Content uploaded by Edoardo Marcucci
Author content
All content in this area was uploaded by Edoardo Marcucci
Content may be subject to copyright.
Behavioral Modeling of Urban Freight
Transport
Testing Non-Linear Policy Effects for Retailers
Edoardo Marcucci and Valerio Gatta
Abstract Decision makers in urban goods movement (UGM) typically need to
assess the impact new policy interventions might have on freight distribution. The
effects of policy changes are inextricably related with the extant regulatory
framework that also influences the relationships among the various actors inter-
acting along the supply chain. The operators commonly considered important,
given the crucial role they play in UGM, are: retailers, transport providers, and
own-account. Notwithstanding the admittedly important role that a detailed
knowledge of these three agent categories has for a correct policy implementation
there is a limited knowledge concerning the specific preferences and behavior of
each agent-type. It is de facto assumed that retailers, own-account and transport
providers have homogenous preferences and can be seamlessly treated. The
upsurge of behavioral models and the acquisition of data necessary to predict
goods and vehicle flows both under the current and, more importantly, under
altered policy/regulatory conditions explains the progressive importance that is
attributed to an agent-based perspective. This research reports the result of a stated
ranking exercise conducted in the Limited Traffic Zone in 2009 in the city center
of Rome focusing on retailers which demand freight transport services and play an
important role in extended supply chains. This paper proposes a comparison
between two different Multinomial Logit model specifications where non-linear
effects for the variations of the levels of the attributes considered are studied and
detected. A meaningful comparison between willingness to pay measures derived
by the two model specifications is proposed so to avoid known scale problems. The
results obtained are very interesting and meaningful from a policy perspective
since they show potentially differentiated effects of the policy implemented in
deep contrast with the, often assumed, homogenous effect hypothesis.
E. Marcucci (&)V. Gatta
DIPES/CREI, University of Roma Tre, Rome, Italy
e-mail: edoardo.marcucci@uniroma3.it
J. Gonzalez-Feliu et al. (eds.), Sustainable Urban Logistics: Concepts, Methods
and Information Systems, EcoProduction, DOI: 10.1007/978-3-642-31788-0_12,
ÓSpringer-Verlag Berlin Heidelberg 2014
227
Keywords Freight operators Limited traffic zone Non-linear effects
Preference heterogeneity Retailers
1 Introduction
Cities have historically, but more so for modern cities, manifested a strong
dependence on freight transport systems to efficiently guarantee the net inflow of
goods and ensure the availability of the necessary resources to fuel economic and
urban growth. Local policy makers have intervened on the articulated contractual
relationships among agents so to achieve the desired policy objectives. The most
important agent-types in urban goods movement (UGM) are: retailers, transport
providers and own-account. Few are the studies that have explicitly investigated
the specific preferences and behavior of each of these agent-types (Stathopoulos
et al. 2012; Stathopoulos et al. 2011) notwithstanding the a priori relevance that is
ascribed to them (Ogden 1992). At the base of this research gap in this field one
can safely put the lack of appropriate data that is, in turn, linked to elicitation costs
and the low interest agent-types usually show when asked to participate in applied
research projects in this field. The capability policy interventions have in pro-
ducing the desired results is inextricably intertwined with the detailed knowledge
policy makers need to have concerning the most likely response the intervention
will produce given the extant regulatory, contractual and consuetudinal relation-
ships that characterize this sector in the given city where the policy is to be
implemented. In other words, we believe that one-size-fit-all policies, implying
policy transferability, are not easy to define nor to implement in accordance to
what has already been underlined by recent research (Stathopoulos et al. 2012).
The results reported and discussed are based on a data set derived from a
research conducted for a Volvo Research Foundation project (2009) that focused
on ex-ante policy mix evaluation for freight transport policies. The study con-
centrated on the freight Limited Traffic Zone (LTZ) in the city center of Rome.
The analysis takes advantage of the data set collected that explicitly differentiates
among three agent-types. The data include a wide range of information including
both specific respondent’s and his/her company’s characteristics as well as the
results of a Stated Ranking Exercise (SRE) where interviewees were asked to rank
alternative policy scenarios. The paper reports the results of two Multinomial
Logit model (MNL) specifications aimed at investigating the non-linear effects of
policy intervention on retailers’ utility functions in a similar vein to Rotaris et al.
(2012).
1
A comparison is performed, via willingness to pay (WTP), between the
potentially distorted scenario evaluations deriving from the assumption of linear
1
Non-linear effects on utility function can be also tested via self-stated attribute cutoff. Please
refers to Marcucci and Gatta (2011) for a detailed description and application.
228 E. Marcucci and V. Gatta
policy effects. Our results allow us to comment on the distorted policy forecasts
that would be produced by simpler and rougher treatment of the information
acquired. On the base of recent evidence (Stathopoulos et al. 2011) we assume that
the relevant policy attributes for retailers are: (1) number of loading and unloading
bays (LUB); (2) probability of finding loading and unloading bays free (PLUBF);
(3) entrance fee (EF) charged to enter the LTZ.
The paper contributes to UGM literature by bridging a specific gap via in depth
investigation of retailers’ preferences. Recent papers have investigated the role of
heterogeneity for both transport providers (Gatta and Marcucci 2013) and own-
account (Marcucci and Gatta 2013) agents with respect to policy intervention
whereas this paper focuses the attention on the presence and magnitude of non-
linear effects given the different levels of the attributes considered. Policy makers
usually intervene and evaluate policies assuming that attribute variations have
linear effects thus hypothesizing there is no dependence on the status quo (SQ)
level of the policy variable and, furthermore, that both increases and decreases
have symmetric effects on agent’s utilities. The results reported show that one
cannot assume linear effects and consequently both the direction of the variation as
well as its magnitude should explicitly be considered when assessing a given
policy change. Having estimated the coefficients for the various attributes and
levels we calculate, via WTP measures, the biases that a linear assumption con-
cerning the effects implies.
The paper is structured as follows. Section 2 reports a short literature review
concerning agent-type analysis for UGM. Section 3 describes the survey instru-
ment developed and the data acquired while Sect. 4 reports and discusses the
econometric results and policy implications. Section 5 concludes and illustrates
future research endeavors.
2 Literature Review
Freight modeling is usually performed via aggregate models thus limiting the
attention dedicated to agent-level considerations that represent the appropriate
level of analysis to investigate if a behavioral approach to the phenomenon is
adopted.
2
This section succinctly summarizes recent literature that testifies the
increasing attention paid to behavioral issues in UGM.
3
Hensher and Figliozzi (2007) underline the weaknesses of the standard
approaches to UGM modelling. In fact, the modified four-step approach (M4SA)
2
UMG literature analysis also reveals a substantial heterogeneity in the approaches adopted
relates to the public or private perspective considered. The former mainly focuses on the
definition of policies for reducing the negative external effects on cities, while the latter
essentially aims at enhancing the efficiency of business operations (Corò and Marcucci 2001;
Marcucci and D’Agostino 2003).
3
For a definition of UGM see for example Ambrosini and Routhier (2004).
Behavioral Modeling of Urban Freight Transport 229
when used to simulate UGM does not adequately consider the complexity char-
acterizing freight movements at different geographical scales. At the same time,
however, it is appropriate to note that not all researchers in the field univocally
share this view. For different positions one could refer to Sonntag (1985); Ogden
(1992); Ambrosini et al. (2008) and the article by Gonzalez-Feliu et al. (2013)in
this book.
This explanatory deficit is particularly relevant since the M4SA is structurally
not capable of explaining potentially relevant preferences for current scenarios
and, even more important, the possible reactions to policy changes. On the con-
trary, models adopting a behavioral approach (BA) to UGM modeling, repre-
senting only part of the larger disaggregate models set, explicitly consider
stakeholders’ utility maximization.
4
BA to UGM presume the researcher is capable
of univocally and correctly identifying key decision makers so to develop an
agent-based micro-simulation approach modeling framework that both describes
and forecasts the behavior of the actors considered (Liedtke and Schepperle 2004).
UGM is, according to a copious and qualified group of eminent researchers (Gray
1982; Wisetjindawat et al. 2006; de Jong and Ben-Akiva 2007; Russo and Comi
2011; Filippi et al. 2010; Comi et al. 2012; Sammi et al. 2009; Chow et al. 2010;
Roorda et al. 2010) an appropriate field of research where the development of
micro agent-based models is most likely going to produce policy relevant results.
In fact, since freight is not moved for its own sake and the underlying motivations
can always be traced back to the profit maximization intent of a given agent,
participating in the process, it appears appropriate to analyze the choices made
according to a well-known and robust theoretical framework that has successfully
been applied in many other fields (also outside transportation) whenever economic
agents’ modeling is deemed appropriate.
Different UGM options are influenced, given the derived nature of freight
transport demand, in their relative convenience for each agent-type considered, by
changes in fuel prices, land use patterns and pricing strategies in the markets that
demand freight transport services. It has been suggested (Puckett and Greaves
2009) that in order to understand the impacts, measured in terms of the market
outcomes that a policy might produce, one should conjointly consider all the
instruments policy makers could use and the relevant attributes capable of
affecting agents’ freight choices.
Policy makers are intrinsically and structurally interested in knowing, before
implementing a given policy, what the most likely reactions will be in terms of
achievement of the desired objectives. As it will be apparent when discussing the
econometric results (Sect. 4) the research proposed can quantify the WTP for the
possible policies implemented with respect to the reference scenario before
the policy is actually put into action in a real-life context. This paper focuses on
4
In fact, the success or failure of UGM initiatives mainly depends on the reaction of
stakeholders to the implementation characteristics’ of policies (see, for instance, Marcucci and
Danielis 2008; Paglione and Gatta 2007; Danielis and Marcucci 2007; Marcucci et al. 2007).
230 E. Marcucci and V. Gatta
the role and preference of retailers that, in the context studied, play a relevant role
(Quack and de Koster 2009).
3 Survey Instrument and Data Description
This paper is based on data acquired in Rome’s LTZ between March and
December 2009 thanks to a project carried out for Volvo Research Foundation
(2009). The LTZ in the city center of Rome was first implemented in the late 80s
over a 5 km
2
area originally banned to non-resident vehicles only. Only Euro 1
and more fuel-efficient vehicles are allowed to enter the LTZ with free access
granted to residents while others (e.g. retailers and freight carriers) pay an access
fee. Cameras and optical character recognition software are used to enforce the
system which operates diurnally with a yearly entrance fee of 565€per number
plate.
Notwithstanding the extensive list of impediments applying generically to all
agents a wide ranging of ad hoc exemptions applies to third party freight operators.
The regulation, after a careful reading of all the exemptions conceded, seems
mostly targeted to discouraging own-account operators.
As it is for the questionnaire development it is important to first define, select,
develop and customize the attributes to be included in the questionnaire which, in
our case, was a SRE since it was considered most appropriate to use a ranking
exercise given the final aim was to unveil agents’ preferences concerning UGM
policies which are not de facto ‘‘chosen’’. The project involved different phases
among which the most important are: (1) advancement from stakeholder consul-
tation to final attribute selection criteria; (2) attribute definition; (3) levels and
ranges selection; (4) progressive design differentiation by agent-type (Stathopoulos
et al. 2011).
The SRE alternatives are characterized by a set of attributes, which can take
several levels. The attributes considered were selected thanks to: (1) literature
survey; (2) previous UGM studies performed in Rome; (3) focus groups with
experts. An in depth review of the literature adopting an agent-based perspective
allowed the identification of a set of eligible attributes that represented potentially
conflicting policy instruments.
5
Previous UGM studies in Rome (STA 2001; Filippi and Campagna 2008)
together with expert and stakeholder focus groups were very useful in guiding the
attribute selection process
6
that were characterized by high and shared support of
5
Nighttime deliveries, for instance, were considered efficiency enhancing by carriers but
considered a mere increase in costs by retailers and were consequently excluded.
6
An important phase of the expert surveys focused on defining the policies considered most
appropriate to mitigate the identified UGM problems (Stathopoulos et al. 2011). Volvo Report
(2010) provides a detailed overview of the link between the stakeholder survey results and the
attributes used in the SRE.
Behavioral Modeling of Urban Freight Transport 231
the stakeholders contacted (Stathopoulos et al. 2011). The attributes were also
validated via a pilot test with real operators. The final list included: LUB, PLUBF,
and EF. All attributes are considered as possible levers of intervention by local
decision-makers and perceived as appropriate measures for possible policy mixes
by stakeholders (Marcucci et al. 2012). Attributes, number of levels, and ranges
are reported in Table 1. Attributes are all characterized by, at least, three levels
thus allowing the test of non-linear effects that represent the core of this paper and
play a special role in the evaluation of policy reactions to policy changes where
different effects can be originated by varying specific levels. Joint stakeholder as
well as local policy-makers meetings were an important source of information
concerning the attribute distribution and range. Based on the ranges provided by
the stakeholders and the comments from local planning changes the minimum and
maximum points of the attribute ranges were defined to achieve realism and
properly mirror plausible policy changes. For LUB and PLUBF the minimum
coincides with the current situation, implying that the policy scenarios only pro-
posed an increase in the levels. EF was defined having a wide range of variation in
both directions with respect to the status quo level since past policy changes have
been quite abrupt.
A SRE is adopted to test currently unavailable options. The alternatives pre-
sented to respondents, who had to rank them, include two policy options plus the
SQ alternative. Table 2reports an example of a SRE task.
In total, 252 interviews were finalized and 229 used after removing pilot
interviews that were utilized to preliminary test the design. The sample of retailers
used for estimation consists of 90 units whose distribution is scattered in nine main
macro-freight sectors, namely: (1) food (fresh, canned, drinks, tobacco, bars, hotels
and restaurants); (2) personal and house hygiene (detergents, pharmaceuticals,
cosmetics, perfumes, watches, barbers, etc.); (3) stationery (e.g. paper, newspa-
pers, toys, books, CDs etc.); (4) house accessories (e.g. dish washers, computers,
telephones, metal products etc.); (5) car accessories (e.g. vehicle components,
Table 1 Attribute levels and ranges used in the SRE
Attribute Number
of levels
Level and range of attribute
(Status Quo underscored)
Loading/unloading bays (LUB): 3 400, 800, 1200
Probability of free l/u bays (PLUBF): 3 10 %,20%,30%
Entrance fee (EF): 5 200€, 400€,600€, 800€, 1000€
Table 2 Example of a ranking task
Policy 1 Policy 2 Status Quo
Loading/unloading bays (LUB): 400 800 400
Probability of free l/u bays (PLUBF): 20 % 10 % 10 %
Entrance fee (EF): 1000€200€600€
Policy ranking hhh
232 E. Marcucci and V. Gatta
vehicles, gasoline, etc.); (6) services (e.g. laundry, flowers, live animals, acces-
sories and animal food, etc.); (7) clothing (cloth, leather, etc.); (8) construction
(e.g. cement, scaffold, chemical products, etc.); (9) other (all those not included in
previous categories). We follow the same classification proposed in Filippi and
Campagna (2008) for comparison purpose, since it represents still today the most
reliable investigation of urban freight data in Rome.
4 Econometric Results and Policy Implications
This section reports the results of the models
7
estimated for retailers based on the
data obtained via the SRE described in the previous section.
The methodological framework is based on random utility theory, where utility
is modeled as a random variable consisting of both deterministic and stochastic
part. The former is a function, linear in its parameters, of the fundamental attri-
butes, while the latter is the random term. Different assumptions about the dis-
tribution of the random term imply different discrete choice models that can be
used to analyze the gathered choice data with the purpose of estimating the
parameters associated with the attributes. In the early 1970s Mc Fadden (1974)
developed MNL which is derived from the assumption that the error terms of the
utility functions are independent and identically Gumbel distributed. MNL has
many interesting and much appreciated advantages (closed form, ease of inter-
pretation, etc.) is also characterized by relevant drawbacks linked to preference
homogeneity assumption across respondents. Even if confounded for the scale, the
estimated parameter represents the marginal utility of each attribute variation and
implies an equal taste for all agents for the given attribute.
8
The first model (M1), employing a MNL specification,
9
utilizes all attributes as
linear and normalized while the second (M2) adopts an effects coding for the
variables in order to investigate potential non-linear effects of the different levels
of the explanatory variables.
M1, reported in Table 3, employing just normalized variables, provides inter-
esting results and also shows a good fit of the model (adj. Rho
2
=0.142; 5 Coeff.).
All the coefficients are statistically significant and with the expected sign with the
exception of the two alternative specific constants (ASCs) for which there was no
strong a priori concerning the sign. In particular LUB and PLUBF have a positive
coefficient since an increase in either the number of loading and unloading bays or
7
The models are estimated using NLOGIT 4.0.
8
For a detailed discussion of the methodological framework and possible applications of
discrete choice models see, for example, Marcucci (2005); Gatta (2006).
9
We just recall that a MNL specification of the model implies an implicit assumption
concerning the independence from irrelevant alternatives. In other words, it is assumed the
unobserved effects homogeneously impact all the alternatives in the same way that is equivalent
to hypothesizing that the error component is identically and independently distributed.
Behavioral Modeling of Urban Freight Transport 233
in the probability of finding them free has a positive impact on retailers’ utility. On
the contrary, an increase in EF has a negative impact on retailers’ utility. M1 also
includes two ASCs for the unlabeled hypothetical cases (ASC_Alt1, ASC_Alt2)
whose coefficients represent the overall alternative impact on retailers’ utility
when all the coefficients of the other attributes have a zero value. In our case,
results show that, there is an a priori evaluation against the SQ (ASC_Alt3 has a
negative sign) and, after conducting a Wald test for ASC_Alt1 and ASC_Alt2, we
cannot reject the null that the difference between the two coefficients is different
from zero. In summary, one can affirm that ASC_Alt1 and ASC_Alt2 have a
positive, but undistinguished between them, effect on utility. Furthermore, it is also
interesting to note that the ASC inclusion in the model not only substantially
increased the model fit but also provided more realistic interpretation of the
parameters.
The normalization adopted for the explanatory variables allows us to compare
the estimated coefficients of the attributes considered. One can notice that tariff
plays the lion part in explaining retailers’ preferences. In fact the EF’s coefficient
is more than double the sum of LUB and PUBF coefficients. This result is further
reinforced by looking at the t-stat of each of the variables considered that testify
EF’s coefficient is, almost for sure (t-stat 16.44), different from zero even if LUB
and PLUBF coefficients are highly significant too (respectively t-stat 5.24 and
6.51).
M2, reported in Table 4, differs from M1 in the treatment of the variables
which, in this case, are effects coded.
10
The different coding aims at detecting
possible non-linearities in the explanatory variables’ effects. In fact, the estimation
of a single parameter for a given attribute will give rise to a linear estimate (i.e.
slope) and we generically refer to these estimates as linear estimates (M1). An
attribute’s impact can be estimated with two dummy (or effects) parameters, which
are usually referred to as a quadratic estimate or higher degree dummy (or effects)
parameters which are also referred to as polynomial of degree L-1 estimates (with
L denoting the number of dummy or effects parameters). In more detail, one can
affirm that the more complex the part-worth utility function, the more advisable is
to move to more articulated coding structures capable of recovering the necessary
data to estimate the more complex non-linear relationships.
Table 3 Econometric results based on M1
Variable Coefficient St.Err. T-Stat Expected Sign
LUB 0.253 0.048 5.24 +
PLUBF 0.347 0.053 6.51 +
EF -0.699 0.042 -16.44
ASC_Alt1 0.824 0.154 5.32 *
ASC_Alt2 0.657 0.136 4.82 *
10
For a clear description of effects coding the explanatory variable please refer to Hensher et al.
(2005), pp 119–121.
234 E. Marcucci and V. Gatta
M2, thanks to the effects coding of the variables, provides more detailed
information and is characterized by a statistically significant better fit
11
with
respect to M1 (adj. Rho
2
=0.154; 9 Coeff.). All reported coefficients are statis-
tically significant. In fact, the LUB2 (e.g. the second level of the variable LUB, i.e.
800) coefficient, not reported in the table, was not statistically significant thus
suggesting agents’ utility is not influenced by a variation of only 400 LUB from
the SQ situation (i.e. 400).
12
As it is for the PLUBF one can notice that there is an evidently non-linear effect
of the variable. In fact, going from a 10 Probability Base Points (PBP) for PLUBF
(i.e. SQ level) to 20 PBP we have a much greater impact on retailers’ utility
[BetaPLUBF2-1 =BetaPLUBF2 (0.246)—BetaPLUBF1 (-0.509) =0.756] than
going from 20 PBP to 30 PBP [BetaPLUBF3-2 =BetaPLUBF3 (0.262)—
BetaPLUBF2 (0.246) =0.016]. EF is the variable that benefited the most from the
adoption of effects coding in detecting non-linearities. This is both due to the
presence of five levels compared to the three levels for the other variables as well
as to their symmetricity with respect to the SQ (i.e. 600€). The analysis of ASCs
leads us to the same conclusions reported for M1.
With reference to Fig. 1, and in line with prospect theory (Kahneman and
Tversky 1979), one can observe that reductions in EF produce positive effects on
utility compared to negative effects induced by opposite variations of similar
amount. Initial variations, in both directions, from the SQ (EF3 =600€) have
bigger effects [Beta
EF3–4
=Beta
EF3
(0.300)-Beta
EF4
(-0.762) =1.062 and
Beta
EF2–3
=Beta
EF2
(0.937)–Beta
EF3
(0.300) =0.637] with respect to subsequent
ones [Beta
EF4–5
=Beta
EF4
(-0.762)–Beta
EF5
(-1.589) =0.828 and Beta
EF1–2
=
Beta
EF1
(1.114)–Beta
EF2
(0.937) =0.176]. In fact, for positive variations (EF
increases; EF4 =800€and EF5 =1.000€) we have Beta
EF3–4
=1.062 [-
Beta
EF4–5
=0.828 and for negative variations (EF reductions, EF2 =400€and
EF1 =200€) we have Beta
EF2–3
=0.637[Beta
EF1–2
=0.176. Furthermore, still
Table 4 Econometric results based on M2
Variable Coefficient St.Err. T-Stat
LUB3 0.215 0.046 4.68
PLUBF2 0.246 0.059 4.15
PLUBF3 0.262 0.068 3.86
EF1 1.113 0.104 10.65
EF2 0.937 0.087 10.67
EF4 -0.761 0.099 -7.68
EF5 -1.589 0.126 -12.54
ASC_Alt1 1.085 0.166 6.51
ASC_Alt2 0.814 0.143 5.66
11
We checked this by performing a log-likelihood ratio test.
12
Therefore, we recoded this variable so that LUB3 =1 when LUB =1,200 and -1 otherwise
(according to the effects coding of the variables).
Behavioral Modeling of Urban Freight Transport 235
in line with prospect theory we find that positive variations of equal amount are
valued less than negative variations and, in our case, this is testified by both inner
variations [Beta
EF3–2
(0.637) \Beta
EF3–4
(1.062)] as well as by outer variations
[Beta
EF1–2
(0.176) \Beta
EF4–5
(0.828)]. Similar considerations also apply to
PLUBF (see Fig. 2).
In order to analyze the impact of different estimation methods, define and
measure the potential biases for policy implementation one can use WTP/WTA to
avoid scale problems that would, otherwise, fraud the comparison.
As it is well documented in the literature (Daly et al. 2010) there are different
methods that can be used to test the statistical significance of the ratio of coeffi-
cients between the desired attribute and the monetary one representing the base of
any WTP/WTA measures.
Testing the statistical significance of the ratios is not only important per se,
since it allows the researcher to infer reliability of the results obtained especially
when using them for simulation purposes, but also because it is reasonable to
assume some heterogeneity in the sample selected. Especially in connection with
this last point and for policy evaluation purposes it is interesting to estimate
monetary confidence intervals rather than using single point estimates.
Among the methods that one can use to construct confidence intervals for these
ratios the most popular are: (1) Krinsky and Robb (Krinsky and Robb 1986,1990);
(2) Bootstrap (Efron 1979; Mooney and Duval 1993; Efron and Tibshirani 1993);
(3) Delta Method (see e.g. Greene 2003). In our case we opted for this last method.
WTP are assumed normally distributed and, thus, symmetrical around the mean.
Delta Method’s estimates of the variance of a non-linear function of two random
variables is obtained by taking a first-order Taylor expansion around the mean
value of the variables and calculating the variance for this expression (Hole 2007).
Our choice is motivated by two main considerations: (1) Delta Method is an
exact method compared to both Krinsky-Robb and Bootstrap where a simulated
distribution for the variable of interest is generated; (2) Shanmugalingham (1982)
has empirically shown that the normality assumption underlying the Delta Method
is, in general, less tenable when the standard deviation of the denominator variable
Fig. 1 Part-worth utilities
for EF
236 E. Marcucci and V. Gatta
is large relative to its mean and this is not the case for our results given that the
cost coefficient is strongly significant and no skewness risks are incurred.
13
Table 5and 6report the WTP estimates respectively for M1 and M2. In both
cases all the reported estimates are statistically significant and, with reference to
M2, non-linear effects are clearly evident.
To interpret the meaning of coefficients’ estimates one has to recall that for
estimation purposes and in order to avoid measurement unit effects (e.g. LUB
absolute numbers –400, 800, 1.200–; PLUBF PBP –10, 20, 30–; EF Euros –200€,
400€, 600€, 800€, 1,200€–), it is advisable to normalize all the variables so to
sterilize the unit of measurement effect.
Notwithstanding the considerations above we deem useful to explain in detail
how the monetary WTP were calculated so to facilitate interpretation. For instance
(with reference to M1), as it is for LUB, departing from a normalized WTP of
0.362 and wanting to know the amount of money the interviewees are willing to
pay for an additional LUB one has to perform the following calculations: 0.362 *
(200€/400LUB) =0.18€/LUB whereas for PLUBF we have 0.496 * (200€/
10PBP) =9.93€/PBP.
At this point from a policy perspective it is interesting to compare two different
policies that guarantee, in alternative ways, equal results. In more detail, one can
compare how much people are willing to pay to have an extra LUB free either via
additional LUB construction or via increased probability of finding a LUB free. In
order to perform the comparison one has to recall that, taking the SQ as a
Fig. 2 Part-worth utilities
for PLUBF
Table 5 WTP estimates with
delta-method (based on M1) Variable Coefficient St.Err. T-Stat
LUB/EF 0.362 0.064 5.64
PLUBF/EF 0.496 0.066 7.47
13
Notwithstanding the above made considerations we think it would be interesting to test under
which conditions each of the three methods provides the best results. We are presently working
on a paper specifically addressing this issue using both simulated as well as real data.
Behavioral Modeling of Urban Freight Transport 237
reference, we need to construct 10 extra LUB to ensure one additional free LUB.
On the other hand, one could obtain the same result by an increase of 0.25 PBPs.
One extra free LUB is evaluated 1.80€if obtained by construction of additional
LUBs whereas the same result would be evaluated 2.48€if achieved by increasing
PBPs of finding a LUB free. The apparently contradictory result could be inter-
preted, on one side, as a lack of trust the interviewees have in the announced extra
LUBs construction policy which has for long been on the local administration
agenda and never materialized and, on the other, as an explicit preference for a
short-term, no-financial-outlay policy that can be simply pursued by an increased
surveillance and repression of illegal parking. The policy implications derivable
from this interpretation are clear and suggest the adoption of light intervention
policy based more on regulation rather than LUB construction with a limited
impact on the public purse.
Table 6 WTP estimates with delta-method (based on M2)
Variable Coefficient St.Err. T-Stat
LUB3/EF1 -0.1938 0.0377 -5.13
LUB3/EF2 -0.2302 0.0511 -4.49
LUB3/EF4 0.2834 0.0621 4.56
LUB3/EF5 0.1358 0.0279 4.86
PLUBF2/EF1 -0.2213 0.0597 -3.7
PLUBF2/EF2 -0.2629 0.0679 -3.86
PLUBF2/EF4 0.3236 0.0911 3.55
PLUBF2/EF5 0.1550 0.0400 3.87
PLUBF3/EF1 -0.2358 0.0569 -4.13
PLUBF3/EF2 -0.2802 0.0716 -3.91
PLUBF3/EF4 0.3448 0.0828 4.16
PLUBF3/EF5 0.1652 0.0408 4.04
Table 7 WTP comparison between M1 and M2
Variable M1 M2
WTP (discrete variation)
LUB +800
145
(95 -195 )
81
(45 -117 )
PLUBF +10
99
(73 -125 )
142
(95 -189 )
+20
198
(147 -251 )
145
(102 -188 )
238 E. Marcucci and V. Gatta
Similar considerations apply to M2 (see Table 6) where we also calculate
different WTP measures since we test and discover non-linear effects for the EF. It
is important to clarify that since we have only ameliorative variations, with respect
to the SQ level, for both LUB and PLUBF in the case of reductions of EF levels, in
order to interpret the meaning of the coefficients one has to imagine that the values
derived represent (in order to have a trade off of some sort) the amount of money
Fig. 3 WTP distributions.
A comparison between M1
and M2
Behavioral Modeling of Urban Freight Transport 239
the agent would be willing to receive for not having potentially gained from the
increase in the level of the beneficial attribute under consideration.
Notwithstanding the interesting analysis just discussed one has to scrutinize the
policy implications derived by using either M1 or M2. An informative comparison
between the WTP estimates (and their respective confidence intervals) of the two
models is reported in Table 7.
We underline that all the results reported use Beta
EF3–4
as a base since this
represents the part-worth utility variation from the SQ (600€) to the next step up
(800€). Moreover, for M2, having effects coded the variables, one has to be careful
in interpreting results especially when it comes to WTP measures. In fact, one
should recall that the WTP to move from the basic level of an attribute to a
different one represents the difference in the corresponding valuations (Collins
et al. 2012). In our case, for LUB3 we have 113€representing the amount of
money interviewees are willing to pay to obtain 800 additional LUB.
The results reported in Table 7show the strong policy impacts that adopting
either a linear or non-linear assumption might have. In fact, comparing the results
of M1 and M2, one observes substantial differences for both LUB and PLUBF. In
particular, Fig. 3shows that M1 tends to overestimate WTPs associated with the
greatest attribute levels and, in these cases, the distributions are much flatter than
those related to M2 (i.e. LUB3 and PLUBF3). On the contrary, M1 underestimates
WTP linked to intermediate level and the distribution is characterized by a little
dispersion around the mean value (i.e. PLUBF2).
From a purely statistical point one should suggest policy maker to have more
faith in M2 results giving its higher explanatory power given its capability to fit the
data.
5 Concluding Remarks
This paper reports the results from an empirical research on UGM policy inter-
vention in the Roman freight LTZ. The research specifically focuses on retailers’
preference analysis for hypothetical policy scenarios. The paper innovates in terms
of questionnaire development and in terms of ex-ante policy-mix evaluation. The
results obtained are relevant both from a theoretical point of view as well as from a
more practical and policy-oriented perspective. It is noticeable that notwith-
standing the often called for agent-level analysis, the literature on UGM policies
has rarely investigated this issue at this specific level. Therefore, the paper rep-
resents a first attempt at bridging the gap between theory, applied research and data
needs.
In more detail, from a methodological stance the results reported show that not
only it is important and interesting to adopt an agent-based point of view but also
to consider potentially non-linear effects of the policy instruments adopted. Data
reveals, in fact, that both with respect to all attributes considered the policy
potentially implemented might have a different effect depending on the attribute
240 E. Marcucci and V. Gatta
level the policy is trying to influence. The results have been analyzed in terms of
WTP so to facilitate interpretation and, under this respect, the robust estimation
conducted on the coefficients’ ratios allowed us to produce monetary confidence
intervals for each of the policy attribute considered. The comparison between M1
(linear effects) and M2 (non-linear effects) shows that potentially relevant biases
could characterize the results obtained if non-linearities in the effects are duly
accounted for. The limited amount of observations available do not suggest
extrapolating the results to a real-life context, however we trust the reader will
appreciate the methodology exposed as useful in providing local policy-makers
with relevant information. Future research will pursue two different but concurrent
objectives. On one side we will perform similar investigations on two other rel-
evant UGM agent-types, namely transport providers and own account, while, on
the other, from a methodological perspective, we will also investigate other
potentially relevant issues such as for instance: (1) various forms of heterogeneity
in preferences (e.g. investigating deterministic, stochastic, as well as both deter-
ministic and stochastic, see Marcucci and Gatta 2012); (2) develop interactive
choice models along the methodological lines proposed by Hensher and colleagues
at ITSL Sydney (Hensher and Puckett 2007; Puckett et al. 2007); (3) adopt
Bayesian estimation methods since they are particularly useful when researchers
are faced with a limited number of observations.
Acknowledgments Acknowledgments: we would like to thank Volvo Research Foundation for
funding the project SP-2007-50 ‘‘Innovative solutions to freight distribution in the complex large
urban area of Rome’’ and all those people who administrated the interviews, in particular Matteo
Russo, Silvia De Silvestris, Vivian Diaferia, Amedeo Nanni, Giacomo Lozzi, Matteo Genovese,
Francesco Di Antonio, Emanuele Barzagli, Amedeo Naponiello.
References
Ambrosini C, Routhier JL (2004) Objectives, methods and results of surveys carried out in the
field of urban freight transport: an international comparison. Trans Rev 24(1):57–77
Ambrosini C, Meimbresse B, Routhier JL, Sonntag H (2008) Urban freight policy-oriented
modelling in Europe. In: Taniguchi E, Thompson RG (eds) Innovations in city logistics, Nova
Science Publishing, New York, pp 197–212
Chow JYJ, Yang CHY, Regan AC (2010) State-of-the art of freight forecast modeling: lessons
learned and the road ahead. Transportation 37(6):1011–1030
Collins AT, Rose JM, Hess S (2012) Interactive stated choice surveys: a study of air travel
behavior. Transportation 39(1):55–79
Comi A, Delle Site P, Filippi F, Nuzzolo A (2012) Urban freight transport demand modelling: a
state of the art. In: European Transport/Trasporti Europei 2012 (51), ISTIEE, Trieste
Corò G, Marcucci E (2001) Le politiche per la logistica. In: Quaderni della Mobilità, Proceeding
della Conferenza Politiche economiche per il sistema dei trasporti e della logistica, Regione
Marche, Ancona, vol 1, pp 41–50
Daly S, Hess S, de Jong G (2010) Calculating errors for measures derived from choice modelling
estimates. ITS working paper, Institute for Transport Studies, University of Leeds
Behavioral Modeling of Urban Freight Transport 241
Danielis R, Marcucci E (2007) L’accettabilità del centro di distribuzione urbana delle merci.
Un’analisi sulle preferenze dei negozianti nella città di Fano. Working paper SIET
De Jong G, Ben-Akiva M (2007) A micro-simulation model of shipment size and transport chain
choice. Transp Res Part B 41:950–965
Efron B (1979) Bootstrap methods: another look at the Jackknife. Ann Stat 7:1–26
Efron B, Tilbshirani R (1993) An introduction to the bootstrap. Chapman and Hall, New York
Filippi F, Campagna A (2008) Indagine sulla distribuzione delle merci a Roma, nell’ambito dello
Studio di settore della mobilità delle merci a Roma, CTL: Centre for studies on transport and
logistics and ATAC: agency of mobility of Rome’s municipality
Filippi F, Nuzzolo A, Comi A, Delle Site P (2010) Ex-ante assessment of urban freight transport
policies. In: Procedia—Social and Behavioral Sciences 2(3). Taniguchi E, Thompson RG
(eds) Elsevier Ltd, pp 6332–6342 doi: 10.1016/j.sbspro.2010.04.042
Gatta V (2006) Valutare la Qualità dei Servizi. Un nuovo approccio basato sulla conjoint
analysis. Statistica LXVI(1):85–113
Gatta V, Marcucci E (2013) Urban freight distribution policies: joint accounting of non-linear
attribute effects and discrete mixture heterogeneity, Res Transp Econ, forthcoming
Gonzalez-Feliu J, Toilier F, Ambrosini C, Routhier JL (2013) Estimated data production for
urban goods transport diagnosis. The Freturb methodology. In: Gonzalez-Feliu J et al (eds)
Sustainable urban logistics: concepts, methods and information systems, Springer, Berlin,
Chapter 7 of the present book
Gray R (1982) Behavioural approaches to freight transport modal choice. Transp Rev
2(2):161–184
Greene WH (2003) Econometric analysis, 5th edn. Englewood Cliffs, Prentice Hall
Hensher DA, Figliozzi MA (2007) Behavioural insights into the modelling in freight
transportation and distribution system. Transp Res Part B 41:921–923
Hensher DA, Puckett SM (2007) Theoretical and conceptual frameworks for studying agent
interaction and choice revelation in transportation studies. Int J Transp Econ 34(1):17–47
Hensher DA, Rose JM, Greene WH (2005) Applied choice analysis: a primer. Cambridge
University Press, Cambridge
Hole AR (2007) A comparison of approaches to estimating confidence intervals for willingness to
pay measures. Health Econ 16(8):827–840
Kahneman D, Tversky A (1979) Prospect theory: an analysis of decision under risk.
Econometrica 47:263–291
Krinsky I, Robb AL (1986) On approximating the statistical properties of elasticities. Rev Econo
Stat 68(4):715–719
Krinsky I, Robb AL (1990) On approximating the statistical properties of elasticities: a
correction. Rev Econ Stat 72(1):189–190
Liedtke G, Schepperle H (2004) Segmentation of the transportation market with regard to
activity–based freight transport modelling. Int. J Logistics: Res Appl 7(3):199–218
Marcucci E (2005) I modelli a scelta discreta per l’analisi dei trasporti. Teoria, metodi ed
applicazioni’’, (a cura di). Carocci, Roma
Marcucci E, D’Agostino Z (2003) Centro e periferia nell’attuale evoluzione del sistema italiano
della logistica e dei trasporti: alcune evidenze empiriche dal Nordest. In: Atti della VI
riunione scientifica della SIET, Trasporti e politiche economiche, Palermo, 13 e 14 novembre,
Numero speciale, Annali della Facoltà di Economia e Commercio di Palermo, pp 253–272
Marcucci E, Danielis R (2008) The potential demand for a urban freight consolidation centre.
Transportation 35(2):269–289
Marcucci E, Danielis R, Paglione G, Gatta V (2007) Centri urbani di distribuzione delle merci e
politiche del traffico: una valutazione empirica tramite le preferenze dichiarate. In: Atti del 7°
congresso CIRIAF, pp 373–378
Marcucci E, Gatta V (2011) Regional airport choice: consumer behaviour and policy
implications. J Transp Geogr 19(1):70–84
Marcucci E, Gatta V (2012) Dissecting preference heterogeneity in consumer stated choices.
Transp Res Part E 48(1):331–339
242 E. Marcucci and V. Gatta
Marcucci E, Gatta V (2013) Urban freight distribution and intra-agent heterogeneity: the case of
own-account. Int J Transport Econ 50(2):267–286
Marcucci E, Stathopoulos A, Gatta V, Valeri E (2012) A stated ranking experiment to study
policy acceptance: the case of freight operators in Rome’s LTZ. Italian J Reg Sci 11(3):11–30
Mc Fadden DL (1974) Conditional logit analysis of qualitative choice behavior. In: Zarembka P
(ed) Frontiers in Econometrics, Academic Press, New York
Mooney CZ, Duvald RD (1993) Bootstrapping: a nonparametric approach to statistical inference,
Sage University Paper Series on quantitative applications in the social sciences, Series no. 07-
095, Sage, Newbury Park
Ogden KW (1992) Urban goods movement: a guide to policy and planning. Ashgate, Aldershot
Paglione G, Gatta V (2007) Le preferenze degli attori economici verso l’installazione di un
Centro di Distribuzione Urbana (CDU): il caso di Pesaro, IX Riunione Scientifica SIET,
http://www.sietitalia.org/siet9/papers/Paglione-Gatta_SIET%202007.pdf
Puckett SM, Greaves S (2009) Estimating policy-induced changes in freight vehicle emissions in
Sydney. Institute of transport and logistics studies working paper, The University of Sydney
Puckett SM, Hensher DA, Collins A, Rose J (2007) Design and development of a stated choice
experiment in a two-agent setting: interactions between buyers and sellers of urban freight
distribution services. Transportation 34(4):429–451
Quack HJ, De Koster R (2009) Delivering goods in urban areas: how to deal with urban policy
restrictions and the environment. Transp Sci 43:211–227
Roorda MJ, Cavalcante R, Mccabe S, Kwan H (2010) A conceptual framework for agent-based
modelling of logistics services. Transp Res Part E: Logistics Transp Rev 46:18–31
Rotaris L, Danielis R, Marcucci E, Sarman I (2012) Testing for nonlinearity in the choice of a
freight transport service. In: Marcucci E, Puckett S (Guest Editors), Special issue on: Freight
transport analysis: new trends and methodologies, European Transport/Trasporti Europei, 50,
pp 1–22
Russo F, Comi A (2011) Measures for sustainable freight transportation at urban scale: expected
goals and tested results in Europe. J Urban Planning Develop 137(2). doi: 10.1061/
(ASCE)UP.1943-5444.0000052, American Society of Civil Engineers (ASCE), 142–152
Sammi A, Mohammadian A, Kawamura K (2009) Integrating supply chain management concept
in a goods movement microsimulation. In: IEEE international conference on service
operations. Logistics and Informatics, Chicago
Shanmugalingham S (1982) On the analysis of the ratio of two correlated normal variables.
Statistician 31:251–258
Sonntag H (1985) A computer model of urban commercial traffic. Transp, Policy Decis Making
3(2)
STA (2001) Quadro conoscitivo del problema della distribuzione delle merci nel Centro Storico
di Roma. STA, Rome
Stathopoulos A, Valeri E, Marcucci E (2012) Stakeholder reactions to urban freight policy
innovation. J Transp Geogr 22:34–45
Stathopoulos A, Valeri E, Marcucci E, Gatta V, Nuzzolo A, Comi A (2011) Urban freight policy
innovation for Rome’s LTZ: a stakeholder prospective. In: Macharis C, Melo S (eds) City
distribution and urban freight transport: multiple perspectives. Edward Elgar Publishing
Limited, Cheltenham, pp 75–100
VOLVO Report (2010) Innovative solutions to freight distribution in the complex large urban
area of Rome, Final report, project SP-2007-50, Mimeo
VOLVO Research Foundation (2009) Innovative solutions to freight distribution in the complex
large urban area of Rome, Volvo Research and Educational Foundations and The Italian
Centre of Excellence CTL—Centre for Transport and Logistics 2008–2009
Wisetjindawat W, Sano K, Matsumoto S (2006) Commodity distribution model incorporating
spatial interactions for urban freight movement. Transp Res Rec: J Transp Res Board
1996:41–50
Behavioral Modeling of Urban Freight Transport 243
