On a date with data-mining

Plastic buckets for Rs 70,000 a unit. Razor blades for Rs 8,000 a piece. And diamonds for Rs 900 a carat. These are not fantasy prices. These are actual prices used by people to launder money camouflaged as ‘international trade’ through exorbitantly highvalued import invoices and ridiculously low-valued export invoices. As GoI steps up its efforts to root out black money from the economy, it needs to create a team of experts that can put together disparate data sets and analyse them cleverly to detect money laundering and tax evasion. Money launderers usually exploit inadequate coordination between various agencies of the government to carry out their activities. To offset this, advances in big-data analysis and data mining must be exploited. This effort requires two important changes in mindset. First, the reliance on anecdotes and ‘theories’ postulated by bureaucrats using their ‘personal experiences’ need to be shelved.

It is dangerous to extrapolate to all of Bureaucrat A’s personal experiences serving in, say, the Kerala cadre or Bureaucrat B’s experiences serving in, say, the Jammu and Kashmir cadre. As Nobel-winning research in behavioural economics by Daniel Kahneman demonstrates, the personal experiences of even the most rational individuals shape their ‘theories of life’. Such small sample evidence can’t substitute for systematic empirical evidence. As Sherlock Holmes said, one should theorise only after all the data are in. Second, information as power acts as a double-edged sword. Information-sharing becomes the victim of jostling for power among the various departments in a bureaucracy. Thus, one department won’t share data with another unless it is forced to do so.

To circumvent this, GoI needs to set up a commission with the mandate and the personnel to demand data from any department; put the disparate data sets together; and analyse them to gain insights. Take the use of data mining to detect money laundering. The use of international trade to move money, undetected, from one country to another is one of the oldest tricks in the book. Assume a criminal wants to launder Rs 100 crore to Pakistan.

He would need to have a Pakistani exporter to collude on the transaction. First, the exporter would purchase 1,000 carpets for Rs 1,000 each, thus costing Rs 10 lakh. Second, he would export the 1,000 carpets to the Indian importer for Rs 10 lakh a carpet, resulting in atotal invoice of Rs 100 crore. Third, the Indian importer receives the 1,000 carpets worth Rs 10 lakh but pays the Pakistani exporter Rs 100 crore. As a result, Rs 100 crore has been moved to Pakistan for a cost of Rs 10 lakh.

An alternative method used is through the undervaluation of domestic exports. Research of money laundering in the US indicates a majority of the money laundered is through undervalued exports. This is the preferred option for two reasons. One, most governments don’t adequately monitor their export transactions. Two, it allows the launderer to avoid the use of financial institutions that may be monitored by government agencies. Detecting such transactions requires careful and systematic analysis of data. This is because the local tax official may be aware of the nature of this business and may turn a blind eye in return for a sizeable bribe. The money launderer converts his illegal money into products by purchasing products for cash at the product’s market price.

These products are then exported to a foreign colluding importer at below market prices. The importer receives the undervalued exports and resells them in the market at the real prices that reflect their true value. To detect such transactions, import-export data at a daily frequency need to be carefully assembled. The quality of the data needs to be checked by undertaking statistical and econometric tests for missing fields, intentional fabrication and modification of data, etc. Once such data is assembled, suspicious transactions can be flagged using prices that are very high or very low by examining the statistical distribution of prices.

Such flags can then be used for welldirected audits. Advances in machine learning and analysis of big data provide several opportunities that the government can use to catch money launderers. What is important is for political and bureaucratic will to exploit these opportunities.