In a first, the Delhi government is all set to use a machine-learning tool, developed by two US-based Indian researchers, to identify bogus firms propped up by traders in the city to evade taxes. Initial estimates suggest that the public exchequer stands to gain up to Rs 300 crore annually upon plugging the leaks.
University of California, Berkeley, Associate Professor Dr Aprajit Mahajan and postdoctoral scholar Dr Shekhar Mittal are developing the tool based on scrutiny of a vast dataset of VAT (Value Added Tax) returns registered in Delhi between 2012-17.
How the tool will help identify tax evasions
Tax Violations are mostly in the form of traders generating bills of illegal transactions with firms that exist only on paper to decrease tax liability. DDC chairperson Jasmine Shah said the study's objective was to develop a tool which would obviate the need for random tax raids, as the application, over time, will throw up a shortlist of potential evaders based on a scrutiny of their tax filing records over a long period. “Merely getting shortlisted won’t cast a shadow on a firm; final action will be taken based on subsequent ground work... It may also turn out that a shortlisted firm was actually not evading taxes. That data will be fed into the tool, which will help increase its accuracy,” Shah said.
The researchers and the Delhi government, which commissioned the study, have concluded that since traders use similar means even to evade GST (Goods and Services Tax), which has replaced VAT, the “future iterations of machine learning will build on GST data,” Jasmine Shah, vice-chairman of Delhi Dialogue and Development Commission (DDC), told The Indian Express.
The researchers are likely to officially submit their first set of findings to the government this month. In a paper, ‘Who is Bogus? Using One-Sided Labels to Identify Fraudulent Firms from Tax Returns’, the researchers contended that their work is the first-ever systematic study of tax evasion in an economy with weak compliance.
“Our results indicate that by using our tool, the tax administration can prevent fraud up to $15-45 million. Given that such data exists in many tax jurisdictions and that anecdotal evidence suggests that such false paper trails are a common problem, our work should have high policy relevance both within India and elsewhere,” says the paper.
The tax department used to carry out surprise inspections and raids to catch defaulters or traders resorting to unfair means to avoid paying taxes. However, in September 2018, the Delhi government disbanded a 25-member team of tax inspectors citing rampant corruption in the entire recovery process and harassment meted out to traders.
“A tax authority usually determines the existence of bogus firms by first filtering down based on a few preliminary indicators, and then undertaking physical inspections that verify if a firm is bogus or legitimate. Given the authority’s limited resources, these inspections are done sporadically. While the true revenue implication of bogus firms is unknown, our conversations with tax officials in Delhi suggest that it might be considerable ($300 million in Delhi alone is a commonly mentioned figure),” the researchers observed in the paper.