Median Zillow Rent Index per square foot in Boston and Cambridge

After seeing this article on BostInno I knew something was not quite right about their reporting. West Cambridge in the top 5 median rents? Brighton rents being a third less than any other neighborhood? Their charts use the median unit price data, which does not distinguish between different types of units. I downloaded the Zillow data myself and prepared a few charts of the Median Zillow Rent Index per square foot, which I believe produces a lot more sensible looking results. And in case you are curious, I have plotted all of the data available since 2010:





Here is the data for March 2015 in a sorted table:

Region Name	City	ZRI per SF
South Dorchester	Boston	1.584
Mattapan	Boston	1.69
West Roxbury	Boston	1.708
Roxbury	Boston	1.726
Roslindale	Boston	1.762
Hyde Park	Boston	1.78
North Dorchester	Boston	1.878
East Boston	Boston	2.038
West Cambridge	Cambridge	2.11
Jamaica Plain	Boston	2.128
Mission Hill	Boston	2.22
North Cambridge	Cambridge	2.22
Aggasiz - Harvard North	Cambridge	2.27
The Port - Area 4	Cambridge	2.388
Cambridgeport	Cambridge	2.502
Wellington-Harrington	Cambridge	2.524
Peabody	Cambridge	2.546
Brighton	Boston	2.652
Riverside	Cambridge	2.654
Charlestown	Boston	2.666
South Boston	Boston	2.684
Allston	Boston	2.702
Mid-Cambridge	Cambridge	2.868
East Cambridge	Cambridge	2.918
West End	Boston	3.094
Downtown	Boston	3.258
South End	Boston	3.276
North End	Boston	3.462
Back Bay	Boston	3.512
Chinatown	Boston	3.622
Kenmore	Boston	3.658
Beacon Hill	Boston	3.726
Fenway	Boston	4.144

I hope that this is more useful than the BostInno presentation.

Two charts showing development in Boston over the decades

I've been playing around with the Assessing data from the city of Boston, here are a couple of charts to start. First, a summary by decade of the total amount of gross floor area, living area, and parcel land developed:


Second, the corresponding Floor Area Ratio (gross floor area divided by land area) of all properties built within each decade:


It's important to remember that the Assessing data only contains properties that continue to exist to this day, so anything demolished would not be recognized in this data set. I also noticed that in older properties, year of construction was sometimes rounded off to the nearest decade. And it's possible there are other errors. I had to correct a few more egregious and obvious ones (like fields being swapped), but more subtle errors could sneak by. Condos are handled by summing up the gross floor space for each unit and linking it to the land area used up by the overall building.

Without spending too much time on analysis (that's for later), I'll note that we live in a turn-of-the-twentieth century city: most of the floor space created and still existing seems to have occurred between 1890 and 1930. The 1930s and 1940s had some fairly obvious reasons for a lull, but construction never really picked back up afterwards at the same rate. My hypothesis: Zoning in its modern form was enacted in the mid-1950s, which has put a heavy damper on construction ever since.

The FAR chart shows that development generally hovered around 1.0 floor area to land area, but started to drop precipitously after 1930, until finally tanking at a miserable 0.21 during the 1960s. Although the amount of floor space developed increased from the 1950s into the 1960s, the amount of land area consumed zoomed up even higher.  Zoning could explain some of it, but it's not clear to me why the 1960s are such an outlier in terms of land area consumed. (Turns out it was MassPort's harbor holdings, at 101 million s.f., which I have cut out from the data). By the 1970s, overall floor area ratios returned to a more historically normal average of 1.0 or so. More recently, overall construction averages have exceeded 1.5 FAR, albeit using incomplete data for this decade.

For the curious reader, here is the result of summing over all the parcels in the database:

• Boston parcels gross area: 651,202,719 s.f.
• Boston parcels land area: 1,268,597,774 s.f.
• Boston parcels FAR: 0.51

Taking a look at Green Line average speeds by hour of the day

Stairs can make boarding
be unnecessarily difficult
for some people.

With the real-time data available for the surface Green Line since mid-October, I thought I'd try some analysis with my gathered database. One thing we can look at is average speed of travel. Although the trains can ostensibly move at up to 25 mph in their dedicated lanes, or even faster off-street, they tend to spend a lot of time waiting at signaled intersections and station stops. In fact, the more passengers that are trying to use the system, the slower trains tend to go: overcrowding results in long dwell times as it becomes harder and harder for people to squeeze on and off the train. This effect is made worse whenever the MBTA forces everyone to use only the front door for boarding and alighting the train. The front door is small, and it has stairs, which makes the process take much longer as everyone has to slowly file in and out of the tiny opening.

The net result is that the Green Line can sometimes feel like it is traveling more slowly than if you walked. Well, perhaps that is hyperbole. However, it is possible to outrun a Green Line train on the "B" branch. A typical human runner, with some training, can sustain about 8-9 miles per hour. So we should expect to see average "B" branch speeds in a similar range.

The first thing we can do is consult the schedule by using a convenient mapping website to plot a trip from Boston College to Blandford Street at 8 o'clock in the morning. We find that the MBTA expects a "B" train to take about 27 minutes to cover the approximately 4 mile distance. That's about 8.9 miles per hour. Does that match up with the findings from the real-time data?

The following zoomable line chart is derived by calculating average speeds for the surface "B" branch during each hour of the day from 5 a.m. through 11 p.m., and breaking it down by week. So, for example, we can see that on the week of November 3rd, average speed at 8 a.m. was about 6.79 MPH and at 5 p.m. (hour 17) the average speed dropped to 6.61 MPH. That corresponds to an approximately 36 minute trip between Boston College and Blandford Street. Note that this chart does not distinguish one direction from another, so it is quite possible that the peak direction is much slower and the off-peak direction moves more quickly.


Plot of "B" branch average speed in MPH for the hours of the day from 5 am until 11 pm, looking at each week in November and December separately.

The weeks with the most rapid travel appear to be the final two weeks of December. This is not surprising. Much of the passenger load on the "B" branch comes from the universities, and those are on break during that time. Slightly surprising is the distant third-place finish for the week of Thanksgiving. Although the performance is better than the "normal" weeks, it is still rather low for a week when most students are traveling.

The "B" branch gets busy during the
off-peak hours as well.

In any case, the shape of the chart overall is telling. The two major valleys correspond to the traditional rush hours, when crowding on the Green Line grows extreme and dwell times increase accordingly. The most snappy performance is found either in the very early morning, or later in the evening. There seems to be a slight bump around 9 p.m. that might correspond to the late dinner and nightlife rush.

The difference between an average speed of 6.61 MPH and 10 MPH doesn't seem like much, but it is actually quite a bit: that's a difference of 12 minutes in end-to-end trip times. If the Green Line can manage an average speed of merely 10 MPH on a consistent basis, then a typical passenger can realize a savings of up to 12 minutes on their commute each way. That's pretty significant. Especially if you are able to increase reliability and cut out the variances that can sometimes lead to 15, 20 or more minute delays.

That's why I find prosaic improvements such as proper station spacing, transit signal priority, and all-door boarding to be vital. They don't have the same flashy pizzazz as building a new section of subway. But they can achieve much of the same benefit at a fraction of the cost.

The SPOT app undermines the Clean Air Act, and therefore our air quality, in Boston

I recently read about the "SPOT app" that allows people to easily rent out an empty parking spot that they own and are not using. Sounds reasonable enough. I'm a fan of making more efficient use of physical resources. It's the opposite of "minimum parking quotas" that force everyone to waste huge amounts of land and money, and yet still fail to meet parking demands.

"SPOT app" (source: BostInno)

But, when I took a look at the map included with the article, it occurred to me that there is something not quite right about this. The app allows you to rent spaces in the Back Bay and downtown Boston. It transforms so-called "accessory spaces" that are attached to particular uses (such as residences) and allows them to be used instead as "commercial spaces" that are available to anyone, for a price.

So what's the big deal?

Well, back in the 1970s, the city of Boston was facing an air pollution problem caused by the creation of all those urban highways that tore through the city, bringing hundreds of thousands of cars spewing exhaust fumes into the air. In order to satisfy provisions of the Clean Air Act, the city of Boston agreed to cap the number of "commercial spaces" that would be available at any one time. It's called the "parking freeze" and it's intended to help preserve our air quality. The downtown Boston parking freeze cap is currently set at 35,556 spaces, and there is no capacity for new spaces at the moment. Yet, the SPOT app is effectively creating new commercial spaces that have not been subject to the parking freeze regulations. That means more cars, more air pollution, and more congestion.

I think it would be appropriate if the company that created the SPOT app were to be proactive about dealing with this air pollution problem. Perhaps they should disable the use of the app within the parking freeze zone until they figure out a way to mitigate the air pollution caused by the additional cars they may be attracting into downtown Boston. Perhaps it should only be available for electric vehicles within that zone. Or perhaps they need an allocation from the freeze "bank" in order to offer spaces in that zone. I don't know what the best solution is. But I do know that it is something that should be addressed. 

I'm also a little disappointed that City Councilor Frank Baker did not consider the implications for the parking freeze, the Clean Air Act, and our air quality, before providing an endorsement of the app.

One thing I did find really interesting is that the company behind this app has been collecting price information about short-term parking in various parts of Boston. This article has a breakdown. Average prices range from $1.75/hour in Allston/Brighton up to $3.75/hour in Back Bay. What's remarkable about these averages is that (a) they're higher than the city-wide set meter price of $1.25/hour, and (b) they're really not that expensive, and quite reasonable when compared with typical meter prices in other American cities. Even the Back Bay's average market-driven price of $3.75/hour is less than meter prices in busy parts of Vancouver, New York, Chicago, Los Angeles, and Seattle. Does it really make sense for places like the Back Bay and the South End to have city meter rates that are comparable to Boulder, CO or Rochester, NY (both $1.25/hour)? Hopefully, this inspires the city to give another look at using smart parking reform to address parking issues, instead of hurting the residents of the city with onerous minimum parking quotas. Those quotas are especially harsh on people who don't even own automobiles and yet are still forced to pay the cost to park other people's cars.

Boston the walking city

Spotted this data visualization by the creators of the "Human app" for the iPhone. I don't have an iPhone so I'll let them explain it:

Human is an iPhone app that runs in the background of your phone and automatically detects activities like walking, cycling, running, and motorized transport. All visualizations are solely based on aggregated data from people using the Human app. Imagery shown does not involve the use of maps, as white pixels were drawn by moving Humans.

Walking in Boston as measured by the app (source)

In addition, it finds that 46% of activity in Boston was walking, compared to 43% motorized, 4% running, and 6% bicycling. I also noticed that the visualization of motorized activity includes the surface portions of all the MBTA rail lines: if you look closely you can see them all traced out pretty clearly where they diverge from other roadways. No tunnels are included in the charts at all, presumably because GPS doesn't work underground. So, it seems that in this case, "motorized activity" also includes riding the T. Boston's high walking rate puts it in the top ten worldwide, and only behind Washington, D.C. and New York City in the United States.

I think that's a nice confirmation of what we already knew from census and survey data. If anything, it may show that those estimates were too low, especially since they often focus only on journey to work, which is usually considered to comprise about 20% of total trips. I think that the main criticism of this work, and the one thing that holds it back from being scientifically useful, is that it is only measuring people with iPhones AND the Human app. And I have no idea how well iPhones are distributed among the population, but I can guess: they probably skew towards the rich and technologically literate. So there are some fairly egregious gaps in data. For example, much of southern Boston is simply ... missing. I couldn't tell you whether that's because walking is rare there, or because iPhones are less common. Based on personal experience, I don't think walking is rare in neighborhoods like Roxbury and Dorchester, so I think it's more likely that there are just fewer people carrying iPhones with the Human app installed.

When's the last T?

After the inauguration of late night weekend T service I cobbled together a quick web site as a convenient portal to find out when the last trip of the night would be passing nearby. I didn't get a chance to polish the interface until this month, but now I think it is usable enough for others to try out. Point your browser, desktop or mobile, at tinyurl.com/lastmbta. If you are on a mobile GPS-enabled phone, then it will attempt to pinpoint your location automatically. Or you can just click or tap a location on the map:

Press the "Submit" button and you will get an easily printable screen that lays out the options:

I hope it is useful. Feedback is welcome.

More fun with maps: walksheds and transit lines

As a follow-up to my first look at the contest data, I have mapped the "transit walkshed" coordinates from the 37 Billion Mile contest. This one took a bit more doing since the shapes were linked only to the names of stations, and those names were formatted differently from the way the MBTA official data publishes them. So some "fuzzy" string matching was in order, plus hand tweaks.


The result is worth it, I think. It shows the rapid transit lines, the Silver Line, and some of the key bus route stop walksheds (others seem to be missing). They overlap quite a bit so the colors combine into intermediate shades. The walksheds seem to be streets within a 10 minute walk of the station stop, or so. It's interesting to compare the shapes in different locations, especially seeing the effects of different block sizes and connectivity levels on accessibility.

A first look at the "37 billion mile" data challenge, with some maps

MAPC has put together a data set from MassDOT that consists of the entire Commonwealth's vehicle registration data (suitably anonymized), including odometer readings, and then summed it up in quarter-kilometer square blocks. I grabbed a copy of the data and plotted a few simple Google Fusion maps to check it out:

Household Vehicle Miles per Day

You can see that the data is presented in blocks that are 250 meters on a side. MAPC prepared an estimate of how many miles per day each household travels in one of their cars. That can range from zero, for households with no cars, all the way up to this one crazy person who apparently drives 646 miles per day. I have chosen a color scheme such that the miles driven is divided up into quintiles based on city of Boston map squares. A fifth of the Boston map squares fall into the category of "0 to 11.98 mi/day", the next fifth are "11.98 to 17.65", etc. Although the map shows regional data (and state-wide data is available in the complete set), I decided to focus the color scheme on Boston because I am most interested in the car-free and car-light households.

Vehicles per Household

The same largely goes for Vehicles per Household, which is obtained by taking the number of households in each map square and dividing it from the number of vehicles registered to an address in each map square. Note that "households" is defined as the number of households counted by Census 2010, and registered vehicles were geocoded into the various map squares but that process was not always successful. There have been some slight adjustments to the numbers, according to MAPC documentation, to account for the vehicles that could not be accurately pin-pointed. I will trust their models for now.

Car-free Household Percentage

The last one just highlights where the households with the least number of cars are found on the map. I don't think there are too many surprises here, especially for people familiar with my map based on Census/ACS data. The effect of the Green Line is pretty pronounced, it's easy to see the trace of Commonwealth Ave, Beacon Street, and Huntington/S. Huntington Aves. In fact, the abrupt end to blue-colored blocks just past Heath Street seem to indicate that the loss of the Arborway trolley has really taken a toll in terms of increased car ownership around Centre Street. Or maybe it was increased car ownership that led to the cutback. Another interesting pattern is around the new Fairmount/Indigo Line stations. Most of them seem to be near pockets of car-light or car-free households, even though the stations are relatively new. No doubt, that was a key motivation in the planning of the line.

Quick note on outliers: sometimes you may spot a square that is radically different from its surrounding squares. It could easily be a matter of small sample size, or just a local facility that is skewing the data. For example, a nursing home. Or the Four Seasons hotel, which seems to have about 170 vehicles registered between it and the Boston Public Garden. Sanity checks are always a good idea.

The need for station consolidation on the MBTA "B" branch, in one chart

Update: MBTA announces station consolidation public meeting, Oct 23rd.

Original Article:
A little while back, BU put together a Transportation Master Plan that included some detailed ridership data about the MBTA Green Line "B" branch. I somehow missed this appendix while looking through their master plan in the past. Thanks go to Eric Fischer for alerting me to my oversight.

The data include hour-by-hour breakdown of boardings, alightings, and line volume (the number of people riding through) for each of the surface stations Blandford Street through Warren Street, taken on a Fall 2010 day. It occurred to me that a station's importance is related to the number of people who use it, versus the number of people who would prefer to ride past it. This is also tied to the relative proximity of the stations. Multiple, closely spaced stations will divide up potential ridership and therefore, will mutually reduce the importance of each such station.

The four stations in the scope of the upcoming Commonwealth Avenue Phase 2A street rebuilding project are very closely spaced. They are some of the most closely spaced stations in the entire system, as they are placed on four consecutive city blocks approximately 730 feet apart on average. That is a 2-3 minute leisurely walk between stations. Perhaps only Back of the Hill and Heath Street are closer.

The diagram above is drawn to scale, and it shows how the stations are spaced relative to one another. Above the station names is a bar graph which shows the relative importance of each station, as measured by the number of people who use the station (boarding, alighting) divided by the number of people who ride through the station. It should come as no surprise to any frequent user of the "B" branch that Harvard Avenue is the most important station, by far, and that the four least important stations are: Babcock Street, Pleasant Street, St Paul Street, and BU West; the last being the most underutilized.

I believe that this diagram makes a strong, visual case for station consolidation during the upcoming Phase 2A project, in addition to station relocation to better accommodate signal priority. I would like to see the MBTA begin a public process, much like they did ten years ago, to work on implementing efficient station spacing for this section of the Green Line.

In addition to improving the experience for about 30,000 people, a shorter round trip time will save resources for the MBTA. If the scheduled round trip time for a train can be reduced by 6 minutes, then that means one fewer train is needed to operate the same peak schedule. That train can then be put to better use in other ways: to split it into more 3-car trains, to help bring back 5 minute frequency, or to send it out for much needed maintenance.

Even more savings are available if the MBTA finally implements all-door boarding and transit signal priority, but that's another discussion.

A quick note on the gap between BU West and BU Central: that space is mostly occupied by the Mass Pike trench, currently. The overpass of the Pike will be reconfigured as part of Phase 2B, and then I believe that BU has some air rights projects in mind. If station consolidation is properly implemented in Phase 2A, and if BU ever builds over the Pike, then it may make sense in the future to relocate BU Central further west. But until then, the most pressing issue is figuring out how to fix the Green Line problems within the scope of Phase 2A.

Walk audit of Union Square, Allston and suggestions for the future

Union Square, Allston. The view from the Twin Donuts shop, facing east.

(Updates in bold posted 12th November 2015).

Let's take a look at Union Square in Allston, a vibrant neighborhood of the city of Boston. This is the convergence of several busy streets, a transit hub for buses, a retail district, with an elementary school and plenty of nearby residents living within a few minutes walk. I will call this a kind of "audit" because it is based on my observations, as an outsider. The audit uses timings from a weekend study. The main difference between a weekend and a weekday here is length of the full cycle: 90 seconds on the weekend, 110 on the weekday. Adjust accordingly.

Union Square, landmarks and street names. All satellite imagery courtesy Google.
Click on it for larger view.

Key points from below the jump:

• The "level of service" given to the Union Square intersection, by the New Balance transportation study, is "D" (35s to 55s average delay/vehicle) during weekday a.m. peak and "E" (55s to 80s delay/vehicle) during p.m. peak and Saturday peak. "D" is considered normal in cities.
• Traffic levels have declined by about 5-6% here since 2002.
• An examination of the signal schematic (below) shows that for cars, no direction of travel has to wait more than 72 seconds to be given a green phase, on a weekend.
• Yet, the minimum time for a pedestrian to cross Union Square (north/south), legally, is about 100 seconds, if you happen to arrive at the intersection at the optimal moment in the cycle.
• For the less lucky, the expected crossing time is about 145 seconds with a worst case of about 190 seconds.
• For the less capable walkers, the worst case is about 258 seconds. If you happen to be elderly, or a young child. the city expects you to wait 4.3 minutes to cross 95 feet. That's a rate of 0.25 mph.
• The green right turn arrow given to traffic moving eastbound from Cambridge Street is in direct conflict with a walk signal for people crossing between the school and the fire dept station (see picture below).
• The 21 seconds of walk phase given to the 110' crosswalk of Cambridge Street between the Jackson Mann school and Twin Donuts is not enough time for a healthy adult, much less a person with disabilities or a child going to the school.
• There is no crosswalk marked at the Hano Street bus stop, despite heavy ridership. Instead, the city put up a new sign telling pedestrians not to cross there, in response to a tragic crash.
• In the long-term, a Poynton-style redesign should be seriously considered, with a shared space design such as a single-lane roundabout (or similar) completely replacing the traffic signal-controlled intersection. All users of the intersection would benefit, as would the surrounding neighborhood.

Thought experiment: how much bus service can you get for the price of a parking garage?

Looking through some old data, I noticed that there's a common theme: the marginal operating cost per weekday of your typical frequent MBTA bus route is about $20,000. For example, the 1 bus is $18,456 while the 39 is $22,206. The 57 comes in at about $19,544. All of them are estimated to recover approximately half of that in fares, so overall marginal subsidy per day is about $8,000 to $12,000.

Mind you, the marginal operating cost only factors in the cost of running a bus per hour, plus the cost of driving it per mile. Maintenance, depreciation, administration and other facilities are not covered. But if they had the equipment lying around, then it might be fair to say that adding a weekday's worth of 57-like bus service costs approximately $10,000.

We know that excavating an underground parking garage can cost from $50,000 to $100,000 per parking space (sometimes more, sometimes less, depending on conditions). Speaking loosely, then, each underground parking space could cover the net cost of approximately 5-10 weekdays worth of key bus route service. Let's just assume for simplicity that every day has the same cost as a weekday. Then a year's worth of key bus route service could be covered for the same cost as 36 to 73 underground parking spaces.

A year's worth of bus service. (photo source)

Does this come up in real life? I think it's fair to expect several hundred parking spaces to be built along every core bus line every year. I can think of several hundred that came online along Comm Ave just last year, including some built at a deep (expensive) level. And another few hundred planned for Washington Street in Brighton. The Fenway is seeing plenty of garage construction despite the parking maximums in the neighborhood. Then a quick browse of the BRA's website shows 216 spaces (aboveground) for Kenmore Square, 236 spaces (3 floors below ground) on Harrison Ave in the South End, and 63 underground spaces on E Street in South Boston, just to pick a few examples.

I have been talking marginal costs up until now, but spaces are typically not added one-by-one. Instead, it's whole levels at a time. For example, a developer might propose 80 dwelling units and 60 parking spaces in an underground level. Some misguided neighbors might demand more parking spaces be added. But the only way to do that would be to go yet another level underground. Perhaps the developer can add another 40 parking spaces by doing that. The garage now costs about $5 to $10 million to excavate. Those costs are passed onto to the eventual tenants, and create more traffic and pollution in the area.

Alternatively, the developer could propose to pay for a year's worth of frequent bus service. The cost of adding that additional underground level could easily cover the marginal yearly cost of running a key bus route. Furthermore, doing so would help everyone in the neighborhood rather than a handful of car-owning tenants, would ease traffic and pollution instead of increase it, and would contribute to a better urban environment. And there's likely more than one such eligible development project per year.

I have heard people propose the option of replacing parking subsidies with transit subsidies. It's also a similar idea to value capture. It seems like it may be feasible after all. Some might say that it's unfair to put the burden of providing the public service on a private project (the usual objection to value capture) but I believe it is also unfair for the government to force people to subsidize parking spaces. If you're going to have parking quotas, which are ultimately harmful to the public, then it behooves you to allow them to be replaced by transit support payments, which are beneficial to the public.

How slow were our buses?

With the key bus route improvement project slowly moving ahead, but slipping its already long delayed schedule, I thought it might be interesting to take a look at average bus speeds from last year's data. Here's all routes under 10 MPH. Key/SL bus routes in bold.

Seeing the 1 and the 66 show up in the worst ten is no surprise. Both of them are massively overcrowded at most hours, which lengthens dwell times. The SL4/5 are also crowded, and they also obtain their infamy from the slow crawl through Chinatown. So much for the vaunted Bus 'Rapid' Transit. Having said that, they do have better on-time percentage than most, hitting the low-80%s, which is itself a sad reminder that most of the bus routes cannot achieve even the low service delivery standard of 75% on-time. I will be interested to see if the key bus route improvement project can do anything about the miserable slowness and on-time performance of the 1, 66, 15 and 23. Expectations are low.

Let's put aside the 114, which barely runs at all. Despite being non-key, the 69 serves significant ridership, and has a fairly straightforward routing. It just appears to be a congested corridor. The 55 runs infrequently, has low ridership, and is slow. I imagine that most of the riders in the Fenway with a choice will walk over to one of the Green Lines. Probably neither one of these has any chance of being improved in the near future (although I did notice some tinkering with stop locations on the non-key 86 route, for some unknown reason).

Another site of interest is Bostonography's bus speed map, but as you can see, their colorization is too coarse, marking anything under 10 MPH as "red." Most of the "green" is due to express segments of bus trips that occur on highways.

Buses carry about 400,000 rides per weekday in the system, and the vast majority of them are averaging under 10 MPH. A large part are averaging under 8 MPH. That's pretty bad. The MBTA along with the various cities and towns could be doing a lot more to fix this, and it would wind up saving money on operating costs. Even without bus lanes (which I encourage) a whole bunch of cheap improvements could be done: stop consolidation, curb-extensions for level boarding, off-board payment, proof-of-payment, signal priority, queue jump lanes, etc. Some of these are in the works for the key bus route improvement project, but it should not take 3 years to design and implement them. The big advantage of buses is supposed to be that it's cheap and easy to roll out and tinker with their routes. Why is that not the case here?

Is fare evasion really a problem?

(continued from previous post)

Something that caught my eye in the MBTA ROC report was that it was the first time that I've seen anyone take a real crack at estimating fare evasion. Usually the T just makes up numbers. In this case, it was CTPS that supplied the data, which is found in Appendix D. The ROC is proposing that the student pass program would eliminate the "need" for the "front door-only policy" that is currently in effect. I don't think that the "front door-only policy" was ever a good idea at all, student passes or not, but let's take a look at the data that we finally get to see.

A few notes: This table claims that FY12 trips on Light Rail were at 52,418,000 and Heavy Rail 128,803,000 which, if true, represents a decline of over 20,000,000 trips in each mode, down from what's shown in FY11 NTD data. That seems strange, even considering the fare hike. On the other hand, Bus and Trackless Trolley ridership is up by a million or so.

The table at the bottom shows "Non-AFC" Light Rail trips as being about 12% of all Light Rail trips, but there's no rationale given for that 12%. It's hard to reconstruct this number without knowing how many of the trips counted came through fare-gates and how many were surface-originating.

There's a similar problem when comparing the Total "Evasion" count to Non-AFC Trips, which is about 32% but I have no idea where that proportion is sourced from either.

The average fare is scaled by a factor of 1.25 -- presumably due to the fare hike of last year, and because the average fare is otherwise being determined from FY11 data.

The top table seems to indicate that the rate of people using the rear door is about 9% of riders. Caveat: this doesn't mean they were not paying, it just means nobody checked -- the MBTA's old "Show'n'Go" policy encouraged folks to use the rear door but they would be counted as "DNP" under this category.

At the bottom, there is a handwritten note that $912,000 was lost "from rear door" and this is quoted in the report as "lost approximately one million dollars per year." Putting aside all the other concerns from above, let's assume this is true. The lost million dollars sounds pretty significant, but let's put it in context. The table claims that there were 52.4 million trips on Light Rail at an average fare of $1.21 per trip. That means a total revenue of $63.4 million. If one million in revenue was lost, then that means a loss rate of 1.5% which is significantly lower than 9%, and even lower than past estimates of 2-4%.

In pretty much any system, a 1.5% fare evasion rate would be regarded as a resounding success. Nothing is perfect, and fare collection has very significant costs. Trying to pursue the last 1% could easily cost more than it's worth. And that's assuming that the revenue was "lost" which is not necessarily true. It could be mostly people with passes.

If these numbers are actually correct (admittedly, I am starting to doubt) then there is no reason to pursue fare evasion as an issue. It's a total waste of time and money that will actually cost more than it will collect. And it also harms the riding public, by slowing down trains, bunching them up, dragging down performance and schedules. And that costs real money: if extra trains are needed then you're talking over $200/vehicle hour just to keep on schedule. And that doesn't count the time-cost to passengers on board, something which T management never seems to think about, because they treats their customers like their time is worthless. Likely a holdover attitude from the bad old days of "managed decline": when the T was supposed to die in favor of universal car ownership.

Even if the fare evasion rate was doubled to 3%, that would still not be significant enough to worry about. When should it be worried about? That's easy: when the cost of enforcement is lower than the cost of loss. And the first step towards understanding that is getting real data.

Moral grandstanding about fare evasion is selfish and foolish. Show me the numbers, show me the money, don't waste my time.

Car-free residents by census tract

I was looking at some census data out of curiosity today and decided to plot the percentage of car-free residents by census tract in and around the Boston metro area. The data come from the 2007-2011 American Community Survey 5-Year Estimates.


The highest rates of car-free living occur about where you'd expect: North End, Beacon Hill, Chinatown, Fenway, Allston/Brighton and Mission Hill. There's also a pretty decent rate scattered across Roxbury and Dorchester. I was a bit surprised at how high the rates were in some outlying areas though. For example, around Malden Center there's a tract where 35% people claim to have no car available. Several Lowell tracts exceed 40%. One corridor that sticks out to me is Commonwealth Avenue through Brighton, which has consistently high rates of car-free residents, and can clearly be seen traced out in darker colors. This, despite the fact that the "B" branch has probably the worst quality T service of the rapid transit lines.

Understanding bus RPI

The MBTA quietly added a feature to their website called "Bus RPI" which you can also access from a link at the bottom of the homepage. These "Route Performance Indicators" are essentially the same data as the Biennial Service Plan but should be updated more frequently. The current RPIs only include a subset of the information about each route that the Service Plan did, but hopefully that will be changed in the future. The intention of this release is to be more open about the service planning and scheduling process that goes on at the T.

Route Number, Route Description: these are fairly self-explanatory, I hope.

Route Type: divided up into Local, Key, Community, Commuter, Express or NA. The main importance of this field pertains to the goals that each type is expected to meet. There is also a difference in "average fare paid per passenger" for the different types. Under pre-FY2013 budgeting, the average fares are as follows:

• Local, Community and Key routes: $0.76
• Commuter routes: $2.15
• Express routes: $3.08

Average fare is computed by dividing the total revenue for all routes under that type by the number of riders of that type. It is smaller than a single ride cost because of numerous discounts, including senior, student and monthly passes. CTPS projections of the impacts of last year's fare increase seem to indicate that the new average fares will be approximately 23-24% higher, e.g. at $0.94 for local routes, but we won't know for sure until more data is released. Note that SL1, SL2 service counts as Rapid Transit for average fare purposes ($1.39) but SL4, SL5 receive the local bus average fare per passenger.

Net Cost per Passenger: This number is calculated by taking the difference between operating cost and revenue and dividing by ridership, for each route. However, revenue for each route is computed by taking the average fare for the route type and multiplying it by ridership, so it is not actually revenue that is specific to that particular route. In other words, Net Cost per Passenger is really operating cost per passenger minus average fare.

A common confusion with this metric is that it represents an amount that each additional passenger costs the T. That's not what this means. Take the popular route 1 bus which has a net cost per passenger of $0.63. If I decide to wander down to Mass Ave and hop on the bus, it does not cost the T an extra sixty-three cents. Instead, the T has invested approximately $18,500 dollars on a typical weekday to operate the full schedule of route 1 service. There are normally about 13,300 riders on a typical weekday, giving the net cost of (18,500 / 13,300) - 0.76 = $0.6309. Now add my trip to that total, and you end up with (18,500 / 13,301) - 0.76 = $0.6308, an ever so slight reduction in net cost per passenger. Bus service is a fixed cost, the net cost per passenger is the fixed amount distributed over all the users of the service, that's all.

The astute may notice that some routes have negative net cost per passenger. For example, the SL5 is ($0.03). This is a good thing, but once again, for the same reason, it shouldn't be treated as if each additional rider is profiting the T by three cents. It should also be noted that there are additional costs besides operating costs, which are computed by a formula which only factors in revenue hours and revenue miles.

The net cost per passenger metric is best used as a comparison between MBTA bus routes to see which ones are efficiently and effectively serving customers.

Service Delivery Policy Standards: A set of requirements for frequency, span, crowding, and cost which are more fully outlined in the Service Delivery Policy. Basically: is the service decently usable, and is it not costing us too much?

Ridership Info: This includes both daily (weekday) ridership, as up-to-date as currently available (seems to be 2010) and average rides per trip. The ridership numbers are not from the farebox, so they should reflect real ridership including people who show their passes and don't interact with the automatic fare collection machinery. The average rides per trip is simply ridership divided by number of (one-way) trips, which is intriguing, but the numbers currently shown are actually junk due to an error in formatting. Hopefully that will be remedied in the next RPI release.

Thanks to the MBTA's scheduling team for getting this data out, hopefully it will be expanded in the near future. I'm also looking forward to the next edition of the Blue Book.

Lead poisoning in Boston

I first noticed some reports of the plausible link (related paper) between lead exposure and crime rates a while ago, but Mother Jones recently ran a well-written article by Kevin Drum that everyone should read, summing up the case so far. We've known for a long time that lead is toxic and harmful to mental health, particularly in children, but it's only within the past few decades that we've come to understand that there is no safe level of exposure at all. This link may be one of the most significant public health and safety discoveries of the past fifty years, and it's a shame that it has received so little attention in the past. Hopefully that will change now.

Massachusetts has been a pioneer in lead abatement. I was curious to see how it was effective in Boston. The only report available that I could find with detailed information seems to be from September 2002, but it has some interesting data about elevated blood lead levels in children, from 1991 through 2001:

I took the data and put it into a chart format as well:

It's pretty easy to see that there's been a massive improvement in just those ten years. However, Dorchester and, for most years, Mattapan had the highest incidence of elevated blood lead levels. I don't know why this was the case. Lead exposure in the 20th century seems to have originated first from the use of lead in housing materials (such as paint), and then later from leaded gasoline used in cars. Dorchester and Mattapan have their fair share of older homes, but so does most of Boston. The Southeast Expressway was constructed in the 1950s, just in time for the gasoline-lead epidemic, and is adjacent to Dorchester and Mattapan. But the North End and Charlestown are also adjacent to large highways, yet they see some of the lowest levels of blood lead in the city. They are also both built environments that date back to colonial times.

I was unable to locate any more specific data on soil or other environmental sources of lead in Boston. There is a state website to find out specific buildings' inspection status, but that seems to be it, at least on public websites. A major flaw with the above data is that it only counts 10 mcg/dL as "elevated" but more recent guidelines suggest 5 mcg/dL as the threshold, and others suggest that any amount is too much. So it's quite possible the picture changes when you add the children who are at-risk with 5 or less mcg/dL of blood lead. Another flaw is the arbitrary choice of (BRA-determined) "neighborhood" as the geographical unit. For example, Dorchester is a very large neighborhood. Even when split (by the BRA) into "north" and "south", it is still a vast and diverse place with a large potential variance in lead levels between areas.

While reading the various articles about lead I kept thinking about a few paragraphs that stood out in the recent Boston Globe series on the area known as Bowdoin-Geneva:

In Bowdoin-Geneva, a 68-block swath of Dorchester set between the rivers of traffic on Columbia Road and Dorchester Avenue, spring brings the usual signs of renewal. But it also brings something more sinister. It is a prelude to summer, which plays out year after year in this neighborhood against a backdrop of danger and a soundtrack of gunfire. 

It is a problem that seems to simply never go away. The rate of shootings here over the last 25 years is four times what it is in the city as a whole — averaging 24 a year in the last decade.

[...] 

No one seems to know exactly why. The drugs that spawned deadly territory fights two decades ago are no longer the primary fuel. Poverty, assumed by many to be a progenitor of violence, is much less severe than it was a generation ago and less prevalent than in some other Boston neighborhoods.

And there is this not-very-well-known fact: Gangs were in Bowdoin-Geneva long before the neighborhood became what it is today — at least as far back as the 1950s, when bands of white kids claimed street corners, wore identifying colors, and assigned themselves names.

They most often fought with their fists then, not with guns. But, as though a legacy handed to succeeding generations, gangs have remained even as the income, race, and social makeup of the neighborhood have changed. Bowdoin-Geneva now is an intricate architecture of alliances and loyalties based on little more than streets where boys and young men live. Slights real and perceived trigger violence between them.

This behavioral description seems to be a tell-tale sign of lead poisoning, especially with its persistence across multiple generations and the socio-economic changes in the neighborhood. From the Washington Post article:

In 2002, Herbert Needleman, a psychiatrist at the University of Pittsburgh, compared lead levels of 194 adolescents arrested in Pittsburgh with lead levels of 146 high school adolescents: The arrested youths had lead levels that were four times higher. 

"Impulsivity means you ignore the consequences of what you do," said Needleman, one of the country's foremost experts on lead poisoning, explaining why Nevin's theory is plausible. Lead decreases the ability to tell yourself, "If I do this, I will go to jail."

Then I found this 2007 Boston Globe article:

According to statistics from the city, the number of new lead poisoning cases declined from 1,300 in 2000 to 460 last year. The cases are mostly found in pockets of Dorchester, Roxbury, Mattapan, and Hyde Park. The highest concentration appears to be in neighborhoods in Dorchester's Bowdoin-Geneva Street area.

In my unscientific opinion, lead poisoning is the prime suspect behind the troubles in this neighborhood. I would be interested to see the data that the city has apparently collected, and whether anyone is pursuing this topic more rigorously.

As the days lengthen and temperatures rise, the city too is looking to Bowdoin-Geneva. It has tried fitfully since the 1970s to find a cure for the troubles here. Now, after a spate of murders last year, the city is trying again, with a vow that this time it will make a lasting impact.

If they want to fulfill that vow, instead of just repeating it every year, then I would recommend a heavy investment in lead abatement---much more than already exists---for this neighborhood. It will take money as well as time, and will probably not feel as satisfying as hiring more police officers. But all the available evidence seems to say that it will work, and more than pay for itself as an investment in the next generation.

Replaying real-time bus data, expanded

About a month ago I posted a demo playback of the real-time data from Oct 11, 2011 for the 57 bus, taken from the MBTA Data Contest sample. Now I've expanded it to include all key bus routes, the Silver Line and the CT routes.

Real-time bus data playback

I've added a bunch of options as well, because displaying all of this information on one map can get very CPU intensive. Disabling the Stop Request/Door Open markers will ease up most of the burden. You can also select which routes to display using the checkboxes on the right, as well as toggling whether to show the lines of the bus routes.

The data take awhile to load so it's been split up into pieces which are loaded in the background while the simulation begins. A progress bar will be displayed on top until it finishes loading, but the map will start showing the bus data that have been already downloaded.

It's easy to forget the scale of the system when you ride the same way every day. Just take a look, as it starts early in the morning with a lone 28 bus, and then explodes with activity as it reaches the morning peak, until the entire map is crawling with buses. And these are only the key routes.

(Tested in Chrome and Firefox; IE users may be out of luck).

114 moving violations in one hour

After yet-another close call with a red-light runner at Harvard and Brighton Avenues I decided to spend an hour on this lovely day recording the number of moving violations I could observe at this intersection.

I allowed for folks who had clearly committed to the intersection as the light turned yellow. If anything, this is an under-count of violations, as I was fairly forgiving of various questionable acts. For example, I did not count drivers who entered the intersection to make a left turn and got stuck. I did not count drivers who entered the intersection within a few seconds of the yellow, or who seemed to be unable to stop (and weren't egregiously speeding). I did not count drivers who ended up in the intersection after the light turned red if it was due to another driver's behavior. I looked specifically for intent -- for instance, a change of velocity indicating that the driver was responding to the yellow light and making an explicit decision to violate the law.

I have broadly categorized the violations as follows, in this summary:

• 114 total moving violations in one hour (12:48 to 13:48)
• 23 instances of drivers blatantly speeding through the red light, even though they had plenty of time to stop.
• 63 instances of drivers choosing to slowly roll into the intersection when they could easily have stopped.
• 28 illegal turns-on-red (No Turn on Red is posted on all directions).

This is a heavily trafficked intersection for people on foot because of all the businesses. I also observed a number of bicyclists who stopped and waited for the red light. I did not see any bicyclist blow through the red light.

Notable violations: a fuel tanker accelerated on Harvard south-bound dangerously and ran the red light. A police officer on a motorbike flagrantly blew through the light going east on Brighton, no lights flashing. A silver hatchback with plates MA 14XZ79 stepped on the gas to get through the red light going west on Brighton, and nearly rammed another vehicle, just stopping short in time. A heavy public works vehicle also ran the red light going south on Harvard.

I noticed (and you can see in the data) that drivers were more likely to roll through the light on Harvard Ave, but much more likely to speed through the red light on Brighton Ave. I suspect this is because traffic moves more slowly on Harvard than on Brighton; the latter functions almost as a 4 lane highway at this point. Also it became apparent that there was a significant difference regarding turn-on-red violations among the different corners: the corners with sidewalk bulb-outs saw fewer turn-on-red violations than the clipped corners. It looks like the clipped corners may actually invite more reckless turning motion.

It's easy to find cases of moving violations here. I counted 114 violations in one hour. There is no justifiable excuse for BPD D-14 to be putting any traffic safety resources they have available towards anything else until they first address the complete lack of enforcement of motor vehicle safety here.

Replaying real-time bus data

A while back, the MBTA released a few days worth of high-quality, real-time bus tracking information as part of a "data contest." The contest is over but the data is still available. So, for fun, I've created a new visualization of the data: a "replay" of a day's worth of real-time data for the 57/57A buses.

Screenshot of the "replay" animation

Each triangle shows the motion of an individual bus. They are colored arbitrarily to help distinguish one from another. The red circles pop up when a stop request is pressed, and the blue circles pop up where the bus opens its doors. You can rewind, pause, or fast-forward through the day, from 5 in the morning until 2 at night.

The positions are all based on real-time GPS data, so you can watch as the schedules break down due to bunching, observe buses slowing down due to heavy traffic, or see a single bus laying over multiple times. Although the rest of the data is also available, I decided to stick with just one route for now, to keep things simple.

You can see it by clicking on this link.

How NextBus changes transit

Someday, I imagine that people will wonder what it was like to be forced to use public transportation without real-time predictions and vehicle locations. Even in its current, imperfect implementation, it already feels completely essential. Some people have written that they dislike the time and money spent on such items, preferring the T to work on its core mission of providing frequent, reliable service; instead of electronic distractions. I think they are underestimating the revolutionary change it brings, and it's not that expensive or distracting. The T is already tracking its buses and some of its trains, why not make that data available to the public as well? I think most folks understand this and find it quite useful.

Here's some ways it helps: real-time predictions make long headways more manageable, at least at the origin of the trip, if not so much at connections. Although, even at connection points, if you know you've got awhile, you can at least step into a store without fear of missing your bus. Increasing frequency is very expensive; although nothing can really substitute for it, bus predictions do help ease anxiety.

Another way predictions can help is, if you know the network, you can use the predictions to choose from multiple routes when you have several options. This helps at the origin of a trip when you have to choose a bus stop to walk to, but it also helps with connections. Today, I opted to ride the 86 bus which is an infrequent route that connects to both the 57 and the Green Line, and it happened to arrive while I was waiting. Since I was headed inbound, ultimately, I knew that I could use either one of these to reach my destination. So, once we turned onto Market Street, I pulled out my phone and requested the nearest 57 stopping times. The prediction showed I would have 3-4 minutes to cross the street and catch the 57, so I opted for that. But supposing that I had just missed the 57, I could have easily stayed on the 86 until I reached the more frequent Green Line. It was nice to know I had those options, and I felt comfortable using the 86 like this. It proved to be much quicker to connect this way than to wait for the direct one-seat bus ride which was still far away. Of course, for this kind of unplanned travel, it is also critical to have a good network map with frequent service clearly indicated. And it's true that you could work something out using old fashioned schedules, but you wouldn't have the same certainty, especially given the T's unfortunate reputation for running behind or dropping trips.

One of my favorite ways to defeat the MBTA's inability to maintain proper headways is to note when bus bunching is occurring by glancing at the NextBus readout. For example, the other day I saw two buses that were a minute apart. Normally, that means the first bus is taking a long time to load and unload -- probably because of an excess of passengers. Sure enough, it pulled up completely full. The other waiting riders crammed themselves onboard, but I knew better. One minute later, a bus pulled up with several seats open. And we arrived at my destination stop only 30 seconds after the crush-loaded bus. Without the GPS tracking, it would still be possible to do this, but only if the buses were within line-of-sight.

Given proper, frequent service and some schedule discipline, the need for real-time bus predictions would be lessened. Sadly, we simply don't have that here, and for the time being, the NextBus system helps make up for that. In fact, it probably saves the MBTA money by making their poorly served routes more effective, and by increasing ridership without paying for more drivers and buses.

Anyone have any related tips or tricks they'd like to share?