Support Us

Who’s Bogus? Catching Fraudulent Firms in Delhi

Institutions & Governance India

Photo credit: Adobe Stock Images

Policy Context

A common VAT tax evasion strategy in low ​compliance​ environments ​involves “bogus” firms set up for the purposes of fraud — they generate false paper trails and sell fake receipts to real firms, so that those “costs” can be subtracted out of what the firms owe the tax authority in VAT taxes. While the scope of revenue loss from this type of fraudulent activity is inherently hard to track, officials in Delhi suggest that the Indian government may be losing as much as $300 million annually. ​Identifying these fraudulent firms would be an important step to helping recover this lost tax revenue, but locating them typically requires visiting the physical site to see if the firm is conducting legitimate business. However, the tax authority cannot physically visit all of the firms that they have flagged as potentially fraudulent.

Study Design

Researchers designed a machine learning tool to more efficiently identify these fraudulent firms in Delhi. The machine learning algorithm is trained on data from firms the Delhi tax officials have already audited, including whether they were found to be fraudulent. The authors then ran a proof of concept test to compare the algorithm to the status quo inspection decisions of Delhi tax officials.

Results and Policy Implications

The researchers found that if the tax administration had used the algorithm to predict which firms were fraudulent to target their physical inspections, they could have prevented enough VAT tax evasion to increase revenue by an estimated $15 – 45 million USD. Given the successful proof of concept, the researchers have received active expressions of interest from the tax authorities to adapt this tool to deploy in two Indian states.

For more information about this research, read the working paper here.

Researchers
Partners
  • The Delhi Revenue Authority
Timeline

2017 — ongoing

Share Now

Copyright 2024. All Rights Reserved

Design & Dev by Wonderland Collective