FOOTNOTES


1. Rate bureaus are private firms or associations through
 which rates may be negotiated collectively among carriers
 or carriers and shipper/receivers. These negotiations are
 granted total or partial immunity from antitrust laws.
 Rate bureaus also provide the service of publishing these
 rates (publication is normally required for regulated
 rates). Rate bureaus may represent one or more carriers in
 rate hearings.


2. An exception to this is released rates, in which all or
 part of the liability is removed from the carrier. Such
 rates are not the norm, and when offered are at a discount.
 In effect, then, released rates are simply reduced rates
 for reduced service.


3. Farm level prices for the 24 different commodities recorded
 were obtained from discussions with personnel at the Flor-
 ida Crop and Livestock Reporting Service. Perishability
 estimates for Florida commodities were obtained from Pav-
 lovic et al (1980).


4. A possible alternative specification would have been to
 include both the F.O.B. price (PRICE) and the inverse of
 the storage life (IPERISH) as well as or in lieu of DAY-
 LOSS. This was not done as DAYLOSS alone was felt to be
 the correct representation of shipping urgency. Later
 experimentation suggested that this specification was cor-
 rect. PRICE and IPERISH are strongly correlated (.74).
 The use of both variables separately would likely result in
 multicollinearity problems, making discernment of the sepa-
 rate effects difficult.


5. The circumstances under which the interviews were conducted
 were conducive to promoting open, non-evasive responses.
 The interviewers made it clear that they were with the
 Florida Department of Agriculture and Consumer Services or
 the University of Florida, and not with a law enforcement
 agency. The carriers appeared to be relaxed and responded
 readily and without hesitation. The negligible refusal
 rate supports these observations.


6. The Goldfield-Quandt test involves ordering the observa-
 tions in ascending order by the variable, X, being tested
 for heteroskedasticity. The first and last N observations
 are employed in separate regressions of the model. The
 ratio of the error sum of squares of the regression for the