Assessing practicalities of Benford’s Law - A study of the law’s potential to detect fraud in transactional data
(2021) NEKH02 20202Department of Economics
- Abstract
- Abstract
In modern anti-money laundering operations, data analysis plays a vital role. One method of detecting fraudulent data is Benford’s law, which predicts the distribution of the first significant digits in logarithmically distributed data. Deviation from Benford’s law in data where it should be present might indicate manipulated data.
We investigate the Chi2 and Kolmogorov-Smirnov tests’ properties to see how their empirical sizes and test powers are affected by different sample and variance sizes. With this knowledge, we set out to evaluate the reliability in common methods of testing data for conformity to Benford law. The included methods are graphical analysis, comparison of statistical moments and employment of the mentioned... (More) - Abstract
In modern anti-money laundering operations, data analysis plays a vital role. One method of detecting fraudulent data is Benford’s law, which predicts the distribution of the first significant digits in logarithmically distributed data. Deviation from Benford’s law in data where it should be present might indicate manipulated data.
We investigate the Chi2 and Kolmogorov-Smirnov tests’ properties to see how their empirical sizes and test powers are affected by different sample and variance sizes. With this knowledge, we set out to evaluate the reliability in common methods of testing data for conformity to Benford law. The included methods are graphical analysis, comparison of statistical moments and employment of the mentioned statistical tests on transactional data sets; one with legitimate data and another including fraudulent activity.
Based on our statistical tests results, we conclude that the variance size does not play a significant role when testing data for Benford conformity. Out of the two, the most reliable statistical test is the Chi2 test since its test power is comparably much greater than the Kolmogorov-Smirnov test.
We conclude that Benford’s law has a place in anti-money laundering processes for transactional data since the law was proven to be reliable when examining the data sets as it was able to correctly discern between the legitimate and fraudulent data sets. However, one has to be careful of trade patterns within data sets to avoid misleading results. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9037415
- author
- Wennberg, Jacob LU and Rosencrantz Ollén, Esaias LU
- supervisor
- organization
- course
- NEKH02 20202
- year
- 2021
- type
- M2 - Bachelor Degree
- subject
- keywords
- Benford, Benford's Law, AML, Fraud, Kolmogorov-Smirnov
- language
- English
- id
- 9037415
- date added to LUP
- 2021-03-11 11:56:01
- date last changed
- 2021-03-11 11:56:01
@misc{9037415,
abstract = {{Abstract
In modern anti-money laundering operations, data analysis plays a vital role. One method of detecting fraudulent data is Benford’s law, which predicts the distribution of the first significant digits in logarithmically distributed data. Deviation from Benford’s law in data where it should be present might indicate manipulated data.
We investigate the Chi2 and Kolmogorov-Smirnov tests’ properties to see how their empirical sizes and test powers are affected by different sample and variance sizes. With this knowledge, we set out to evaluate the reliability in common methods of testing data for conformity to Benford law. The included methods are graphical analysis, comparison of statistical moments and employment of the mentioned statistical tests on transactional data sets; one with legitimate data and another including fraudulent activity.
Based on our statistical tests results, we conclude that the variance size does not play a significant role when testing data for Benford conformity. Out of the two, the most reliable statistical test is the Chi2 test since its test power is comparably much greater than the Kolmogorov-Smirnov test.
We conclude that Benford’s law has a place in anti-money laundering processes for transactional data since the law was proven to be reliable when examining the data sets as it was able to correctly discern between the legitimate and fraudulent data sets. However, one has to be careful of trade patterns within data sets to avoid misleading results.}},
author = {{Wennberg, Jacob and Rosencrantz Ollén, Esaias}},
language = {{eng}},
note = {{Student Paper}},
title = {{Assessing practicalities of Benford’s Law - A study of the law’s potential to detect fraud in transactional data}},
year = {{2021}},
}