Rejoinder

Donald C. Shoup

Carl Buttke and Eugene Arnold argue that nothing is wrong with the Institute of Transportation Engineers' (ITE) *Trip Generation*
and *Parking Generation.*
In part, their confidence may derive from their assumption that "the intended users . . . are transportation professionals trained in mathematics, statistics, traffic engineering, and planning fundamentals and who possess engineering judgment." But the actual users are a much broader and more diverse group. The ITE itself says, "*Trip Generation*
is an educational tool for planners, transportation professionals, zoning boards, and others who are interested in estimating the number of vehicle trips generated by a proposed development" (ITE 1997, vol. 3, p. ix). Many of these people are *not*
trained in mathematics, statistics, and traffic engineering. Zoning boards are rarely trained in anythingthey are elected or appointed to their positions, perform their duties as volunteers, and rely heavily on references such as *Parking Generation*
and *Trip Generation.*
They will not realize that the reported rates are often statistically insignificant and refer only to suburban sites with ample free parking and no public transit.

I would like to address three issues that Buttke and Arnold raise, and make a recommendation.

ITE's convention of rounding every parking and trip generation rate to two digits after the decimal point blurs the distinction between precision and accuracy. Buttke and Arnold agree that the two-digits-after-the-decimal-point convention leads to inappropriate precision in some instances, but then say,

There are also many instances in *Trip Generation*
where rates presented with two decimal places are appropriate at that level of precision (e.g., as a rate of 0.57 pm peak-hour trips per occupied room of a business hotel, or 7.27 weekday trips per occupied room).

But *Trip Generation's*
estimate of 7.27 weekday trips per occupied room of a business hotel is based on only one
observation.^{1}
It illustrates perfectly the statistical insignificance and inappropriate precision of many parking and trip generation rates.

An estimate always has some associated uncertainty. The number of significant digits used to express an estimate should reflect this uncertainty. The least significant digit in a number is the one farthest to the right, and the accuracy of any number is usually assumed to be ± 1 of the least significant digit, unless stated otherwise. In a typical engineering context, one would assume that an estimate expressed with five significant digits had been measured more accurately than an estimate expressed with only two significant digits. Because the number of significant digits used to express an estimate should be related to the uncertainty surrounding the estimate, the ITE's automatic two-digits-after-the-decimal-point convention is inappropriate and unscientific.

Buttke and Arnold note that the Trip Generation Committee wrestled with the issue of decimal placement in preparing the first edition of *Trip Generation*
in 1976, and decided to be consistent in reporting all rates with two digits after the decimal
point.^{2}
Accuracy is more important than digits-after-the-decimal-point consistency, however, and one should not use more (or less) precision than is warranted simply for the sake of uniformity. Precision refers to the number of significant digits, not to the number of digits after the decimal point.

Statistically sophisticated users understand the extreme uncertainty of trip generation rates and can ignore the false precision. But many users are not statistically sophisticated. To them, ITE's trip generation rates are *the*
relationship between transportation and land use. Some zoning codes explicitly specify ITE's trip generation rates as the basis for making land-use decisions and as the basis for assessing traffic impact fees, regardless of the sample size or statistical significance of the rates.

In Signal Hill, California, for example, the traffic impact fee is $66 per daily vehicle trip generated by a development project. The number of trips is calculated by multiplying the size of the project times its trip generation rate "as set forth in the most recent edition of the Traffic [*sic*
] Generation manual of the Institute of Transportation
Engineers."^{3}
The sixth edition's trip generation rate for a fast food restaurant is 496.12 trips per 1,000 square feet, so Signal Hill's traffic impact fee is $32.74 per square foot of restaurant space. The uncertain trip generation rates thus determine cities' tax rates.

Buttke and Arnold conclude that "ITE's intent is to provide a helpful resource that will guide transportation professionals in their decisionmaking." Spurious precision is not a real impediment for this purpose, although it is
misleading.^{4}
The real problem with *Parking Generation*
and *Trip Generation*
is that they measure the peak parking demand and the number of vehicle trips *at suburban sites with ample free parking and no public transit.*
Using these precise but poorly understood parking and trip generation rates as a guide to planning leads to bad transportation and land-use decisions. *Parking Generation*
and *Trip Generation*
are helpful resources in designing cities where everyone will drive everywhere they go and park free when they get there.

What can be done to make the ITE reports more reliable? The British counterpart to *Trip Generation*
suggests some possible improvements. The "Trip Rate Information Computer System" (TRICS) gives full information about the characteristics of every surveyed site and its
surroundings.^{5}
Users can thus estimate a trip generation rate based on sites comparable to the one under consideration. In addition to counts of vehicles, TRICS also includes counts of all the people (pedestrians, cyclists, public transport users, and car occupants) who arrive at and depart from a site. By including more than vehicle trips, TRICS takes a broader view of transportation. When all modes are included, the *person*
trip rates are often much higher than the *vehicle*
trip rates.

With its narrow focus on counting cars at suburban sites with free parking, *Trip Generation*
presents a precise but uncertain, skewed, and incomplete measure of the relationship between transportation and land use in the United States. Fortunately, the ITE's Parking and Trip Generation Committees seek to improve each successive edition of *Parking Generation*
and *Trip Generation.*
In future editions, they should settle for less precision, and strive for more accuracy.

^{1.} ITE (1997, vol. 1, p. 543). The estimate of 0.57 pm peak-hour trips per occupied room is based on only four studies.

^{2.} The first (1976), second (1979), and third (1983) editions of *Trip Generation*
report some rates with no digits after the decimal point and other rates with one or two digits after the decimal point. The fourth (1987) edition reports all rates with three digits after the decimal point. The fifth (1991) and sixth (1997) editions report all rates with two digits after the decimal point.

^{3.} Section 21.48.020 of the Signal Hill Municipal Code. The code is available online at
http://www.ci.signal-hill.ca.us/homepage.php.

^{4.}
Even if everyone who refers to *Parking Generation*
and *Trip Generation*
were an engineer or statistician, that does not excuse unjustified precision. Journalists do not casually break grammar and spelling rules just because intelligent readers might be able to figure out what they mean anyway. The burden of clarity and accuracy falls on the writerit cannot be shifted to the reader, no matter who one supposes the reader to be.

^{5.}
The TRICS database is available online at
http://www.trics.org/.