How AI can reveal corporate tax avoidance

Are the phrases utilized in annual stories a key to unlocking the secrets and techniques of company tax avoidance?

In terms of recognizing company tax dodgers, phrases could be as helpful as numbers. Current analysis from Texas McCombs finds that cautious studying of textual content can provide new insights into how firms are attempting to keep away from taxes—actions that might not be obvious from monetary numbers alone.

Dean and Accounting Professor Lillian Mills and her co-author, Kelvin Legislation of Nanyang Technological University, examined 18 years’ price of U.S. multinational firms’ annual stories, ones that mentioned their enterprise actions in international international locations, together with tax havens. The researchers coated a complete of 183,061 stories.

The group used pure language processing (NLP) to investigate the textual content and determine patterns and phrase selections which may reveal what sort of actions firms have been conducting in tax havens. The pc evaluation uncovered clues about these actions.

For instance, suppose a U.S. pharmaceutical firm has developed a profitable drug for treating coronary heart illness, producing a excessive revenue margin. The corporate owns mental property (IP) for the particular formulation of the drug and signifies it has “established a subsidiary in Panama to handle manufacturing and production,” utilizing the patented formulation. By routing income from the sale of the guts illness drug by using IP in a rustic identified for low tax charges, the corporate is ready to pay decrease taxes by the subsidiary within the tax haven.

The phrase “manufacturing” is certainly one of about 80 phrases the pc seems for to recommend operations that is likely to be avoiding taxes. Others embrace “purchasing,” “importing,” “warehouses,” and “distributors.”

Though there isn’t any positive strategy to detect all cases of tax avoidance, Mills says, shut consideration to phrase selections in an annual report can reveal a number of sorts of insights that numbers won’t:

New metrics. A brand new set of measures within the research assesses not solely whether or not an organization has a subsidiary in a tax haven nation, however whether or not it is an energetic subsidiary. The brand new measures are 3 times as efficient as present ones for predicting that an organization is avoiding taxes.

Undisclosed operations. Machine studying methods can determine firms which will have tax haven operations however don’t disclose them in annual stories.

Greater tax avoidance. Nondisclosers flagged by machine learning have decrease efficient tax charges than different firms.

“Using AI to analyze text data could be a powerful tool for both regulators and investors to detect corporate tax avoidance,” Mills says.

“That information could especially help regulators other than the IRS, who don’t have access to companies’ tax returns. It could guide them in looking at publicly available data to find companies that might be using abusive profit-shifting strategies in tax havens.”