This issue recently popped up on Twitter in a short conversation with @sandypsj.
One of the frustrating things about trying to put transit ridership into the context of total road use is that auto volumes and transit ridership usually aren’t reported in the same way.
When you look up traffic data, you get a point volume, usually the average number of cars passing a point on the roadway every day. Sometimes, you can also find the AM and PM peak hour volumes or peak 15-minute volumes in each direction, which are what traffic engineers use to time the traffic lights during periods of heaviest demand. When you look up transit ridership data, you usually get a total number of boardings for the entire line.
So, for the road you have the number of vehicles using only that segment, while for the transit line, you have everyone using any segment. For example, a daily count on Venice east of La Cienega showed 41,428 vehicles per day, while Metro ridership data shows 13,259 riders on Route 33 and 12,311 riders on Route 733, the bus routes serving Venice. If you assume an average vehicle occupancy of 1.2, that’s 49,713 people passing that point in cars. However, it’s not the case that (13,259 + 12,311)/(13,259 + 12,311 + 49,713) = 34% of all users on Venice east of La Cienega are using transit! Many 33 and 733 riders get on and off without going past La Cienega.
To figure out the proper comparison, you need to figure out the transit line volume for the same segment of roadway you have traffic volumes for. To do that, ideally, you need both boardings and alightings at each stop in each direction, perhaps even broken down by time of day. The number of boardings at each station is frequently available for rail lines, less often for bus lines. Data on alightings is not often available for rail or bus, though that’s slowly changing. For example, BART and the MBTA publish ridership data that includes not only boardings and alightings at each rail stop, but also each origin-destination pair. Since every boarding in one direction usually corresponds to an alighting at the same stop in the opposite direction, at a minimum you can get by with boardings at each stop in each direction.
For example, consider a hypothetical feeder bus route serving a rail transit station at Stop A, as shown below. There are ten stops, with the highest number of boardings at the transfer at Stop A, and secondary peaks in demand at Stops C and D, a subsidiary commercial node and transfer point.
We have boarding data in each direction at each stop. Since no alighting data is available, let’s assume alightings at each stop are equal to boardings in the opposite direction. We can therefore calculate the route volumes in each direction, i.e. the number of bus riders on each segment of the line in each direction, by setting up a simple table.
Northbound volume between Stops A and B is 2,000, since 2,000 riders board at Stop A and no one has had a chance to alight. At Stop B, 100 people board and 200 alight, so the route volume is 2,000 + 100 – 200 = 1,900. At Stop C, 800 board and 600 alight, so the route volume is 1,900 + 800 – 600 = 2,100, and so on. Southbound volumes are calculated the same way, by working up the column. Between Stops K and J, route volume is 400. At Stop J, 500 people board and 20, alight, so route volume is 400 + 500 – 20 = 880, and so on.
Note that because of our assumption about alightings, route volume in each direction is the same on each segment. Also note that the highest demand segment is between Stops C and D, not at the highest demand stop, Stop A. Lastly, note that while daily volumes are likely to be equal in each direction, demand throughout the day will probably be unbalanced. For example, since this is a feeder bus, we’d expect southbound volumes to be larger than northbound volumes in the morning, and vice versa in the afternoon.
Ok, now let’s suppose that daily traffic on the roadway segment between Stops D and E is 15,000 vehicles. Assuming an average occupancy of 1.2 passengers per car, that’s 18,000 people in cars. Therefore, between Stops D and E, the portion of total use being served by transit is 3,960/(18,000 + 3,960) = 18%. Note that if you compared total transit line boardings, 8,800, to the traffic volume between Stops D and E, you would significantly overestimate the portion of demand being met by transit. This example looks at daily demand; if you had traffic and transit data by the hour, you could do a more refined analysis.
It might be tempting to ignore this method, because it reveals the transit share to be smaller, but this is the right way to do the comparison. Frequent readers already know that this blog is certainly pro-transit, but also dedicated to honest analysis. When I present something, I want the backup to be airtight, so that transit opponents with ulterior motives can’t shoot it down on technical merit.
Case Study: the 24, the 680, the 242, and the 4 Compared to BART’s Bay Point Line
Twitter user @asmallteapot brought up the Caldecott Tunnel and BART’s Bay Point Line as a potential comparison between transit ridership and freeway volumes. Features like the Caldecott Tunnel offer particularly good reference points, since the tunnel creates a bottleneck where the only two options serving those trips are the freeway and the transit line.
BART provides full origin-destination ridership data, and Caltrans has good freeway volume data. In this example, we’ll compare the BART Bay Point Line between Rockridge and Pittsburg/Bay Point to the competing freeways, the 24, the 680, the 242, and the 4. The comparison between the 24 through the Caldecott Tunnel and BART between Rockridge and Orinda will be most accurate; for the rest of the line there are other alternatives that we can’t account for. This is especially true from Walnut Creek east, where the freeways are also serving trips not in competition with BART.
Here’s the origin-destination data, simplified to look at only the Bay Point Line from the Caldecott Tunnel east. Blue shading indicates westbound trips; red shading indicates eastbound trips.
Here’s the data tabulated into westbound and eastbound volumes, along with comparison to the appropriate freeway segment and BART mode share (assuming 1.2 passengers per car). As you can see, there’s little difference between the volumes in each direction. If we’d only had directional boardings, and assumed alightings equal boardings in the opposite direction, the results would be about the same.
Through the Caldecott Tunnel, BART is handling about 26% of total demand – not bad at all considering the fairly crappy off-peak headways and the fact that the freeway has four tunnels to BART’s one.
In the past, recording and compiling detailed boarding and alighting data would have been an inordinately time-consuming task, but with modern fare cards and automatic passenger counter technology, it should be quite easy, even in 15-minute intervals or at the individual vehicle level. Hopefully, more agencies will make this data available so that planners and activists can put it to good use.