Archive | Transportation Economics

A Primer on Trip Generation

As a key factor in determining rates for Portland’s proposed street utility fee, the Institute of Transportation Engineers’ (ITE, henceforth) Trip Generation Manual has gotten a lot of love lately among local transportation wonks. It is worthwhile, then, to take a quick trip through the weeds of the manual to better understand where the opportunities and complications lie when it comes to using this data as the backbone of our fee structure.

First, a bit of context. The art/science/guesswork of predicting the trip generation of various land uses is the first of four steps undertaken in transportation forecasting. Along with the ensuing three steps—determining where, generally, the trips will originate and end; choosing which mode they will utilize; and identifying the time, route choice, and other properties of the individual trips—the goal is to understand the future needs of the transportation system based on current land use and development patterns. Because most jurisdictions have a concurrency requirement—a stipulation that roads and intersections must have adequate capacity to accommodate new demand concurrent with a proposed development—trip generation is of particular interest to folks in my line of work so that we might identify what developers must do to meet this requirement, blissfully ignoring any notion of induced demand.

To this end, ITE has been aggregating and disseminating trip generation data since the first edition of the manual was published in 1976 (the current edition is the ninth). The context manifests in the data in myriad ways. For any of the 172 land uses listed in the manual, the robustness of the dataset is likely a function of both how often that particular land use arises, and how much NIMBY-ism it’s likely to inspire. The well-known suburban bent of the data owes to the fact that most development over the last 40 years has occurred in the suburbs or exurbs, so this is where the vast majority of studies have been conducted.

The trips described by the manual are one-way trips, so what one might colloquially describe as “a trip to the grocery store” is actually two trips: the trip from home to the store, and the return trip home. This is important for analysis purposes—the “trip to the grocery store” will indeed traverse an intersection along the way twice—but results in quantities that are twice as high as what one might intuit. This means that in a closed system each trip is double counted, as both the home and the grocery store would be credited with generating both an inbound and an outbound trip in this example.

The manual provides two general mechanisms for determining the trip generation of a given land use. The first is a mathematical function derived through what’s called regression analysis, which attempts to fit the cleanest possible curve to a set of disparate data points. The second is the trip rate, often expressed as a number of trips per thousand square feet. But it’s important to recognize that square footage is often not the best (or even a viable) independent variable for predicting trip generation. For example, student enrollment is a much better predictor of trip generation for schools, employee count is a better predictor for offices, and the number of ‘fueling positions’ is a better predictor for gas stations.

Interestingly, there’s another predictor for trips generated by a gas station that works better than floor area: the amount of traffic using the street that it’s located upon. That’s because gas stations generate a large number of pass-by trips, which are trips that are ultimately headed to another destination (this destination is credited with generating the primary trip) but stop at a business located directly along the way. A similar type of trip—a diverted trip—is also a trip ultimately headed elsewhere, but in this case the pit stop entails a small amount of out-of-direction travel. There are also internal capture trips, which describe trips that take place entirely on roads and facilities located within a mixed use development. Note the suburban bias there; in the city, this would likely take the form of a person parking once and visiting several locations on foot, using the public streets.

And here we have arrived at the biggest failing of the Trip Generation Manual with regard to our purposes: The manual implicitly considers only vehicular trips. Assuming that only a nominal number of trips are non-automotive might work for the ‘burbs, but this often causes the stated trip rates to be wildly inaccurate in the city. Recent work by Professor Kelly Clifton’s research group at Portland State University confirms what we might have suspected: The more “urbany” a built environment, the more inaccurate the assumption that all trips are automotive is likely to be.

Thus, there is a need to distinguish between the vehicular trips quantified by the ITE manual and what are generally called person trips, which include all trips regardless of mode. Fortunately, as Clifton verifies, the latter seems to be relatively consistent regardless of the built environment. So perhaps by considering only vehicular trips, but doing so primarily in auto-centric locations, the manual has inadvertently provided a good proxy for estimating the differences in person trips generated from one land use to the next. There are exceptions, of course—the manual will understate the person trips generated by a school compared to other land uses, for example, due to the prevalence of buses in travel to and from schools. But it seems that basing a potential street fee on person trip rates inferred from Trip Generation Manual data is defensible and keeps with the spirit of the residential fee in being mode-independent. Basing the fee on vehicular trips, by contrast, would be far more complicated to implement and would leave unsolved many of the issues with the gas tax that Chris Smith wrote about last week.

The manual offers a lot of utility with regard to predicting trip generation, but really it’s just one piece of a puzzle that fits together differently from one land use to the next, and one business to the next. To accurately model the trip generation of a particular business requires a heck of a lot more than the published trip rate, which does not consider countless predictors and is often derived from a small sample size. While using these rates as the basis of the street fee would hardly be the first or most egregious misuse of the data, it seems inevitable that it will result in some businesses substantially overpaying and others substantially underpaying. As luck would have it, that seems to fit with the spirit of this fee quite well.

Oregonian still tone-deaf on Port of Portland


The Oregonian continues to advocate paving over scarce urban greenspace for new port facilities rather than re-using Portland’s brownfields, such as the former site of the Atofina Chemicals plant near Linnton. Photo: Alexander B. Craghead, 2012.

The Oregonian continues to demonstrate its tone-deafness regarding the Port of Portland. Last week, the newspaper revealed its “editorial agenda” for 2014, one plank of which is titled “Portland’s industrial lands scavenger hunt.” The title is misleading. The editorial’s real thrust is to complain bitterly about the city’s policy towards economic development, relying on the cancellation of the Port of Portland’s West Hayden Island development as exhibit number one.

Maritime transportation facilities are of vital importance to the city and the region. Yet the paper seems to believe that the only way that the region’s maritime trade can grow is to pave over natural resources:

City planners hope to lean heavily on brownfield restoration to replenish the supply of industrial lands. In concept, it’s a good plan. Take land that currently is an environmental nuisance and has little value. With a mix of public and private money, clean it up and put businesses on the sites. But making those projects pencil out, particularly for industrial uses, might well prove as difficult as finding a way to balance the Port’s needs and environmentalists’ concerns on West Hayden Island.

Allow me to translate: The Oregonian thinks cleaning up brownfields is too costly and difficult, and therefore would have us pave over greenspaces like West Hayden Island instead.

Leaving aside whether or not the Port of Portland’s development of West Hayden Island would really have been the job creator that the paper claims, this is simply bad land use and transportation policy.

Worse, a bigger question remains: is the Port of Portland’s wish for more marine terminals being driven by regional needs, or by unnecessary and destructive inter-port competition? Does the Lower Columbia really need to be served by six different commercial portsEven Seattle and Tacoma, once hated rivals, are beginning to cooperate. Even if the region does need more marine terminals, the only reason to build on West Hayden Island, versus some other location, was because West Hayden Island belonged to the Port of Portland, versus some other port authority. Put another way, if Hayden Island were on the Washington side of the state line, nobody would have been talking about turning it into a port facility, given the ease on constructing westward along the river’s north shore, as the Port of Vancouver, USA is in fact doing. West Hayden Island, simply put, was needless inter-governmental competition at its worst.

Marine transportation has as vast and direct land use impact, perhaps as much as automobile transportation, perhaps more so. It’s time that the port authorities along the Lower Columbia began to cooperate, act together more efficiently, and make fewer wasteful land use decisions. Maybe the cancellation of the West Hayden Island port plans will open the door to a broader public debate on the matter. If so, then its cancellation will have proven to be not only a good thing for the environment, but also a good thing for the future of rational maritime transportation in the region. But somehow, I doubt it. It seems far more likely that we will continue to get “chicken little” op-eds out of the Big O, rather than meaningful debates.

The inevitable end of container traffic at the Port of Portland

In recent months, the Port of Portland’s probable loss of the Hanjin shipping company has been in the news. Local media reported on the event, largely painting it as a minor tragedy. Chris, here at Portland Transport, provided his own take, noting how that those businesses using Terminal 6 would now have to truck their goods to Puget Sound, an increase not only in cost but also in carbon emissions.

I have a slightly different take from Chris: container traffic at the Port of Portland is doomed. It is only a question of when.

Hanjin’s departure has sparked a lot of silly analysis. An Oregonian editorial, for example, blamed most of the matter on increased costs from labor disputes, an issue that the Port of Portland claims was a factor in Hanjin’s decision. (The official press release does not mention this factor, nor does the Hanjin letter to shipping customers [PDF] obtained by the Oregonian.)

Meanwhile, the Portland Business Journal sloppily threw a bunch of statistics at the matter, attempting to make the case that Hanjin in specific, and container import-export at the Port of Portland in general, were crucial to the metropolitan economy.

Hanjin’s decision comes as the region executes on a plan to increase exports, which are a significant contributor to the Portland metro economy. A recent Brookings Institution study found that exports accounted for $33.9 billion in regional economic activity in 2012, driving in large part by technology exports. (See the region’s top-five exported goods.)

In addition, a recent Portland Business Alliance study found that Oregon companies made and exported $16.5 billion worth of goods to countries worldwide in 2012, creating 490,000 jobs.

Not all of those good were moved by Hanjin, but as the largest carrier calling on Portland, Hanjin certainly accounted for a good deal of that traffic.

Reading this take from the Portland Business Journal, one might conclude that Hanjin was doing massive amounts of business out of T6, and the the Port of Portland was some kind of big time player in container shipping. Yet this is not at all the case. Terminal volume at T6 for the year 2012 (the most recent year for which data was available) stands at 183,203 TEU. (TEU means “twenty-foot equivalent units,” representing a standard twenty-foot shipping container.) Moreover, T6 has never handled more than 340,000 TEU in a year, and that was ten years ago, in 2003. While this may seem like a lot, let’s put it into perspective, by comparing the Port of Portland’s container traffic to the other major west coast terminals at Los Angeles, Long Beach, Seattle, Oakland, and Tacoma. The following graphic, based on 2010 volumes, gives you some sense of proportion:


Portland is the smallest “major” U.S. container port. Tacoma, the next larger, is more than eight times bigger. As a player in the global container shipping market, Portland doesn’t even exist.

To understand why we need to understand the special nature of containerized cargo. This is the type of stuff that is not so time sensitive or small that it can economically fly, but is of sufficient value and lightness that it can be shipped relatively cheaply and still be competitively sold at its destination. Inbound containerized cargo to the United States is often consumer goods manufactured in Asia. Outbound cargo is typically specialty products or niche materials. In Oregon, examples might be specialty cedar boards for Japanese sauna construction, pallets of Christmas trees headed to Hawaii, or a couple truckloads of Hazelnuts going to China.

Containerized cargo, because it is relatively light and of higher value, is very mobile. Outbound loads will go to the port where shipping companies can offer the fastest transit time to the final destination for the least cost, and that recipe usually means that containers end up shipping in and out where other containers already are, since competition breeds lower costs. It also means that containers tend to go to those ports that are closest to their final destinations, thus reducing transit times. This is why goods heading to Europe typically are trucked or carried by railways to Eastern ports, while goods heading to Asia typically go west to Pacific ports.

So far, you’d think, so good for Portland. Since we are a Pacific port, we’re closer to Asia. But so are all the other ports of the U.S. West Coast. This is were port competitiveness begins. Note that when Hanjin announced its pull out from Portland, the Puget Sound Business Journal treated it as good news: container traffic lost in Portland would likely relocate to Puget Sound ports. In the American shipping world, what is bad for one port is good for all the others.

To compete, each port has its unique characteristics. Portland, along with there ports of the lower Columbia River, has a geographic advantage by being at the mouth of the Columbia River Gorge, the only water-level route through the Cascade-Sierra divide. This makes these ports naturally strong for handling bulk materials, where mobility is heavily restricted by weight and a high shipping cost. According to statistics published by the Port of Portland, Portland Harbor — which includes both the Port of Portland and several private terminals along the Willamette River — is “the largest wheat export port in the United States, the largest mineral export port on the U.S. West Coast, and the 4th largest export tonnage port on the U.S. West Coast.”

Container traffic is, however, highly mobile. Portland is located 100 miles away from the ocean, up a river with a relatively shallow depth (43 feet) and that must be dredged, and on the other side of one of the most treacherous river bars in the world. Put another way, to serve Portland, you have to have a reason to spend the extra time, money, and risk to reach it. All that grain can’t easily or cheaply move elsewhere, but containers — especially if they trucking in anyway and are not being filled directly from Portland producers — can fairly easily and cheaply end up at other ports.

And over the last half-century, that’s exactly what has happened. Railroads, truckers, and shipping lines have all contributed to the development of major container traffic at every major port of the U.S. Pacific Coast–except for Portland. No labor agreement or policy change is going to alter this historic trend.


Breaking down the economics of bus vs MAX

There’s been a bunch of stuff in the comments of the SW Corridor: Transportation Bundles article about the relative (operational) efficiencies of bus versus light rail. The discussion mirrors a debate that occurs in the wider community. Two dueling theories are commonly articulated:

  • That MAX is more efficient, operationally, than bus–that in busy corridors, it can move far more people than can the equivalent number of busses, and requires a lot less subsidy. TriMet frequently articulates this position in its public messaging.
  • That MAX is economically inefficient–an unwise investment at best, and a rip-off at worst; diverting resources from (and starving) the bus system, and a major contributor to the agency’s budget problems. Many agency critics make this charge–ranging from conservative/libertarians broadly opposed to any capital investment in transit, to poverty advocates on the left, to “good government” types suspicious of any large capital expenditure, to many in ATU757, to riders of “marginal” bus routes whose service is imperilled by the agency’s budget woes.

Who’s right? As usual, things are a bit more complicated than a bumper-sticker slogan can capture. After the jump, we get into the numbers. I’m only focusing (for the most part) on operational expenses–the question of capital costs are ignored. The article also focuses on bus service vs light rail–the Streetcar, WES, and LIFT are discussed, but in far less details.

The raw data
A few resources which are the primary sources for this article; all are published by (or based on data published by) TriMet.

There are several different ways to approach the numbers; we consider each in turn:

The big picture: Looking at the budget (pp20-22), we find that TriMet has allocated $314M for its Operations Division, by far the biggest chunk of its total operating budget of $473M. (Other expense items include $16.5M for non-grant-funded capital projects, $47.8M for pensions/OPEB, $48.2M for debt service, and $47M for various administrative functions). Of that $314 M, we find the following:

  • $157.1M for bus service ($100.9M for operations, and $56.8M for maintenance)
  • $52.5M for MAX ($16.4M for operations, $15.7M for ROW maintenance, and $20.4M for equipment maintenance
  • $9.3M for Streetcar expenses, including both TriMet’s subsidy and pass-through funding from the City of Portlandsubsidies to the Portland Streetcar
  • $47.5M for Accessible Transportation programs ($31.M for LIFT and $16.4M for TriMet’s Medical Transportation Program)
  • $6.4M for WES
  • $14.3M for facilities
  • $24.5M for dispatch, supervisors, planning, and other support.

Total ridership on TriMet (excluding Streetcar) in FY2012 was as follows: There were a total of 102.2 million boarding rides on the system–42.2M on MAX, 59.6M on bus, and 481k on WES. (For those unfamiliar with terminology, a “boarding ride” is counted whenever someone steps on a transit vehicle–a journey with two transfers counts as three boarding rides). 41.3% boardings were on MAX, 58.3% on bus, and 0.41% on WES.

Total passenger revenue in FY2012 was–coincidentally–$102.2 million, giving an average system-wide revenue per boarding ride of $1.00. This figure is noticeably less than the full-price fare, due to various factors like discounts, transfers, Free Rail Zone (which was still active during FY12), and fare evasion. TriMet had a total Farebox Recovery Ratio (FRR) of 27.8%, vs total system cost. Mode-specific FRRs were: MAX, 44.2%, Bus, 24.8%, WES 6.9%, LIFT 4.8%.

The details: The Service and Ridership table above gives lots of interesting data, including historical data for the past fourteen service years. Complete data is at the link above, and some of the more interesting stuff (for this discussion) has been summarized here. The table below excludes WES and LIFT (they’re expensive, no surprise), and only includes every third year, to increase the chances it looks nice on your device. Apologies for the crude formatting.







Bus Sys Costs






Bus Op Costs






Bus Veh Miles






Bus Veh Hrs






Bus Rev Hours






Bus Boarding Rides






Bus Revenue/Ride






Bus In-Svc %






Bus Op Cost/Hr






Bus Sys Cost/Hr






Bus Op Cost/Ride






Bus Sys Cost/Ride






Bus Subsidy/Ride






Bus Rides/Veh Hr






Bus FRR (System)






Bus Break-even Rides/Hr







MAX Sys Costs






Max Op Costs






MAX Veh Miles






MAX Veh Hours






MAX Rev Hours






MAX Boarding Rides






MAX Revenue/Ride






MAX In-Svc %






MAX Op Cost/Hr






MAX Sys Cost/Hr






MAX Op Cost/Ride






MAX Sys Cost/Ride






MAX Subsidy/Ride






MAX Rides/Veh Hr






MAX FRR (System)






MAX Break-even Rides/Hr







Additional breakdowns on per-service subsidy are also made available by TriMet. In particular, the operating subsidy on the following modes/routes in FY12 was as follows:

  • Blue Line: $0.55
  • Red Line: $0.90
  • Green Line: $0.92
  • Yellow Line: $1.32
  • NS Streetcar: $1.32
  • Frequent bus: $1.54
  • “Standard” bus: $2.73.
  • WES: $18.55
  • LIFT: $27.93

The figures in the Service and Ridership table don’t line up exactly with line items in the published budget, as many budget items are allocated between bus, MAX, and other modes; and still others are not assigned to any particular mode’s cost center at all.

More details on the cost breakdown. Portland Afoot gives the following breakdown for the per-vehicle-hour costs of operating both bus and MAX. The figures cited here also don’t line up exactly with any of the numbers published for a given year (they may be budgeted figures rather than actuals), but they are sufficiently close that we will repeat them here. According to Portland Afoot, a bus costs $120.95 per hour to operate, as follows:

  • $9.84 for fuel and tires
  • $59.24 for driver labor ($29.29 wages, $30.95 benefits)
  • $20.25 for maintenance and supplies (excluding tires)
  • $8.73 for non-vehicle maintenance (I assume primarily stops, shelters, signage, etc).
  • $26.78 for administration and “other”.

For MAX (total $328.11):

  • $13.21 for electricity
  • $77.66 for driver labor ($33.79 for wages, $43.87 for benefits)
  • $61.21 for vehicle maintenance
  • $73.86 for non-vehicle maintenance (tracks, wires, signalling, stations)
  • $178.35 for administration and other (including “other wages”).

Rail’s direct vehicle costs (energy, consumables, labor, and maintenance) are more than that of bus($152 vs $89), but rail has significantly higher non-vehicle and administrative costs. Much of this is due to the need to maintain rail’s extensive physical plant–tracks, power lines and power systems, stations, etc.; as well as to operate the various train control systems, which have no analog on the bus. (Busses travel on public rights of way maintained by the state or by municipal governments; it should be noted that TriMet pays no weight-mile tax to help maintain the roads, despite the fact that busses–with very large axle loads–are among the worst offenders out there at chewing up asphalt). Many of these expenses, however, depend more on the amount of tracks installed, as opposed to the number of trains running–as a result, cutting (or increasing) rail service has less of an effect on TriMet’s bottom line than you might expect.

Some analysis

From a per-rider perspective, MAX looks great: The per-rider subsidy last fiscal year was almost a third on MAX vs the bus ($0.67 vs $1.96), and the farebox recovery ratio almost 20% higher (44% vs 25%). From a per-vehicle-hour perspective, on the other hand, MAX looks terrible: It costs nearly three times as much ($269 vs $98) to keep a MAX train on the tracks as it does to keep a bus on the road.

What is going on?

A key parameter that explains the difference, is ridership: MAX gets over four times the ridership per vehicle hour compared to the bus system. There are many possible reasons for this, pro and con:

  • Capacity: A 2-car LRT can carry 4-5 times as many passengers as a 40′ bus (and about 3 times as many passengers as a 20m articulated bus or a Streetcar). There are many times during the service day that vehicles are full, and passing up passengers as a result. The bus “break-even rides/hour”, which gives the number of hourly boardings required to break even, is greater than the capacity of a bus, whereas all 357 hourly boardings required for MAX to break even, actually could ride together on a MAX train.
  • Corridor strength: MAX has generally been deployed on corridors with either excellent existing ridership, or good untapped potential; there are no “social service” routes on the MAX system. Many bus lines, on the other hand, are “coverage” routes with low ridership ’round the clock; these drag down the numbers of the bus system tremendously. Unfortunately, I don’t have more detailed data on TriMet’s frequent service bus routes–though some of them come close to MAX in terms of financial performance (and may even exceed the lower-performing MAX lines).
  • Service quality: It’s long been claimed by rail supporters, that rail is more attractive (to riders) than bus. There are many anecdotes of yuppies who won’t be caught dead on a bus but will happily ride a train. While TriMet’s data doesn’t contradict this (rail ridership is much higher), it doesn’t say anything about why–there are many factors at play. In addition to the allure of steel wheels, there’s the issue of frequency, speed, and reliability; somewhat better amenities (particularly at stations–though nobody will confuse a MAX train with an Amtrak sleeping car, or even WES). There are many ways in which MAX is a better product for the transit rider than local bus service–assuming, of course, that it goes where you need. Of course, many of MAX’s desirable attributes have nothing to do with bus-vs-rail; a good-quality BRT line could conceivably offer similar operating parameter. OTOH, the N/S streetcar is an excellent performer despite being slow and unreliable–that said, it’s got a lower fare and runs in a dense urban neighborhood. (The CL streetcar, on the other hand…)
  • MAX-centric service configuration: One common complaint about MAX is that many suburban transit trips require a transfer to MAX to get downtown, whereas prior to the service being installed, one could ride a bus all the way in. No bus that serves Hillsboro or Aloha, for instance, reaches downtown; the only transfer-free service between Tualatin and Portland runs during peaks. When MLR completes, many Clackamas County bus riders will likewise get to transfer to MAX in Milwaukie, as many of the current busses that run from downtown to Milwaukie TC will likely be curtailed in favor of LRT. Certainly the 33 will terminate in Milwaukie; though I expect the 70 to keep running. While TriMet has a nice grid system in Portland itself, out in the ‘burbs it uses more of a hub-and-spoke system; this is particularly an issue in suburbs with rail. (Of course, riders in Tigard see the same issue with forced transfers to the 12…)
  • Transit-oriented development: While the impact of this is perhaps exaggerated by those touting its benefits (as well as by anti-urbanists panicking about “Portland creep”, as though Oak Grove is going to somehow turn into the Pearl District), much upzoning has occurred around the MAX line–and many developers, homeowners, and renters consider proximity thereto to be a valuable amenity. Out in Washington County, now that the housing collapse is apparently over, oodles of new construction are occurring along SW Baseline, all of it within walking distance of a MAX station. (Much of it is single-family housing, but packed in like sardines single-family housing). Where development has occurred, development patters along MAX tend to be denser that development patterns in transit-poor neighborhoods; increased density drives ridership.

The wide gulf between “standard” bus and frequent bus is further evidence that ridership is indeed a big part of the story; there is little technical difference between the two types of service. (Non-frequent routes are more likely to get older busses, and frequent routes on busy corridors more likely to experience overloaded conditions). One might naively think that frequent bus would be less efficient, given that there are more runs and thus more expense occurred, but that’s not the case–good quality service is more likely to attract ridership; and that principle applies to rail just as much as to better bus service. Of course, there has to be sufficient potential ridership along a route for this to apply; turning the 84 into Frequent Service wouldn’t make it into an efficient, high-ridership run.

Of course, none of this answers the questions as to whether continued investment in MAX remains a good idea. Capital dollars are getting scarce; likely putting the breaks on future projects beyond MLR and/or the CRC. And with labor expenses continuing to grow (largely due to increasing health-care costs–it’s not as though operators and retirees are getting richer; it’s that a specific benefit is gotten considerably more expensive), operational squeezes will likely continue. While rail is cost-effective in the corridors it serves, it can’t be deployed everywhere, and in places where it is deployed, TriMet is committed to provide a certain level of service as a condition of funding grants. This commitment, along with rail’s higher fixed cost base, and greater pressure to cut low-margin routes during a budget crisis, does tend to concentrate most of the weight of the budget axe on the bus system.