Virginia Tech Department of Economics
Tony Dziepak Academic Webpage front page My Curriculum Vita My Research My Dissertation Abstract Courses and Evaluation Economics Links

Metro Delays: Problems and Solutions

Tony Dziepak
October 2004

I would like for Metro to improve their information regarding delays. What exactly does it mean when they are experiencing a 10-minute delay? The problem with Metro is similar to one which we had in the FAA before we had the Air Traffic Organization--the delay infromation is operations-oriented, not customer-oriented.

For instance, the Metrorail customer could care less about whether or not the trains are running behind schedule. What matters to the customer is how long they have to wait on the platform for the next train, and whether or not that train is going to go at full speed or less-than-full speed while they are in transit. They do not care which particular train they ride. If trains are 30 minutes behind schedule--so what, as long as they are running.

What exactly does "experiencing a 10 minute delay" mean? It could mean one of several things.

1) A delay is currently accruing, and it has been 10 minutes so far since the next train should have served the station. However, this means nothing to the customer because what counts is how long they have to wait for the next train.

2) Trains are 10 minutes behind schedule, but running. This means very little to the customer since it doesn't matter to the person on the platform which particular train he/she rides. However, it is a warning to the customer that the trains may be more crowded than the no-delay case.

3) Something inbetween 1) and 2) a delay has accrued, trains are running, but slowly due to congestion or due to manual operation, single-track sharing, or other reasons.

In all cases, the quantification of the delay in minutes is not very meaningful to the customer. The customer translates this "delay factor" to an increased misery index, which can be any linear combination of additional car acrowding, extra platform waiting time, and extra transit time.

SOLUTION: We need a customer-based delay information system.

1) What do customers want to know? They want to know
a) the ETA of the next train on their departure platform,
b) the speed with which the train is expected to travel after departure (as a percentage of full speed), and
c) the estimted capacity of the next train to handle passengers after people get off at the station at which the customer is departing.

2) When and where do they want the information?
As soon as possible, as accurate as possible. Thus the system of delay detection and information dissemination should be as automatic as possible, not requiring operators or station managers to input information.

I propose: switch from voice communication to push-button warnings, which have two functions--to alert station managers and the coordination center, automatically adjust traffic flow, and adjust the automated delay infromation. For example, the train operator presses a button for debris in the track, or for a medical emergency.

I also propose: Every train has a position sensor. There is a computer program that automatically estimates ETAs and future speed based on current position, speed, and past performance.

And finally: A relatively inexpensive crowdedness sensor: an infrared beam passes across the cars through the windows, near the top as it leaves a station. A computer program will calculate estimated train capacity by the proportion of infrared reflection. A more crowded car, with more people standing, will absorb more infrared light.


Uniform simple table displays, available at the station and online at website, and available as an email message. Each terminal station has one table message per rail. Examples:

Smithsonian (Westbound)
line   arrive speed load
Orange 0.9min  94%   83%
Blue   3.3min  93%   65%
Orange 6.1min  93%   85%
Blue   10 min  95%   63%

Rhode Island Ave (Northbound)
destination arrive speed load
Silver Spr  0.9min 101%   89%
Glenmont    3.3min  99%   95%
Silver Spr  6.1min 100%   88%
Glenmont    10 min  98%   97%
Note: Speed is a percentage of average on-schedule full speed. It is averaged over the last several stations so there is no noticeable dip in speed when the train stops at a station--unless they stay at the station for longer than average time. Average speed can exceed 100% if the operator is shortening platform time in an attempt to recover from a previous delay.

Capacity is an estimated percentage of car acpacity, which might be when all seats are taken and aisles are full of standing passengers with comfortable personal space. Capacity may exceed 100% when passengers are packed like sardines--an indication that additional passengers may not be able to get on. The estimated capacity should take into account how many passengers are expected to exit the train at stations up to and including the departure station. This is based on day of week, time of day, and other data that anticipate how many exit particular stations.

All text is in amber except estimated arrival times, speed, and load. When arrival times are less than 3 minutes peak and less than 6 minutes nonpeak, ETA is green. When the ETA of the first train of a particular line or destination is at least 6 minutes peak and 12 minutes nonpeak, it is red.

All estimated speeds at least 95% are green. Speeds under 80% are red. Estimated loads under 80% peak, 60% nonpeak are green; and loads at least 95% peak, 90% nonpeek are red.

Virginia Polytechnic Institute and State University Department of Economics at Virginia Tech Send an email to Tony Dziepak

Website address:>