• TABLE OF CONTENTS
HIDE
 Front Cover
 Title Page
 Table of Contents
 List of Tables
 List of Figures
 Main
 Reference
 Back Cover














Group Title: Bulletin - University of Florida. Agricultural Experiment Station ; 866
Title: The role of service and trip characteristics in carrier and load selection
CITATION THUMBNAILS PAGE IMAGE ZOOMABLE
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00027162/00001
 Material Information
Title: The role of service and trip characteristics in carrier and load selection
Series Title: Bulletin Agricultural Experiment Station, University of Florida
Physical Description: 56 p. : ill. ; 23 cm.
Language: English
Creator: Beilock, Richard
Garrod, Peter V ( Peter Vince ), 1943-
Miklius, Walter, 1928-
Publisher: Agricultural Experiment Station, Institute of Food and Agricultural Sciences, University of Florida
Place of Publication: Gainesville Fla.
Publication Date: 1987
 Subjects
Subject: Freight and freightage -- Rates   ( lcsh )
Genre: government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
non-fiction   ( marcgt )
 Notes
Bibliography: Bibliography: p. 53-56.
Statement of Responsibility: Richard Beilock, Peter Garrod, Walter Miklius.
General Note: "March 1987."
Funding: Bulletin (University of Florida. Agricultural Experiment Station) ;
 Record Information
Bibliographic ID: UF00027162
Volume ID: VID00001
Source Institution: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: aleph - 000927689
oclc - 16261916
notis - AEN8412
issn - 0096-607X ;

Table of Contents
    Front Cover
        Front cover
    Title Page
        Page i
    Table of Contents
        Page ii
    List of Tables
        Page iii
    List of Figures
        Page iv
    Main
        Page 1
        Page 2
        Page 3
        Page 4
        Page 5
        Page 6
        Page 7
        Page 8
        Page 9
        Page 10
        Page 11
        Page 12
        Page 13
        Page 14
        Page 15
        Page 16
        Page 17
        Page 18
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
        Page 35
        Page 36
        Page 37
        Page 38
        Page 39
        Page 40
        Page 41
        Page 42
        Page 43
        Page 44
        Page 45
        Page 46
        Page 47
        Page 48
        Page 49
        Page 50
        Page 51
        Page 52
    Reference
        Page 53
        Page 54
        Page 55
        Page 56
    Back Cover
        Back cover
Full Text
March 1987

March 1987
*i.


Bulletin 866 (technical)


The Role of Service

and Trip Characteristics

in Carrier and Load Selection





Richard Beilock
Peter Garrod
Walter Miklius














Agricultural Experiment Station
Institute of Food and Agricultural Sciences
University of Florida, Gainesville
J.M. Davidson, Dean for Research












THE ROLE OF SERVICE AND TRIP CHARACTERISTICS
IN CARRIER AND LOAD SELECTION







Richard Beilock
Peter Garrod
Walter Miklius

















The authors are, respectively, associate professor of Food and
Resource Economics at the University of Florida, associate
professor of Agricultural and Resource Economics at the
University of Hawaii, and professor of Economics at the
University of Hawaii.







TABLE OF CONTENTS
Pane

LIST OF TABLES iii

LIST OF FIGURES iv

INTRODUCTION 1

THEORETICAL DEVELOPMENT 3

Traditional Explanation of Freight
Rate Variations 3
Cross Commodity Variations 3
Variations Across Destinations 4
Alternative Explanations of Freight
Rate Variations 5
Cross Commodity Variations 5
Variations Across Destinations 7
The Conceptual Model 7
Conjoint Measurement 11

DATA COLLECTION 14

The First Broker Experiment 16
The Second Broker Experiment 17
The Carrier Experiment 17

EMPIRICAL RESULTS 34

CONCLUSIONS 49

FOOTNOTES 51

REFERENCES 53







LIST OF TABLES
Table Page

1 Freight Charges for Selected
Commodities Shipped from
Florida to New York, June 1984 2

2 Freight Charges per Truckload for
Florida Watermelon for Selected
Destinations, June 1984 2

3 First Broker Experiment 18

4 Second Broker Experiment 23

5 Carrier Experiment 29

6 First Broker Experiment Estimated
Parameters 37

7 Second Broker Experiment Estimated
Parameters 38

8 Carrier Experiment Estimated Parameters 39

9 First Broker Experiment Imputed
Value of a 1-Unit Decrease in Time,
Spread, and Variance 41

10 Value of a 1-Hour Decrease in Time
and Range 42

11 Carrier Experiment Value of a
1-Unit Decrease in Queue Length 42






LIST OF FIGURES


Figure Page

1 Partial Demand for 1-Hour
Decrease in Transit Time 44

2 Partial Demand for 1-Hour
Decrease in Transit Time
by Price Expectations 46

3 Partial Demand for 1-Hour
Decrease in Transit Time
by Cargo Value 47

4 Partial Supply Curve for a Queue
with 1-Hour Longer Wait (origin) 48

5 Partial Supply Curve for Queue with
1-Day Longer Wait (destination) 50






INTRODUCTION


Transportation charges per truckload often vary, depending
upon the commodity, even from the same origin to the same
destination. For regulated freight these variations have been
formalized into the class rate system. Under this system,
commodities are grouped into various classes depending upon
their physical characteristics and their value. Commodities in
higher numbered classes bear higher per unit weight freight
charges than those in lower numbered classes. There is
considerable evidence that class rate structures persist even
after the transport of the commodities is deregulated. Such
variations are also evident in never-regulated freight, such as
that for produce.1 This is demonstrated in Table 1, which
shows freight charges per truckload in effect during the third
week of June 1984 from Florida to New York for six different
commodities.

In addition to rate variations across commodities, freight
rates from the same origin may vary across destinations by
more than would be expected on the basis of direct costs.
This type of variation is demonstrated in Table 2, which
shows freight charges per truckload of watermelon during the
third week in June 1984 to 10 destinations.

In the transportation literature, price discrimination is the
most frequent explanation given for rate variations in freight
charges, particularly for variations across commodities.2 In
this bulletin, it will be demonstrated that rate variations
across commodities and/or across destinations are consistent
with a more fully specified model of decision making in the
truck transport markets which does not rely upon price
discrimination. In the next subsection, the traditional
explanations for rate variations are discussed, as well as
recent theoretical and empirical findings which suggest that
there are more reasons for rate differentials than previously
thought. This is followed by the presentation of the
conceptual model employed in this study. In the ensuing
section, this model is transformed to facilitate empirical
estimation. The section includes a brief discussion of the
statistical technique employed, conjoint measurement. In the
next section, the data collection procedures are described.
This is followed by the presentation and discussion of the






Table 1. Freight Charges for Selected Commodities
Shipped from Florida to New York, June 1984.
Freight chargea F.O.B. Florida0
Commodity per truckload ($) cargo value ($)
Watermelon 1,733 2,150
Grapefruit 1,513 2,958
Sweet Corn 1,758 6,655
Tomatoes 2,100 10,750
Green Peppers 2,300 18,920
Ornamentals 2,589 c

aRate information derived from a survey of 104 randomly
selected truckers at the Florida Agricultural Inspection Station
on U.S. 1-95 (northbound). Average rates reported for each
commodity.
bFederal-State Market News Services (1984). Truckloads are
assumed to be 43,000 pounds.
CValues for truckloads of ornamentals are highly variable but
generally exceed those for produce.

Table 2. Freight Charges per Truckload for Florida
Watermelon for Selected Destinations, June 1984.
Freight chargesb Freight charges
Destinationa truck load ($) per mile ($)
South Carolina 949 1.97
North Carolina 1,050 1.54
Virginia 1,336 1.61
Washington, D.C. 1,700 1.81
Maryland 1,316 1.33
Pennsylvania 1,494 1.38
New Jersey 1,513 1.29
New York 1,733 1.47
Massachusetts 1,875 1.36
Ontario 3,000 1.94

aAssumed to be the largest city in each state.
bSee footnote a in Table 1 above.
cMileage calculated from Orlando, Florida.






results. In the final two sections, the results are summarized
and conclusions drawn.


THEORETICAL DEVELOPMENT

Traditional Explanation of Freight Rate Variations

Cross Commodity Variations
In general, cross commodity freight rate differentials are
positively related to the value of the cargos, ceteris paribus.
This is clearly the case for regulated freight. The class to
which a commodity is assigned depends on two criteria:
physical characteristics which affect the cost of loading and
hauling, and value. Both of these are positively related to the
class number and, hence, to the freight rate. A classic
example of the role of value is the fact that ivory cue balls
are in a higher class than are plastic cue balls.

Rate structures which are positively related to the value of
the cargos are known as value-of-service rate structures.
For over a century, value-of-service rate structures have been
branded by economists as a form of price discrimination (e.g.,
Olson and Wilson, Chapter 3). The rationale behind this is
based upon Marshall's third condition for inelastic indirect or
derived demands -- that the price of the factor in question
represents only a small part of the expenses of the final
commodity (Marshall, p. 319). Demands for transportation of
freight are clearly derived from the demands for the final
products in which the transport services are inputs. Assuming
that the elasticities of demand for these final products are
reasonably comparable, or at least, that they are not inversely
related to the prices of the final products, then a value-of-
service rate structure is consistent with price discrimination.3

Due to the atomistic structure of most segments of the
trucking industry, it had been commonly thought that the
existence of value-of-service rate structures depended upon
the pricing discipline afforded by regulation.






Without the existence of regulation the motor
carrier industry would appear to be one of the
best examples of a perfectly competitive in-
dustry.... Common carriers may publish their own
rate lists (called tariffs) subject to ICC approval
but most do so through a tariff publishing agent.
The most common agent is a motor carrier
conference composed of a number of member
carriers, usually organized on the state or
regional basis. The conference rates are used by
almost all carriers operating in the applicable
area, even by nonmembers. It is the existence of
these rate conferences along with the regulatory
powers of the Commission which allow rates to be
set at other than perfectly competitive levels.
(Olson, pp. 395-396)

Since Olson wrote the above passage in 1972, regulations have
generally been relaxed, and in some jurisdictions such as
Arizona and Florida they have been removed entirely (excep-
ting safety regulations). Evidence has emerged which
indicates the persistence of value-of-service rate structures in
these jurisdictions (Beilock and Freeman, Blair, Kasserman, and
McClave) and even in cases of never-regulated freight. This
evidence has cast doubt on the universal applicability of the
traditional viewpoint that "value of service equals regulation-
abetted price discrimination."


Variations Across Destinations
Regulatory authorities have generally adopted the stance
that the charging of different rates per mile to different
destinations is discriminatory, particularly if the destinations
are similar distances. An example of this is the fact that
virtually all regulatory bodies will not sanction or will rarely
sanction differential rates between two points, depending upon
the direction of travel. Another example is that regulatory
bodies generally prohibit higher total rates for shorter than
for longer hauls. As the most elementary precondition for
price discrimination is the ability to separate markets, the
premise that differential charges based upon destinations is
discriminatory seems reasonable. Again, however, empirical
evidence indicates that this explanation is not valid in all







cases. For example, several thousand carriers service the
unregulated market for Florida produce.4 It is difficult to
imagine that they could successfully conspire to discriminate
against various destinations. Yet as was evidenced in Table
1, rates per mile differ markedly across destinations.

In the next subsection alternative explanations for cross
commodity and cross destination rate variations will be
presented. Some of these will then be incorporated into a
model which does not depend upon the presence of price
discrimination to explain these variations.


Alternative Explanations of Freight Rate Variations

Cross Commodity Variations
In recent years two alternative explanations for the
existence of value-of-service rate structures have been
developed. One explanation, which follows from a model
proposed by Salop and Stiglitz and extended by Wilde and
Schwartz and Sadanand and Wilde, still ascribes value-of-
service rate structures to price discrimination, but the
mechanism for it does not require explicit or even implicit
collusion among the suppliers. Rather, all that is required are
nontrivial search costs and differing intensities of search.
Under these conditions, different vendors may sell their goods
or services at different prices. Those with lower prices will
sell to the high-intensity searchers and to the low-intensity
searchers who are lucky enough to happen upon them. Those
with higher prices will sell to the low-intensity searchers who
fail to locate a lower price vendor.

Satterthwaite adds the interesting observation that search
costs will be particularly high if the good or service is
differentiable to those searchers whose demands are sensitive
to the characteristics which are differentiable. For example,
if transportation services differ according to some characteris-
tic such as reliability, then searchers who value reliable
service must obtain information regarding reliability as well as
rates. Frequently the only good source for information regard-
ing the quality of a vendor's product or service are the
observations of those who have already used the vendor. If
the number of vendors is high, then the probability of finding






a user of a particular vendor is low. Therefore, if there is a
large number of sellers and purchasers of a good or service
with variable quality, search costs will likely be high for
those who value high quality.

Those shipping more valuable cargos would be expected to
be the most sensitive to the quality of the transportation
services, ceteris paribus. It follows, then, that these shippers
would tend to rely upon known carriers, even if they charge
more than other carriers with unknown reputations.

An alternative explanation for value-of-service rate
structures, proposed by DeVany and Saving, also assumes
differentials in service qualities but does not rely upon price
discrimination. They suggest that the observed rate differen-
tials may be accounted for by differences in the qualities or
characteristics of the service. If the differences in the
qualities of service.are positively correlated with cargo value,
then a value-of-service rate structure would be the expected
result of a competitive market bringing costs and prices into
alignment.

Cargo liability coverage is an obvious candidate as a cargo
value-related service quality. Clearly carriers of highly valued
cargos bear more liability exposure than do those hauling
low-valued cargos. Fauth, however, found that in the case of
general freight carriers, loss and damage expenditures did not
explain a significant portion of the variation in class freight
rates. Another service characteristic, suggested by DeVany
and Saving, is the speed or promptness of the service. With
regard to mail service, the success of firms such as Federal
Express and Airborne attests to the high premiums which
expedited services can command.

Regardless of the relevant service characteristicss,
however, it seems reasonable that those with higher-valued
cargos would, ceteris Daribus, demand higher quality services.
If the rates and costs for these services approximate one
another, then a value-of-service rate structure would be the
result.






Variations Across Destinations
The variation of freight charges across destinations may be
attributable to variations in the availability of complementary
hauls from those locations. Jara-Diaz, Miklius and DeLoach,
Harris, Wilson, Baesmann and Daughety, and others have
shown that transportation services from point A to point B
are compliments in production to services in the opposite
direction. The production of a transportation service in one
direction creates empty capacity in the reverse direction or,
indeed, to any point from the original destination. Without
complementary hauls from a destination, the total cost of a
movement must include all or some of the costs involved in
repositioning the vehicle for the ensuing revenue-generating
movement. In other words, the carrier must consider the
opportunity costs associated with ending a movement at a
particular destination. If those costs are high, it should be
reflected in the rates. Rate differentials based upon these
considerations are not discriminatory. No empirical estimates
exist, however, regarding the importance of complementary
hauls in rate setting.


The Conceptual Model

The model of decision making we postulate employs the
nondiscriminatory mechanisms described above to explain
freight rate variations and the usual assumption of profit-
maximizing behavior on the part of the decision makers
(shippers, carriers, and receivers). Buyers and sellers of the
transportation service base their decisions on the full cost of
the movement to themselves (M. Johnson). From the view-
point of the shippers and the receivers, the full cost of
service (FTCs/r) includes not only the freight rate (R) but
also the costs associated with the quality of service (C(QUA-
L)):

(1) FTCs/r = R + C(QUAL).

The quality-associated cost elements include those related
to search and arrangement, the variability of pickup and
delivery times, cargo handling, and speed. The last three
elements are related to changes in the value of the cargo due
to product deterioration or price variations at the destination.






From the point of view of the carriers, the full cost of
service (FTCcar) includes not only the cost of carrying the
load (CHAUL) but also costs carriers will incur at the origin
and destination while arranging and waiting for a load
(CWAIT) as well as the opportunity cost (OPP), measured in
terms of future revenues foregone, which are incurred by
accepting carriage to the specified destination:

(2) FTCcar = CHAUL + CWAIT + OPP.

The demand equation for transportation can be derived
from the profit equation of the shipper/receiver. The relevant
costs and benefits in the context of determining the demand
for transport are the freight rate (RATE), storage costs (SC),
delay and variability costs (DC), and the change in the value
of the cargo as it moves from the origin to the destination
(CHVAL):

(3) PROFITs/r = f(RATE, SC, DC, CHVAL)

where RATE = (rate ,0,...,ratei,j);

SC = (sco,...,sci);

DC =(dc0,1,...,dcij);

CHVAL = (chvall,0,...,chvali,j);

i = (number of commodities (1 through I);

j = (number of destinations (0 through J)
(0 = origin).


PROFIT is maximized subject to the constraint that the total
quantity shipped cannot exceed the supply at the origin (Q,):


(4) E QTRANSij < Q1


where QTRANS = the quantity of transport services
required.






The origin (j = 0) is included as a destination because one
alternative is to sell the commodity at the origin. The
(derived) demand equation for transport for commodity i to
destination j is obtained by taking the negative of the
derivative of PROFIT with respect to ti,j:

(5) QTRANSi,j = -fRATE(i,j), (RATE, SC, DC, CHVAL).

Of course, QTRANS must be non-negative, and the constraint
that the amount shipped cannot exceed supply must hold. As
is the case for many agricultural activities, it is probably
more appropriate to invert the demand equation, making RATE
a function of quantity:

(6) RATEij = f-1 (QTRANS, SC, DC, CHVAL)

where f-I is the inverse of f with respect to RATEij.

If any of the variables in the profit equation differ across
commodities, the demand for transport will also vary across
commodities. If transportation services are homogeneous and
if those with differing demand characteristics can be segrega-
ted, then price discrimination might occur. On the other
hand, if transportation services are nonhomogeneous, then the
differences in demand may result in the purchase of different
levels and mixtures of service characteristics. The latter is
particularly likely if the market structure precludes effective
segregation of demand groups and/or pricing discipline among
the vendors, and if the barriers to entry are low.

The supply of transport services can be derived in similar
fashion. The expected profits of carriers are the discounted
present values of their net earnings from their next trips and
from all following trips. Their profit equations include freight
rates (RATE), operating costs (OPCST), costs related to
queuing for their next shipments (QU), and expected search
costs at destinations and the probability of not obtaining an
outbound load (EREV):

(7) PROFITcar = g(RATE, OPSCT, QU, EREV)






RATE = ratel,l,...,ratek,m;


OPSCT = opsctl,1,...,opsctk,m;

QU = qul,l,...,quk,m;

EREV = erevl,1,...,erevk,m;

k = number of commodities the carrier is
able to transport (k = I through K);

m = number of possible destinations
(1 = 1 through m).


The carrier's maximization problem is subject to the constraint
that the number of trucks leaving from the origin at time t
cannot exceed the available supply (S):


(8) E QTRANSi,j St.
ij


This availability, in turn, is a function of past decisions by
the carriers to carry goods or to deadhead (i.e., move empty)
to the origin:


(9) St = E S t-
ij 1,J


The supply of transport offered for a specific commodity-
destination pair (i,j) is obtained by taking the derivative of
carrier profits with respect to RATE..:
1,J
(10) QTRANSi,j = gRATE(i,j) (RATE, OPSCT, QU, EREV).

Thus the supply of trucks for any given commodity-destination
pair will depend not only on the rate offered and direct costs
incurred, but on the expected wait at the origin for the load


where






and the earnings stream the carrier can expect to obtain at
the destination.

As specified, the above model is theoretically estimable.
However, empirical estimates for many of the key variables
would be difficult to obtain. Among the variables which
would be difficult to quantify are search and queuing costs for
both carriers and shipper/receivers, expected revenues
associated with servicing various locales, and the imputed
costs shippers place on uncertainty in delivery times. If such
data were available, the values that carriers and ship-
per/receivers place on various attributes could be determined
by employing a revealed preference approach. As this is not
the case, an alternative approach is resorted to in which the
decision makers are presented with alternative hypothetical
situations. Each situation is defined with respect to the
relevant attributes. The decision makers order the alterna-
tives in terms of their desirability, and the attribute weights
are inferred from these orderings. This type of analysis can
be carried out by performing an experiment using a panel of
decision makers and employing a technique called conjoint
measurement.


Conjoint Measurement

Conjoint measurement is a technique for inferring attribute
utilities from an individual's ranking of a product or service
profile. It was developed and first employed in the fields of
mathematical psychology in psychometrics. Increasingly in
recent years it has been employed in marketing studies (see
the survey by Cattin and Wittnick). The procedure has also
been used in transportation research (Davidson, and Garrod
and Miklius). It requires three steps. The first involves
collecting information from individual decision makers about
their ordinal preferences. This is accomplished by a survey
where respondents are presented with a number of alternatives
characterized by different attributes and are asked to rank
the alternatives in order of their preference.

The second step is to estimate ordinal preference functions
for each individual surveyed on the basis of his/her reported






ranking of the alternatives. This is based on the following
two assumptions regarding that individual's utility function:

a. The utility function is of the form


(11) U = U(ql,q2,...,qn)


where U = the ordinal preference level;
i
q = the level of the ith attribute.

b. The utility function is strongly separable under some
monotonic transformation (h) such that h(U) can be written:


(12) h(U)= E W(gk(q1 ...n)).
k


If each gk is a function of only qk, then the Wk's can be
readily interpreted as utility or preference weights. As
ordering is unchanged, the monotonic transformation, h(U), is
an equivalent ordinal function to U; i.e., both functions
produce the same ordering.

The final step is to estimate the parameters of equation
(12) using the information obtained in the first step. This
involves estimating the parameters of the following equation
for each respondent:


(13) Rj = E Wk(gk(gkj))
k

where k = 1,..., n (n = number of attributes);

j = 1,..., ALT (ALT = number of alternative
scenarios presented);

R. = the rank of the jth alternative.
J







Equation (13) is estimated subject to the constraint that the
relative ordering of the estimated rankings (Rj^) agrees with
those revealed in Step 1. That is, for all i and j between 1
and ALT, the sign of Ri^ RJ^ must be the same as for R1-
Rj.

Most monotonic regression procedures choose W to
minimize a loss function of the following type:


(14) L = p dij (Ri^ Rj^)k
ij


where di,j = 1 if (Ri Rj)*(Ri^ Rj^) < 0

= 0 otherwise.


When k = 0, this can be accomplished with integer programm-
ing (Pekelman and Sen) or nonlinear programming (Garrod,
Garrod and Miklius). When k = 1, linear programming may be
employed (Srinivasan, Srinivasan and Shocker, and Pekelman
and Sen). If k = 2, iterative procedures (R. Johnson) or the
minimization of stress (Kruskal) are feasible estimation
approaches.

An alternative to the above is to convert the reported
rankings to binary data and then use a probability model such
as logit or probit, to estimate W. This involves the following
steps:

a. Form the following ALT*((ALT-1)/2) equations.


(15) Ri Rj = Wk(gk(qk,1)) Wk(gk(qk,1)) for all i > j.


b. Next, transform the righthand side into a Bernoulli
variable that has the value of 1 if R. R. > 0 and 0 otherwise.
1 j







c. Estimate the transformed relation using a probability
model such as logit or probit. This approach has the ad-
vantage over the other procedures that statistical tests of the
parameters are possible.

An empirical problem that occurs when trying to solicit
rank ordering is the number of possible alternatives to be
considered. For example, if there are k attributes in (11) anJ
if each one appears at N different levels, then there are N
possible alternatives to consider. This can easily be a very
large number, making it difficult for a respondent to rank
consistently. One solution, suggested by Green, is to use a
factorial experimental design to limit the number of alterna-
tives considered. Another solution applies in the case where
only two attributes are included in a set of alternatives to be
ranked and the researcher has a priori knowledge of the
expected signs of the weights. The solution then is to
arrange the alternatives such that across the scenarios the
attribute levels move in the opposite direction in terms of
desirability. In this way the respondent is not presented with
the obvious first and last choices but constantly has to make
tradeoffs between the two attributes. When this procedure
is used, the number of alternatives to be considered reduces
to the number of levels of the attribute with the largest
number of levels.

This strategy was employed in the current study. For each
set of choices the entire set of attributes was described, but
only the levels of two attributes were varied in the above
manner. In addition, iterations of the same sets of alterna-
tives were presented with the range of levels for one of the
attributes successively increased. This procedure was employ-
ed to avoid corner solutions in which the respondent appears
to be ordering based upon one attribute alone because the
extent of the changes for the other attribute is not sufficient
to reveal the point at which the respondent is willing to make
tradeoffs between the two.



DATA COLLECTION

Ideally, to obtain estimates of the relative weights given to
the different attributes of the commodities, transport situa-







tions and destinations, preference data should be collected
from all the participants in the transport decision- making
process. For the produce-shipping segment in Florida, this
usually would include carriers and receivers (almost all
produce is shipped f.o.b., origin, Pavlovic et al.). However,
due to the fact that receivers are located in several widely
dispersed geographic areas, it was beyond the means of the
project to conduct interviews with them. Instead, interviews
were carried out with carriers and truck brokers. Truck
brokers act as agents in arranging and carrying out the
transportation function. As such, they are extremely familiar
with how receivers deal with the tradeoffs they face regarding
transportation alternatives. In the interviews, the brokers
were asked to order the alternatives offered as the "average
receiver" would, rather than selecting the alternatives that
would be in a broker's best interests. The use of brokers is
particularly appropriate in Florida, as they arrange about 60
percent of all produce and ornamental shipments (Beilock and
Fletcher).

Truck service between Florida and the rest of the nation
and Canada is provided by both independent truckers and
fleets, private and for-hire. For produce shipments, the
majority of these services are provided by independent owner-
operators (Beilock and Fletcher). As owner-operators are the
most important type of carrier and are the least likely to
engage in long-term contracts for carriage (and, thus, the
most likely to engage frequently in arbitrage), they were
chosen as the subject for the carrier interviews.

Three experiments were conducted: two with panels of
brokers and one with carriers. The broker interviews were
carried out at their places of business. Brokers located
throughout the Florida Peninsula were interviewed. Owner-
operators were interviewed at the northbound Florida Agricul-
tural Inspection Station on U.S. Interstate 95. It cannot be
asserted that either the broker or the carrier panels are
random or representative of the industry as a whole, though
there is no evident reason to suspect that they are not.
However, the intent of this research is to give an indication
of the importance of the attributes examined and the potential
of the approach employed. As such, individual, rather than
composite estimates of the parameters will be presented.







The First Broker Experiment
In the first broker experiment, the value to receivers of
the following were examined:

1. The time elapsed before delivery; that is, the total
time from when transport is desired until delivery.

2. The variability of delivery times.

3. Expected price changes for the cargo at the point of
delivery.

This was accomplished by an experiment design which required
the respondents to make tradeoffs between the freight rate
and levels of these attributes. The a priori expectation was
that receivers would be willing to pay more for faster transits
and more dependable delivery times, and that the tradeoff
between cost and total time to delivery would be greater when
prices were expected to fall than when they were expected to
rise. Alternatives were designed to avoid the obvious choice of
the lowest cost, fastest, and most reliable service.

The interview consisted of four questions. The instructions
and basic scenario are presented in Table 3a. This was read
to the respondent and reviewed as required. Each respondent
was reminded repeatedly that they were to answer as the
"average" receiver they dealt with would, rather than to base
their responses on their own preferences. The brokers
appeared to have little or no difficulty in this regard.

The four questions and the sets of alternatives to be
ordered are presented in Table 3b. Three sets of alternatives
for each question were used to more closely identify the
tradeoff points. Each alternative was typed onto an unnum-
bered index card. The order in which the sets of alternatives
were presented to each respondent as well as the order of the
cards in each set was random. This was made evident to the
respondents by shuffling the sets of alternatives and then the
cards in each set in their presence. In this way no proper or
expected ordering of the answers was implied.

The first experiment was conducted in March 1984. The
ranges of values used were selected to be representative of







those which were currently applicable. The panel consisted of
28 brokers.


The Second Broker Experiment
An implication of the postulated model is that the effects
of total transportation time and delivery reliability on the
amounts receivers will pay for transport are functions of
commodity characteristics. The second broker experiment was
designed to test for such differences in receiver (i.e., broker)
responses. A priori it would be expected that both rapid and
reliable delivery would be more important to receivers of
higher-valued and more perishable commodities.

To test this expectation, questions 1 and 4 from the first
broker experiment were each asked twice: once for a relative-
ly high-valued and perishable commodity (sweet corn), and
once for a relatively low-valued and long-lived commodity
(watermelons). The basic situation, questions, and alternatives
are presented in Tables 4a and 4b. Differences between this
and the first broker experiment with respect to freight rates
and commodity values reflect differences in these prices
between the time that each survey was fielded. The second
broker survey was conducted in June 1984. The second panel
was composed of 13 brokers.



The Carrier Experiment
The carrier experiment was conducted with a panel of
owner-operators. The questions and the basic situations were
designed to allow the respondents to determine their full cost
of providing service. Three facets of the full cost of service
were examined:

1. Queuing or waiting costs for a load at the origin
(Florida);

2. Expected queuing or waiting costs for a complemen-
tary load from the destination point, assuming stable
freight rates at the origin (to which the carrier will
return);






3. Expected queuing or waiting costs for a complemen-
tary load from the destination point, assuming
declining freight rates and load availability at the
origin (to which the carrier will return).

Three basic questions were asked, one for each of the factors
listed above. The basic situations and the questions are
presented in Tables 5a and 5b.

As with the broker surveys, the basic situations were read
to the respondents, with repetitions given as needed. The
participants appeared to have little difficulty ordering the
alternatives. Several, in fact, commented that they enjoyed
the process. The carrier experiment was conducted in March
1984. The panel consisted of 21 owner-operators.



Table 3a. First Broker Experiment: Instructions to
Respondents and Basic Situation

We're going to play a type of game. I'm going to describe
a situation and then present you with five choices. I'll then
ask you to order them according to preferences of the average
receiver you represent. Keep in mind that I'm asking you to
put yourself in the receiver's shoes -- to give his preferences,
rather than yours. Finally, feel free to ask questions as we
go along.

Basic Situation

It's 9:00 a.m. Monday morning. You are a receiver located
at the Hunt's Point Market in New York City. You have a
truckload of grapefruit in south Florida ready for loading. The
fruit has a value of $5,000 to $6,000.







Table 3b. First Broker Experiment: Questions and Sets of
Alternatives

1. It's 9:00 a.m. Monday morning. You are a receiver located
in New York City (Hunt's Point) and have a truckload of
grapefruit in south Florida ready for loading. The
truckload is valued at about $5,000 to $6,000. Trucks are
currently scarce, and the following represents the best
offers you have been able to obtain:

Alternatives

Set 1.

Tuesday midnight delivery and rate = $1,800
Wednesday 6:00 AM delivery and rate = $1,700
Wednesday 6:00 PM delivery and rate = $1,600
Wednesday midnight delivery and rate = $1,500
Thursday 6:00 AM delivery and rate = $1,400

Set 2.

Tuesday midnight delivery and rate = $2,200
Wednesday 6:00 AM delivery and rate = $2,000
Wednesday 6:00 PM delivery and rate = $1,800
Wednesday midnight delivery and rate = $1,600
Thursday 6:00 AM delivery and rate = $1,400


Set 3.

Tuesday midnight delivery and rate = $2,600
Wednesday 6:00 AM delivery and rate = $2,300
Wednesday 6:00 PM delivery and rate = $2,000
Wednesday midnight delivery and rate = $1,700
Thursday 6:00 AM delivery and rate = $1,400







Table 3b. (cont.)

2. Similar to the last situation, you, a New York City
receiver, are interested in getting a load of grapefruit you
have purchased shipped north. Your preferred delivery
time is Wednesday morning, say around 10:00 AM, so you
have tried to arrange the delivery time. These are the
best offers you have received:

Alternatives

Set 1.


Delivery within 30 minutes of 10:00
and rate = $1,800

Delivery within 1 hour of 10:00 AM
rate = $1,700

Delivery within 4 hours of 10:00 AM
rate = $1,600


AM Wednesday


Wednesday and


Wednesday and


Delivery sometime on Wednesday (6:00 AM-Midnight)
and rate = $1,500

Delivery sometime between Tuesday afternoon and
Thursday morning and rate = $1,400

Set 2.

Delivery within 30 minutes of 10:00 AM Wednesday
and rate = $2,000

Delivery within 1 hour of 10:00 AM Wednesday and
rate = $1,800

Delivery within 4 hours of 10:00 AM Wednesday and
rate = $1,600

Delivery sometime on Wednesday (6:00 AM-midnight)
and rate = $1,400






Table 3b. (cont.)


Delivery sometime between Tuesday afternoon and
Thursday morning and rate = $1,200

Set 3.

Delivery within 30 minutes of 10:00 AM Wednesday
and rate = $2,300

Delivery within 1 hour of 10:00 Wednesday and rate =
$2,000

Delivery within 4 hours of 10:00 AM Wednesday and
rate= $1,700

Delivery sometime on Wednesday (6:00 AM-midnight)
and rate = $1,400

Delivery sometime between Tuesday afternoon and
Thursday morning and rate = $1,100


3. Again, you are a receiver who has bought a load of
grapefruit in south Florida. The only difference now is
that you think that there is a better than even chance
that grapefruit prices in New York will rise by 10 to 15
percent sometime over the next 4 days.

Alternatives

Set 1.

Tuesday midnight delivery and rate = $1,800
Wednesday 6:00 AM delivery and rate = $1,700
Wednesday 6:00 PM delivery and rate = $1,600
Wednesday midnight delivery and rate = $1,500
Thursday 6:00 AM delivery and rate = $1,400






Table 3b. (cont.)


Set 2.

Tuesday midnight delivery and rate = $2,200
Wednesday 6:00 AM delivery and rate = $2,000
Wednesday 6:00 PM delivery and rate = $1,800
Wednesday midnight delivery and rate = $1,600
Thursday 6:00 AM delivery and rate = $1,400

Set 3.

Tuesday midnight delivery and rate = $2,600
Wednesday 6:00 AM delivery and rate = $2,300
Wednesday 6:00 PM delivery and rate = $2,000
Wednesday midnight delivery and rate = $1,700
Thursday 6:00 AM delivery and rate = $1,400



4. Just as before, it's Monday morning and you are a New
York City receiver who has bought a load of grapefruit in
south Florida. You anticipate a possible 10 to 15 percent
decline in grapefruit prices in New York sometime over the
the next 4 days.


Alternatives

Set 1.

Tuesday midnight delivery and rate = $1,800
Wednesday 6:00 AM delivery and rate = $1,700
Wednesday 6:00 PM delivery and rate = $1,600
Wednesday midnight delivery and rate = $1,500
Thursday 6:00 AM delivery and rate = $1,400






Table 3b. (cont.)

Set 2.

Tuesday midnight delivery and rate = $2,200
Wednesday 6:00 AM delivery and rate = $2,000
Wednesday 6:00 PM delivery and rate = $1,800
Wednesday midnight delivery and rate = $1,600
Thursday 6:00 AM delivery and rate = $1,400

Set 3.

Tuesday midnight delivery and rate = $2,600
Wednesday 6:00 AM delivery and rate = $2,300
Wednesday 6:00 PM delivery and rate = $2,000
Wednesday midnight delivery and rate = $1,700
Thursday 6:00 AM delivery and rate = $1,400





Table 4a. Second Broker Experiment: Instructions to
Respondents and Basic Situation


We're going to play sort of a game in which I will describe
a situation and ask you to rank five alternatives according to
your preferences. The point of all this is to get a better
understanding of how transportation decisions are made.

What we're really interested in is decision making from the
standpoint of the receiver, the one who orders the transporta-
tion service. So please put yourself in his shoes; that is,
answer according to what you think he would say.


Basic Situation

For all of the questions, pretend that you are a New York
City area receiver. It's Monday, around 9:00 AM, and you
have a load of produce in south Florida ready to be shipped
to you. You anticipate a level market in New York with prices






Table 4a. (cont.)


neither rising nor falling appreciably over the next several
days. All of the transportation options will refer to an
enclosed, refrigerated trailer with top ice available should you
desire. Again, please put yourself in the receiver's shoes.


Table 4b. Second Broker Experiment: Questions and Sets
of Alternatives

1. It's 9:00 AM Monday and you (the New York City area
receiver) have a load of sweet corn in south Florida ready
for shipment. The load is valued at about $5,000. Trucks
are currently scarce and the following represents the best
offers you have been able to obtain.


Alternatives

Set 1.

Tuesday midnight delivery and rate = $2,100
Wednesday 6:00 AM delivery and rate = $2,000
Wednesday 6:00 PM delivery and rate = $1,900
Wednesday midnight delivery and rate = $1,800
Thursday 6:00 AM delivery and rate = $1,700

Set 2.

Tuesday midnight delivery and rate = $2,500
Wednesday 6:00 AM delivery and rate = $2,300
Wednesday 6:00 PM delivery and rate = $2,100
Wednesday midnight delivery and rate = $1,900
Thursday 6:00 AM delivery and rate = $1,700

Set 3.

Tuesday midnight delivery and rate = $2,900
Wednesday 6:00 AM delivery and rate = $2,600
Wednesday 6:00 PM delivery and rate = $2,300
Wednesday midnight delivery and rate = $2,000
Thursday 6:00 AM delivery and rate = $1,700






Table 4b. (cont.)


2. Again, we have the same situation as before. It's 9:00 AM
Monday morning and you have a load to be shipped to you.
The only difference is that now the load is watermelons.
The load is valued at about $2,200. The following repre-
sents the best offers you have been able to obtain.

Alternatives

Set 1.

Tuesday midnight delivery and rate = $1,800
Wednesday 6:00 AM delivery and rate = $1,700
Wednesday 6:00 PM delivery and rate = $1,600
Wednesday midnight delivery and rate = $1,500
Thursday 6:00 AM delivery and rate = $1,400

Set 2.

Tuesday midnight delivery and rate = $2,200
Wednesday 6:00 AM delivery and rate = $2,000
Wednesday 6:00 PM delivery and rate = $1,800
Wednesday midnight delivery and rate = $1,600
Thursday 6:00 AM delivery and rate = $1,400

Set 3.

Tuesday midnight delivery and rate = $2,600
Wednesday 6:00 AM delivery and rate = $2,300
Wednesday 6:00 PM delivery and rate = $2,000
Wednesday midnight delivery and rate = $1,700
Thursday 6:00 AM delivery and rate = $1,400


3. Similar to the previous situations, you are interested in
getting a load shipped to you from south Florida. The load
is sweet corn, valued at about $5,000. Because things are
somewhat hectic at your New York area loading docks, you
would prefer delivery on Wednesday morning, say around
10:00 AM. The following represents the best offers you
have been able to obtain.






Table 4b. (cont.)


Alternatives

Set 1.

Delivery within 30 minutes of 10:00 AM Wednesday
and rate = $2,400

Delivery within 1 hour of 10:00 AM Wednesday and
rate = $2,300

Delivery within 4 hours of 10:00 AM Wednesday and
rate = $2,200

Delivery sometime on Wednesday (6:00 AM-midnight)
and rate = $2,100

Delivery sometime between Tuesday afternoon and
Thursday morning and rate = $2,000


Set 2.

Delivery within 30 minutes of 10:00 AM Wednesday
and rate = $2,600

Delivery within 1 hour of 10:00 AM Wednesday and
rate = $2,400

Delivery within 4 hours of 10:00 AM Wednesday and
rate = $2,200

Delivery sometime on Wednesday (6:00 AM-midnight)
and rate = $2,000

Delivery sometime between Tuesday afternoon and
Thursday morning and rate = $1,800







Table 4b. (cont.)

Set 3.

Delivery within 30 minutes of 10:00 AM Wednesday
and rate = $2,900

Delivery within 1 hour of 10:00 AM Wednesday and
rate = $2,600

Delivery within 4 hours of 10:00 AM Wednesday and
rate = $2,300

Delivery sometime on Wednesday (6:00 AM-midnight)
and rate = $2,000

Delivery sometime between Tuesday afternoon and
Thursday morning and rate = $1,700


4. Again, you are looking for a 10:00 AM Wednesday delivery
or as close as possible. This time, however, the load is
watermelons, valued at around $2,200. The following
represents the best offers you have been able to obtain.


Alternatives

Set 1.

Delivery within 30 minutes of 10:00 AM Wednesday
and rate = $2,100

Delivery within 1 hour of 10:00 AM Wednesday and
rate = $2,000

Delivery within 4 hours of 10:00 AM Wednesday and
rate = $1,900






Table 4b. (cont.)
Set 1 (cont.)

Delivery sometime on Wednesday (6:00 AM-midnight)
and rate = $1,800

Delivery sometime between Tuesday afternoon and
Thursday morning and rate = $1,700

Set 2.

Delivery within 30 minutes of 10:00 AM Wednesday
and rate = $2,300

Delivery within 1 hour of 10:00 AM Wednesday and
rate = $2,100

Delivery within 4 hours of 10:00 AM Wednesday and
rate = $1,900

Delivery sometime on Wednesday (6:00 AM-midnight)
and rate = $1,700

Delivery sometime between Tuesday afternoon and
Thursday morning and rate = $1,500

Set 3.

Delivery within 30 minutes of 10:00 AM Wednesday
and rate = $2,600

Delivery within 1 hour of 10:00 AM Wednesday and
rate = $2,300

Delivery within 4 hours of 10:00 AM Wednesday and
rate = $2,000

Delivery sometime on Wednesday (6:00 AM-midnight)
and rate = $1,700

Delivery sometime between Tuesday afternoon and
Thursday morning and rate = $1,400






Table 5a. Carrier Experiment: Instructions to Respondents
and Basic Situation

We're going to play a game. I'm going to describe a
situation to you and then present you with five choices. I'll
then ask you to order these choices according to your
preferences. Please try to do this as though you were
actually faced with the situation.

As we go along, feel free to ask questions.

Basic Situation

It's 9:00 AM on a Tuesday morning. You are in south
Florida and looking for a northbound load. You would prefer
a load to the Northeast, where you think you'll be able to get
another load for the return run to Florida. You anticipate
that return loads will pay around $1,300.


Table 5b. Carrier Experiment: Questions and Sets of
Alternatives

1. It's 9:00 AM on a Tuesday morning. You're in south
Florida and you're looking for a northbound load,
preferably to the Northeast where you know you'll be able
to get another load bound for Florida. You figure that the
load you'll get out of the Northeast will pay about $1,300.
After making several phone calls to brokers and shippers in
the area, you have received five offers. Each is to haul a
mixed load to New York. Each requires two pickups within
a short distance of one another and one drop at Hunt's
Point. As far as you can tell, these are the best offers
you'll be able to find.

Alternatives

Set 1.

No wait to load and rate = $1,800






Table 5b. (cont.)

Set 1 (cont.)

Load late this afternoon (wait of about 8 hours) and
rate = $1,900

Load tomorrow morning (24-hour wait) and rate =
$2,000

Load day after tomorrow (48-hour wait) and rate =
$2,100

Load in 4 days (96-hour wait) and rate = $2,200

Set 2.

No wait to load and rate = $1,800

Load late this afternoon (wait of about 8 hours) and
rate = $1,950

Load tomorrow morning (24-hour wait) and rate =
$2,100

Load day after tomorrow (48-hour wait) and rate =
$2,250

Load in 4 days (96-hour wait) and rate = $2,400

Set 3.

No wait to load and rate = $1,800

Load late this afternoon (wait of about 8 hours) and
rate = $1,850

Load tomorrow morning (24-hour wait) and rate =
$1,900

Load day after tomorrow (48-hour wait) and rate =
$1,950






Table 5b. (cont.)
Set 3 (cont.)

Load in 4 days (96-hour wait) and rate = $2,000

2. Again, it's a Tuesday morning and you're in south Florida
looking for a load. After calling around, you have received
five offers to haul a mixed load to five different northeast
cities. Each city is about 1,300 miles distant. All of the
shipments are ready to be immediately loaded and all
require two pickups and one drop.

After making delivery in the Northeast you'll be wanting to
return to Florida. In each city, if you get a backhaul to
Florida, you figure that it will pay between $1,000 and
$1,300. In some of the cities, backhauls to Florida are
more available than in others.

Alternatives

Set 1.


Northbound rate = $2,700.

Northbound rate = $2,400.


Northbound rate = $2,100.



Northbound rate = $1,800.


Northbound rate = $1,500.


No backhaul to Florida available.

Maximum expected wait for a
return load to Florida = 5 days.

Maximum expected wait for a
return load to Florida = 2 1/2
days.

Maximum expected wait for a
return load to Florida = 1 day.

No wait, return load ready when
you are.






Table 5b. (cont.)


Set 2.


Northbound rate = $2,400.

Northbound rate = $2,350.


Northbound rate = $2,300.



Northbound rate = $2,250.


Northbound rate = $2,200.


No backhaul to Florida available.

Maximum expected wait for a
return load to Florida = 5 days.

Maximum expected wait for a
return load to Florida = 2 1/2
days.

Maximum expected wait for a
return load to Florida = 1 day.

No wait, return load ready when
you are.


Set 3.


Northbound rate = $2,400.

Northbound rate = $2,300.


Northbound rate = $2,200.



Northbound rate = $2,100.


Northbound rate = $2,000.


No backhaul to Florida available.

Maximum expected wait for a
return load to Florida = 5 days.

Maximum expected wait for a
return load to Florida = 2 1/2
days.

Maximum expected wait for a
return load to Florida = 1 day.

No wait, return load ready when
you are.


3. This situation is very similar to the last one. The only
difference is that now I want you to pretend that it's the
end of May or early June. Produce shipments from Florida
are at their peak, but you know that over the next few
weeks they'll be dropping off very sharply.







Table 5b. (cont.)

Just to review the situation, it's Tuesday morning and
you're in south Florida looking for a load. After calling
around, you have gotten offers for loads to five different
northeast cities. Each city is about 1,300 miles distant.
The loads are very similar -- all mixed loads, and all
requiring two pickups and one drop.

You'll be wanting to return to Florida. In each city, if
you get a backhaul to Florida, you figure that it will pay
between $1,000 and $1,300. In some of the cities backhauls
to Florida are more available than in others.


Alternatives

Set 1.


Northbound rate = $2,900.

Northbound rate = $2,600.


Northbound rate = $2,300.



Northbound rate = $2,000.


Northbound rate = $1,700.


Set 2.


Northbound rate = $2,600.

Northbound rate = $2,500.


No backhaul to Florida available.

Maximum expected wait for a
return load to Florida = 5 days.

Maximum expected wait for a
return load to Florida = 2 1/2
days.

Maximum expected wait for a
return load to Florida = 1 day.

No wait, return load ready when
you are.



No backhaul to Florida available.

Maximum expected wait for a
return load to Florida = 5 days.







Table 5b. (cont.)


Set 2 (cont.)


Northbound rate = $2,400.



Northbound rate = $2,300.


Northbound rate = $2,200.


Maximum expected wait for a
return load to Florida = 2 1/2
days.

Maximum expected wait for a
return load to Florida = 1 day.

No wait, return load ready when
you are.


Set 3.


Northbound rate = $2,600.

Northbound rate = $2,550.


Northbound rate = $2,500.



Northbound rate = $2,450.


Northbound rate = $2,400.


No backhaul to Florida available.

Maximum expected wait for a
return load to Florida = 5 days.

Maximum expected wait for a
return load to Florida = 2 1/2
days.

Maximum expected wait for a
return load to Florida = 1 day.

No wait, return load ready when
you are.


EMPIRICAL RESULTS

For each alternative the five ranks were converted to
binary responses and the probit algorithm in SHAZAM (White)
was employed to estimate the parameters. The following
models were estimated:






First Broker Experiment

I=f(COST,TIME*DO,SPREAD or

VAR,COST*Du,TIME*Du,COST*DD,TIME*DD);


Second Broker Experiment

I=f(COST,TIME,SPREAD,COST*DH,TIME*DH,SPREAD*DH)


Carrier Experiment


I=f(RATE ,QUEU1,RATE2,QUEU2,RATE3,QUEU3);


where


COST =

TIME =

SPREAD =

VAR =

DO =


DU =

D =
D


D =
RATE =
RATEI


the transport charge or rate;

the time elapsed before delivery;

the range in possible delivery times;

the variance in possible delivery times;

a dummy variable equal to 1 when prices
were not expected to change;

a dummy variable equal to 1 when prices
are expected to rise;

a dummy variable that is equal to 1 when
prices are expected to fall;

a dummy variable that is equal to 1 for the
high-valued perishable commodity;

the rate the carrier would receive from
south Florida to New York;






QUEU1 =


RATE2 =



RATE =


QUEU2 =


QUEU3 =


the wait before loading for a load to New
York;

the rate the carrier would receive between
south Florida and a northeastern city in
late April;

the same as RATE2 but in late May or
early June;

the expected wait for a backhaul to Florida
in late April;

the expected wait for a backhaul to Florida
in late May or early June;


I = an index which is a monotonic transforma-
tion of the assigned ranks.


The model was estimated for each respondent.6 The
criteria used for selecting which explanatory variables to
include in each equation was the degree of concordance.7 A
pair of ranks is defined as concordant if the sign of Ri Rj
agrees with the sign of Ri" R-. Insignificant variables
were dropped if removing them did not decrease the degree of
concordance. This is equivalent to maximizing Kendall's Tau
statistic. Kendall's Tau is a measure of rank-order correlation
and is linearly related to the degree of concordance (DC), Tau
= 2*DC/100 1 (Snedecor and Cochran, pp. 192-193).

The results for each respondent are reported in Tables 6,
7, and 8. Also reported in these tables is the degree of
concordance and the likelihood ratio. The likelihood ratio can
be used to test a null hypothesis that the binary data were
distributed binomially, which is equivalent to a null hypothesis
that the coefficients of the explanatory variables are all zero.
The likelihood ratio is distributed as a Chi Squared statistic
with degrees of freedom equal to the number of explanatory
variables included in the equation. Based on the likelihood
ratio test, the null hypothesis would be rejected at the .001
probability level for all the estimated relations except for




Table 6. First Broker Experiment (Estimated Parameters).

Broker Time*Do Cost Spread Variance Time*Du Cost*Du Time*DD Cost*DD In(LR) DC


1
2
3
4
5
6
7
8
9
10
11
12
S 13
S 14
15
16
17
18
19
20
21
22
23
24
25
26
27
28


-.08771 -.00487
-.04636* -.00703
-.03361* -.00462
-.02968** -.00693
-.08131 -.00731
-.37451 -.00682
-.44744 -.02448
.03251** -.00441
-.00425
-.05334* -.00157*
-.24453 -.00539
-.24791 -.00570
-.01987
-.00444
-.05893 -.00498
.03560* .00183
-.13489 -.00985
-.00790
-.01988
-.01582
-.00818
-.08904 -.00209
-.01107
-.00986
-.01190
-.05407 -.00319
-2.16860 -.03068
-.07300 -.00186


-.23503
-.36065

-.38765
-.10830

1.73830
-.68424*
-.09220
2.56760


-.27106
1.26780
-.08618
-.10286

-.22563



-.18770

-.25072
-.42329
-.05666*


-.16622 -.00863* -.00003**
.35791** -.44288
-.01266 .12751 -.00706 -.03127*
.07573** -1.37360*
.00876** .00445* -.16078
-.19486 -.16782 -.00168**
-.72208* -.01537** -.09748
.33382 -.67546
.21137 -.19886
.55126* -.69929*
-.16868 .35468* -.87147*
-.07800 .03900** -.62787
.02485** -.33417*
.36888 -.14746
-.12750* -.00724* -.34168
.48017 -1.05090
-.11538 -.16660
.27698* .01033
-.05086 .13632** -.22524
-.29095
-.05397 -.11294
.37113 -.11757
-.27972 -.94909 -.03058**
.16780* -.02490**
.42851** .03098**
.01556** -.11844
-.44737 -.77053 -.00269** -2.16860
-.17333 .00137** -.30312


* 1.65 > asymptotic t ratio > 1.
** asymptotic t ratio < 1.


-.00486 54.53
-.00725* 117.63
71.73
-.02104* 123.96
-.00034 79.73
.00913** 91.86
.02219 112.15
-.03139 123.69
-.00366* 94.96
128.50
-.00986** 128.40
121.83
-.01340 146.62
.00093** 118.26
91.63
-02511 106.29
104.30
.00685 105.45
.01188* 136.56
152.63
.00411 104.76
-.00043** 78.94
.01207 113.50
.00691 113.12
.01294 108.23
-.00107** 46.88
154.85
81.16











Table 7. Second Broker Experiment Estimated Parameters.


Broker Cost Time Range Cost*DH Time*DH Range*DH ln(LR) DC


1 -.00749 -.15462 -.82643 -.00338 -.33577 109.49 92.5
2 -.00564 -.10277 .00359 -.05445* -.12153 48.79 78.3
3 -.00367 -.06515 -.14397 .00156 27.89 68.3
4 -.00935 -.78245 -.26573 129.16 92.5
5 -.00348 -.07643 -.19742 -.00247* -.09414 -.19193 50.39 79.2
6 -.00350 -.06638 -.10790 -.01179 -.07772* -.57987 72.37 85.0
7 -.01067 -.36418 .00810 -.09397 .19391 77.56 83.3
8 -.02734 -.60127 159.97 99.2
9 -.00160 -.03621* -.14830 .00180 .02762** .10783* 17.08 60.8
10 -.01033 -.14299 -.28970 .00475* -.06217** .01056** 78.96 81.7
11 -.00686 -.26210 -.51837 -.00105** -.99577** -.08904** 90.04 86.7
12 -.00536 -.12257 -.29251 -.00673* -.79015 -.06421** 87.98 85.0
13 -.00462 -.22319 -.02538** -.00139** -.03339** -.06421** 85.0


* 1.65 > asymptotic t ratio > 1.
** asymptotic t ratio < 1.






Table 8. Carrier Experiment Estimated Parameters.


Carrier RATE1


1
2
3
4
5
6
7
8
9
10
S11
12
13
14
15
16
17
18
19
20
21


.01222
.02936
.02402
.17596
.01576
.03525
.02518
.01434
.04360
.04238
.01615*
.04332
.01329
.02464
.05534
.04867
.01707
.06534*
.02247
.02346
.04626


* 1.65 > asymptotic t > 1.
** 1 asymptotic t < 1.


QUEU1


-.11165
-.35240
-.25761
-1.84660
-.07157
-.27284
-.20770
-.10112

-.40140
-.25418
-.38572
-.67933
-.21589
-.73536
-.48285
-.10616
-.58475
-.30321
-.21117
-.43629


RATE


-.00048**
.00183
-.00134**
.00627
.00736
.00356
.00213
.00703
.01745
.00174
.01512
.00181

.00229
.00193
.00355
-.00165*
.00786

.00189
.01766


QUEU2


-.8698*
-.19624
-.10128
-.62381
-.37907
-.29099
-.23051
-.36265
-.35880
-.19681
-2.43090
-.19457
1.92460
-.24541
-.20692
-.28985
-.06411*
-.40326
-4.09130
-.19165
-.36345


RATE


-.00199*
.00110*
-.00287*
.00238
.00146*
.00227
.00237
.00114*
.00158*

.00536*
.00109*

.00255
.00119*
.00226
-.00134*
.00106*

.00057*
.00160


UUEU.J


-.05337**
-.15101
-.02506**
-.24515'
-.13097
-.19583
-.18959
-.14813
-.19361
-.11681
-1.66540
-.14946
-1.92460
-.02359
-.16090
-.19755
-.01280**
-.14654
-4.09130
-.09774
-.19611


IlklK)TI


50.96
61.55
66.87
90.59
53.11
67.11
58.07
56.52
80.30
62.78
113.24
60.21
104.39
60.90
66.94
72.75
44.74
75.40
120.79
48.94
78.69


IniLK.)


DC

82.2
83.3
82.2
92.2
83.3
86.7
83.3
82.2
90.0
81.1
100.0
82.2
95.6
84.4
85.6
88.9
80.0
87.8
100.0
74.4
90.0







broker number 9 in the second experiment; in this case the
null hypothesis would only be rejected at the .01 probability
level.

The results are quite good. None of the respondents
ranked all of the alternatives on the basis of cost or rate
alone. In every case, there was some evidence that the level
of the other factors affected the rate brokers (and presumably
receivers) would be willing to pay or the rate the carriers
would be willing to accept.

The implied tradeoffs between the levels of the factors and
the freight rate the respondent would be willing to accept,
based on the estimated ordinal preference functions, are given
in Tables 9 to 11 for the first and second broker experiments
and the carrier experiment, respectively. The results indicate
a wide range of imputed values for the different respondents.
Because preferences do vary between individuals and individual
situations and perspectives are far from identical, the wide
range of values is not unexpected when individual preference
data is being examined.8 With one exception, in the few
cases where the results were counter to a priori expectations,
the estimated parameters are not significantly different from
zero. The exception is broker 16 in the first experiment; many
of his apparent choices are unexpected and significant.

The postulated model suggests that receivers are willing to
trade off an increase in freight rates for a decrease in total
delivery time and the spread of variance of the delivery time.
This was the observed results for all but brokers 8 and 16.
For broker number 8 the estimated standard error of time was
greater than the estimated parameter.

It is possible to trace out a partial demand curve for the
brokers in the first experiment for a service which is faster.
If the cost of the fast and slow service were equal, the
survey results indicate that the faster service would be chosen
(except for brokers 8 and 16). However, as the cost of the
faster service increases or, equivalently, as the difference in
cost between the faster and slower service increases, some
brokers will start to shift to the slower service. With regard
to the data in Table 9, the first broker to shift to the slower
service would be broker number 4, followed by number 2, then






Table 9. First Broker Experiment Imputed Value of a 1-
Unit Decrease in Time, Spread, and Variance.

Time, given prices
are expected to
Broker Time Spread Variance Rise Fall

1 18.03 48.31 12.32 13.96
2 6.59 51.29 -50.90 31.00
3 7.27 2.74 52.40 6.76
4 4.28 55.95 -10.93 49.12
5 11.13 14.82 -3.07 21.02
6 54.93 28.58 19.75
7 18.28 71.01 18.12 42.55
8 -7.38 155.21 -75.72 18.87
9 21.67 -49.68 25.11
10 33.93 1,633.33 -350.67 444.84
11 45.34 31.28 -65.76 57.12
12 43.49 13.68 -6.84 110.14
13 13.64 -1.25 51.63
14 285.80 -83.16 42.09
15 11.83 17.30 10.44 68.60
16 -19.46 -56.24 262.56 45.14
17 13.69 11.71 16.91
18 28.55 -35.05 -9.80
19 2.56 -6.86 28.16
20 18.40
21 6.59 27.71
22 42.68 89.96 -177.88 46.76
23 25.27 85.74 30.58
24 25.42 -17.01 8.44
25 35.56 -36.00 29.82
26 16.96 17.77 -4.88 27.81
27 70.68 14.58 23.09 70.68
28 39.19 93.04 -.74 162.71






Table 10. Value of a I-Hour Decrease in Time and Range.
Low-value good High-value good

Broker Time Range Time Range

1 20.64 110.30 45.08 75.98
2 18.23 23.65 59.50
3 17.73 39.19 30.86 68.19
4 83.73 28.43 83.73 28.43
5 21.97 56.74 28.66 65.43
6 18.98 30.85 9.43 44.99
7 34.14 36.61 66.34
8 21.99
9 22.58 92.49 -42.83 -201.95
10 13.84 28.04 36.76 50.02
11 38.21 75.57 32.40 76.77
12 22.87 54.58 75.52 51.77
13 48.33 5.50 42.70 14.91



Table 11. Carrier Experiment (Value of a 1-Unit
Decrease in Queue Length).

Location of Queue
Destination

Early in Peak of
Carrier Origin the season the season


9.14
12.00
10.73
10.49
4.54
7.74
8.25
7.05

9.47
15.74
8.90
51.11


-180.41
107.51
-75.54
99.49
51.47
81.64
108.23
51.56
20.56
113.35
160.82
107.60


-26.84
136.75
-8.74
102.90
89.65
86.22
80.02
129.94
122.42

310.88
137.13






Table 11. (cont.)

Location of Queue
Destination

Early in Peak of
Carrier Origin the season the season

14 8.76 107.12 79.95
15 13.29 107.01 134.77
16 9.92 81.64 87.43
17 6.22 -38.87 -9.55
18 8.95 51.31 137.88
19 13.50
20 9.00 101.27 171.78
21 9.43 20.59 122.23


3, etc. Plotting these points on a graph where the vertical
axis is the difference in cost and the horizontal axis is the
number of brokers using the fast service, a relationship such
as that illustrated in Figure 1 is defined. This is a partial
demand relationship, as the demand for transport is taken as
given, and the curve only illustrates the demand for a faster
service given a specified slower service. Also, the curve ends
at N*, the number of brokers in the panel. Thus, for a given
difference in rates, some brokers would use the faster, more
expensive service and others the slower, cheaper service.
Similar relations for spread and variance in time in transit
could be derived in like fashion.

When prices are expected to rise, it is expected that the
imputed value of a 1-unit decrease in time before delivery
would be less than when prices are expected to decline. This
was the observed result in all but three cases. In two of
these cases, the estimated parameters were insignificant
(brokers number 3 and 23), and the other was broker number
16. The different signs of the estimated values of a 1-unit
decrease in time when prices are expected to rise indicate the
different preferences of the individual brokers. Those with
negative signs apparently preferred a slower shipment in order










50



40



S30



S20
010


10



10



-10


1 2 3 4 5 6 7 8 9 10 11 12 13 14
Number of respondents

Figure 1. Partial demand for 1-hour decrease in transit time.
NOTE: Does not include broker 16 (first panel)






to take advantage of the expected price increase. The others
(save for the three exceptions) still would prefer faster to
slower shipment, though the values they place on speed are
less than when prices are expected to fall. This is depicted
in Figure 2. The portions of the partial demand curves for
reduced transit time which are in the negative range for rate
differentials reflect demands for slower service. If a rise in
commodity prices at the destinations is anticipated, this is a
theoretically reasonable result.

The relative importance of time elapsed before delivery and
range of delivery times to cost in the receiver's transport
decision for two different products, a high valued perishable
product (sweet corn) and a lesser valued, more durable
product (watermelons). It was expected that the value of a 1-
unit decrease in time and/or range would be greater for the
high-valued good. This was the observed result in 9 out of 13
cases for time and 11 out of 13 cases for range. The cases
where the results ran counter to expectations were all
associated with insignificant parameters.

Using the same procedure as was employed in Figure 1
yields two partial demand curves, one for low-valued goods
and one for high-valued goods (Figure 3). The demand curve
of shippers of low-valued goods for faster service lies to the
left of the demand of shippers of high-valued goods. That is,
brokers (and, presumably, receivers) shipping high-valued
goods are willing to pay more for faster service than when
they are shipping a less valuable commodity. Similar partial
demand curves could be derived for range.

The results for carriers indicate that both queue lengths
and the opportunity costs associated with expected changes in
load availabilities and/or rates are important considerations in
the decision to offer or not to offer service. The choices of
all of the carriers revealed a willingness to trade off shorter
queues at the origin (Florida) for lower freight rates. The
partial supply curve, derived from Table 11, is presented in
Figure 4. Similar results were obtained for all but three
carriers with respect to the impact of expected queue lengths
(for a return load) at the destination. All but the three
carriers (carriers numbers 1, 3, and 17) placed a positive value
on shorter expected queues at the destination. The estimated









400 -


300 -


200 -


100
Commodity price rising





- 1 0 Commodity price falling


-200


-300


-400
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

Number of respondents

Figure 2. Partial demand for 1-hour decrease in transit time by price expectations.
NOTE: Does not include broker 16 (first panel)







80 -


70 -


60 -

0
50 -


40 Higher value (Corn)


30 -
Lower value (Melons)

20 -


10
10 -




1 2 3 4 5 6 7 8 9 10 11
Number of respondents
Figure 3. Partial demand for 1-hour decrease in transit time by cargo value.
NOTE: Does not include broker 9 (second panel)





55-

50

45

40

35

30

S25

00 20

15

10

5



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

Number of respondents


Figure 4. Partial supply curve for a queue with 1-hour longer wait (origin).






parameters associated with the three exceptions were insig-
nificant at conventional levels. As expected, the partial
supply curve in the peak of the season lies to the left of that
for midseason (Figure 5), suggesting that carriers are cog-
nizant of the changes in opportunity costs associated with
waiting due to changing market conditions regarding future
loads.



CONCLUSIONS

The results suggest that the observed variation in freight
charges among commodities is not inconsistent with competi-
tive markets, or, alternatively, that the correlation of rates
with the value of the commodity does not necessarily imply
price discrimination. The estimated weights in the preference
functions indicate that the receivers are willing to pay more
for faster delivery when the commodity being shipped is a
high-valued, perishable item than when it is lower-valued and
less perishable. Thus, in a competitive market, the shippers
with high-valued and/or perishable commodities would bid up
the price of carriage to obtain trucks when trucks were
scarce and would also be willing to pay more for prompt,
dependable delivery when prices were expected to decline.

Furthermore, the results suggest that variation in rates
among destinations that is not justified by direct costs is also
consistent with competitive markets; or alternatively, that
variations in freight charges not explainable by direct costs do
not provide sufficient evidence of price discrimination. The
estimated weights in the carrier's preference functions and the
corresponding estimated tradeoffs between destinations and
rates carriers would be willing to accept indicate that carriers
take the imputed cost of waiting for the next load at the
destination into account when choosing which load to accept.
Thus, in a competitive market, truckers would be more willing
to accept a lower rate to a market where backhauls are more
likely to be available than they would to a destination where
they would have to wait longer for a backhaul or where a
backhaul would not be available.










300 -




200

End of season


S100 -
SMidseason



0




-100




-200 I I I I I 1 I I I I I I I I I I I
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
Number of respondents


Figure 5. Partial supply curve for a queue with 1-day longer wait (destination).







FOOTNOTES


1. See Beilock and Freeman and Blair, Kasserman, and
McClave for evidence of class rates after deregulation in
Arizona and Florida, and Beilock for evidence of a class-
like rate structure in the never-regulated trucking market
for produce.

2. For a discussion of price discrimination in transportation
markets based on commodity value, see Olson.

3. For a derivation of the proof for the above, see Layard
and Walters (pp. 259-267) or Wilson (Chapter 1).

4. In the last year produce haulers and brokers were regis-
tered in Florida (1979), there were over 22,000 carriers and
over 200 brokers. Also, the Florida Department of
Agriculture and Consumer Services maintains a list of over
2,000 active receivers.

5. Ordinal functions such as (13) and (15) and their estimates
are not unique. At best, they are unique up to a factor of
proportionality. However, it is possible that there are
several functions which will yield the same orderings.
Based on trying several models and estimating techniques,
the problem of non-uniqueness appears to be present only
when one or more of the estimated parameters have very
large estimated standard errors.

6. The models were estimated using three different techniques,
one programming approach -- both linear and nonlinear,
and two probability models. The results, while not
identical, did not differ in any significant fashion. In
terms of the degree of concordance, no single approach
demonstrated a clear advantage over the others. However,
both probability models provided more information in that
estimated asymptotic standard errors were available and the
algorithms were faster and chapter to run. It was decided
to report the results of using the profit algorithm for the
very simple and pragmatic reason that it provided the most
useful summary statistics.







7. One problem with using probability models to estimate
preference ranking results from the objective criteria.
Most algorithms for probability models maximize the
likelihood function appropriate to the model. This is not
the same as maximizing the degree of concordance,
although in most cases, it is a very close approximation.
It is possible, however, for the degree of concordance to
actually decrease as the maximum of the likelihood function
is approached. This apparently is a problem only when the
slope of the likelihood function in the area of the maxi-
mum is very flat with respect to some parameter. This is
indicated by large estimated variance for that parameter.
Our solution to this problem was a pragmatic one. We
simply examined the degree of concordance at each
iteration of the maximization process and stopped the
procedure when the maximum degree of concordance had
been obtained.

8. Other possible reasons include the non-uniqueness of
weights in ordinal preference functions, misspecification of
the preference function, and possible flaws in the ex-
perimental design. The non-uniqueness of weights,
however, should not be a problem when the substitution
between change in the level of an attribute and cost is
examined. The form of the preference equations used are
consistent with a maintained assumption of transitivity, and
it is possible that the attributes considered did not enter
into the respondents' decision-making function in a
transitive fashion. However, this is unlikely as all the
factors would affect profits. It is possible that some
respondents may not have really understood how the game
is played or may have intentionally given either wrong or
misleading answers. This is probably what happened with
broker number 16 in the first experiment. This broker's
responses were unexpected on the first and last questions,
and the responses to the second and third questions were
inconsistent with the reponses to the first.







REFERENCES


Baessman, R., and A. Daughety. "Complementarity in Production
and Empty Backhauls in Transportation." Presented at
Western Economics Association, 1977.

Baumol, W. Economic Theory and Operations Research, 4th ed.
New Jersey: Prentice-Hall, Inc., 1977.

Beilock, R. "Is Regulation Necessary for Value-of-Service
Pricing?" Rand Journal of Economics 16,1(1985):93-102.

Beilock, R., and J. Freeman. The Impact of Motor Carrier
Deregulation on Freight Rates in Arizona and Florida. U.S.
Department of Transportation, 1985.

Beilock, R.P., and G. Fletcher. "Exempt Agricultural Commodity
Hauler in Florida." Proceedings of the Transportation
Research Forum 24(1983):444-450.

Beilock, R.P., and S. Shonkwiler. "Modeling Weekly Truck Rates
for Perishables." Southern Journal of Agricultural Econom-
ics 15(1983):83-87.

Blair, R., D. Kasserman, and J. McClave. The Economic Effect
of Deregulation of Interstate Trucking in Florida. Report
prepared for the Florida Attorney General's Office, 1984.

Cattin, P., and R. Wittnick. "Commercial Use of Conjoint
Analysis: A Survey." Journal of Marketing 46(1982):44-53.

Davidson, J.D. "Forecasting Traffic on STOL." Operations
Research Quarterly 24,4(1973):561-569.

DeVany, A.S., and T.R. Savings. "Product Quality, Uncertainty,
and Regulation: The Trucking Industry." The American
Economic Review 67(1977):583-594.

Fauth, G.R. "The Role of Commodity Value in Motor Carrier
Class Rate Structure." Paper presented at the Conference
of Regulatory Reform in Surface Transportation, Syracuse
University, Syracuse, New York, March, 1983.







"Florida Conditions and Supply Report on Fresh Fruits and
Vegetables." Federal-State Market News Service. Winter
Park, Florida, 1984 (weekly), various issues.

Garrod, P. "A General Program for Monotonic Regression."
Unpublished paper, Department of Agricultural and Re-
source Economics, University of Hawaii, 1980.

Garrod, P., and W. Miklius. "Estimated Value of Improvements
in Transport Services." Transportation Research 17A(1983)-
:33-37.

Green, P.E. "On the Analysis of Interactions in Marketing
Research Data." Journal of Marketing Research 10(1973):4-
10-420.

Harris, R. "Economies of Traffic Density in the Rail Freight
Industry." Bell Journal of Economics 8(1977):556-564.

Jara-Diaz, S. "Transportation Production and Cost Functions."
Transportaion Science 16,4(1982):522-539.

Johnson, M. "Estimating the Influence of Service Quality on
Transportation Demand." American Journal of Agricultural
Economics 58(1976):496-503.

Johnson, R. "A Simple Method for Pairwise Monotone Regres-
sion." Psychometrika 40(1975):163-168.

Judge, G., R. Hill, W. Griffiths, H. Lutkepohl, and T. Lee.
Introduction to the Theory and Practice of Econometrics.
New York: John Wiley & Sons, 1982.

Kruskal, J.B. "Analysis of Factorial Experiments by Estimating
Monotone Transformations of the Data." Journal of the
Royal Statistical Society. B. 27(1965):251-263.

Layard, P., and A. Walters. Microeconomic Theory. New York:
McGraw-Hill, 1978. Marshall, A. Principles of Economics,
9th Ed. London: MacMillan & Co., Ltd, 1964.






Miklius, W., and D.B. DeLoach. "A Further Case for Nonregula-
ted Trucking." Journal of Farm Economics 47(1965):933-
947.

Olson, J. "Price Discrimination by Regulated Motor Carriers."
American Economic Review 62(1972):395-402.

Pavlovic, K., G. Long, F. Reaves, and T. Maze. Domestic
Transportation for Florida Perishable Produce. Transporta-
tion Research Center, University of Florida, 1980.

Pekelman, D., and S.K. Sen. "Mathematical Programming Models
for the Determination of Attribute Weights." Management
Science 20(1974):1217-1229.

Sadanand, A., and L. Wilde. "A Generalized Model of Pricing
for Homogeneous Goods Under Imperfect Information."
Review of Economic Studies 49(1982):229-240.

Salop, S., and J. Stiglitz. "Bargains and Ripoffs: A Model of
Monopolistically Competitive Price Dispersion." Review of
Economic Studies 44(1979):492-510.

Satterthwaite, M. "Consumer Information, Equilibrium Industry
Price, and the Number of Sellers." Bell Journal of
Economics 10(1979):483-502.

Snedecor, G., and W. Cochran. Statistical Methods, 7th ed.,
Ames: Iowa State Press, 1980.

Srinivasan, V., and A. Shocker. "Linear Programming Techniques
for Multidimensional Analysis of Preferences." Psvchome-
trika, 38(1973):337-369.

Srinivasan, V. "Estimating Weight for Multiple Attributes in a
Composite Criterion Using Pairwise Judgments." Psvchome-
trika 38(1973):473-493.

White, K.J. "A General Computer Program for Econometric
Methods SHAZAM." Econometrica 46(1978):239-240.






Wilde, L., and A. Schwartz. "Equilibrium Comparison Shopping."
Review of Economic Studies 46(1979):543-552.

Wilson, G. Economic Analysis of Intercity Freight Transporta-
tion. Bloomington: Indiana University Press, 1972.





































t UNIVERSITY OF FLORIDA 1




This publication was produced at a cost of $1,554.45, or 78 cents
per copy to provide information about the role of service and trip
characteristics in carrier and load selection.


All programs and related activities sponsored or assisted by the Florida Agricultural
Experiment Station are open to all persons regardless of race, color, national origin,
age, sex, or handicap.


ISSN 0096-607X




University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs