By Miguel A. Fonseca and Shaun B. Grimshaw
We live our lives ever more online: we shop, we socialize, and get our news through the internet. Slowly, our interactions with government and public services have also moved into the digital sphere. In addition to lowering the costs of accessing some public services, digital systems allow for tax authorities to make it easier for taxpayers to pay their taxes correctly and on time. But these advances are not without potential pitfalls. In this post, we review some of the behavioral research done by TARC on digitalizing tax administration.
Pre-populating tax forms
Digital systems allow for tax administrations to combine (within legal privacy constraints) data from different sources to simplify filing processes. In practice, this means using third-party data such as employers, banks or pension companies to prepopulate tax forms. This is already a reality in the field: the state of California trialled prepopulates elements of its state tax returns with the Ready Return program; in addition, tax return pre-population happens to varying degrees in more than ten European Union countries and Australia.
The primary benefit of pre-populating is the reduction of errors in filing. Errors in filing decisions account for a significant portion of noncompliance: the U.K. National Audit Office estimates that it loses £6.5 billion (19%) of its tax revenues due to filing errors. Andreoni and co-authors estimate that 7% of U.S. taxpayers make mistakes when filing their tax returns. The financial case for helping taxpayers is overwhelming.
However, there are important ethical and operational concerns with pre-population of tax forms. The primary source of concerns lies in the possibility of error in the pre-population process.
One possibility is that the tax authority under-estimates taxpayers’ tax liabilities. Taxpayers, upon viewing a pre-populated form, may accept the incorrectly pre-populated values because of status quo bias, inattention, or simply because they trust the tax authority’s assessment to be correct. Such an event could lead taxpayers open to penalties in the event an audit takes place, as they would still bear legal responsibility for filing their taxes correctly.
Another possibility is that the tax authority over-estimates tax liabilities. In that case, leaving taxpayers financially worse off. In either case, there are ethical concerns given the duty of care that public administrations must have towards citizens and companies. Such errors would likely result in increase in audits, as well as a public relations blow and a potential drop in trust in government.
Behavioural inertia from pre-population is a real concern. Research on Finnish tax filing data following the implementation of pre-populated tax forms shows that taxpayers who received a prepopulated return were more likely to report the items that were prepopulated and less likely to report deductions that were not prepopulated.
TARC behavioural research on the effects of pre-population confirms that pre-population is a powerful behavioural tool; but it should only be used if the underlying data is highly reliable. Setting default levels that underestimate taxpayers’ true tax liability lead to significant drops in compliance and tax revenue.
Digital systems allow for us to move beyond the old static interface that taxpayers have used for decades to relay their tax affairs to the government, best exemplified by the old paper form. Software interfaces allow for the form itself to respond in real time, and capture errors before they are submitted. Behavioral science has long argued for the use of smart, reactive nudges, which respond in real time to inputs of users. Personalising the nudge makes it more salient, and therefore more effective than static nudges that may fade into the cognitive background.
TARC research shows that nudges that respond to user inputs in electronic tax forms are significantly more effective at raising compliance than static nudges with the same content.
Feedback report pre-filing
A data-driven approach is important, not only to determine where taxpayers make errors, but also to determine how to design the informational content of the nudge itself. The increasing digitization of public services has made available vast datasets that open the door to bespoke behavioral interventions. This is possible through the use of more granular data, coming from individual entries on forms, which could enable public administrations to detect and predict common errors and design information nudges to mitigate such errors.
One such intervention
The effectiveness of information nudges has been demonstrated in a survey RCT in Finland. Following a tax rule change, researchers included information about that tax rule change in a survey to a random sample of small business owners, but not to a different random sample. By matching survey responses to tax returns data, the authors found a significant reduction in filing errors among informed business owners relative to the uninformed group. This suggests that embedding targeted information nudges within the filing process could be beneficial to taxpayers.
TARC research looked at the effectiveness of information nudges, whereby automated reminders about filing or payment obligations were sent to taxpayers in response to filing behavior in specific fields in the tax form. The nudges pointed taxpayers to publicly available information following the submission of a tax return. Relative to a baseline no-nudge, condition, nudges led to substantial re-checking of tax forms. However, nudges shown inappropriately to individuals lead to an increase in filing errors, while correctly shown nudges lead to substantial reductions in errors only if they are prescriptive in nature; generic nudges lead individuals to replace one filing error with another.
Digitalization has opened vast new opportunities for tax administrations to engage more directly and effectively with taxpayers. The potential to reduce non-compliance (due to error or otherwise) is there to be taken. However, there are challenges with such technologies which means tax administrations should be careful before adopting new measures. Pre-population of tax forms is already a reality in many countries, although the evidence suggests it should be restricted to extremely high-quality third-party data, as the downside risk is high. Dynamic nudges that react to user input offer great promise as low-cost interventions that may reduce the amount of non-compliance.