Journal of Financial Risk Management
2013. Vol.2, No.1, 29-31
Published Online March 2013 in SciRes ( DOI:10.4236/jfrm.2013.21004
Assessing Money Laundering Risk of Financial Institutions with
AHP: Supervisory Perspective
Ke Jia1,2, Xi Zhao1, Ling Zhang3
1School of Management, Tianjin University, Tianjin, China
2Anti-Money Laundering Section, The People’s Bank of China, Tianjin Branch, Tianjin, China
3Human Resources Department, Tianjin University, Tianjin, China
Received January 6th, 2013; r evised February 7th, 2013; accepted February 14th, 2013
This paper proposed a risk assessment model with which supervisory authorities can calculate the money
laundering risk (MLR) level of financial institutions and make comparisons among multiple institutions.
The model is based on the Analytic Hierarchy Process (AHP) and decomposes MLR into two second-tier
criteria, i.e. Inherent Risk & Control Risk. AHP pair wise comparisons made by the experts from various
fields are processed through AHP software to get the weight of each factor. Using this model, MLR of
each financial institution could be obtained and certain comparison among them could be carried out.
Keywords: Money Laundering Risk; Assessment; AHP
Money laundering risk (MLR) is newly recognized as a serious
risk endangering the financial sector as well as the whole society ,
and is drawing increasing attention in recent decades on both
regulation and supervision. (Fer werda, Kattenberg, Chang, Unger,
Groot, & Bikker, 2013; Stoke s, 2012; Kishor & Lescuyer, 2012)
To appropriately apply the risk-based approach recommended in
International Standards on Combating Money Laundering and
the Financing of Terrorism & Proliferation by The Financial
Action Task For ce on Money Laundering ( FATF) and efficiently
allocate supervisory resources, national supervisory authorities
need to accurately assess the MLR levels of financial institutions.
MLR of an institution could be affecte d by many facto rs, in-
cluding institution size, internal rules, management attitude, and
so on. In China, the assessment of MLR are mostly carried out
by certain reviewers grouped with supervisors and specialists
simply giving marks considering some factors (Cai & Liu,
2011). However, as risk factors are distinct from each other in
their natures and weights (Wang & Yang, 2007), an overall ac-
curate assessment could not be obtained using this method,
consequently it is hard to make comparison within institutions.
Given that reviewers could not only raise the examining fac-
tors, but also point out the inherent relationship of these factors,
which could then be analysed using Analytic Hierarchy Process
(Saaty, 1990), the weight of each factors as well as reasonable
marks could be obtained.
Locating MLR factors and building-up MLR structure have
been challenging assessors and researchers in the worldwide in
that the elements composing MLR are complicated (IIROC,
2010). This research created a MLR assessment model which
enables reviewers to evaluate and compare the MLRs of finan-
cial institutions. The core task is to find the most significant
risk factors and establish a logical MLR assessing model.
AHP theory was proposed by Thomas. L. Saaty in 1970s, by
which complex issues can be structured and analyzed by hier-
archical division, and subjective decision according to objective
conclusions would be made. In this analysis, AHP software
with version 0.5.2 was used to obtain the weights of index
Decomposing MLR into a Hierarchy of Factors
In reality, during the process of assessment, supervisory and
management department always divide MLR (A) of a financial
institution into two components, Inherent Risk and Control
Risk, which could be deemed B1 and B2, respectively, as the
second level of this AHP model.
The hierarchical structure of MLR is shown as Figure 1.
Figure 1.
MLR structure.
Copyright © 2013 SciRes. 29
Inherent Risk (IR, B1)
IR is the susceptibility of a financial institution to money
laundering occurred given inherent and environmental charac-
teristics, but without regard to the internal control structure. IR
comprises a number of elements among which the following
three are the most significant.
The size of the institution (C1). A multi-national bank has a
higher possibility of being misused in laundering money
than a local saving bank (Reserve Bank of New Zealand,
2011). Although the measurements of institution size are
various (e.g. by asset, capital, revenue, profit, employee
number or branch number, etc.), the number of customers is
the most relevant indicator in analyzing the interaction be-
tween size and MLR of an institution because all money
launderings are eventually committed by “customers”, and
thus could be used here to define the size of institution.
The geographic location of the institution (C2). This ele-
ment actually concerns where the customers come from. In-
stitutions operating in the regions with high crime rate
would face more potential money-launderers and thus have
higher MLR (Federal Financial Institutions Examination
Council, 2010).
The business nature of the institution (C3). Institutions with
high proportion of cash deposit or withdrawal, cross-border
wire transfer and non-face-to-face businesses are normally
more vulnerable to money laundering. (Council of Europe,
Control Risk (CR, B2)
CR is the risk that money laundering may occur and not be
prevented or detected on a timely basis by the internal control
structure of the institution. CR is determined by the factors
inside an institution and can be controlled by the institution.
This paper identified the following seven fundamental factors
which directly affect CR level and from which other inside
factors are derived (Ma, 2009).
Management attitude and knowledge (C4). Reviewers can
assess the senior executives’ attitude and knowledge about
AML by interviewing the executives as well as the em-
ployees or by checking the written responsibilities of the
Procedures and measures (C5). Reviewers can assess the
validity of the AML procedures and measures in an institu-
tion by off-site reviews.
Computer system (C6). The two core roles that the computer
system is expected to play in the AML structure of an insti-
tution are storing customer identification information and
transaction records and analyzing abnormal transactions.
On-site test is needed to assess the efficiency of the AML
computer system in an institution.
Resources allocated (C7). The resource allocated in AML
can be measured by the total working hours of all AML
staff in the institution or the amount of funds spent on
Performance of customer due diligence (C8). On-site in-
spection is needed to assess whether the performance of
customer due diligence regulatory requirements or internal
procedures are fully implemented within an institution, in-
cluding identifying and verifying the identity of the cus-
tomer and the beneficial owner, recording the basic identity
information of the customer, and so on.
Performance of suspicious transactions report (C9). On-site
inspection is needed to assess whether STR regulatory re-
quirements or internal procedures are fully implemented
within an institution, including analyzing abnormal transac-
tions, filing reports and making them to the financial intel-
ligence unit.
Trainings (C10). To be assessed by interview or examina-
Making Pair-Wise Comparisons and Obtaining the
Judgmental Ma t ri x
After building AHP model, the priorities have been decided.
Elements are compared pair-wise and judgments on compara-
tive attractiveness of elements are captured using the traditional
9 rating scale, with 9 indicating “extreme importance”, 7 indi-
cating “very strong or demonstrated importance”, 5 indicating
“strong or essential importance”, 3 indicating “fairly impor-
tance”, 1 indicating “equal importance” when give the intensity
of importance. Scores of 2, 4, 6, 8 demonstrate intermediate
values and reciprocals show inverse comparison.
16 experts were invited to give the relative importance, and
for the convenience of calculation, the average value is round
Results and Discussion
Pair wise comparisons are carried out with AHP software
and the result is shown as Table 1.
As indicated by the table, with less one third contributed by
IR (B1, 30.2%) and most proportion determined by CR (B2,
69.8%), the ML is basically “controllable” provided that the
institution has a strong internal control system. Reviewers
should thus focus more attentions on the CR control of a finan-
cial institution. Regarding the lowest hierarchy of factors, the
primary task for a financial institution in mitigating MLR is to
strictly conduct Performance of customer due diligence (C8,
19.1%) and Performance of suspicious transactions report (C9,
17.3%) measures, and supports from management (C4, 9.4%) is
also considerable important followed by valid internal rules (C5,
9.1%). Although not controlled by the AML arrangement of the
institution, the size of the institution also plays a significant role
(C1, 15.9%) in determining the MLR of the institution.
Table 1.
The weights of MLR factors.
hierarchy 2nd
hierarchy Weights
to 1st 3rd
hierarchy Weights
to 2nd Weights
to 1st
C1 0.525 0.159
C2 0.200 0.060 B1 0.302
C3 0.275 0.083
C4 0.134 0.094
C5 0.131 0.091
C6 0.072 0.050
C7 0.074 0.052
C8 0.273 0.191
C9 0.248 0.173
B2 0.698
C10 0.068 0.047
Copyright © 2013 SciRes.
Copyright © 2013 SciRes. 31
Reviewers or assessors can mark a financial institution on
each factor, multiply the marks by the weights of factors in
Table 1 and then add the products up to obtain the final
weighted MRL mark of the institution. By comparing the
weighted MLR mark of one institution with others, ranking of
institutions could be obtained.
In this study, MLR was decomposed into a hierarchy of ele-
ments whose weights are computed by AHP. The MLR struc-
ture established and the weights of the elements obtained en-
able reviewers to calculate the MLR level of a financial institu-
tion and guide financial institutions to manage their MLR to
some extent. The MLR in financial sector, however, constantly
changes as a result of the development in society and economy,
which requires a dynamic MLR assessment model. Adjustment
on the MLR structure and the weights of risk elements should
be on an ongoing basis.
Ling Zhang thanks the financial support from The Ministry
of education of Humanities and Social Science Research Fund
Plan/Youth Fund/Self-financing project (11YJC870036) as well
as Open Fund IT2012006 of the ISTIC-Thomson Reuters Joint
Lab for Scientometrics Research.
Cai, Y. L., & Liu, Z. M. (2011). Establishment of Chinese anti money
laundering supervision mode drawing on the experience of UK & US.
Research on China’s anti-money laundering (pp. 374-376). Beijing:
China Financial Publishing House Press.
Council of Europe (2010). Guidelines on bank risk analysis aimed at
preventing money laundering an d t errorism financing.
Federal Financial Institutions Examination Council (2010). Bank se-
crecy act/anti-money laundering examination manual.
Ferwerda, J., Kattenberg, M., Chang, H.-H., Unger, B., Groot, L., &
Bikker, J. A. (2013). Gravity models of trade-based money launder-
ing. Applied Econom ic s , 45, 3170-3182.
IIROC (Investment Industry Regulatory Organization of Canada).
(2010). Anti-Money Laundering Com pli an ce Gu id a nc e. Toronto.
Kishor, N., & Lescuyer, G. (2012). Controlling illegal logging in do-
mestic and international markets by harnessing multi-level govern-
ance opportunities. International Journal of the Commons, 6, 255-
Ma, J. (2009). Financial institutions anti-money laundering practical
manual (pp. 268-276). Beijing: China Financial Publishing House
Reserve Bank of New Zealand (2011). Sector risk assessment for reg-
istered banks, non-ba nk de po si t t ak e rs, and life insurers.
Saaty, T. L. (1990). How to make a decision: The analytic hierarchy
process. European Journal of Ope ra tional Research, 48, 9-26.
Stokes, T. (2012). Virtual money laundering: The case of Bitcoin and
the Linden dollar. Information and Communications Technology Law,
21, 221-236. doi:10.1080/13600834.2012.744225
Wang, S. N., & Yang, J. G. (2007). A money laundering risk evaluation
method based on decision tree. The 6th International Conference on
Machine Learning and Cybernetics, 1, 283-286.