Cities with severe job surpluses have the primary responsibility to increase housing, not cities with housing surpluses.
Job location externality costs. The old hyper-growth paradigm does not consider externality costs created by the location of a job and would never impose controls on increases in jobs in order to prevent a worsening of those externalities. We need to consider three issues: the costs and benefits of jobs considering locational externalities, how to quantify the amount of housing needed for a solution, and policies that would prevent externalities from getting worse.
1) Estimating job location costs and benefits. Jobs can have high externality costs because of their location. More jobs in an area already having a severe job surplus will cause increases in housing prices, commuting time, and air pollution. These costs increase geometrically as local housing supply and road capacity are more and more burdened over their designed capacities. Such jobs may have higher regional costs than benefits because of their location. Commonly recognized benefits of a job in a severe job surplus area need to be judged against usually unmeasured costs, and compared with the costs and benefits of the same growth in an area with surplus workers.
We need to quantify the benefits of a job and the costs of its externalities, but there is no established methodology for the analysis. The benefit estimate could be the annual total compensation for the job and profit expected by the employer, ignoring multiplier effects. Measuring the three costs is more complicated but could also ignore multiplier effects.
1. The cost estimate requires an estimate of the increase in housing costs within a reasonable commuting area of the job surplus area. For example, home prices in Silicon Valley can be compared with prices in an average metro region for a similar house and similar average commute time to a similar job. The higher housing price in (and near) the job surplus city would be attributed to the failure of that city (and its near neighbors) to build enough housing.
2. The cost estimate also requires gauging the annual value of time lost by commuters compared to a median commute time. Commute time value is difficult to analyze because workers do not accept jobs with an unacceptable commute, and once a worker has made a locational decision considering commute duration, the time is valued essentially at zero. The major grievance is felt by those who, after their locational decision, find a once acceptable duration degraded by more traffic. A reasonable estimate, then, can be based on societal median commute durations compared to those for similar jobs in the job surplus area.
3. Finally, the cost estimate requires quantification of air pollution and other commute-related externalized pollution costs created by the job surplus area, both from length of commute and from congestion. Silicon Valley, for example, has more long, congested drive-alone commutes than San Francisco, so each job is associated with more air pollution.
A benefit cost ratio can then be estimated.
We can look at a hypothetical illustration. If the annual costs per new job were something like $10,000 in pollution, $10,000 in wasted time, and $70,000 in housing costs, the externalized costs per job could total $90,000, probably more than the benefit of the job.
Policy makers today, following the dominant paradigm, do not consider these three costs, or any other regional externalities . The company creating the job in the severe surplus area has no responsibility for the problems it creates. The company benefits, because it externalizes the costs.
The city approving the land use where the job exists has no responsibility either. The city reaps sales tax and real estate tax revenue while avoiding the expenditure of serving worker housing. The regional agencies are run by locally elected officials who know little about economic analysis and are more committed to local power than to local responsibility. They are elected by developer and business contributions from an affluent class committed to the myths of unsustainable growth. The media and the economists also show no interest in these questions. The problem at this point is not so much the lack of an answer as the lack of asking the question. The sustainability paradigm is too weak to get the question on the table.
2) Quantifying housing need. How to quantify the amount of housing needed to eliminate the externalities? The estimates below look at local balances of employed residents and jobs and at the capacity of transportation infrastructure to deliver workers. How can we distinguish between productive agglomeration economies and a severe job surplus with high external costs?
A simple count of jobs and employed residents in a given area, the "land use balance," is too simplistic. A large area tends to have a better job-housing land use balance but can easily include overly long commutes. A small area is likely to be very imbalanced without causing systemic problems. Economic productivity is increased by concentrations of related jobs, i.e., agglomeration economies. Therefore, the job housing balance should consider the ability to commute without undue costs.
The freeway tipping point. Probably the best criterion is the capacity of transportation infrastructure to deliver workers to the jobs without excessive commute distance, congestion, or duration. This criterion can be operationalized as the freeway capacity "tipping point" at which an increase in traffic of just a few cars causes speed instabilities and big slowdowns. The impacts are disproportionate to the number of cars because the slowdown does not affect just the added cars, but everyone using the freeway. A small number of commutes above the tipping point inflicts high and geometrically increasing systemic costs in lower vehicle throughput, delay and pollution.
The tipping point can be quantified; it occurs when speeds drop below about 35 miles per hour and when flow reaches volumes of about 1800 to 2000 vehicles per freeway lane per hour. At this speed and volume so close to carrying capacity, very small increments of traffic cause temporary slowdowns and speedups, experienced by most drivers as the yo-yo or slinky effect of bunching up and stringing out without apparent cause. As travel demand increases even more, average speeds go down and stay down.
Bay Area Alliance for Sustainable Development recognizes the relationship among job surpluses, housing shortages, and freeway capacity. Its Draft Compact . . . of July 2000, uses, as an indicator of regional progress, "Housing units needed in job surplus areas to alleviate severe congestion."
An analysis of job-housing imbalance should focus on housing supply but could also focus on increasing the capacity of transportation infrastructure. For reasons discussed elsewhere, expanding freeways is uneconomic and unsustainable. We have been expanding freeways for decades, resulting in longer and faster commutes and no reduction in commute congestion or duration. Expanding public transit has more intuitive appeal, and may apply in some cases where densities along corridors support it, but long distance transit can also be uneconomic compared with simply reducing commute distances, which results from increasing local housing supply. More local housing also helps short distance transit with higher housing densities, shorter distances to work, and improved access to transit. Transit usually moves more slowly than a car, but, if it does not have as far to go, its duration can be competitive with driving alone.
Focusing on housing, we need to relate housing supply to the freeway tipping point. There are two tools for doing this, computer models and a rougher kind of estimate using a spreadsheet.
1. First, we will discuss computer models of land use and transportation. These models can be used to estimate job surpluses/housing shortfalls. The Metropolitan Transportation Commission (MTC) does advanced quantitative analysis of travel in the Bay Area. The MTC model (MTC BAYCAST-90) and private models (like EMME/2, TP+/Viper, and MINUTP) show travel volumes and times in huge trip tables based on small geographic areas called travel analysis zones. The MTC table has 1,099 zones, forming a 1099x1099 matrix for the nine county region. The zones are all connected to each other by a network of roads and transit lines, also in the model.
The models consider land uses, auto ownership, certain costs of travel, mode choice, highway and transit networks, trip volumes, and speeds. The models estimate traffic on all of the thousands of links of the network. In the "base year" the models simulate and replicate actual land uses, networks, and travel counts. Assumptions can be changed to estimate alternative scenarios.
The capacity tipping points for bottleneck freeways serving the severe job surplus centers are well known. The trip table or a "screenline" at freeway bottlenecks can report the geographic location of "productions" and "attractions." A screenline is an imaginary line across a link that identifies it for reporting trips, speeds, and level of service. Modelers define a "production" as the location where a round trip typically originates. For example, the home is the production location for a "home-based work trip" whether the trip goes from home to work or from work to home. The trip table shows the home as the production zone for both trips. Similarly, the "attraction" is the work place for both trips.
The models can report the number of trips in peak direction during peak hour and how many are above the tipping point, which then indicates the putative number of houses that would need to be closer to work to replace the excess freeway trips. The longer the trip, the greater the externalities, so the longest trips are of greatest interest. We can identify fairly precisely the job and housing locations connected by long commutes that cause an exponential increase in external costs to the region. The greatest gain in sustainability would be to somehow move the most distant housing to the job center. In modeling lingo, the most distant productions would be moved to the attraction zone, creating an intra-zonal trip probably not using a freeway. Eliminating the longer commutes would have the greatest benefit for the whole system, taking cars off many links of the freeway system before they reach the screenline.
Thus, we can quantify job and house locations and the number of related long commutes that cause an exponential increase in external costs to the region. The modeling of pricing changes, land use changes, transit improvements, and mode choice also allow a fairly precise definition of a solution for both land use and mode.
No one has yet done this kind of research in the region and perhaps the world. The models have always been used to determine how much more pavement is needed to meet land use imbalances. These powerful tools have not been used to reason backwards from highway capacity to land use problems and to quantify housing responsibilities.
2. Second, we will discuss a rough estimate of housing need/job surplus using a spreadsheet and MTC superdistricts. The Bay Area has nine counties, which are too big, and 101 cities, which are too many and too diverse to study easily. MTC also divides the Region into 34 superdistricts of roughly comparable size. Each superdistrict usually has a roughly reasonable commute shed consisting of itself and its adjacent superdistricts. The basic data are available on line at ABAG's FTP site where MTC data can be downloaded.
Our estimate is based on very approximate commute distances and lacks the precision of a model-based analysis. The estimate is not based on freeway capacity, but on distance, looking for job surpluses that cannot be covered by "reasonable" commutes. For example, the distance from San Rafael to downtown San Francisco is 20 miles, which I considered reasonable. From Novato to downtown San Francisco is 28 miles, which I accepted but considered to be on the outside edge of reasonable. The methodology could be applied to any region and for different ways of defining "reasonable."
An estimate of these numbers based on superdistrict-defined commute sheds is in Table 1 for 2000 and Table 2 for 2020. Table 1, "Job Surpluses by MTC 34 Superdistrict 2000," uses ABAG's Projections 2000, the latest available. The Table lists the superdistricts, their employed residents, employment (jobs), worker surplus, adjustments of worker surpluses based on allocations from adjacent superdistricts, adjusted surplus as a percent of workers, and a description of the adjustments. Superdistricts with job surpluses after adjustment are bolded.
Some adjustment to the roughly reasonable commute shed is needed because it underestimates the number of workers who can get to work in a reasonable distance. The adjustment estimates "cascade flows." A cascade flow allows workers from a job surplus city to commute to an even bigger job surplus nearby and have its jobs covered by employees from employee surplus cities further away. For example, Daly City/San Bruno has a job surplus which can be considered more than covered by the employee surplus from San Mateo/Burlingame, allowing Daly City/San Bruno employed residents to be allocated to the big job surplus in San Francisco. Similarly, workers can flow from St. Helena to Napa to Vallejo to Richmond to Oakland to San Francisco, all with reasonable commutes. The cascade adjustments do not describe what actually happens; they are a "what if" approximation of how short commutes might work, with the jobs not reachable by reasonable commutes being the excess causing problems.
Twelve of 34 superdistricts have land use job surpluses, that is, a surplus of total employment over employed residents within the superdistrict. Nine of these, however, including Superdistrict 15, disappear when adjusted for short commutes from adjacent superdistricts. The surplus of 10,000 jobs in Superdistrict 15 Livermore/Pleasanton is misleading. Data on nearby Tracy, if treated as an adjacent superdistrict, would show the Livermore/Pleasanton surplus is covered by employed residents from Tracy.
The three superdistricts remaining have severe job surpluses beyond reach by reasonable commutes. Two of them constitute Silicon Valley: Superdistrict 8 with Palo Alto and Superdistrict 9 with Sunnyvale, Mountain View, Santa Clara. These four cities, out of the 101 in the region, stick out like sore thumbs in regional statistics. The third superdistrict is downtown San Francisco. In 2000 "The City" had an adjusted surplus is 66,000 jobs, while Silicon Valley had an adjusted surplus of 108,000 jobs. Silicon Valley and San Francisco have imbalances many orders of magnitude bigger than any others and should be the focus of any serious discussion. Not surprisingly, they have colossal housing prices and horrible commutes. Silicon Valley is the bigger problem because of its larger size, predominance of drive-alone commutes, dispersion of destinations, and lack of transit infrastructure. San Francisco has lower environmental and commute externalities, making it easier for The City to meet housing supply or transit access goals.
ABAG Projections 2000 estimates that current trends will make matters worse. See Table 2. The City job surplus goes from 66,000 to 145,000 jobs. Silicon Valley goes from 108,000 to 133,000 jobs. Converting the job figures to households would give us the number of houses needed.
While a computer estimate would be more precise, the spreadsheet estimate
gives an idea of the magnitude of the problem. What should be done about
Below: Superdistrict Map
home back to top of this section To Table 1 To Table 2 forward to next text