The power of data
For many tax authorities, the digital journey starts with a re-evaluation.
The Australian Taxation Office’s Smarter Data Program, established in July 2015, put one office in charge of data management, analytics, intelligence and risk assessment, and serves as a platform to test new technologies and data management techniques.
But authorities also want to quickly get to the first steps in boosting collections, and they do it by using data to improve the audit process.
A database of tax returns can help to quickly spot the atypical ones — a corporation with a significantly higher type or rate of deduction than competitors, for example.
A recent OECD survey found that of 16 member countries participating, 15 use advanced analytics to prioritize audits.
‘A closer look’
“It’s no longer just a person peeling through the paperwork and deciding who to audit,” says EY Americas Tax Policy Leader Cathy Koch. “It’s an electronic review of the data to flag companies outside the norm. Then they’ll take a closer look.”
The next step for tax authorities has typically been to harmonize digital platforms across government agencies, allowing tax authorities to use data from all of them to spot trends and patterns.
This is useful to predict which taxpayers are most likely to underpay, miscalculate or fail to file a return at all.
Tax offices then use these tools to warn potential problem taxpayers through advance communication such as messages through social media, and influence behavior through targeted policy adjustments.
“Most tax administrations are trying to expand or broaden their thinking far beyond what they need to have in place to deal with a taxpayer who hasn’t filed or who hasn’t paid,” says Thomas Brandt, Head of the OECD’s Tax Administration unit.
“They’re thinking about whether a taxpayer would be more likely to respond to a letter or some other type of contact, and they’re thinking about the right timing to do it.”
The Canada Revenue Agency, for example, has moved from a prediction model based on a single tax cycle to a ranking system in which taxpayers are evaluated digitally on an ongoing basis for the risk of noncompliance.
Risk-scoring corporations and high net worth individuals has resulted in a 33% increase in revenue, from C$9.4 billion in the 2012–13 fiscal year to C$12.5 billion in 2015–16, the agency says.
Spotting problems
Computer models can also be designed to spot specific types of issues within a tax return and supporting information:
For corporate taxpayers, it’s important to mimic this process and spot in advance what elements of their tax returns could draw attention, says EY’s Sanger.
Tax authorities can dig deeper into tax returns using artificial intelligence.
In the US, the IRS used the work of researchers at the Massachusetts Institute of Technology and a Washington-based company to boost its oversight of partnerships, one of the more common forms of flow-through tax entities, which corporations have used to avoid paying tax twice on the same income.
The researchers found a mushrooming of this structure from 2005 to 2015 and an estimated US$91 billion in underreported income as a result.
They designed an algorithm to analyze partnerships and simulate the transactions between them.
They found that specific combinations of partnership structure and transaction type often indicate incorrect deductions, suggesting that any tax return containing them is a good audit candidate.
This information was absorbed into a database managed by the Office of Tax Shelter Analysis, which was created in 2000.
As of late 2015, the team behind this innovation was exploring whether other areas of tax law are suitable for this artificial-intelligence approach.
Corporate tax concerns
For corporate payers, the focus on value-added tax (VAT) is particularly relevant.
Authorities in France, Mexico, the Netherlands, Norway and elsewhere have all reported building analytical models to uncover problem claims.
Singapore, Malaysia and New Zealand use social network analysis — the visual display of connections between entities — to detect VAT carousel fraud, which exploits VAT-free treatment of cross jurisdictional sales.
Social network analysis can spot links between taxpayers and joint bank accounts, or shared telephone numbers.
This data-driven approach is changing the profile of employees in tax administrations.
In October 2015, the Netherlands’ tax authority, Belastingdienst, simultaneously announced 5,000 redundancies along with 1,500 new postings for data scientists.
Information exchange
Alongside this ongoing process of bulking up intelligence and analysis capabilities, tax authorities are increasingly sharing information about specific taxpayers and tactics to understand the data with their peers worldwide.
There is not a long history of sharing at this level between tax authorities, in part because hard copy data is not as easily shared as electronic versions. Going digital makes it possible. “If you submit something electronically, it will be shared in real time with many other governments,” EY’s Thomas says.
While countries can act on their own initiative — the Canada Revenue Agency reports sharing based on its double-taxation treaties — the OECD is also a leading facilitator of exchange. Its Forum on Tax Administration was created in 2002, and the 46 current member countries account for more than 85% of global economic output. A members’ meeting on advanced analytics in Dublin in 2011 increased existing demand for the regular trading of information, experiences and innovations.