Risk Analysis - 2021 - de La Cruz López - Assessing The Risk of Robbery in Bank Branches To Reduce Impact On Personnel

Risk Analysis, Vol. 42, No. 2, 2022 DOI: 10.1111/risa.
13761
Assessing the risk of robbery in bank branches to reduce

impact on personnel
María Pilar de la Cruz López ,1 Juan José Cartelle Barros ,2,∗ Alfredo del Caño
Gochi ,1 María Concepción Garaboa Fernández,3 and Jesús Blanco Leis3
According to existing literature, bank robberies can have a considerable impact on the people
involved (employees, customers, and police officers), even if the direct economic losses are
negligible. Consequently, this article presents a model to assess the risk of bank robbery, with
the aim of reducing the impact on the people and prioritizing the investments in security
measures. It is based on the MIVES (Spanish acronym for the Integrated Value Model for
Sustainability Assessment) method and it was combined with Monte Carlo simulation as a
way of taking into account the uncertainty. Correlations were also modeled, for simulation
purposes. Indicators for addressing issues related to security features, employees, operational
procedures, and physical and social environment were defined. The model was applied to two
fictitious but realistic sets of cases. The first simulation provides a quick overview of the risk
level of a fictitious bank, before collecting the full set of data from hundreds or thousands
of branches. The second simulation analyzes the risk variation of a specific bank branch over
time. The model was also used to assess the risk index of 636 real branches belonging to a
Spanish bank. All the results are presented and discussed in depth. The model allows the user
to identify the weak points of a branch, so that corrective measures can be taken.
KEY WORDS: Commercial bank branches; MCDM method; Monte Carlo; risk analysis; robbery risk
1. INTRODUCTION: LITERATURE REVIEW ing at historical data from banks or scientific publi-
AND MAIN OBJECTIVES cations (European Banking Federation [EBF], 2010,
2011; Gill, 2000; Reilly, Rickman, & Witt, 2012). This
In recent years, the amount of money stolen dur-
is due, among other things, to the fact that banks have
ing bank robberies in developed countries has de-
made a great effort to improve their anti-theft sys-
creased. This conclusion may be reached by look-
tems (EBF, 2010, 2011). For example, money storage
devices are becoming increasingly safer with mecha-
1 Departamento de Ingeniería Civil, Escuela Politécnica Superior
nisms for hindering robberies, such as delayed open-
(EPS), Universidade da Coruña (UDC), Campus de Esteiro, ing (Dugato, 2014). On the other hand, electronic
C/Mendizábal s/n, Ferrol (A Coruña), Ferrol, 15403, Spain. payment systems, e-banking services, and Automated
2 Escuela Politécnica Superior (EPS), Universidade da Coruña
Teller Machines (ATMs) have become more com-
(UDC), Campus de Esteiro, C/Mendizábal s/n, Ferrol (A monplace (Dugato, 2014; EBF, 2010, 2011). Conse-
Coruña), Ferrol, 15403, Spain.
3 ABANCA Corporación Bancaria S.A., C/ Olmos 26, Coruña, A quently, bank offices do not currently need to have
Coruña, 15003, Spain. large amounts of cash. Thus, in a considerable num-
∗ Address correspondence to Juan José Cartelle Barros, Escuela
ber of bank robberies, no cash is actually taken
Politécnica Superior (EPS), Universidade da Coruña (UDC), and, in almost all the remaining cases, the amount
Campus de Esteiro, C/Mendizábal s/n, 15403, Ferrol (A Coruña),
of money stolen is minimal. Furthermore, convicted
Spain; juan.cartelle1@udc.es
bank robbers fare badly in a harsh penal system and,
385 0272-4332/22/0100-0385$22.00/1 © 2021 The Authors. Risk Analysis pub-

lished by Wiley Periodicals LLC on behalf of Society for Risk Analysis.
This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in
any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
15396924, 2022, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/risa.13761 by Readcube (Labtiva Inc.), Wiley Online Library on [23/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
386 de la Cruz López et al.
in most developed countries, police forces usually O’Donnell, 1996; Samavati, 2006). These include,
solve 60–70% of these cases. As the risk–benefit ratio among other aspects, gun availability, motivation
is not favorable (Dugato, 2014), professional crim- group dynamics of robbers, social issues, bank pro-
inal activity is moving toward less risky objectives cedures, location of the branch, escape routes, or
with greater opportunities (Dugato, 2014; EBF, 2010, previous bank robberies (once a bank branch has
2011; Haran & Martin, 1977). In this sense, cyber been robbed, the evidence suggests that the proba-
bank robberies are becoming more frequent, which bility of repetition increases up to a certain number
has attracted the attention of a considerable num- of times).
ber of researchers such as Hole, Moen and Tjostheim Other authors examine robbery typology, modus
(2006), Lesk (2011), Sood and Enbody (2013), Gor- operandi, criminals’ characteristics, or even the mo-
ton (2014) or Damenu and Beaumont (2017), among ment at which the crimes were committed (Abraham
others. & Baldassaro, 2001; Borzycki, 2003; EBF, 2010, 2011;
Despite the displacement effect that has taken Federal Bureau of Investigation [FBI], 2003, 2004,
place, many robbers are still willing to take the risk 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013,
of robbing a bank branch even if the financial bene- 2014, 2015, 2016, 2017, 2018; Samavati, 2004). These
fits are reduced (Morrison & O’Donnell, 1996). On studies refer to a city, a region or, at most, a country,
the other hand, the impact that bank robberies can but there is no study on Spain.
have on everyone involved—employees, customers, Some works analyze the differences that ex-
and police officers—can be very severe, or even fa- ist between different cities or countries (Dugato,
tal. While direct economic losses may be negligi- 2014; Levine, 2007; Matthews, Pease, & Pease, 2001;
ble, people can sustain serious psychological after- Samavati, 2006; Wang, 2002). Security measures also
effects, physical injuries, and even death. The possi- receive attention in the literature, although many
ble psychological after-effects have been studied by works are quite old and outdated (Ozenne, 1974).
a range of authors; in general, directly experiencing However, some of them are useful to get an idea of
a bank robbery often has dire consequences (Con- the security devices that robbers have been most con-
verso & Viotti, 2014; Fichera et al., 2015; Frans, Åhs, cerned about (Büchler & Leineweber, 1991; Hannan,
Bihre, & Åhs, 2018; Giorgi, Fiz Perez, et al., 2015; 1982; Kube, 1988; Nugent, Burns, Wilson, & Chap-
Giorgi, Leon Perez, Montani, Courcy, & Arcangeli, pell, 1989).
2015; Hansen, Armour, & Elklit, 2012; Hansen, Ar- Other research is on the security measures them-
mour, Shevlin, & Elklit, 2014; Hansen & Elklit, 2013; selves, with the aim of improving them or developing
Hansen, Hyland, & Armour, 2016; Hansen, Las- new ones. Szczodrak and Szwoch (2013) propose us-
gaard, & Elklit, 2013; Johnston, 1978; Jones, 2002; ing thermal camera images for bank robbery detec-
Jones & Jones, 1998; Kamphuis & Emmelkamp, 1998; tion. Similarly, Khera and Verma (2014) look at an
Leymann, 1985, 1988; Miller-Burke, Attridge, & Fass, autonomous control system for bank vaults to detect
1999; Mucci, Giorgi, Perez, Iavicoli, & Arcangeli, and record suspicious movements. Kotus, Łopatka,
2015; Paes-Machado & Nascimento, 2006). Similar Czyżewski and Bogdanis (2016) propose a new con-
impacts occur when other establishments, such as trol system capable of distinguishing between nor-
pharmacies (Fichera, Sartori, & Costa, 2009), super- mal sounds and the ones—like shots and screams—
markets, jewelry shops, or tobacconists (Setti et al., caused by threatening circumstances. By simulating
2018), are targeted by robbers. a dangerous situation in a real bank branch, the au-
Despite the fact that there is a considerable thors tested their system. Similarly, Gupta, Kumar
body of literature addressing the psychological con- and Malhotra (2015) develop a hand gesture recog-
sequences of bank robberies, the number of stud- nition method for bank employees during a robbery.
ies analyzing the risk of robbery in banking is lim- In relatively recent times, one study looks at the
ited. Dugato (2014) says that academia has not paid robbery procedures used in Italy (De Leo, Volpini, &
enough attention to this issue, and that most of the De Gregorio, 2006). The authors highlight the most
existing studies use out-of-date information, adopt- relevant security measures to prevent robberies and
ing a descriptive approach instead of an analytical how people must act during such crimes to dimin-
one. ish the possible harmful effects. Weisel (2007) and
Some publications deal with factors that Braga (2008) carry out similar studies in the United
may influence robbers’ decisions (Dugato, 2014; States on a comparable scale. All of these stud-
Hochstetler, 2001; Levine, 2007; Morrison & ies have a common thread: the effectiveness of the
Risk of robbery in bank branches 387
security measures is analyzed in a qualitative way measures that an office needs to achieve an adequate
based on the authors’ opinions or interviews with level of risk. Further problems with this method can
criminals. be found in Cox (2008) (subjectivity and ambiguity
To the best of the authors’ knowledge, there is in estimating threat, vulnerability, and consequence
only one quantitative model to manage the risk of figures; problems with correlations, among others).
robbery in bank branches: Guazzoni and Ronsivalle Three security staff members of a Spanish bank
(2009). It uses an Artificial Neural Network (ANN), (ABANCA Corporación Bancaria S.A., from now on
which takes into account both endogenous (concern- ABANCA, or the bank), with extensive experience
ing the bank office and its security systems) and ex- in security and safety issues, have participated in cre-
ogenous (associated with the geographic location, ating the model. In terms of bank robbery-related
population density, and the crime rate) factors. Both matters, their combined experience is over 60 years.
come into play when a Global Robbery Risk Index One is a criminology expert with 37 years of expe-
(GRRI) is being estimated for each bank branch. rience. Moreover, other experts were consulted, one
The authors rely on all the available historical data of whom was a former police inspector with over a
of bank robberies in Italy to train and validate the decade of security experience in the banking sector.
ANN. Consequently, if the ANN is employed to esti- Trade unions health and safety representatives were
mate the GRRI of the bank branches located in a dif- also involved.
ferent country, the associated historical data must be On the one hand, through two Monte Carlo sim-
collected to train the model first. This can be a time- ulations, the model was used to generate a wide range
consuming and cost-extensive activity, especially if of fictitious case studies. On the other, ABANCA
there are no available databases. Furthermore, it may performed a deterministic risk assessment of its en-
be impossible to collect certain data. If this is the tire network of bank branches, carrying out improve-
case, the ANN training cannot be successfully exe- ment projects until all the bank branches complied
cuted, and the results would be untrustworthy. with management’s requirements. For obvious rea-
In view of the above, the main objective of this sons, it is not possible to reproduce real data associ-
article is to present an up-to-date model that makes ated with a specific office here, but the general results
it possible to quantitatively assess the risk of rob- derived from using the complete model to assess all
bery in bank branches. The model is based on the the ABANCA branches are presented and discussed.
MIVES method (Spanish Acronym of Integrated The reader should bear in mind that the model
Value Model for Sustainability Assessment method). goes beyond the national legislation in terms of
To deal with uncertainty, it was combined with Monte banks’ security issues.
Carlo simulation. Correlations between risk indica- This article is organized as follows. In Section 2,
tors were also taken into account. The assessment the methodology is explained, while the final model
model does not need historical data for its appli- is presented in Section 3. The simulation case studies
cation, although certain historical information was are included in Section 4. Results are presented and
considered during its conception. The results can be discussed in Section 5. Finally, the main conclusions
used to identify the weak points of a branch. Con- are outlined in Section 6.
sequently, investments to increase security can be
made. This serves to reduce both the risk of robbery
and also the possible impacts on people. To the best 2. MATERIALS AND METHODS
of the authors’ knowledge, there is no model like the
one presented here in the existing literature.
2.1. Multi-Criteria Decision Making (MCDM)
The characteristics, advantages, and disadvan-
Methods: MIVES
tages of a total of 15 methods that could have been
applied to this work were analyzed. It was concluded Most of the decisions to be taken in real life are
that MIVES was the best suited to the bank’s needs. of a multicriteria nature, and assessing the risk of
In particular, for example, the traditional method of robbery in bank branches is not an exception. Multi-
estimating the level of risk as a function of threat, vul- Criteria Decision Making (MCDM) methods are of
nerability, and consequences, has several problems. great help at the time of facing multi-criteria prob-
One is the nonadditivity of the risks estimated by lems.
this method. On the other hand, this technique is not A wide range of MCDM methods exist such as
suitable for adequately allocating the new security the analytic hierarchy process (AHP) (Saaty, 1980,
Table I. Some Recent Studies Using MIVES Method for Solving Problems in a Wide Range of Fields
Field of Application Source
Energy sector (Cartelle Barros et al., 2015; Cartelle Barros, Lara Coira, de la Cruz López, & del
Caño Gochi, 2016; Cartelle Barros, Lara Coira, de la Cruz López, del Caño Gochi,
& Soares, 2020)
Construction sector (Casanovas-rubio et al., 2019; Habibi, Pons Valladares, & Peña, 2020; Josa, de la
Fuente, Casanovas-Rubio, Armengou, & Aguado, 2021; Zubizarreta et al., 2019)
University education (Pons, Franquesa, & Hosseini, 2019)
Industrial sector (Cartelle Barros, Lara Coira, de la Cruz López, & del Caño Gochi, 2018)
Prioritization of investments in public services (Pardo-Bosch, Aguado, & Pino, 2019; Pujadas, Pardo-Bosch, Aguado-Renter, &
Aguado, 2017)
Natural disasters and extreme events (Gandini, Garmendia, Prieto, Alvarez, & San-José, 2020)
Emergency and post-disaster problems (Hosseini, Pons, & de la Fuente, 2018; Hosseini, Yazdani, & de la Fuente, 2020)
Project management (Zubizarreta, Ganzarain, Cuadrado, & Lizarralde, 2021)
2006); the analytic network process (ANP) (Saaty

& Vargas, 2006) or MIVES (Cartelle Barros, Lara
Coira, de la Cruz López, & del Caño Gochi, 2015;
Casanovas-rubio, Pujadas, Pardo-bosch, Blanco, &
Aguado, 2019; de la Cruz, Castro, del Caño, Gómez,
Lara, & Cartelle, 2014; Zubizarreta, Cuadrado, Orbe,
& García, 2019), among others. The reader can find in
Hajkowicz & Collins (2007) and in Shao et al. (2020)
additional information about some of these and other
MCDM methods used in the scientific literature.
The results provided by different MCDM meth-
ods are similar (Zamani-Sabzi, King, Gard, &
Abudu, 2016), at least at the time of selecting the
best alternatives (Chitsaz & Banihabib, 2015). In fact,
the discrepancies that do exist are mostly related to Fig 1. Example of a theoretical requirement tree.
the weights of the model, which can be corrected
(Kou, Lu, Peng, & Shi, 2012). However, not all the
MCDM methods present exactly the same advan- other discussed throughout this article, have led to
tages. MIVES is one of the alternatives with the best the use of MIVES in this work. MIVES uses require-
complexity-performance ratio (Cartelle Barros et al., ment trees and value functions.
2015). By way of example, MIVES makes it possi- A requirement tree is a scheme that usually con-
ble to consider potential nonlinearities in the assess- sists of an overall index with three breakdown lev-
ment of both quantitative (continuous) and qualita- els: requirements, criteria, and indicators. The first
tive (discrete) indicators. If necessary, it also inte- two levels facilitate the understanding of the prob-
grates AHP for establishing weights (Cartelle Bar- lem to be solved, and the calculations to be made.
ros et al., 2015). Moreover, it can be easily combined The indicators are the quantitative or qualitative as-
with the Monte Carlo simulation (de la Cruz, Castro, pects that are going to be assessed through the use
del Caño, Gómez, Lara, & Cartelle, 2014) or fuzzy of value functions. The overall index is used to make
arithmetic (de la Cruz, Castro, del Caño, Gómez, an assessment that takes into account all the indica-
Lara, & Gradaille, 2014) in order to consider uncer- tors of the model, as an aid to decision making. Fig. 1
tainty. MIVES is a flexible method that can be used shows an example of a theoretical requirement tree
to solve problems in a wide range of fields. In fact, with three requirements, six criteria, and nine indica-
to date, more than 60 articles on MIVES have been tors. The requirement tree conceived in this work is
published in scientific journals. The reader can find shown in Table II.
in Table I some of the most recent studies using this Value functions are mathematical tools that
method. The reasons summarized here, along with serve to transform the different units for the
Table II. Requirement Tree for the Model with the Weights for the Requirements (α ind ), Criteria (β ind ) and Indicators (γ ind ). The Input
Values of the Qualitative Indicators as well as their Corresponding Levels of Risk (Vind ) are Also Included
b
α ind Requirements β ind Criteria γ ind Indicators
a
RI 26% 1. Branch security 10% 1.1. Alarm reception 100% 1.1.1. Communication channels:
features
• One channel (PSTN), Vind = 1
• One channel (IP), Vind = 0.85
• Two channels (PSTN/GSM), Vind = 0.5
• Two channels (one of them IP), Vind = 0.4
• Three channels (IP, PSTN, GSM), Vind = 0
40% 1.2. Storage device 50% 1.2.1. Safes:

location
• Visible and located in the main service area, Vind = 1
• Not visible and located in the main service area, Vind
= 0.25
• Located in the archive, Vind = 0.05
50% 1.2.2. Cash box:
• Visible from the main service area, Vind = 1

• Not visible from the main service area, Vind = 0
50% 1.3. Security devices 75% 1.3.1. Devices for cash storing and handling:
• One safe and one cash box, Vind = 1

• One safe and one cash dispenser, or cash recycler,
Vind = 0.83
• One safe, one cash dispenser or cash recycler, and
one cash box, Vind = 0.66
• Two safes and one cash box, Vind = 0.5
• Two safes and one cash dispenser, or cash recycler,
Vind = 0.33
• Two safes, one cash dispenser or cash recycler and
one cash box, Vind = 0.16
• Two safes, more than one cash dispenser or cash
recycler, and one cash box, Vind = 0
25% 1.3.2. Security cameras:
• Main entrance, Vind = 1

• Main entrance and cashier desks, Vind = 0.8
• Main entrance, cashier desks, and main service area
Vind = 0.4
• The previous three options and the archive, Vind =
0.15
• The previous four options and out of the branch,
Vind = 0
c
21% 2. Branch employees 60% 2.1. Deterrent 100% 2.1.1. Number of employees :
capacity
• Pind,min = 8
• Pind,max = 0
• nind = 4
• mind = 0.5
• Aind = 3
(Continued)
Table II. (Continued)

b
40% 2.2. Response 30% 2.2.1. Staff composition:

capacity
• Combination of personnel with more and less of two
years of experience, Vind = 1
• Only personnel with more than two years of
experience, Vind = 0.20
70% 2.2.2. Training:
• At least one worker without training, Vind = 1

• All personnel with training but one of them with
only online training, Vind = 0.4
• All personnel with classroom training, Vind = 0.12
• All personnel with classroom and online training,
Vind = 0.05
26% 3. Operational 30% 3.1. Cash storing 100% 3.1.1. Cash storing during noncommercial hours:
procedures during
noncommercial • Out of storing locking-devices, Vind = 1
hours • In one locking device, Vind = 0.66
• In two locking devices, Vind = 0.33
• In more than two locking devices, Vind = 0
70% 3.2. Cash handling 95% 3.2.1. Cash handling during commercial hours:
during
commercial hours • Bunker, Vind = 1
• Security lock, Vind = 0.95
• Cash box, Vind = 0.9
• Cash dispenser or cash recycler, Vind = 0.2
• Other, Vind = 0
5% 3.2.2. Amount of money:
• Level 3 (less than 100,000 €), Vind = 0

• Level 2 (between 100,000 and 200,000 €), Vind = 0.5
• Level 1 (more than 200,000 €), Vind = 1
12% 4. Physical 100% 4.1. Physical 45% 4.1.1. Location:

environment environment
• Rural area (less than 10,000 inhabitants), Vind = 1
• Urban area, Vind = 0.95
• Suburbs, Vind = 0.9
• Others, Vind = 0.5
35% 4.1.2. Surrounding buildings:
• Abandoned building, upper, or lower floor, Vind = 1

• Uninhabited building, upper or lower floor; or
garage and basement, Vind = 0.95
• There are no surrounding buildings or floors, Vind =
0.5
• Without information about the surrounding
buildings, Vind = 0.3
• Inhabited surrounding-building, upper floor, or
lower floor, Vind = 0.05
20% 4.1.3. Escape route:
• Easy and close to the branch (vehicle), Vind = 1

• Easy and close to the branch (on foot), Vind = 0.9
• Difficult but close to the branch, Vind = 0.25
• Difficult and distant to the branch, Vind = 0.1
(Continued)
Table II. (Continued)

b
15% 5. Social 70% 5.1. Potential of 50% 5.1.1. Number of robberies in the province:
environment conflict
• Level 3 (more than 10), Vind = 1
• Level 2 (between 5 and 10), Vind = 0.4
• Level 1 (less than 5), Vind = 0.1
50% 5.1.2. Number of robberies in the branch (5 years):
• More than 3, Vind = 1

• Between 1 and 3, Vind = 0.7
• Less than 1, Vind = 0.2
30% 5.2. Police 100% 5.2.1. Response time

operations
• More than 30 minutes, Vind = 1
• Between 10 and 30 minutes, Vind = 0.8
• Less than 10 minutes, Vind = 0.1
a RI is the risk index. It falls within the interval [0,1], being 0 and 1 the maximum and minimum levels of satisfaction (or the minimum and
maximum levels of risk), respectively.
b The possible input values (answers) that the qualitative (discrete) indicators can take are listed below their names. The number that follows
each input value is the level of risk (Vind ). Once again, this number falls within the interval [0,1], being 0 and 1 the maximum and minimum
levels of satisfaction (or the minimum and maximum levels of risk), respectively.
c This indicator can be treated as a quantitative (continuous) one. The values that take the different value function parameters are indicated
below its name. These are the same parameters that appear in Equation (2). It could also have been possible to treat this indicator as a
discrete one.
indicators into a common and dimensionless param- three ABANCA employees with 37, 14, and 10
eter called value (Vind ). In this work, this parame- years of experience in bank robbery risk, and
ter will be related to the risk level. The performance three university researchers with more than 25,
of each indicator is assessed through a value func- 20, and seven years of experience in the fields of
tion that allows the user to consider possible non- risk assessment and management, and decision
linearities. MIVES also allows the establishment of support methods.
filters to cancel the evaluation if an indicator does • Meetings were held for the project team to dis-
not reach a minimum level. More details about the cuss how to assess each indicator included in the
MIVES model presented here, including the require- preliminary MIVES model.
ment tree and the corresponding value functions, are • An inquiry form was used to collect real in-
discussed in Section 3.1. Moreover, the reader can formation from the bank branches. The objec-
find more information about the MIVES method in tive of this step was to determine the extent to
Cartelle Barros et al. (2015) and in de la Cruz, Castro, which real information could feed the prelim-
del Caño, Gómez, Lara and Cartelle (2014). inary model. Furthermore, the bank branches
were asked to provide comments on the prelimi-
nary model so that modifications could be made.
2.2. General Methodology • Statistical data related to robberies that oc-
The followed methodology entailed these main curred in the bank were gathered. The available
steps, listed chronologically: historical data were taken into account at the
time of creating the model. Nevertheless, as pre-
• A preliminary proposal was made for a require- viously stated, the user does not need to collect
ment tree with all the indicators that had to be historical information for using the model pre-
considered to assess the risk of robbery in bank sented here.
branches. Both academia and security staff from • Extensive consultations were made with the
the bank were involved in this first task. The trade unions. They could also make com-
group (from now on, the project team) included ments on and suggestions about the preliminary
model. This and the two previous steps were mon. The case here studied is sufficiently intricate for
conducted simultaneously. using three levels.
• With the information collected in the pre- The number of requirements, criteria, and indi-
vious two phases, the preliminary model was cators is not established beforehand. It is the result
amended. After this step, the final model, with of the reflection process undertaken for creating the
24 indicators, was thoroughly defined and ready model. In this work, the requirements and criteria
to use. were selected during brainstorming meetings aimed
• The final model was implemented to assess the at compiling all aspects that can influence the risk of
risk of robbery of the bank branches. robbery. At those moments, requirements and crite-
• The results obtained were used to create a pri- ria served, among other things, (i) to better under-
oritized list of investment projects for improving stand the problem and its specific topics; (ii) to group
security systems. the generated ideas in an orderly manner, including
• A final phase was carried out by the authors: in a single criterion all indicators belonging to a spe-
a sensitivity analysis on the 24-indicator-model cific topic, and then doing the same with criteria and
was performed. With this step, it was possible to requirements; and (iii) to identify and eliminate rep-
reduce the number of indicators in the model, etitions.
eliminating those that did not generate signifi- On the other hand, if only two breakdown lev-
cant differences in the results. Thus, a second, els (requirements and indicators) are considered, or
final model with 17 indicators was defined. Cor- even if only one level is defined (all the indicators
relations were also established between the indi- belonging to the same block), some problems may
cators. The latter model is the one presented in arise at the time of establishing the weights. For ex-
Section 3.1 and used in the two simulation cases ample, if the model in Table II consisted of only
for this work. one level with the 17 indicators, their weights would
have to be defined by comparing 17 very uneven pa-
rameters. In this way, it is easy to lose sight of the
overall picture, and the resulting weights are not al-
3. MODEL ways consistent. The problem increases when em-
ploying 24 indicators. Even when using AHP there
could be problems. By employing requirements and
3.1. MIVES Model
criteria, the model developers only have to compare
Table II presents the requirement tree, with its the importance of a limited number of parameters, in
corresponding weights, from the final model men- each tree branch. Moreover, the requirement tree al-
tioned above. It consists of 17 indicators grouped into lows the generation of subindexes, or partial indexes,
five thematic blocks or requirements. The first block which can be very useful. In this work, it allows to
relates to branch security features; it takes into ac- evaluate not only the office as a whole, but also its
count five different indicators. Three of them belong different aspects (employees, operating procedures,
to the second block in which the consequences that social environment, etc.). In the same way, for in-
the branch employees have on the risk of robbery are stance, subindexes of environmental, social, and eco-
assessed. The branch operational procedures were nomic sustainability can be calculated when assessing
evaluated by the use of three indicators included in sustainability employing life-cycle analysis and the
the third requirement. Finally, another set of three MIVES method (Cartelle Barros et al., 2015).
indicators was used to assess the physical and social All the indicators included in Table II are qual-
environment. itative (discrete), with only one exception: indicator
A glossary of terms was included as Supplemen- 2.1.1 (Number of employees). Each qualitative indi-
tary material. cator presents some possible input values (listed be-
As previously indicated, requirement trees usu- low its name in Table II), or answers to a specific
ally consists of three levels: requirements, criteria, question. Each input value or answer is a semantic
and indicators. There can be trees with less than three label associated with a specific level of risk (Vind ).
levels, for very simple problems in which it is not nec- This number falls within the interval [0,1], being 0
essary to assess a great number of indicators. Simi- and 1 the minimum and maximum levels of risk, re-
larly, there can be very complex problems that may spectively (or the maximum and minimum levels of
need more than three levels, although it is not com- satisfaction). The risk (Vind ) associated with each of
Fig 2. Example of qualitative value function: indicator 1.1.1. Fig 3. Value function related to the indicator 2.1.1. (Number of
(Communication channels). employees).
ing value functions. Moreover, throughout the use of

the potential values taken by an indicator was estab- this equation, the geometry of a value function can
lished by the project team, taking into account all the be adapted for being more or less demanding at the
available information mentioned earlier. These can time of assessing a specific indicator (more or less
be found in Table II. Fig. 2 includes an example of concave, for instance). Consequently, Equation (1) is
qualitative value function, in particular, for the indi- usually adequate for the majority of MIVES mod-
cator related to communication channels. It should els, allowing the user to define many different ge-
be noted that this value function is slightly nonlin- ometries. Nevertheless, there can be specific cases in
ear. On the other hand, this indicator presents five which the use of Equation (1) is not valid. For ex-
possible input values with their corresponding lev- ample, value functions for thermal comfort can have
els of risk (Vind ). For example, as can be deducted parabolic geometries, among other options (Alarcon,
from Fig. 2, the input value “1 channel (PSTN)” is Aguado, Manga, & Josa, 2011). However, Equation
the worst option, since it is linked to the maximum (1) is suitable for the case presented in this study.
level of risk (Vind = 1). As can be seen in Fig. 3, the influence on risk of
The level of risk (Vind ) associated with the con- the number of employees present in the branch is as-
tinuous indicator 2.1.1. (Number of employees) is sessed through a decreasing value function in which
calculated using the following equation: a higher input value presents a lower level of risk. In
other words, a large number of employees will nor-

|Pind −Pind,min | Aind mally be considered as a deterrent by the robber. By
1 − exp −mind · nind
way of example, if a specific bank branch has four em-
Vind = , (1)
|Pind,max −Pind,min | Aind ployees, the level of risk (Vind ) will take a value close
1 − exp −mind · nind to 0.4 as can be deduced from Fig. 3. The reader can
obtain this value by introducing in Equation (1) the
where Pind is the input value to the value function of corresponding values for all the parameters.
the alternative under assessment. Pind,min is the input Using the model presented in Table II, it is pos-
value that returns the minimum level of risk (Vind = sible to obtain the Risk Index (RI). It is a number
0). Similarly, Pind,max is the input value that generates that falls within the interval [0,1], the maximum and
the maximum level of risk (Vind = 1). Aind , mind , and minimum levels of satisfaction (or the minimum and
nind are shape parameters used to generate different maximum levels of risk), respectively. The RI is cal-
geometries. The values that these parameters adopt culated from Equation (2):
are included in Table II. They generate the nonlin-
ear geometry shown in Fig. 3. Equation (1) makes it

17
possible to generate linear, concave, convex, and S- RI = αind · βind · γind · Vind , (2)
shaped geometries both for decreasing and increas- ind=1
where α ind , β ind , and γ ind are the weights for the re- fies national regulations, although there is great room
quirements, criteria, and indicators, respectively. The for improvement.
sum of the weights of all the elements belonging to
a certain branch of the requirement tree must be
3.2. Sensitivity Analysis
equal to 100 %. For example, the first requirement
(1. Branch security features) consists of three criteria: The final model employed by the bank was made
1.1. Alarm reception, 1.2. Storage device location, up of 24 indicators. However, it is not desirable to
and 1.3. Security devices. The sum of their weights handle a large number of indicators. Each additional
(10%, 40% and 50%, respectively) gets a value of parameter to be assessed requires information as
100%. well as extra mathematical operations. Obtaining in-
There are different options for establishing the formation can be a complex, even impossible, task.
weights for the requirements, criteria, and indica- Therefore, whenever possible, a sensitivity analysis
tors. The simplest alternative is direct allocation. As must be performed to remove unnecessary indica-
its name suggests, it involves using numerical val- tors. These indicators are the ones that hardly cause
ues directly established by experts in the field. Its major changes in the results (RI), regardless of the
use is appropriate when a reduced number of ele- value adopted. There can be indicators that always
ments have to be weighted (normally, between 2 and assume the same value (or a similar one) for all the
4), as long as there are no significant discrepancies alternatives. In such a case, those indicators can also
among the experts. The proportional method is an- be eliminated.
other option. In this case, one of the elements that One of the strengths of the MIVES method is
must be weighted is selected as reference, and a rel- that it uses value functions. As previously indicated
ative importance is assigned to it (for example 100). in Section 2.1, they make it possible to consider non-
The weights for the remaining elements that belong linearities when assessing with both continuous and
to the same branch of the tree are established by di- discrete indicators (please see Figs. 2 and 3). Thus, a
rect comparison with the reference requirement, cri- sensitivity analysis should be carried out under dif-
teria, or indicator. The process ends with a normal- ferent situations. Four different scenarios were con-
ization stage, so that the sum of the weights gets a sidered in this analysis. In the first one, all the indi-
value of 100%. This method can be appropriate when cators adopted the input value associated with the
the number of variables to be weighted varies be- minimum level of satisfaction (maximum risk) as a
tween three and seven. It is also useful when there reference. The second case was the opposite of the
is some discrepancy among the people involved in first one. In the third scenario all the indicators took
the weighting process, even if the number of pa- the input value closest to the medium level of risk
rameters is below five. If there are significant dis- (0.5). If two input values were at the same distance
crepancies among the experts or if the number of from a 0.5 risk level, the worst one, or higher risk,
parameters to be weighted exceeds seven, the use was chosen. If an indicator could only adopt two val-
of AHP is recommended (Saaty, 1980, 2006). The ues, the worst one was selected unless the best one
weights included in Table II were defined with the looked very close to a 0.5 risk level. The last case was
help of these three methods, in several meetings similar to the third one with two differences. Now, if
with all the members of the project team. The fi- two input values were at the same distance from the
nal proposal was reviewed and approved by the key medium level of risk, the best one (lower risk) was se-
staff of the bank and its branches, and by the trade lected. Furthermore, if there were only two possible
unions. input values, the best one was used, unless the worst
It should be noted that RI is not the probability one adopted a value very close to 0.5.
of robbery that each bank branch faces. In fact, any For the four scenarios, the same process was car-
branch, irrespective of its security systems, can be tar- ried out. The RI was calculated with all the indicators
geted by criminals. Therefore, an RI of 0 is not equiv- adopting the reference value. After that, the value
alent to a zero probability in terms of being robbed. adopted for each one of the indicators was modified
It must be understood as a minimum level of risk, cor- separately, from the best to the worst possible input
responding to an excellent office, taking into account value, while all the other indicators continued to have
the available statistical data and the opinions of the the same reference value. The maximum and mini-
experts that participated in the project. In the same mum RIs were calculated and compared with the ini-
line, an RI equal to 1 suggests that the branch satis- tial one. This provided a vision of how each indicator
could affect the RI. This process was repeated for all The relation between these two aspects is clear. It is
the indicators independently. After that, the indica- not possible to put the cash into two locking-devices
tors with less influence on the RI were identified and if the office only has one. Nevertheless, the opposite
removed, obtaining a final MIVES model with 17 in- can happen: that is, cash can be stored out of locking-
dicators (Table II). devices even if there are two or more of those de-
By way of example, indicators such as the num- vices. This last case can be linked to an inadvisable
ber of panic buttons or the number of different po- practice by employees. Therefore, during the Monte
lice forces were removed. In the first case, if people’s Carlo simulation, if indicator 1.3.1. adopted the third,
safety is considered paramount, as is the case here, fourth, fifth, sixth, or seventh input values of Table II,
the general idea is to avoid using the panic button, indicator 3.1.1. could adopt any of the possible input
since this could lead to a violent situation. Conse- values. Nevertheless, if indicator 1.3.1. adopted the
quently, the number of panic buttons during a rob- first or the second values, indicator 3.1.1. had to adopt
bery is irrelevant. In the second one, police forces one of the first three input values.
work in a coordinated way. Thus, the response time There was also a correlation between indicators
is the determining factor. The other indicators that 1.3.1. (Devices for cash storing and handling) and
were removed after the sensitivity analysis are: se- 3.2.1. (Cash handling during commercial hours). In
curity level of the premises regarding reception of this case, the correlation was imposed by the exis-
alarms (Levels I or II; very soon after the model was tence or nonexistence of a cash box, cash dispenser,
completed, all the branches were at Level II); manda- or cash recycler. In other words, if indicator 1.3.1.
tory security measures, required by legislation (this adopted the first or fourth input values of Table II,
was merely a compliance check indicator; the bank all the possible input values are valid for indicator
had always gone far beyond such measures); custody 3.2.1. with the exception of the fourth one. If indi-
of the branch keys and procedures for opening the cator 1.3.1. took the second or the fifth input value,
heavy security devices during noncommercial hours indicator 3.2.1. could not adopt the third input value.
(the corresponding weights were very low); and type In all other cases, indicator 3.2.1. could take any input
of branch (regular, temporary or stand for specific value.
events, among others; very low weight). A correlation was also modeled between in-
dicators 2.1.1. (Number of employees) and 3.2.2.
(Amount of money). The number of employees was
3.3. Correlations between Indicators
assumed to be an integer number. A noninteger num-
Probabilistic MIVES models very often present ber of employees could be valid, if one or more of the
whatever type of relation between two or more in- workers had part-time contracts. Nevertheless, this
dicators. In other words, there can be indicators that assumption was not considered here. On the other
cannot adopt an independent random input value. If hand, it was reasonable to assume that a higher num-
the input value of one indicator depends on the one ber of employees was linked to a higher level of activ-
adopted by other indicator, it is possible to say that ity (payment transactions, bill collection) and, as a re-
these two indicators are correlated. Correlations can sult, to the need to have more cash in the office. If the
be covered in different ways. In this work, an analyt- number of employees was less than or equal to two,
ical approach was taken. indicator 3.2.2. adopted the first input value (Level
Correlations were only needed when a proba- 3) of Table II. If it was higher than 8, the amount of
bilistic case was analyzed. In other words, they were money was over 200,000 € (Level 1). Level 2 was the
only taken into account for the two fictitious simu- input value adopted in the remaining cases.
lation cases presented in Section 4. Nevertheless, if
a real branch had been studied at a particular mo-
4. SIMULATION STUDIES
ment, correlations and simulations would not have
been required, since the user of the model would Two different simulation cases were studied in
have needed to introduce only one input value for this manuscript. The goal of the first one is to provide
each indicator, in particular, the one linked to the real a quick overview of the risk level of a whole bank; in
bank branch. this case, a fictitious one. Before collecting the full
A correlation was established between indicators set of data from hundreds or thousands of branches,
1.3.1. (Devices for cash storing and handling) and a simulation can be made to give a general idea of the
3.1.1. (Cash storing during noncommercial hours). bank’s situation. To do this, the bank’s security staff,
who are well aware of the condition of most agencies, count. Consequently, only the indicators linked to
will set up distribution functions for the model’s indi- employees (indicators 2.1.1., 2.2.1., 2.2.2., 3.1.1., and
cators. Here, a probability was assigned to each one 3.2.1.) were defined as probabilistic. The remaining
of the possible input values for the qualitative (dis- indicators were considered deterministic, adopting
crete) indicators (Table III, Simulation 1). Despite only one possible input value (probability equal to 1
the fact that the considered commercial bank did not in Table III). The reasons for treating certain indica-
exist, those values were established with the help of tors as probabilistic are the same as those mentioned
the main ABANCA experts, taking into account real in Simulation 1. Furthermore, the reader should bear
situations in different Spanish banks. The sum of the in mind that offenders usually perpetrate a robbery
probabilities for the different answers must be equal when the benefit is supposed to be higher or when,
to 1, as can be deduced from Table III. For the num- according to their perception, the risk is lower (Mor-
ber of employees (indicator 2.1.1), a closed triangu- rison & O’Donnell, 1996). With these ideas linked,
lar distribution was defined (Table IV). As previously the time of day in which the number of employees
mentioned, the number that the triangular distribu- is at its lowest—with fewer witnesses and possible
tion provides in each iteration was rounded to the heroes to interfere—can be perceived by the robber
nearest integer. as the best moment to act.
Therefore, results of Simulation 1 provide quick, In Simulation 2, all the indicators were defined
preliminary information about the real level of risk as qualitative. Therefore, the number of employees
that the different bank branches can have. From (indicator 2.1.1.) was transformed into a discrete pa-
these results, after analyzing the most unfavorable rameter with only three possible input values: (i) two
cases, the security specialists can quickly identify the employees, (ii) three employees, and (iii) four em-
real offices with the highest risk of robbery, in order ployees. A probability was defined for each possible
to make immediate decisions. For instance, invest- answer (Table V). This is still consistent with the con-
ing more in training employees, changing the storage tinuous value function displayed in Fig. 3. In other
devices’ locations, replacing the current devices with words, the levels of risk (Vind ) for the three possible
safer ones or increasing the number of safety devices, answers were obtained by using the continuous value
among many other options. function. Their values are 0.8302, 0.6350, and 0.4008,
The reader should bear in mind that, as a re- respectively.
sult of the Monte Carlo simulation, the number of Indicator 3.2.2. (Amount of money) is not in-
possible situations (that is, the number of RIs) can cluded in Table III. Defining the probabilities for the
be higher than the number of bank branches that a different input values of this parameter is not neces-
commercial bank can have. Nevertheless, real bank sary, since it is correlated with the number of employ-
branches are dynamic, since the staff and number of ees (indicator 2.1.1.).
employees are not always the same. People go on
holiday, take time off, or are even moved to another
branch. They also carry out work-related tasks out of 5. RESULTS AND DISCUSSION
the office or go out to have a break. The bank of-
This section is divided into two different sub-
fice can be refurbished to modernize facilities. Even
sections. In the first, the results for the two simulation
migratory movements can have an impact on certain
case studies are presented and discussed. Section 5.2
indicators, such as the amount of money (indicator
summarizes the general results obtained once the 24-
3.2.2). This can be the case of a community that has
indicator model to the ABANCA branches was ap-
undergone a population loss due to an economic cri-
plied.
sis. With a reduced number of inhabitants, less money
is needed for fewer economic transactions. The oppo-
site can also happen.
5.1. Results for the Simulation Case Studies
As real bank branches were dynamic, the authors
found it necessary to carry out a second simulation, Table VI includes the statistical parameters of
to examine how uncertainty can affect a specific bank the RI for the two simulation case studies, after the
branch. Simulation 2 provided probabilistic informa- Monte Carlo method was applied. Fig. 4 contains the
tion about a fictitious one. In this case, the term “dy- cumulative probability curve as well as the frequency
namic” was associated with staff. In other words, pos- histogram for Simulation 1. Fig. 5 provides analogous
sible reforms of the office were not taken into ac- information for Simulation 2.
Table III. Probabilities Associated with the Different Input Values of the Qualitative (Discrete) Indicators for the Two Simulations
Indicators Simulations
Simulation 1 Simulation 2
a a
Probability Probability
1.1.1. Communication channels:

• One channel (PSTN) 0.06 0
• One channel (IP) 0.12 0
• Two channels (PSTN/GSM) 0.17 0
• Two channels (one of them IP) 0.59 1
• Three channels (IP, PSTN, GSM) 0.06 0
1.2.1. Safes:
• Visible and located in the main service area 0.06 0
• Not visible and located in the main service area 0.47 1
• Located in the archive 0.47 0
1.2.2 Cash box:

• Visible from the main service area 0.09 0
• Not visible from the main service area 0.91 1
1.3.1. Devices for cash storing and handling:

• One safe and one cash box 0.04 0
• One safe and one cash dispenser or cash recycler 0.04 0
• One safe, one cash dispenser or cash recycler and one cash box 0.46 1
• Two safes and one cash box 0.04 0
• Two safes and one cash dispenser or cash recycler 0.04 0
• Two safes, one cash dispenser or cash recycler and one cash box 0.36 0
• Two safes, more than one cash dispenser or cash recycler and one cash box 0.02 0
1.3.2. Security cameras:

• Main entrance 0.05 0
• Main entrance and cashier desks 0.49 1
• Main entrance, cashier desks, and main service area 0.39 0
• The previous three options and the archive 0.05 0
• The previous four options and out of the branch 0.02 0
2.2.1. Staff composition:

• Combination of personnel with more and less of two years of experience 0.71 0.10
• Only personnel with more than two years of experience 0.29 0.90
2.2.2. Training:
• At least one worker without training 0.06 0
• All personnel with training but one of them with only online training 0.65 0.91
• All personnel with classroom training 0.26 0.09
• All personnel with classroom and online training 0.03 0
3.1.1. Cash storing during noncommercial hours:

• Out of storing locking devices 0.01 0
• In one locking device 0.14 0
• In two locking devices 0.52 0.70
• In more than two locking devices 0.33 0.30
(Continued)
Table III. (Continued)
Indicators Simulations
Simulation 1 Simulation 2
a a
Probability Probability
3.2.1. Cash handling during commercial hours:

• Bunker 0.02 0
• Security lock 0.02 0
• Cash box 0.31 0.23
• Cash dispenser or cash recycler 0.63 0.77
• Other 0.02 0
4.1.1. Location:
• Rural area (less than 10,000 inhabitants) 0.39 0
• Urban area 0.39 1
• Suburbs 0.20 0
• Others 0.02 0
4.1.2. Surrounding buildings:

• Abandoned building, upper or lower floor 0.02 0
• Uninhabited building, upper, or lower floor; or garage and basement 0.04 0
• There are no surrounding buildings or floors 0.02 0
• Without information about the surrounding buildings 0.30 0
• Inhabited surrounding building, upper floor, or lower floor 0.62 1
4.1.3. Escape route:

• Easy and close to the branch (vehicle) 0.48 1
• Easy and close to the branch (on foot) 0.18 0
• Difficult but close to the branch 0.24 0
• Difficult and distant to the branch 0.10 0
5.1.1. Number of robberies in the province:

• Level 3 (more than 10) 0.07 0
• Level 2 (between 5 and 10) 0.26 0
• Level 1 (less than 5) 0.67 1
5.1.2. Number of robberies in the branch (5 years):

• More than 3 0.04 0
• Between 1 and 3 0.22 0
• Fewer than 1 0.74 1
5.2.1. Response time

• More than 30 minutes 0.03 0
• Between 10 and 30 minutes 0.69 1
• Less than 10 minutes 0.28 0
a The sum of the probabilities of all the possible input values for an indicator is equal to 1.
Unfortunately, the results obtained in this work the need for collecting information from all branches.
cannot be compared with other from the existing lit- Consequently, it seems reasonable to obtain a sig-
erature. As explained in previous sections, there are nificant difference between the minimum and max-
no similar or comparable pieces of research. As indi- imum levels of risk, as is the case in this study (0.1556
cated in Section 4, Simulation 1 can provide a quick and 0.7508, first row of Table VI). A commercial
overview of the risk level of a whole bank, without bank can have modern branches equipped with the
Table IV. Model Input Values for Indicator 2.1.1. (Number of employees) for Simulation 1
Simulation Distribution Function Distribution Parameters
Minimum Mode Maximum

a
Simul. 1 Closed triangular 1 3 10
a In Simulation 1, a closed triangular distribution function was used during the Monte Carlo simulation for indicator 2.1.1.
Table V. Probability Values for Indicator 2.1.1. (Number of Employees) for Simulation 2
a
Simulation Distribution Function Distribution input Values and their Probabilities
2 Employees 3 Employees 4 Employees
Simul. 2 Discrete 0.3 0.4 0.3
a InSimulation 2, indicator 2.1.1. was treated as a discrete indicator with three possible input values: (i) two employees, (ii) three employees,
and (iii) four employees. The level of risk associated to each one of the input values was obtained by using Equation (2) with the parameters
defined in Table II and being Pind equal to 2, 3, and 4. The levels of risk are 0.8302, 0.6350, and 0.4008, respectively.
Table VI. Statistical Parameters of the RI for the Two Simulation Cases
Simulation Statistical parameters
Minimum Maximum Mean Modal Interval (MI) Frequency of MI SD
Simul. 1 0.1556 0.7508 0.4180 [0.3,0.4) 37.32% 0.087

Simul. 2 0.3446 0.5776 0.4315 [0.4,0.5) 47.57% 0.056
Fig 4. Cumulative probability curve and frequency histogram for the RI of the first simulation analysis. Interval 1 corresponds with (0.0,
0.1). The range for each interval is equal to 0.1.
best security features, located in nonconflictive ar- of risk (between 0.3 and 0.4, the longest bar in the
eas, with a considerable number of experienced and frequency histogram of Fig. 4) and the mean value
well-trained employees. Nevertheless, the opposite (0.4180 in Table VI) are far from this latter scenario
can also happen, at least in terms of its location. How- of high risk level. This is because commercial banks
ever, if the reader analyzes the results as a whole, in developed countries go beyond the security legis-
he or she realizes that both the more probable level lation. The reader should bear in mind that it is not
Fig 5. Cumulative probability curve and frequency histogram for the RI of the second simulation study. Interval 1 corresponds with (0.0,
0.1). The range for each interval is equal to 0.1.
possible to obtain a level of risk equal to zero, since the same time, is not conducive to the robbers’ easy
even the best branch from a security point of view escape.
can be the target of a robbery. Nonetheless, a level of In addition to the issues related to the previous
risk below 0.5 can be considered a high-performing simulation (Simulation 1), it is also interesting to un-
result, especially taking into account that the model derstand how the level of risk of a specific branch
presented here is quite demanding. The percentage can vary over time (Simulation 2). This is particularly
of branches with a RI over 0.5 is under 20% as can true when changes are more likely to happen in the
be deducted from the cumulative probability curve short term, that is, the ones linked to the staff. This
in Fig. 4. In cases such as these, the bank staff from is discussed in the second simulation (Simulation 2).
the security department must study the specific char- Before talking about it, a brief comparison can be
acteristic of each one of the branches separately, to made between its results with those for Simulation
detect and correct weak points. 1.
Possible corrective measures may entail: (i) im- The variability affecting a specific branch (Sim-
proving branch security features by, for instance, re- ulation 2) is lower than the one affecting all the
placing current devices with safer ones, or increasing branches of a commercial bank (aspect related to
the number of safety devices; (ii) promoting train- Simulation 1). Consequently, in the second simula-
ing courses on security issues for the employees, and tion, a smaller number of indicators are defined as
(iii) establishing control procedures for ensuring that probabilistic, as explained in Section 4. Obviously,
the staff follow guidelines in terms of cash handling the greater the variability (uncertainty) affecting the
and storing, among many others. On the other hand, inputs to the model, the greater the variability affect-
some indicators that can contribute to an increased ing its results. This is the reason why the difference
level of risk, in particular the ones linked to require- between the maximum (0.5776, Table VI) and mini-
ments 4 and 5 (physical and social environment, re- mum (0.3446, Table VI) levels of risk in this second
spectively), are less readily corrected. Where an un- simulation is lower than in the first one (maximum
acceptable level of risk has been achieved, the com- and minimum values of 0.7508 and 0.1556, respec-
mercial bank must consider the option of closing the tively, Table VI). This is also the case when frequency
branch and opening a new one that can meet the histograms from Figs. 4 and 5 are compared. The fre-
needs of the same population group but, this time, quency histogram for the second simulation (Fig. 5)
with a lower RI. In other words, at the time of open- presents RIs in a smaller number of intervals. That is,
ing the new branch, it is important to find a space with the branch in Simulation 2 obtained RIs belonging
no abandoned or uninhabited surrounding buildings, to [0.3,0.4), [0.4,0.5), and [0.5,0.6) (Fig. 5), while in
located in an area that police response is swift and, at Simulation 1, the results fall within the following
intervals: [0.1,0.2), [0.2,0.3), [0.3,0.4), [0.4,0.5), ertheless, the bank or any commercial bank may find
[0.5,0.6), [0.6,0.7), and [0.7,0.8) (Fig. 4). it useful to employ the complete model, even if doing
Taking these factors into account, it seems rea- so only has a slight impact on the results. Stated an-
sonable to say that, in Simulation 2, there is a reduced other way, the complete model provides a more com-
number of “possible real snapshots” in comparison prehensive definition of each real branch.
with the first one. On the other hand, in Simulation The 24-indicator model was used to estimate, in
2, the RI function leaps as can be seen in Fig. 5. This a deterministic way, the RI of 636 real bank branches
is due to the fact that only a limited number of quali- belonging to ABANCA. They were classified into
tative indicators were treated as probabilistic. three groups depending on the level of risk: (i) low
Regarding the results of this second simulation, it risk (RI ≤ 0.5), (ii) medium risk (0.5 < RI ≤ 0.65),
can be concluded that the level of risk is acceptable, and (iii) high risk (RI > 0.65). A medium or high-
with a mean value under 0.5 (0.4315, Table VI). Nev- risk branch means that the office is far from security
ertheless, it should be noted that the level of risk of excellence, although it fulfils national legislation in
a specific branch can vary considerably (from 0.34 to terms of bank security issues. 567 branches obtained
0.58, minimum and maximum values in Table VI) de- a RI equal to or under 0.5 (low risk), while 54 and
pending on the number of employees and their level 15 branches obtained RIs belonging to the medium
of training, as well as on whether or not they follow and high-risk levels, respectively. An action plan with
guidelines for cash handling and storing. The reader preventive and corrective measures was designed for
should not forget that professional robbers normally the 69 branches with medium and high-risk levels.
study their targets with great care. Consequently, if These measures were classified into four groups de-
the employees of a specific branch do not respect the pending on different priority levels. The aim was to
bank’s procedures by, for instance, failing to place take steps in the short term with the branches with
cash in the storage devices, the robber will see that the highest RIs. Consequently, in the six months fol-
branch as a more attractive target. In the same vein, lowing the assessment, 18 branches reduced their RI
if robbers realize that there are specific moments in considerably. Among the measures carried out were
the day when the number of employees is reduced, relocating storage devices, increasing the number of
they will take advantage of this situation. devices for cash storing, and handling, improving, or
changing communication channels and installing new
security cameras.
5.2. General Results for a Real Bank
At the time of preparing this manuscript, all the
The security personnel of the bank used the 24- branches that have been assessed now present a low
indicator model by means of a software developed by or medium level of risk. In fact, the corrective mea-
researchers from the Technical University of Catalo- sures implemented in some branches meant that they
nia and the University of A Coruña (Technical Uni- could go from a high level to a low level of risk. As
versity of Catalonia, 2016). Once the model is defined previously indicated, an RI between 0.4 and 0.5 is
in the corresponding software, assessing the robbery linked to a high-performing branch for risk of rob-
risk of a specific bank branch is a simple process. The bery.
user only has to collect the real information related On the other hand, it has been found that the in-
to the model indicators. These data are then intro- vestment made to improve the branches with high-
duced into the MIVES software, which automatically est RIs did manage to hinder robberies. In fact, since
provides the risk index of the branch under consider- the improvement plan was carried out, the number
ation. A beta version of the MIVES software as well of raids and attempted robberies has dropped con-
as its user manual can be downloaded from the web- siderably. Furthermore, the model also helped pro-
site of the Technical University of Catalonia (Tech- mote a safety and security culture among bank em-
nical University of Catalonia, 2016). Deterministic ployees; they know about and respect the security
models can be implemented on electronic spread- procedures. Their perception has also changed. They
sheets, but this is not possible when probabilistic are now aware that ABANCA’s priority is to protect
analysis is needed (in this case additional software is people. Since the implementation of the improve-
required). ment plan, no sick leave related to robberies has been
The results obtained with the 24-indicator model taken. Consequently, the trade unions supported this
must be similar to the ones that the final 17-indicator- model, which has had a positive impact on this com-
model can provide, as explained in Section 3.2. Nev- mercial bank.
The application of the model to a real bank storing and handling, among others, were taken
served to confirm its usefulness at the time of clas- in the short term (six months after the as-
sifying a branch according to its risk of robbery. The sessment) to reduce the RI of 18 of the 636
bank security personnel compared all the results with branches.
the real situation in the different bank branches, and • This kind of model is useful in fostering a secu-
found no contradictory results. This served to val- rity and safety culture among employees. It also
idate the model. This work also served to identify serves to change the perception of the employ-
quickly and accurately the measures that most re- ees in terms of bank’s priorities.
duce the risk index. This allows the bank to manage
This model should be regularly updated to take
more efficiently the money spent on improving the
into account new legal requirements as well as secu-
security of its branches. As previously indicated, af-
rity devices. Despite the fact that this model was de-
ter adopting the corrective measures identified with
signed for a specific commercial bank, it can be ap-
the help of the model here presented, the number of
plied to other banks, at least in the same country. In
attempted robberies has decreased significantly.
some specific cases, minor changes may be necessary.
Furthermore, both the methodology and a part of the
6. CONCLUSIONS model can also be used to assess the risk of robbery
for different types of commercial establishments if
In this article, a model for assessing the risk of the corresponding modifications are introduced.
bank robberies was presented with the aim of reduc- Regarding future applications, the model could
ing the risk, the impact on people and prioritizing in- be used to assess the risk of other commercial banks,
vestment to improve security. The model consists of both in Spain and in other countries. This would al-
a wide range of indicators in which the following is- low comparisons to be made. Furthermore, it could
sues are addressed: (i) security features, (ii) employ- also be modified so that it can be applied to other
ees, (iii) operational procedures, (iv) physical envi- commercial establishments.
ronment, and (v) social environment. The model is
based on the MIVES method and it was combined
with the Monte Carlo simulation to take uncertainty ACKNOWLEDGMENTS
into account. Correlations between indicators were The present research work was supported and
also established. Two fictitious but possible sets of funded by ABANCA Corporación Bancaria S.A.
real cases were considered, employing the Monte Funding for open access charge: Universidade da
Carlo simulation method. The model was also used Coruña/CISUG. The authors would also like to
to assess, in a deterministic way, the risk of 636 real thank Prof. Antonio Aguado (Technical University
bank branches belonging to a Spanish commercial of Catalonia, Spain) as well as the anonymous re-
bank (ABANCA). The most important conclusions viewers for some interesting comments that served to
of this work are: improve the article. Two of the authors are employ-
ees of the bank (ABANCA).
• The model provides valuable results, easy for
bank employees to understand. It allows the
REFERENCES
user to identify the weak points of a branch.
• Uncertainty and variability can play a key role. Abraham, B. R., & Baldassaro, P. J. (2001). Leaving robbers bar-
The level of risk for a specific branch can signif- ren. Security Management, 45(2), 42–46.
Alarcon, B., Aguado, A., Manga, R., & Josa, A. (2011). A
icantly vary over time, even if the office is not value function for assessing sustainability: Application to indus-
renovated. trial buildings. Sustainability, 3, 35–50. https://doi.org/10.3390/
• Considerable levels of risk can be achieved if su3010035
Borzycki, M. (2003). Bank robbery in Australia. Australian Insti-
employees do not respect the internal guidelines tute of Criminology, 253, 1–6.
for cash handling and storing. Braga, A. A. (2008). Problem-oriented policing and crime preven-
• The model was validated through its application (2nd edition). New York: Criminal Justice Press.
Büchler, H., & Leineweber, H. (1991). The escape behavior of
tion to the previously alluded bank. It served to bank robbers and circular blockade operations by the police.
identify the bank branches with medium or high In E. Kube & H. U. Störzer (Eds.), Police Research in the
levels of risk. Federal Republic of Germany (pp. 199–208). Berlin, Germany:
Springer. https://doi.org/10.1007/978-3-642-74176-0_16
• Steps like relocating storage devices, installing Cartelle Barros, J. J., Lara Coira, M., de la Cruz López, M. P., &
new security cameras and using devices for cash del Caño Gochi, A. (2015). Assessing the global sustainability
of different electricity generation systems. Energy, 89, 473–489. Federal Bureau of Investigation (FBI). (2006). Bank crime statis-
https://doi.org/10.1016/j.energy.2015.05.110 tics 2006. Washington, DC: FBI.
Cartelle Barros, J. J., Lara Coira, M., de la Cruz López, M. P., & Federal Bureau of Investigation (FBI). (2007). Bank crime statis-
del Caño Gochi, A. (2016). Probabilistic life-cycle cost analysis tics 2007. Washington, DC: FBI.
for renewable and non-renewable power plants. Energy, 112, Federal Bureau of Investigation (FBI). (2008). Bank crime statis-
774–787. https://doi.org/10.1016/j.energy.2016.06.098 tics 2008. Washington, DC: FBI.
Cartelle Barros, J. J., Lara Coira, M., de la Cruz López, M. P., & Federal Bureau of Investigation (FBI). (2009). Bank crime statis-
del Caño Gochi, A. (2018). Sustainability optimisation of shell tics 2009. Washington, DC: FBI.
and tube heat exchanger, using a new integrated methodology. Federal Bureau of Investigation (FBI). (2010). Bank crime statis-
Journal of Cleaner Production, 200, 552–567. https://doi.org/10. tics 2010. Washington, DC: FBI.
1016/j.jclepro.2018.07.266 Federal Bureau of Investigation (FBI). (2011). Bank crime statis-
Cartelle Barros, J. J., Lara Coira, M., de la Cruz López, M. P., & del tics 2011. Washington, DC: FBI.
Caño Gochi, A., & Soares, I. (2020). Probabilistic multicriteria Federal Bureau of Investigation (FBI). (2012). Bank crime statis-
environmental assessment of power plants: A global approach. tics 2012. Washington, DC: FBI.
Applied Energy, 260, 114344. https://doi.org/10.1016/j.apenergy. Federal Bureau of Investigation (FBI). (2013). Bank crime statis-
2019.114344 tics 2013. Washington, DC: FBI.
Casanovas-rubio, M., Pujadas, P., Pardo-bosch, F., Blanco, A., & Federal Bureau of Investigation (FBI). (2014). Bank crime statis-
Aguado, A. (2019). Sustainability assessment of trenches in- tics 2014. Washington, DC: FBI.
cluding the new eco-trench : a multi-criteria decision-making Federal Bureau of Investigation (FBI). (2015). Bank crime statis-
tool. Journal of Cleaner Production, 238, 117957. https://doi.org/ tics 2015. Washington, DC: FBI.
10.1016/j.jclepro.2019.117957 Federal Bureau of Investigation (FBI). (2016). Bank crime statis-
Chitsaz, N., & Banihabib, M. E. (2015). Comparison of different tics 2016. Washington, DC: FBI.
multi criteria decision-making models in prioritizing flood man- Federal Bureau of Investigation (FBI). (2017). Bank crime statis-
agement alternatives. Water Resources Management, 29, 2503– tics 2017. Washington, DC: FBI
2525. https://doi.org/10.1007/s11269-015-0954-6 Federal Bureau of Investigation (FBI). (2018). Bank crime statis-
Converso, D., & Viotti, S. (2014). Post-traumatic stress reaction tics 2018. Washington, DC: FBI.
in a sample of bank employees victims of robbery in the work- Fichera, G. P., Fattori, A., Neri, L., Musti, M., Coggiola, M., &
place: The role of pre-trauma and peri-trauma factors. Medicina Costa, G. (2015). Post-traumatic stress disorder among bank
Del Lavoro, 105(4), 243–254. employee victims of robbery. Occupational Medicine, 65, 283–
Cox, L. A. (2008). Some limitations of “ risk = threat × vulner- 289. https://doi.org/10.1093/occmed/kqu180
ability × consequence ” for risk analysis of terrorist attacks. Fichera, G. P., Sartori, S., & Costa, G. (2009). Post-traumatic Stress
Risk Analysis, 28(6), 1749–1761. https://doi.org/10.1111/j.1539- Disorder following robbery at the workplace: A pilot study on
6924.2008.01142.x 136 pharmacy workers. Medicina Del Lavoro, 100(2), 97–108.
Damenu, T. K., & Beaumont, C. (2017). Analysing information Frans, Ö., Åhs, J., Bihre, E., & Åhs, F. (2018). Distance to threat
security in a bank using soft systems methodology. Information and risk of acute and posttraumatic stress disorder following
and Computer Security, 25(3), 240–258. https://doi.org/10.1108/ bank robbery: a longitudinal study. Psychiatry Research, 267,
ICS-07-2016-0053. 461–466. https://doi.org/10.1016/j.psychres.2018.06.050
de la Cruz, M. P., Castro, A., del Caño, A., Gómez, D., Lara, M., Gandini, A., Garmendia, L., Prieto, I., Alvarez, I., & San-José,
& Cartelle, J. J. (2014). Comprehensive methods for dealing J.-T. (2020). A holistic and multi-stakeholder methodology for
with uncertainty in assessing sustainability part 1: The MIVES- vulnerability assessment of cities to flooding and extreme pre-
Monte Carlo method. In M. S. García-Cascales, J. M. Sánchez- cipitation events. Sustainable Cities and Society, 63, 102437.
Lozano, A. D. Masegosa, & C. Cruz-Corona (Eds.), Soft com- https://doi.org/10.1016/j.scs.2020.102437
puting applications for renewable energy and energy efficiency Gill, M. (2000). Commercial robbery. London, UK: Perpetuity
(pp. 69–106). Hershey, PA: IGI Global. Press.
de la Cruz, M. P., Castro, A., del Caño, A., Gómez, D., Lara, Giorgi, G., Leon Perez, J. M., Montani, F., Courcy, F., & Arcangeli,
M., & Gradaille, G. (2014). Comprehensive methods for deal- G. (2015). Distress and job satisfaction after robbery assaults:
ing with uncertainty in assessing sustainability part 2. In M. S. a longitudinal study. Occupational Medicine, 65, 290–295.
García-Cascales, J. M. Sánchez-Lozano, A. D. Masegosa, & C. https://doi.org/10.1093/occmed/kqv051
Cruz-Corona (Eds.), Soft computing applications for renewable Giorgi, G., Fiz Perez, F. S., D ’ Antonio, A. C., Mucci, N., Ferrero,
energy and energy efficiency (pp. 107–140). Hershey, PA: IGI C., Cupelli, V., & Arcangeli, G. (2015). Psychometric properties
Global. of the impact of event scale-6 in a sample of victims of bank rob-
De Leo, G., Volpini, L., & De Gregorio, E. (2006). La rapina in bery. Psychology Research and Behavior Management, 2015(8),
banca aspetti criminologici, problemi di sicurezza e prospettive 99–104. https://doi.org/10.2147/PRBM.S73901
di prevenzione.pdf. Rassegna Penitenziaria e Criminologica, 2, Gorton, D. (2014). Using incident response trees as a tool for risk
105–119. management of online financial services. Risk Analysis, 34(9),
Dugato, M. (2014). Analyzing bank robbery in Italy. In S. Canep- 1763–1774. https://doi.org/10.1111/risa.12195
pele & F. Calderoni (Eds.), Organized crime, corruption and Guazzoni, C., & Ronsivalle, G. B. (2009). An artificial neural net-
crime prevention (pp. 115–125). Berlin, Germany: Springer. work for bank robbery risk management: The OS.SI.F web on-
European Banking Federation (EBF). (2010). 18th report on bank line tool of the ABI anti-crime department. In E. Corchado,
robberies & other raid types. Brussels, Belgium: EBF. R. Zunino, P. Gastaldo, & A. Herrero (Eds.), Proceedings of
European Banking Federation (EBF). (2011). 19th report on bank the international workshop on computational intelligence in se-
robberies & other raid types. Brussels, Belgium: EBF. curity for information systems CISIS’08 (Vol. 53, pp. 1–10).
Federal Bureau of Investigation (FBI). (2003). Special report: Berlin, Germany: Springer. https://doi.org/10.1007/978-3-540-
Bank robbery in the United States. Washington, DC: FBI. 88181-0_1
Federal Bureau of Investigation (FBI). (2004). Bank crime statis- Gupta, A., Kumar, Y., & Malhotra, S. (2015). Banking secu-
tics 2004. Washington, DC: FBI. rity system using hand gesture recognition. Paper presented
Federal Bureau of Investigation (FBI). (2005). Bank crime statis- in 2015 International Conference on Recent Developments
tics 2005. Washington, DC: FBI. in Control, Automation and Power Engineering (RDCAPE)
(pp. 243–246). Noida, India: IEEE. https://doi.org/10.1109/ Kamphuis, J. H., & Emmelkamp, P. M. G. (1998). Crime-related
RDCAPE.2015.7281403 trauma: Psychological distress in victims of bank robbery. Jour-
Habibi, S., Pons Valladares, O., & Peña, D. (2020). New sustain- nal of Anxiety Disorders, 12(3), 199–208. https://doi.org/10.
ability assessment model for Intelligent Façade Layers when 1016/S0887-6185(98)00009-7
applied to refurbish school buildings skins. Sustainable Energy Khera, N., & Verma, A. (2014). Development of an intelligent
Technologies and Assessments, 42, 100839. https://doi.org/10. system for bank security. Paper presented at 5th International
1016/j.seta.2020.100839 Conference on Confluence 2014: the Next Generation Informa-
Hajkowicz, S., & Collins, K. (2007). A review of multiple criteria tion Technology Summit (pp. 319–322). Noida, India: IEEE.
analysis for water resource planning and Management, Water https://doi.org/10.1109/CONFLUENCE.2014.6949339
Resources Management, 21, 1553–1566. https://doi.org/10.1007/ Kotus, J., Łopatka, K., Czyżewski, A., & Bogdanis, G. (2016).
s11269-006-9112-5 Processing of acoustical data in a multimodal bank operating
Hannan, T. H. (1982). Bank robberies and bank security precau- room surveillance system. Multimedia Tools and Applications,
tions. The Journal of Legal Studies, 11(1), 83–92. https://doi.org/ 75, 10787–10805. https://doi.org/10.1007/s11042-014-2264-z
10.1086/467693 Kou, G., Lu, Y., Peng, Y., & Shi, Y. (2012). Evaluation of classifi-
Hansen, M., Armour, C., & Elklit, A. (2012). Assessing a dyspho- cation algorithms using MCDM and rank correlation. Interna-
ric arousal model of acute stress disorder symptoms in a clin- tional Journal of Information Technology and Decision Making,
ical sample of rape and bank robbery victims. European Jour- 11(1), 197–225. https://doi.org/10.1142/S0219622012500095
nal of Psychotraumatology, 3, 1–10. https://doi.org/10.3402/ejpt. Kube, E. (1988). Preventing bank robbery: Lessons from inter-
v3i0.18201 viewing robbers. Journal of Security Administration, 11(2), 78–
Hansen, M., Armour, C., Shevlin, M., & Elklit, A. (2014). 83.
Investigating the psychological impact of bank robbery: a Lesk, M. (2011). Cybersecurity and economics. IEEE Security
cohort study. Journal of Anxiety Disorders, 28, 454–459. and Privacy Magazine, 9(6), 76–79. https://doi.org/10.1109/MSP.
https://doi.org/10.1016/j.janxdis.2014.04.005 2011.160
Hansen, M., & Elklit, A. (2013). Does acute stress disorder predict Levine, N. (2007). Crime travel demand and bank robberies: Us-
posttraumatic stress disorder following bank robbery? Journal ing CrimeStat III to model bank robbery trips. Social Sci-
of Interpersonal Violence, 28(1), 25–44. https://doi.org/10.1177/ ence Computer Review, 25(2), 239–258. https://doi.org/10.1177/
0886260512448848 0894439306298923
Hansen, M., Hyland, P., & Armour, C. (2016). Does highly symp- Leymann, H. (1985). Somatic and psychological symptoms after
tomatic class membership in the acute phase predict highly the experience of life threatening events: A profile analysis. Vic-
symptomatic classification in victims 6 months after traumatic timology, 10(1–4), 512–538.
exposure? Journal of Anxiety Disorders, 40, 44–51. https://doi. Leymann, H. (1988). Stress reactions after bank robberies: psycho-
org/10.1016/j.janxdis.2016.04.008 logical and psychosomatic reaction patterns. Work and Stress, 2,
Hansen, M., Lasgaard, M., & Elklit, A. (2013). The latent factor 123–132. https://doi.org/10.1080/02678378808259156
structure of acute stress disorder following bank robbery: test- Matthews, R., Pease, C., & Pease, K. (2001). Repeated bank rob-
ing alternative models in light of the pending DSM-5. British bery: theme and variations. In G. Farrell & K. Pease (Eds.), Re-
Journal of Clinical Psychology, 52, 82–91. https://doi.org/10. peat victimization, crime prevention studies (Vol. 12, pp. 153–
1111/bjc.12002 164). Monsey, NY: Criminal Justice Press.
Haran, J. F., & Martin, J. M. (1977). The imprisonment of bank Miller-Burke, J., Attridge, M., & Fass, P. M. (1999). Impact
robbers: The issue of deterrence. Federal Probation, 41(3), 27– of traumatic events and organizational response: A study of
30. bank robberies. Journal of Occupational and Environmen-
Hochstetler, A. (2001). Opportunities and decisions: Interactional tal Medicine, 41(2), 73–83. https://doi.org/10.1097/00043764-
dynamics in robbery and burglary groups. Criminology, 39(3), 199902000-00001
737–764. https://doi.org/10.1111/j.1745-9125.2001.tb00939.x Morrison, S. A., & O’Donnell, I. (1996). An analysis of the
Hole, K. J., Moen, V., & Tjostheim, T. (2006). Case study: On- decision-making practices of armed robbers. National Criminal
line banking security. IEEE Security & Privacy, 4(2), 14–20. Justice, 167532, 1–30.
https://doi.org/10.1109/MSP.2006.36 Mucci, N., Giorgi, G., Perez, J. F., Iavicoli, I., & Arcangeli, G.
Hosseini, S. M. A., Pons, O., & de la Fuente, A. (2018). A combi- (2015). Predictors of trauma in bank employee robbery vic-
nation of the Knapsack algorithm and MIVES for choosing op- tims. Neuropsychiatric Disease and Treatment, 11, 2605–2612.
timal temporary housing site locations: a case study in Tehran. https://doi.org/10.2147/NDT.S88836
International Journal of Disaster Risk Reduction, 27, 265–277. Nugent, S., Burns, D., Wilson, P., & Chappell, D. (1989). Armed
https://doi.org/10.1016/j.ijdrr.2017.10.013 robbery from an offender’ s perspective: Implications for preven-
Hosseini, S. M. A., Yazdani, R., & de la Fuente, A. (2020). Multi- tion. Canberra, Australia: Australian Institute of Criminology.
objective interior design optimization method based on sus- Ozenne, T. (1974). The economics of bank robbery. The Journal of
tainability concepts for post-disaster temporary housing units. Legal Studies, 3(1), 19–51.
Building and Environment, 173, 106742. https://doi.org/10.1016/ Paes-Machado, E., & Nascimento, A. M. (2006). Bank money
j.buildenv.2020.106742 shields: work-related victimisation, moral dilemmas and crisis
Johnston, D. A. (1978). Psychological observations of bank rob- in the bank profession. International Review of Victimology, 13,
bery. American Journal of Psychiatry, 135(11), 1377–1379. 1–25. https://doi.org/10.1177/026975800601300101
https://doi.org/10.1176/ajp.135.11.1377 Pardo-Bosch, F., Aguado, A., & Pino, M. (2019). Holistic model
Jones, C. A. (2002). Victim perspective of bank robbery trauma to analyze and prioritize urban sustainable buildings for public
and recovery. Traumatology, 8(4), 191–204. https://doi.org/10. services. Sustainable Cities and Society, 44, 227–236. https://doi.
1177/153476560200800402 org/10.1016/j.scs.2018.09.028
Jones, I. H., & Jones, A. L. (1998). Psychological consequences of Pons, O., Franquesa, J., & Hosseini, S. M. A. (2019). Integrated
armed hold up. Australian Family Physician, 17(6), 447–450. value model to assess the sustainability of active learning ac-
Josa, I., de la Fuente, A., Casanovas-Rubio, M. del M., Armen- tivities and strategies in architecture lectures for large groups.
gou, J., & Aguado, A. (2021). Sustainability-oriented model to Sustainability, 11, 2917. https://doi.org/10.3390/su11102917
decide on concrete pipeline reinforcement. Sustainability, 13, Pujadas, P., Pardo-Bosch, F., Aguado-Renter, A., & Aguado, A.
3026. https://doi.org/10.3390/su13063026 (2017). MIVES multi-criteria approach for the evaluation, pri-
oritization, and selection of public investment projects. A case
study in the city of Barcelona. Land Use Policy, 64, 29–37. Technical University of Catalonia. (2016). MIVES tool: Method-
https://doi.org/10.1016/j.landusepol.2017.02.014 ology, downloading, error collection procedure and mod-
Reilly, B., Rickman, N., & Witt, R. (2012). Robbing banks: els. Retrieved from https://deca.upc.edu/es/proyectos/mives/
crime does pay - but not very much. Significance, 9(3), 17–21. herramientas
https://doi.org/10.1111/j.1740-9713.2012.00570.x Wang, J. Z. (2002). Bank robberies by an Asian gang: an assess-
Saaty, T. L. (1980). The analytic hierarchy process. New York: Mc- ment of the routine activities theory. International Journal of
Graw Hill. Offender Therapy and Comparative Criminology, 46(5), 555–
Saaty, T. L. (2006). Fundamentals of decision making and priority 568. https://doi.org/10.1177/030662402236740
theory with the analytic hierarchy process. Pittsburg, PA: RWS Weisel, D. L. (2007). Bank robbery. Problem-oriented guides for
Publications. police. Washington, DC: Center for Problem Oriented Policing.
Saaty, T. L., & Vargas, L. G. (2006). Decision making with the ana- Zamani-Sabzi, H., King, J. P., Gard, C. C., & Abudu, S. (2016).
lytic network process (1st ed.). Berlin, Germany: Springer. Statistical and analytical comparison of multi-criteria decision-
Samavati, H. (2004). Bank robbers’ blues. Business Horizons, making techniques under fuzzy environment. Operations Re-
47(1), 59–63. https://doi.org/10.1016/j.bushor.2003.11.009 search Perspectives, 3, 92–117. https://doi.org/10.1016/j.orp.2016.
Samavati, H. (2006). Economics of crime: panel data analysis of 11.001
bank robbery in the United States. Atlantic Economic Journal, Zubizarreta, M., Cuadrado, J., Orbe, A., & García, H. (2019).
34, 455–466. https://doi.org/10.1007/s11293-006-9033-y Modeling the environmental sustainability of timber structures:
Setti, I., van der Velden, P. G., Sommovigo, V., Ferretti, M. S., a case study. Environmental Impact Assessment Review, 78,
Giorgi, G., O’Shea, D., & Argentero, P. (2018). Well-being and 106286. https://doi.org/10.1016/j.eiar.2019.106286
functioning at work following thefts and robberies: a compar- Zubizarreta, M., Ganzarain, J., Cuadrado, J., & Lizarralde, R.
ative study. Frontiers in Psychology, 9, 1–13. https://doi.org/10. (2021). Evaluating disruptive innovation project management
3389/fpsyg.2018.00168 capabilities. Sustainability, 13, 1–22. https://doi.org/10.3390/
Shao, M., Han, Z., Sun, J., Xiao, C., Zhang, S., & Zhao, Y. (2020). su13010001
A review of multi-criteria decision making applications for re-
newable energy site selection. Renewable Energy, 157, 377–403.
https://doi.org/10.1016/j.renene.2020.04.137 SUPPORTING INFORMATION
Sood, A. K., & Enbody, R. J. (2013). The art of cyber bank rob-
bery: stealing your money through insidious attacks. CrossTalk, Additional supporting information may be found on-
9–16. line in the Supporting Information section at the end
Szczodrak, M., & Szwoch, G. (2013). An approach to the detec- of the article.
tion of bank robbery acts employing thermal image analysis. In
Signal processing—Algorithms, architectures, arrangements, and
applications (SPA) (pp. 297–301). Poznan, Poland: IEEE. Supplementary Material

Risk Analysis - 2021 - de La Cruz López - Assessing The Risk of Robbery in Bank Branches To Reduce Impact On Personnel

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Risk Analysis - 2021 - de La Cruz López - Assessing The Risk of Robbery in Bank Branches To Reduce Impact On Personnel

Uploaded by

Copyright:

Available Formats

Risk Analysis, Vol. 42, No. 2, 2022 DOI: 10.1111/risa.

Assessing the risk of robbery in bank branches to reduce

385 0272-4332/22/0100-0385$22.00/1 © 2021 The Authors. Risk Analysis pub-

Field of Application Source

2006); the analytic network process (ANP) (Saaty

40% 1.2. Storage device 50% 1.2.1. Safes:

50% 1.2.2. Cash box:

• Visible from the main service area, Vind = 1

• One safe and one cash box, Vind = 1

25% 1.3.2. Security cameras:

• Main entrance, Vind = 1

Table II. (Continued)

40% 2.2. Response 30% 2.2.1. Staff composition:

70% 2.2.2. Training:

• At least one worker without training, Vind = 1

5% 3.2.2. Amount of money:

• Level 3 (less than 100,000 €), Vind = 0

12% 4. Physical 100% 4.1. Physical 45% 4.1.1. Location:

35% 4.1.2. Surrounding buildings:

• Abandoned building, upper, or lower floor, Vind = 1

20% 4.1.3. Escape route:

• Easy and close to the branch (vehicle), Vind = 1

Table II. (Continued)

50% 5.1.2. Number of robberies in the branch (5 years):

• More than 3, Vind = 1

30% 5.2. Police 100% 5.2.1. Response time

ing value functions. Moreover, throughout the use of

1.1.1. Communication channels:

1.2.2 Cash box:

1.3.1. Devices for cash storing and handling:

1.3.2. Security cameras:

2.2.1. Staff composition:

3.1.1. Cash storing during noncommercial hours:

Table III. (Continued)

3.2.1. Cash handling during commercial hours:

4.1.2. Surrounding buildings:

4.1.3. Escape route:

5.1.1. Number of robberies in the province:

5.1.2. Number of robberies in the branch (5 years):

5.2.1. Response time

Simulation Distribution Function Distribution Parameters

Minimum Mode Maximum

2 Employees 3 Employees 4 Employees

Simul. 2 Discrete 0.3 0.4 0.3

Simulation Statistical parameters

Minimum Maximum Mean Modal Interval (MI) Frequency of MI SD

Simul. 1 0.1556 0.7508 0.4180 [0.3,0.4) 37.32% 0.087

You might also like