Ribbon

Sunday, October 26, 2014

The IRS and Data Science




Today, the NYT publishes an article about the IRS seizing bank accounts on suspicion.
In short, the IRS intends to battle drug cartels by flagging accounts with a suspicious activity. The quote from the officer gives us an insight about their methodology:


based on my training and experience, the pattern “is consistent with structuring.”
In other words, Drug Cartels have a higher propensity to make deposits below $10K (legal ceiling before the bank reports it by law) thus the IRS flagged all of the accounts fitting this criterion.

However, the stats reported by Institute for Justice are reported to be mediocre:
Only one in five was prosecuted as a criminal structuring case
 If innocent, the victims are left to prove it with all the red-tape it involves.


The problem:


This method simply does not work well for the following reasons:
1- A False Positive (Seizing an account that did not belong to Drug Organization) is very costly, from a reputation perspective (making it to the headlines as proved today) and from a victim's perspective (with no cash, a company will likely go bankrupt). In this case, False Positive are plentiful (75% of the cases flagged)! This low performance could have been identified (and even back-tested) before scaling it up to the entire country
2- In addition, the IRS did not have a good remediation process in place

Another way to phrase it is that this pattern is not a sufficient to prove a relationship with drugs.

The solution:


In brief, the IRS should hire Data Scientists. It would have appeared that, in those challenging "flagging" problems, it makes sense to focus on Relevance (out of 100 accounts flagged, how many are actually owned by Drug Cartels) and less on total size of the assets seized. 
Nothing is more heartbreaking than a Pop's and Mom's business running into cash problems because of a undeserved seizure. And that surely would trigger national outrage reprimanding Big Government crushing the little people,
The model should achieve 95%+ relevance in order to stay credible. A long way from the 20% reported.


No comments:

Post a Comment