Cashing Out on Bus Reliability

Everyone who’s taken the bus knows the frustration of standing in line to board or sitting at a stop while people tap their cards and pay their fares. Cash use on MBTA buses not only slows down the boarding process, it also adds variability to bus runs that could make it harder for buses to stay on schedule. In future fare payment systems that the MBTA and other agencies are considering, cash would no longer be accepted on-board buses as off-board payment options would be greatly expanded.

In this post, we look at the breakdown of fare media used to board the MBTA’s key bus routes, and we compare cash use on those routes with bus reliability data from the MBTA Performance Dashboard. Looking at off-peak periods in April 2016, the data suggest that the number of late bus stops could be reduced by more than 10% if all cash transactions happened off-board. And that is in addition to the direct benefits of no cash on board: an easier boarding process, faster bus travel, and potentially more frequent service with the same number of buses by decreasing the time it takes each bus to complete a trip.

Cash Use

If you ride MBTA buses, you know that the slowest boardings use cash — either direct cash payments or CharlieCard “top-ups” (using the farebox to add cash to the stored value on a CharlieCard). The graphs below summarize boardings by fare media for April 2016 on the MBTA’s key bus routes. Figure 1 illustrates that cash payments make up a larger share of off-peak / weekend boardings than peak period boardings (where the peak period is 6:30-9:30am and 3:30-6:30pm on weekdays). Figure 2 shows quite a bit of variation in fare media across the MBTA’s key bus routes.

Fare Media used by Time Period
Fare Media on key bus by route

Much of the focus on the benefits of reducing or ending cash usage on buses is on reducing the dwell time (the time buses spend at stops). That improved speed means not only quicker travel times for passengers, but also the potential to run higher frequency service with the same number of buses.


There may be another benefit to moving cash use off-board — reliability. When you’re waiting for a bus, have you ever thought about how fare payment is affecting your wait time? Different boarding speeds caused by cash payments is one of many sources of variability that make it difficult for buses to run on schedule and sometimes cause buses to bunch. Reducing this variability would result in shorter wait times and less guesswork about travel times.

While it’s difficult to know exactly how much reliability would be improved if there were no cash on board, we can get some hints by looking at reliability data on the MBTA Performance Dashboard. The dashboard provides daily reliability for each bus route and time period (peak and off-peak). How reliability is measured depends on the frequency of the bus.

Traffic is a big factor in reliability during peak periods, which adds a lot of noise when we’re trying to look at a smaller factor like cash boardings. So let’s focus on off-peak periods (including weekends) in April 2016. For each key bus route, we can compare average off-peak reliability with the share of all off-peak boardings that involved cash. Figure 3 below shows some correlation — bus routes with higher shares of cash boardings tended to have worse reliability. Of course, that doesn’t necessarily mean cash boardings cause that pattern — cash boardings might just be correlated with other things that affects reliability (like roadway congestion).

Reliability by share of boardings using cash by route

One way to get closer to the actual impact of cash boardings on reliability is to look at variation within each bus route (instead of across bus routes). On any given day for a specific route, the share of cash boardings probably isn’t related to other factors affecting reliability. Each point on Figure 4 below represents the off-peak period for one day on one of the key bus routes; each color represents a different bus route. It’s really a more detailed version of Figure 3 above (which showed averages across all days). The lines are separate trend lines for each bus route.

The result isn’t totally clear. For most of the key bus routes, the trend lines (the pattern within each bus route) slope down; for these routes, cash boardings in April 2016 during the off-peak period were associated with lower reliability, as expected. But in a few cases the opposite is true — the trend lines slope up, and cash boardings were associated with higher reliability. So while there seems to be an overall pattern, there is also a lot of variation in reliability that is not explained by cash boardings alone.

Reliability by share of boardings by day

To make an overall observation about how much cash boardings might affect reliability during off-peak times, we can estimate the average slope of all the individual bus route trend lines. We do this by subtracting the average reliability and average cash boardings share for each bus route (shown in Figure 3) from the daily points in Figure 4 above. Then we can find a single trendline for all the points. This is equivalent to estimating a simple fixed effects linear regression model: Reliabilityit = a + b CashShareit + BusRoutei, where i = bus route, t = day in April 2016, and BusRoutei is a set of indicator variables for each bus route.

Figure 5 below shows the results, with a 95% confidence interval shaded around the trendline. As you can see, the trend is small and noisy. It’s highly unlikely to be zero, but the true slope — the impact on reliability of increasing the share of cash boardings by 1% — could range from -0.05% to -0.67% (the 95% confidence range).

De-meaned Reliability vs. Share of Cash Boardings by Route and Day

What does it mean?

If we use the middle of the range of estimates above (the average estimate represented by the dotted line in Figure 5), then we would expect reducing the cash share of boardings by 1% to increase reliability by 0.36%. We saw in Figure 1 that average cash use on key bus routes in the off-peak period was 9% in April 2016, so we might expect average off-peak reliability on those routes to improve by 3.24% (0.36%*9) if there were no cash on board.

Any statistical estimate like this is uncertain and requires certain assumptions. For example, our estimate might not apply to such a large change in the share of cash boardings, since the range of off-peak cash use within any key bus route was only about 6%. More sophisticated statistical modeling with more data could provide more precise estimates.

But as a first analysis, it looks like there would be noticeable improvements to off-peak key bus reliability if all cash transactions happened off-board (for example at ticket vending machines). What does that mean for bus riders? To measure reliability for key bus routes, the MBTA checks the time between the previous bus and the current bus at a few bus stops along each route (including the ends of the route). Key bus reliability is then calculated as the percentage of those checks where the gap between buses was no more than 3 minutes later than scheduled; if a bus that’s supposed to come every 10 minutes hits one of those check-points 14 minutes after the previous bus, then it’s considered late.

Average off-peak reliability across all the key bus routes in April 2016 was 76%, meaning that 1 in 4 times that buses hit one of the reliability check-points, they were more than 3 minutes late. A 3 percentage point improvement in reliability from moving all cash off-board would bring those odds down to 1 in 5. That might not sounds like much of a difference, but it’s more than a 10% reduction in the number of times the bus is running late. More people show up during long gaps between buses than during short gaps, so it’s an even bigger reduction in the chance that you’ll be stuck waiting for a bus that’s running more than 3 minutes late. And these reliability benefits are all on top of the direct benefits of no cash on board — an easier boarding process, faster trips, and the potential to run more frequent bus service using the same number of buses.