Log in / Register
Home arrow Business & Finance arrow Fraud and fraud detection
< Prev   CONTENTS   Next >


We can check to see if the data set conforms to Benford's Law. For Benford's Law, we will test on where TRANSACTION_TYPE are P or payments and on positive values.

Benford's Law First Digit Test for Payment Amounts

FIGURE 8.4 Benford's Law First Digit Test for Payment Amounts

For the high-level first digit test, the resulting graph shows high conformity in Figure 8.4. We can then select the first two digits for a better precision analysis.

There is acceptable conformity in Figure 8.5. While it is not unusual for the first two digits test for all the digit pairs not to meet the expected count, on the whole the data set conforms. However, it would be prudent to examine some of the contents of the highly suspicious and suspicious numbers.

For the last two digits test shown in Figure 8.6, there is close conformity for almost all the numbers with the exception of 0 and 50, which in reality is normal, especially for numbers ending with zero.

Since the data set as a whole conforms we can accept the Benford's Law results or we can apply the tests on specific individual agencies. If a specific agency is selected for a

Benford's Law First Two Digits Test on Payment Amounts

FIGURE 8.5 Benford's Law First Two Digits Test on Payment Amounts

Benford's Law Last Two Digits Test

FIGURE 8.6 Benford's Law Last Two Digits Test

detail review, then all the general tests can be reapplied just to the payments associated with that specific agency.


This data set contains many high-dollar-payment amounts. There are 1,590 records where the paid amounts were $1 million or more. To see these high-dollar records, we simply index the PAYMENT_AMOUNT field in descending order. The largest amount of almost $49 million was paid to a Protected Information vendor name. Many of the top amounts exceeding $20 million were paid to JPMorgan Chase Bank for the purchases of securities and other investments. It is unlikely in our general review that these high-dollar amounts would be of significant interest, unless we were auditing investment transactions or focusing on payments to banks.

What we are interested in are unusually high payments to individual vendors. We could employ the Top Records feature of IDEA to extract a user-specified number of the top-most amounts for each vendor. However, we would have to examine all the vendors. There are many vendors in this payment ifile. Our objective is to review, by vendors, where the largest amount paid is significantly higher than the next highest amount. High differences can signify errors or fraud.

To make reviewing details easier, in the resulting Relative Size Factor test file, we use the Defined Action Field feature on the VENDOR_NAME linked to the payment file. We can then just click on any of the vendor names to obtain an extraction preview for the selected vendor. As an example, we will look at the transactions of the incredibly high RSF ratio of the first vendor displayed. All records had payments of $3.00 each while the largest records showed a payment of $42,500.00, shown in Figure 8.7.

Relative Size Factor Test of Payments by Vendor Names

FIGURE 8.7 Relative Size Factor Test of Payments by Vendor Names

You can then select other high-RSF ratios, such as Central Rural Electric in record number 31, to examine in more detail. There were 107 transactions with average amounts of $156.88 if the largest number of $216,805.24 was excluded.

Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
Business & Finance
Computer Science
Language & Literature
Political science