the agency and set up a process to use other measures as a check and balance.
Mary Sue O’Melia is president of TransTrack Systems®, Inc., a business intelligence solution that transforms volumes of data into meaningful information. David Brandauer is the chief operating officer for BLIC North America, a transportation technology consulting firm. Visit our authors at www.transtrack.net and www.blic.us.