Scholarly article on topic 'Simulating Urban Freight Flows with Combined Shopping and Restocking Demand Models'

Simulating Urban Freight Flows with Combined Shopping and Restocking Demand Models Academic research paper on "Civil engineering"

CC BY-NC-ND
0
0
Share paper
OECD Field of science
Keywords
{"Urban freight transport" / "city logistics" / "demand model" / "shopping trip" / "shop type choice"}

Abstract of research paper on Civil engineering, author of scientific article — Antonio Comi, Agostino Nuzzolo

Abstract This paper presents the development of a modelling system to simulate urban freight flows and preliminary results of a survey carried out to investigate end-consumer behaviour. The set of models involves the simulation of end-consumer choices in relation to type of retail outlet (e.g. small, medium or large) since such choices undoubtedly impact on freight distribution flows. Indeed, the characteristics of the restocking process are strictly related to the type of retail activities to be restocked in terms of delivery size, delivery frequency, freight vehicle type and so on. The paper details each model for shopping mobility origin-destination matrix estimation and investigates the main variables affecting behaviour in relation to trip generation and type of retail outlet.

Academic research paper on topic "Simulating Urban Freight Flows with Combined Shopping and Restocking Demand Models"

Available online at www.sciencedirect.com

ScienceDirect

Procedía - Social and Behavioral Sciences 125 (2014) 49 - 61

8th International Conference on City Logistics

Simulating Urban Freight Flows with Combined Shopping and

Restocking Demand Models

Antonio Comia*, Agostino Nuzzoloa

aDepartment of Enterprise Engineering, University of Rome Tor Vergata, via del Politécnico 1, Rome 00133, Italy

Abstract

This paper presents the development of a modelling system to simulate urban freight flows and preliminary results of a survey carried out to investigate end-consumer behaviour. The set of models involves the simulation of end-consumer choices in relation to type of retail outlet (e.g. small, medium or large) since such choices undoubtedly impact on freight distribution flows. Indeed, the characteristics of the restocking process are strictly related to the type of retail activities to be restocked in terms of delivery size, delivery frequency, freight vehicle type and so on. The paper details each model for shopping mobility origin-destination matrix estimation and investigates the main variables affecting behaviour in relation to trip generation and type of retail outlet.

© 2014 The Authors. Published by Elsevier Ltd.

Selection and peer-review under responsibility of the Organising Committee of the 8th International Conference on CityLogistics.

Keywords: Urban freight transport; city logistics; demand model; shopping trip; shop type choice

1. Introduction

Urban freight flows are mainly comprised of two components related to shopping and restocking. Indeed, surveys carried out in some European cities (Schoemaker, Allen, Huschebek & Monigl, 2006; Gonzalez-Feliu, Ambrosini, Pluvinet, Toilier & Routhier, 2012) reveal that, considering only urban freight mobility, about 69% of urban distances (veh-km) covered each day by motorized vehicles consist of shopping trips, 24% of restocking

* Corresponding author. Tel.: +39 06 7259 7059; fax: +39 06 7259 7053. E-mail address: comi@ing.uniroma2.it

1877-0428 © 2014 The Authors. Published by Elsevier Ltd.

Selection and peer-review under responsibility of the Organising Committee of the 8th International Conference on City Logistics. doi: 10.1016/j.sbspro.2014.01.1455

trips and the remaining 7% result from urban management (e.g. building sites, waste collection, network maintenance).

End-consumer choices in relation to type of retail outlet (e.g. small, medium or large) undoubtedly impact on freight distribution flows: the characteristics of the restocking process are strictly related to the type of retail activities to be restocked in terms of delivery size, delivery frequency, freight vehicle type and so on. For example, delivery size and freight vehicle dimensions tend to increase with the dimension of retail activities, while delivery frequency tends to decrease, with considerable effect on the total distance travelled by freight vehicles. Therefore, end-consumer choices among small, medium and large retail businesses affect restocking characteristics and the total freight vehicle distance travelled.

Furthermore, end-consumer shopping location choices depend on the location of commercial supply with respect to residence and on end-consumer behaviour, which in turn depends on some characteristics such as age, income, family dimension and lifestyle. Further, end-consumer choices of retail type can also depend on the accessibility of shopping areas; thus if accessibility changes (for example, as a consequence of shopping demand travel management), type of shop and/or transport mode can also change. Then, if the characteristics of end consumers, residential and commercial land-use distribution, and/or accessibility to the commercial area change, the freight restocking characteristics may also too. Similarly, some city logistics measures can reduce the restocking accessibility of an area and induce re-allocation of retail businesses.

In this context, a city logistics scenario (i.e. set of measures), implemented to improve urban sustainability and reduce the impacts of these two freight transport components (i.e. shopping and restocking), can affect one of the two components with impacts on the other, too. Therefore, a study of urban freight transport and the relative methodology to assess a city logistics scenario should consider both components jointly.

Although several models have been proposed in the field of shopping mobility and restocking mobility (de Jong, Vierth, Tavasszy & Ben-Akiva, 2012; Comi, Delle Site, Filippi & Nuzzolo, 2012), traditionally these two demand segments have been independently handled. Shopping mobility has been studied as a component of passenger demand through the relationships between travel behaviour, the built environment (e.g. land use allocated for different business activities, density) and socio-economic characteristics (Nuzzolo & Coppola, 2005; Ewing & Cervero, 2010). Few studies have analysed it as a component of freight mobility and considered that actions impacting on purchasing behaviour of end consumers (e.g. location of retail outlet, transport mode to use for shopping) can also affect restocking mobility (Bronzini, 2008; Miodonski, & Kawamura, 2012; Wygonik, Bassok, Goodchild, McCormack & Carlson, 2012).

With regards to restocking, various freight demand models have been proposed, many of which are multi-stage models (Taniguchi, Thompson, Yamada & van Duin, 2001; Comi, Delle Site, Filippi & Nuzzolo, 2012; Anand, Quak, van Duin & Tavasszy, 2012) that can be classified in relation to the reference units used: quantity, delivery, tour and vehicle. Quantity allows us to point out the mechanism underlying the generation of freight transport demand: freight transport is generated by the requirement of end consumers to satisfy their needs for goods and services (Gonzalez-Feliu, Toilier & Routheir, 2010; Russo & Comi, 2012). Quantity-based models are more specific for assessing strategic action on transportation flows, such as those impacting on the location of warehouses and retail activity. Delivery is the unit used by transport and logistics operators, allowing us to investigate in greater depth the logistic process of restocking (Muñuzuri, Cortés, Onieva & Guadix, 2012). Using delivery-based models, assessment may be made of the impacts on the transport service type used for restocking (e.g. on own account or by third party), and on shipment size. Tours can be used to investigate delivery in relation to departure time, vehicle type, number and sequence of stops. Finally, vehicle flows, interacting within the assignment model, allow us to obtain link flows and to estimate and evaluate the transport performance and impacts of a given city logistics scenario.

Although restocking flows are generated to satisfy end-consumer demand and restocking models consequently have to take account of end-consumer choices, few have proposed joint modelling frameworks (Oppenheim, 1993; Russo & Comi, 2010; Gonzalez-Feliu, Ambrosini, Pluvinet, Toilier & Routhier, 2012), showing that further work needs to be done in this field.

Given the desirability of a joint modelling framework, this paper presents a modelling system which takes into account some factors of end-consumer behaviour, such as the choice of retail outlet type, and links shopping and

restocking mobility. It consists of four model sub-systems to estimate the shopping mobility O-D matrices, the restocking quantity O-D matrices, the delivery O-D matrices and the restocking vehicle O-D matrices. This modelling system was developed in the authors' previous works (Nuzzolo & Comi, 2013a).

The paper delves into the shopping model sub-system and presents the results of an analysis of factors that influence freight mobility, namely trip generation and choices of retail outlet type. The study is based on surveys conducted in Rome, where more than 300 households were interviewed.

2. Modelling framework

The joint modelling framework considers both the outlet type chosen by the end-consumer and restocked by

carriers, comprising four model sub-systems (Fig. 1):

• shopping mobility sub-system; this simulates the end-consumer's behaviour vis-à-vis shopping, and estimates the modal O-D matrices and freight flows attracted by each traffic zone per outlet type; this step allows us to point out the effects due to implementations of long-term actions (e.g. urban land-use governance) on the location of retail establishments and place of residence; at the tactical or operational level they could assess effects on the choices of type and location of shops for purchasing goods and the transport mode used;

• quantity model sub-system; this allows us to estimate the quantity origin-destination (O-D) matrices by freight type and restocked outlet type; this step highlights the effects due to implementation of strategic actions (e.g. urban land-use governance) on the locations of logistic facilities (e.g. warehouses and distribution centres) and retail businesses (e.g. local shops or shopping centres);

• delivery model sub-system; this can convert quantities into delivery O-D flows per outlet type; the delivery flows are also split in terms of transport services used (e.g. retailer on own account, wholesaler on own account and carrier); this step serves to define restocking journeys in terms of transport services and shipment size (i.e. tactical level);

• vehicle model sub-system; this allows us to obtain the restocking vehicle O-D flows, satisfying the given delivery O-D matrices, and investigate the tours undertaken to restock the study area; the tours are characterized by departure time, number of stops, vehicle used and sequence of delivery locations.

SHOPPING RESTOCKING - Quantity

Fig. 1. Urban freight modelling framework (Nuzzolo and Comi, 2013b)

2.1. Shopping flow modelling

Shopping may be considered a major trip purpose as it forms part of the lifestyle of the population. Shimazaki, Kazunori & Shihana (1994) mentioned that shopping is the second most frequent type of urban travel. Nevertheless, most of the current transport literature focuses on studying the characteristics of worker trips, with little emphasis being placed on studying non-worker travel patterns, such as shopping trips (Mokhtarian, 2004; Mokhtarian, Ory & Cao, 2007; Cao, Douma, Cleaveland & Xu, 2010).

In general, the focus of transportation research is mainly on trip generation, distribution and mode steps within the well-known four-step models (Cascetta, 2009). Although researchers have increasingly argued the high incidence of multi-stop trips in empirically observed behaviour (e.g. Ingene & Ghosh, 1990; Thill, 1992; Dellaert, Arentze, Bierlaire, Borgers & Timmermans, 1998; Popkowski Leszczyc, Sinha & Sahgal, 2004), the commonly used modelling structure refers to round trips. It can easily be extended to trip chains.

Following what was proposed by Russo & Comi (2010) and assuming that the end consumer is in zone o, the choice dimensions involved are: the number of trips (x) for shopping, the type of shop (k; e.g. small, medium, large) and destination (d), the transport mode (or sequence of modes; m). The global demand function can be decomposed into the product of sub-models, each of which relates to one or more choice dimensions. The sequence used is the following:

D'od [skm] =D'o [s] • p' [dk / so] ■ p' [m / dkso] (1)

• D'od [skm] is the weekly average number of trips with origin in zone o undertaken by the end consumer belonging to category i for purchasing goods of type s in retail outlet type k located in zone d by using transport mode m;

• D'o[s] is the weekly average number of relevant trips undertaken by end consumers belonging to category i for purchasing goods of type s with origin in zone o, obtained by a trip generation model;

• p'[dk/so] is the probability that users, undertaking a trip from o, travel to destination zone d for purchasing at reail outlet (shop) type k (small shop, supermarket, hyper-market), obtained by a shop type and location model;

• p'[m/dkso] is the probability that users, travelling between o and d for purchasing in shop type k, use transport mode m, obtained by a modal choice or split model.

Finally, the quantities required by each zone to satisfy end-consumer needs can be obtained by introducing a quantity purchase model. This model gives us the probability that the end consumer, arriving in zone d, purchases something of a certain dimension (dim). Therefore, the quantity of freight type s sold by retail outlets k in zone d, QI.d[sk], can be calculated as:

Qld .]= XQIi '] = Z Z Di [skm\p'[dim / mks\dim (2)

i ' o, rn

• QI'.d [sk] is the quantity of goods bought/sold in retail outlet k in zone d given by the demand of end consumers belonging to category i living/working in a zone within the study area;

• dim is the dimension of purchases, expressed in kg;

• p'[dim/mks] is the probability that a trip concludes with a purchase of dimension dim conditional upon undertaking a trip to retail outlet k for a purchase of goods type s using transport mode m.

It should be noted that the freight quantity required in a zone d also depends on the end consumers who purchase there and live/work in a zone external to the study area (QE.d[sk]). This rate can be determined through direct estimates based on surveys carried out at the border of the study area.

2.2. Restocking flow modelling

Referring to the general modelling framework proposed by Nuzzolo & Comi (2013a), the following sections describe the model sub-system that can be used to estimate quantity (delivery and freight vehicle) O-D flows and tours for restocking. Although it can refer to different freight types, for simplicity of notation the class index s will be understood unless otherwise stated.

2.2.1. Quantity model sub-system

Let Qod[k] be the average quantity of freight flows moved from zone o (e.g. warehouse location zone) to the outlets of type k of zone d, obtained by the following equation:

Qd [k]=Qd [k]-p[o/dk] (3)

• Qd[k] is the average freight quantity attracted by (i.e. to be delivered in) zone d and retail outlet k , obtained by an attraction model;

• p[o/dk] is the probability that freight attracted by zone d and retail outlet k comes from zone o; it represents the acquisition share obtained by an acquisition model.

In the literature (de Jong, Vierth, Tavasszy & Ben-Akiva, 2012; Comi, Delle, Filippi & Nuzzolo, 2012), we find different models developed to estimate the quantity attracted by each traffic zone (Q.d) but they neglect the dependence on outlet type k.

Furthermore, in the proposed joint modelling framework, we have to define a congruence between sold and bought quantities that is synthesized by the following relation:

Qd [k]=QTd[k] = QId [k] + QEd [k].

Once Q.d[k] has been estimated, the delivery and the vehicle O-D matrices can be obtained by using the other two modelling sub-systems (Fig. 1) as described in Nuzzolo & Comi (2013b; 2013c), to which readers can refer for more detail.

3. A shopping model sub-system

Some shopping models are developed using the results of surveys carried out in Rome, where more than 300 households were interviewed, considering both home-based and non-home-based shopping trips. The survey allowed us to investigate purchasers' behaviour, providing details on the trip generation and the choices of retail outlet type.

3.1. The dataset

Only interviewees older than 14 years were asked for details on purchase trips (concluded with a purchase of more than € 30.00). The survey questions covered shopping behaviour during the previous week and the interviews were structured into three sections:

• data identifying the household, structured into further sub-sections and designed to collect data on: number of household members, location, vehicle availability (e.g. car and motorcycle), socio-economic data of each member (e.g. sex, age, driver licence availability, type of job);

• data on purchase trip; this section was specified for each household member and enabled data to be collected on shopping trips and purchases (e.g. freight types, trip origin and destination, day, type of retail outlet where the purchases was made, value of purchased goods);

• e-commerce; this allowed data to be collected on e-shopping.

3.2. The shopping trip general characteristics

Of the sampled households, 74% have more than three members, with an average of 3.37, and on average 3.12 members are older than 14 years. The interviewees who made at least a notable purchase during the survey week were 49% female and 51% male, while only 7% were under 19. The last census data revealed that, in Rome, households comprised an average of 2.42 members, 47% of residents were male, and 17% under 19 (ISTAT, 2001).

Only 3% of the interviewees made a trip-chain (i.e. visit to more than one zone during the same shopping trip). 22% of interviewees stated they had made further shopping trips during the week (average of 1.34 weekly trips) without buying anything (also if she/he left to spend more than € 30). 68% of shopping trips start from home (Table 1), with higher percentages for foodstuffs.

Table 1. Distribution of weekly shopping trips by activity at origin

Foodstuffs Hygiene and household products Cloth Other Average

Home 75% 72% 64% 55% 68%

Work 13% 12% 13% 20% 14%

Personal service and leisure 10% 12% 14% 12% 12%

Study 2% 4% 9% 13% 6%

Total 100% 100% 100% 100% 100%

Average 36% 19% 27% 18%

As expected, the daily distribution of shopping trips revealed that more than 25% of weekly trips are undertaken on Saturday to large retail outlets (Fig. 2) and 50% of trips start from the origin between 3pm and 7pm (Fig. 3).

50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0%

llllilinlll.1

^ ^ J& & ^

■ All retail outlet types □ Small retail outlet ■ Medium retail outlet ■ Large retail outlet

Fig. 2. Distribution of weekly shopping trips by day of the week and retail outlet type

Foodstuffs ■ Hygiene and household products ■ Other

Fig. 3. Distribution of weekly shopping trips by time of day

Of the sample members, 13% stated they did e-shopping during the survey week and 86% of e-shoppers did so for reasons of time.

3.3. The shopping trip generation model

According to the classification used by the Italian Institute of Statistics (ISTAT, 2001), we identified four main classes of freight: foodstuffs, hygiene and household products, clothing and shoes, and other. Fig. 4 reports the distribution of purchases according to this freight classification. Most trips (77%) were undertaken to buy only one type of freight, 17% for two freight types and 6% for more than three.

Fig. 4. Distribution of weekly shopping trips according to freight type

From Table 2, we can see that the number of weekly trips undertaken for foodstuffs is quite constant for households with more than two members, unlike the other classes which vary with respect to the number of household members.

Table 2. Average number of weekly trips by household size

Number of components Foodstuffs Hygiene and household products Cloth Other Average

1 0.73 0.55 0.73 0.64 0.88

2 1.13 0.67 0.74 0.33 0.96

3 1.27 0.60 1.04 0.49 1.13

more than 3 1.18 0.71 1.18 0.76 1.27

Average 1.19 0.66 1.06 0.61 1.17

In relation to age, it may be noted that young people (under 29 years) undertake more trips to purchase "other" goods, while elderly people travel more for foodstuffs (Table 3). The highest number of trips per week is made by housewives who undertake more than one weekly trip for foodstuffs (Table 4).

Table 3. Average number of weekly trips by age

Age Foodstuffs Hygiene and household products Cloth Other Average

less old than 29 years 0.18 0.17 0.63 0.37 0.45

30 - 64 years old 0.80 0.43 0.51 0.28 0.67

older than 65 years 0.97 0.28 0.25 0.31 0.60

Average 0.60 0.33 0.54 0.31 0.89

Table 4. Average number of weekly trips by member's occupation

Job Foodstuffs Hygiene and household products Cloth Other Average

Occupied 0.67 0.37 0.50 0.33 0.62

Housewife 1.19 0.52 0.63 0.08 0.81

other 0.33 0.23 0.56 0.36 0.49

Average 0.60 0.33 0.54 0.31 0.89

Although, the number of shopping trips is also related to the characteristics of the zone of residence, as a first approximation we can express them only in relation to the main socio-economic characteristics of the household (e.g. number of members) that can be easily obtained from the Italian Statistics database (ISTAT). Therefore, the weekly average number of relevant trips undertaken by end consumers (household or member) for purchasing goods of type s with origin in zone o, Do[s], can be expressed as:

Do [s] = ns [o] • ms [o] = ns [o] • • Xj (4)

• ns[o] are the number of inhabitants older than 14 years resident in the traffic zone o;

• ms[o] is the average number of weekly trips undertaken by inhabitants older than 14 years; it is expressed as a linear function in the coefficients Pj of attributes Xj^

The considered attributes Xio are the mean values of socio-economic variables such as the number of household members. The estimated parameters are reported in Table 5. We can see that the parameters are statistically significant as demonstrated by t-st values. As confirmed by survey statistics, parameter analysis shows that housewives undertake more trips than those in other types of employment, while young people travel mainly for other types of goods.

Table 5. Average number of weekly shopping trips made by residents (older than 14 years): parameter estimation

Freight type Number of household components Young people Age between 30 - 64 years Occupied Housewife Female R2

Type of variable 0/1 0/1 0/1 0/1 0/1

Foodstuffs 0.46 (6.64) 0.14 (1.94) 0.45 (4.09) 0.32 (6.80) 0.53

Hygiene and household products 0.10 (2.33) 0.37 (10.23) 0.12 (2.82) 0.31

Cloth 0.53 (10.23) 0.42 (9.83) 0.19 (3.65) 0.43

Other 0.07 (2.42) 0.11 (10.45) 0.20 (2.50) 0.27

All freight 0.11 (2.59) 0.77 (4.62) 1.21 (6.73) 0.23 (2.00) 0.65 (3.79) 0.30 (3.45) 0.77

(-) t-st value

3.4. The shop type choice model

As stated above, urban freight transport is undoubtedly impacted by the end consumer's choices. The characteristics of the restocking process are strictly related to the type of businesses to be restocked in terms of delivery size, delivery frequency, freight vehicle type and so on. All these characteristics depend on the type of retail outlet: small shops require small frequent deliveries, while large retail outlets require large non-frequent deliveries. Therefore, the characteristics of end-consumer trips are analysed below in relation to retail outlet type. From survey data, it emerged that the choice of retail outlet to make purchases depends on freight types. As explained by Table 6, 68% of trips for foodstuffs are directed towards medium-size retail outlets (e.g. supermarkets). Even if the average shares among the three types of retail outlets are similar (Table 7), our data revealed that young people (under 29 years old) prefer larger retail outlets more than older consumers (over 65 years). Small retail outlets are equally chosen by different types of employment status (Table 8), while medium ones are preferred by housewives and large by other classes (i.e. employees and other).

Antonio Comi and Agostino Nuzzolo / Procedia - Social and Behavioral Sciences 125 (2014) 49 - 61 Table 6. Distribution of weekly shopping trips by type of retail outlet

Foodstuffs Hygiene and household product Cloth Other Average

Small 20% 26% 51% 49% 35%

Medium 68% 46% 1% 1% 34%

Large 12% 28% 48% 50% 31%

Total 100% 100% 100% 100% 100%

Table 7. Distribution of weekly shopping trips by type of retail outlet and age

Age Small retail outlet Medium retail outlet Large retail outlet Total

under 29 years 43% 14% 43% 100%

30 - 64 years old 32% 40% 28% 100%

over 65 years 36% 49% 15% 100%

Average 35% 34% 31% 100%

Table 8. Distribution of weekly shopping trips by type of retail outlet and occupation

Job Small retail outlet Medium retail outlet Large retail outlet Total

employee 33% 35% 32% 100%

housewife 36% 47% 18% 100%

other (e.g. retired) 39% 25% 36% 100%

Total 35% 34% 31% 100%

In terms of purpose, 30% of interviewees choose the retail outlet type in relation to proximity, while 45% in relation to price and quality of services offered (Table 9 and 10). The small retail outlet is mainly preferred for its quality of service, the supermarket (i.e. medium retail outlet) for its proximity and the large outlet for price and quality of service. End consumers tend to go shopping alone or with one other person (Table 11).

Table 9. Distribution of weekly shopping trips by type of retail outlet and purpose

Small retail outlet Medium retail outlet Large retail outlet Total

low traffic and parking availability 0% 2% 5% 2%

price 19% 29% 30% 26%

service and quality 43% 14% 30% 29%

proximity 23% 46% 20% 30%

other 16% 10% 14% 13%

Total 100% 100% 100% 100%

Table 10. Distribution of weekly shopping trips by type of retail outlet and travel time

Small retail outlet Medium retail outlet Large retail outlet Total

less than 10 minutes 32% 57% 18% 36%

11 - 20 minutes 35% 29% 34% 33%

21 - 60 minutes 29% 12% 45% 28%

more than 60 minutes 4% 2% 3% 3%

Total 100% 100% 100% 100%

Table 11. Distribution of weekly shopping trips by type of retail outlet and group size

Small retail outlet Medium retail outlet Large retail outlet Total

alone 49% 61% 34% 48%

1 35% 31% 40% 35%

2 10% 7% 16% 11%

3 4% 1% 6% 3%

more than 4 2% 0% 4% 2%

Total 100% 100% 100% 100%

In the sphere of city logistics, in order to reduce the number of vehicles due to freight transport, measures related to user transportation mode have to be implemented. Such action should push end consumers towards mass transit, which means that the factors influencing the choice of mode need to be investigated. Analysis of survey data showed that 74% of users use private vehicles and only 7% use transit (e.g. bus, metro; Table 12). Younger users (i.e. between 15 and 29 years old) present the highest share of those using transit (Table 13). The average travel time spend travelling by car was about 24 minutes, by transit about 35 minutes and on foot about 14 minutes.

Table 12. Distribution of weekly shopping trips by type of retail outlet and transport mode

Private vehicle (e.g . car) Transit On foot Total

Small retail outlet 66% 11% 23% 100%

Medium retail outlet 68% 4% 28% 100%

Large retail outlet 91% 5% 4% 100%

Average 74% 7% 19% 100%

Table 13. Distribution of weekly shopping trips by age and transport mode

under 29 years 30 - 64 years over 64 years Average

Private vehicle (e.g. car) 67% 78% 71% 74%

Transit 14% 4% 2% 7%

On foot 19% 18% 27% 19%

Total 100% 100% 100% 100%

Based on this analysis, a multinomial logit model for the choice of retail outlet type was calibrated. The systemic utilities (Vk) to buy in a retail outlet k (small, medium or large retail outlet) were expressed as follows:

V „ =-0.059-TH +0.221- OL - 0.117- NC + 0.513-OW - 0.229- NG + 2.133

small —2.7 0.7 -1.5 2.3 -2.8 4.4

V =-0.646-OY - 0.291- OC + 4.560- HF + 4.560-HY - 0.535- NG - 2.457

medium _2A _L4 UJ 9.4 ^tJ

V = 0.295- HO + 0.251- HY + 0.223- VA + 0.571- WK

l ge 1.6 1.1 1.9 3.7

p2 = 0.27

• TH is the hour of the day when shopping trip is undertaken;

• OL is a dummy variable equal to 1 if the user is older than 64, 0 otherwise;

• NC is the number of household components;

• OW is a dummy variable equal to 1 if the user is an housewife, 0 otherwise;

• NG is the number of components of group for shopping;

• OY is a dummy variable equal to 1 if the user is younger than 29, 0 otherwise;

• OC is a dummy variable equal to 1 if the user is employee, 0 otherwise;

• HY is a dummy variable equal to 1 if the user buys hygiene and personal products, 0 otherwise;

• HF is a dummy variable equal to 1 if the user buys foodstuffs, 0 otherwise;

• HO is a dummy variable equal to 1 if the user buys other products, 0 otherwise;

• VA is a number of different types of products purchased;

• WK is a dummy variable equal to 1 if the trip is undertaken on week-end, 0 otherwise.

The signs of all model coefficients are intuitively correct and the t-st value confirms their statistical significance. As it emerged from previous analysis, the housewives prefer to buy in a small and close retail outlets. The probability to buy in large retail outlets increases in the week-end, if different types of goods are bought and if the trip is undertaken with more than one person. These models were implemented for the assessment of land-use scenarios in the city of Padua (Italy; Nuzzolo, Comi & Papa, 2013) where the variation of external costs due to a relocation of urban retail outlets and freight facilities were estimated. The preliminary results showed that the strategy to locate large retail outlets out of downtown and in the first urban ring could be winning.

4. Conclusions

This paper presented advances in the modelling framework for estimating urban freight transport flows, under the assumption that goods attracted by an urban area are strictly related to end-consumer choices. Some analyses based on a survey administered to end consumers in Rome were presented. The paper investigated the factors that mainly influence freight mobility in relation to trip generation and choices of shop type. Such choices are influenced by both freight type and socio-economic household characteristics (e.g. age, number of component, employment status). Further analyses are in progress to improve these first results, including zone and level-of-service attributes (e.g. accessibility), and to develop other models, such as mode choice models, that can be used to simulate this segment of urban mobility in the ex-ante assessment of city logistics scenarios.

References

Anand, A., Quak, H., van Duin, R. & Tavasszy, L. (2012). City logistics modeling efforts: trends and gaps - a review. Procedia - Social and Behavioral Sciences, 39, 101-115.

Bronzini, M. (2008). Relationships between land use and freight and commercial truck traffic in metropolitan areas. Special Report 298, Transportation Research Board.

Cao, X., Douma, F., Cleaveland, F. & Xu, Z. (2010). The Interactions between E-Shopping and Store Shopping: A Case Study of the Twin

Cities. Research Report CTS 10-12, August, Center for Transportation Studies, University of Minnesota. Cascetta, E. (2009). Transportation Systems Analysis: Models and Applications. Springer.

Comi, A., Delle Site, P., Filippi, F. & Nuzzolo, A. (2012). Urban Freight Transport Demand Modelling: a State of the Art. European Transport 2012 (51), ISTIEE, Trieste, Italy.

De Jong, G., Vierth, I., Tavasszy, L. & Ben-Akiva, M. (2012). Recent developments in national and international freight transport models

within Europe. Transportation 40 (2), 347-371. Dellaert, B. G. C., Arentze, T. A., Bierlaire, M., Borgers, A. W. J. & Timmermans, H. J. P. (1998). Investigating Consumers' Tendency to

Combine Multiple Shopping Purposes and Destinations. Journal of Marketing Research, 35, 177-188. Ewing, R. & Cervero, R. (2010). Travel and built environment - Meta analysis. Journal of American Planning Association 76(3), 265-294. Gonzalez-Feliu, J., Ambrosini, C., Pluvinet, P., Toilier, F. & Routhier, J. L. (2012). A simulation framework for evaluating the impacts of urban

goods transport in terms of road occupancy. Journal of Computational Science 3 (4), 206-215. Gonzalez-Feliu, J., Toilier, F. & Routhier, J.L. (2010). End-consumer goods movement generation in French medium urban areas. Procedia -

Social and Behavioral Sciences 2 (3), 6189-6204. Ingene, C. A. & Ghosh, A. (1990). Consumer and Producer Behavior in a Multipur-pose Shopping Environment. Geographical Analysis, 22(1), 70-93.

ISTAT (2001). Census 2001. Warehouse of statistics. Italian Institute of Statistics, Rome, Italy,

Miodonski, D. & Kawamura, K. (2012). Effects of Built Environment on Freight Consumption. Procedia - Social and Behavioral Sciences, 39, 74-88.

Mokhtarian, P. L. (2004). A conceptual analysis of the transportation impacts of B2C e-commerce. Transportation 31(3), 257-284. Mokhtarian, P. L., Ory, D. T. & Cao, X. (2007). Shopping-Related Attitudes: A Factor and Cluster Analysis of Northern California Shoppers.

Proceeding of the 11th WCTR, Berkeley (CA), USA. Muñuzuri, J., Cortés, P., Onieva, L. & Guadix, J. (2012). Estimation of Daily Vehicle Flows for Urban Freight Deliveries. Journal of Urban Planning and Development 138 (1), 43-52.

Nuzzolo, A. & Comi, A. (2013a). Urban freight demand forecasting: a mixed quantity/delivery/vehicle-base model. Transportation Research Part E, forthcoming.

Nuzzolo, A. & Comi, A. (2013b). City logistics planning: demand modelling requirements for direct effect forecasting. In E. Taniguchi & R. G.

Thompson (Eds.), Proceedings of the 8th International Conference on City Logistics. Bali, Indonesia. Nuzzolo, A. & Comi, A. (2013c). Tactical and operational city logistics: freight vehicle flow modelling. In Ben-Akiva, M., Meersman, H. &

Van de Voorde, E. (Eds.), Freight Transport Modelling (pp. 433-451). Bingley, U.K., Emerald Group Publishing Limited. Nuzzolo, A. & Coppola, P. (2005). S.T.I.T.: a system of mathematical models for the simulation of land-use and transport interactions.

Proceedings of European Transport Conference, Edinburgh, Scotland. Nuzzolo, A., Comi, A. & Papa, E. (2013). Simulating the effects of shopping attitudes on urban goods distribution. Procedia — Social and

Behavioral Sciences, Proceedings of the 16th Meeting of the EURO Working Group on Transportation (EWGT 2013), Elsevier. Oppenheim, N. (1993). A combined, equilibrium model of urban personal travel and goods movements. Transportation Science, 27(2), 161173.

Popkowski Leszczyc, P. T. L., Sinha, A. & Sahgal, A. (2004). The Effect of Multi-purpose Shopping on Pricing and Location Strategy for

Grocery Stores. Journal of Retailing, 80, 85-99. Russo, F. & Comi, A. (2010). A modelling system to simulate goods movements at an urban scale. Transportation, 37(6), 987-1009. Russo, F. & Comi, A. (2012). The Simulation of Shopping Trips at Urban Scale: Attraction Macro-Model. Procedia - Social and Behavioral Sciences, 39, 387-399.

Schoemaker, J., Allen, J., Huschebek, M., & Monigl, J. (2006). Quantification oof Urban Freight Transport Effects I. BESTUFS Consortium, www.bestufs.net.

Shimazaki T., Kazunori, H. & Shihana, S. M. (1994). Comparative study of transportation modal choice in Asian countries. Transportation Research Record 1441, 71-83.

Taniguchi, E., Thompson, R. G., Yamada, T. & van Duin, R. (2001). City Logistics - Network Modelling and Intelligent Transport Systems. Pergamon.

Thill, J. C. (1992). Spatial Duopolistic Competition with Multipurpose and Multistop Shopping. Annals of Regional Science 26, 287-304. Wygonik, E., Bassok, A., Goodchild, A., McCormack, E. & Carlson, D. (2012). Sustainable urban goods movement: emerging research agendas. Proceedings of Workshop Innovation in Urban Freight, Seattle, WA, USA.