The Best-Laid Plans. Randal O'Toole
10 percent of all property taxes paid in California go to TIF, and every state but Arizona allows cities to use TIF for urban renewal.2 From one point of view, TIF is just a scheme by city officials to divert taxes to favored developers. But TIF-supported urban-renewal projects are always backed up with lofty documents written by urban planners. What makes planners think they can know that a particular neighborhood or district would be better off with condos instead of offices, mixed-use developments instead of single-family homes, or skyscrapers instead of low-rise buildings?
As law professor Bernard Siegan points out, land-use planners must consider “questions of compatibility, economic feasibility, property values, existing uses, adjoining and nearby uses, traffic, topography, utilities, schools, future growth, conservation, and environment” for each parcel of land. Just to determine the feasibility of one use for one site at one time “would require a market survey costing possibly thousands of dollars.”3 Planning requires data, and the amount of data needed to ensure that a plan is both efficient and equitable is simply overwhelming.
A small suburb may have thousands of parcels of land; a major city may have hundreds of thousands; a large urban area may have millions. Any given parcel might be put to dozens of different uses: single-family residential at various lot sizes; multifamily residential at various densities; small-, medium-, and big-box retail; low-, medium-, or high-rise office space; light, medium, or heavy industrial use; developed and undeveloped open space; and so forth.
No planning agency has the budget needed to do a market analysis of all the possible uses for all the parcels in their jurisdiction. Yet planners claim they can determine the optimal uses for most or all sites in an entire region—not just for today but for many years into the future, and not just considering market factors but considering social, environmental, and other nonmarket factors as well.
If the data needs for land-use planning are daunting, transportation data requirements are impossible. Transportation planners must deal not with chunks of inanimate land but with people whose preferences and tastes can vary widely. As an expression of that variation, most people make several trips each day to different work, school, shopping, home, recreation, and other locations. Imagine a million people leaving roughly 400,000 different homes and going to a similar number of different destinations several times each day. No one could possibly collect the data needed to understand all those trips.
Many planners rely largely on information provided by their hierarchical bureaucracy. As information filters up the bureaucracy, much of it is necessarily left out. The models that the planners build contain only a tiny fraction of the information available at the lowest levels of the bureaucracy. Nor is the information provided by the bureaucracy unbiased. At each level, officials with their own goals tend to pass on information that they think will promote those goals. As Anthony Downs notes, “Each official tends to distort the information he passes upward into the hierarchy, exaggerating those data favorable to himself and minimizing those unfavorable to himself.”4
Planners simplify their data collection problems by lumping things into a few classes and averaging the data for each category. Land-use planners writing the 50-year plan for the Portland urban area lumped land uses into just 10 classes. Only two of these classes, “inner neighborhoods” and “outer neighborhoods,” represented all the variation of residential land in the urban area.5 The detail that is lost from such lumping is staggering. Yet Portland’s plan has had an enormous effect on the lives of Portland-area residents since planners began implementing it in 1995.
The Forecasting Problem
During World War II, Kenneth Arrow—who would later win a Nobel Prize in economics—was ordered to help with long-range weather forecasts for the Army Air Corps. His group soon realized that their forecasts were no better than numbers pulled out of a hat, and they asked to be assigned to more useful work. “The Commanding General is well aware that the forecasts are no good,” they were told in reply. “However, he needs them for planning purposes.”6
In addition to needing data about the present, planners need data about the future. According to one planning advocate, planners “can take account of processes which are occurring so slowly, or will begin to occur so far in the future, that no single producer could be aware of their existence.”7 Economist Gerald Sirkin scoffs, “What is this mysterious prophetic vision that comes to a man when he sits at a desk in the central authority but not when he sits at a desk in a business or university?”8
Planners typically write plans for the next 10 years or more. Some plans are even ostentatiously called “20-year plans” or “50-year plans.” For such plans to be worthwhile, planners must be able to accurately answer such questions as
• What technologies will be available in the future?
• How much will land, energy, and other resources cost?
• How will individual tastes and preferences change?
• How will people earn their incomes?
None of these questions can be answered with any degree of confidence. Yet any plan that is based on inaccurate answers to even one of these questions is likely to be drastically wrong, locking people into expensive but unnecessary policies and programs. Despite the absolute need for accurate predictions, planners must contend with the rather unstartling Law of the Future:
Planners have no better insight into the future than anyone else.
So government plans written today are unlikely to make any sense to the people in the future that the plans are supposed to benefit. Imagine, for example, writing a 50-year plan for your city in 1950. In 1950,
• few people had ever flown, and no one had ever flown in a commercial jet;
• few people had ever worked with computers, and not even the most far-seeing science-fiction writers had predicted microcomputers or the Internet;
• few people could afford to regularly make long-distance phone calls, and no one had ever made a direct-dial long-distance call;
• few married women worked, and the highest-paid jobs were all held by men; and
• few other countries could match the United States as a manufacturing powerhouse, and no one had ever imported a transistor radio from Japan, Korea, or China.
With the information available in 1950, your plan for the year 2000 would make the airport too small and the train station too big. You would assume that high telecommunication costs would force jobs to cluster tightly together. Because you would assume that few married women would work, the homes you would plan would have one-car garages. You would plan for the wrong ratio of blue-collar to white-collar jobs, and you would never imagine that large numbers of people would want a home office.
Although planners cannot truly know the future, they rely on forecasting to provide projections about future populations and demand for various goods and services. Forecasting is generally based on projecting current trends into the future while taking into account demographics—for example, aging baby boomers—and other changes that planners think they can foresee.
Planning forecasts can be very intimidating. But “the technical complexity of forecasts is in fact quite misleading,” says University of California planning professor Martin Wachs. “While equations, computers, and enormous data bases give the forecasts an aura of ’science,’ which invests them with certain authority in the political arena, the most critical data needed to make a forecast often consists of assumptions about the future.” These assumptions, Wachs admits, “can never be known with certainty.”9
Wachs adds, “Forecasters are usually drawn from the ranks of social scientists, engineers, and planners whose education and professional identities are based primarily on technical methodological skills.” While we can train these people to run complicated computer models, “we don’t—and probably can’t—educate them to make better assumptions.”10
Since