This year’s IRS/TPC Joint Research Conference on Tax Administration examined a central question for the tax system: How can the IRS improve taxpayer compliance—and close the tax gap—without relying only on more audits?
Across four sessions, researchers’ findings pointed to an answer: Tax compliance improves when the IRS makes it easier to comply, has better information before returns are filed, uses audits more effectively, and learns from taxpayer behavior over time.
In other words, the tax gap is not just about taxpayer honesty. It also reflects administrative design.
Help me help you: Nudging taxpayers into compliance
The first session showed why taxpayer service and information matter before enforcement ever begins.
Vishal Baloria, of the University of Connecticut School of Business, focused on the effects of IRS staffing cuts following a December 2010 hiring freeze: When IRS capacity falls, filing can fall, too. His analysis suggested that both enforcement and taxpayer services matter for compliance. Access to tax examiners or preparers appeared to mitigate the negative effect of staffing cuts, while high- and middle-income taxpayers drove much of the filing decline.
Ali Ekmen, formerly with the Turkish Tax Inspection Board, made a related point by contrasting self-reporting models with system-reporting models. Under a self-reporting model, taxpayers collect records, classify income, file returns, and are checked after the fact. Under a system-reporting model, information from employers, banks, platforms, e-invoicing, withholding, and third-party reporting can pre-fill or anchor the tax return.
The session leaned into the idea that compliance should be built into the filing process, as tax administration increasingly shifts from reactive enforcement to proactive compliance support.
From more audits to better audits
The second session focused on audit effectiveness, audit selection, and post-audit behavior. Alex Turk, of the Research Applied Analytics and Statistics Division of the IRS, addressed the problem of imperfect detection: Auditors may not always uncover the true amount of underreported tax.
Brandon Anderson, of the IRS Research Division, examined how often audit-selection models should be retrained when taxpayer behavior and reporting patterns shift. Together, the research pointed to a practical conclusion: Audit systems need to adapt.
That matters because audits can do more than raise revenue from a single return. They can also shape how taxpayers behave later. Better auditor performance measurement and adaptive selection models can increase compliance without necessarily increasing audit volume.
Enforcement effects can last, but not always
Carlos Scartascini, of the Research Department of the Inter-American Development Bank, closed the second session by exploring how taxpayers respond after being audited. Audits may increase compliance among evaders but have little effect on compliers (who may sometimes become less compliant). The results underscored one of the day’s recurring themes: Taxpayer responses vary depending on perceived chances of audit.
The third session extended that theme by examining the longer-term and indirect effects of enforcement.
Justin Nave, of MITRE, focused on how audits affect taxpayer filing behavior after the audit occurs. The main finding was striking: Audits led taxpayers to report roughly $1,000 more in taxes each year for the first four years after the audit. That suggests audits may generate compliance benefits beyond the immediate audit adjustment by changing subsequent reporting behavior.
Zizhuo Chen, of the University of Minnesota Twin Cities, examined spillover effects from a high-intensity tax enforcement action: the 2015 New York City zapper crackdown. Chen’s work suggested that enforcement attention may spill over through social networks, affecting compliance even outside the directly targeted area.
Tom Hertz, of the Research, Applied Analytics, and Statistics Division of the IRS, presented an analysis of the IRS’ Automated Underreporter (AUR) program which identifies discrepancies between individual tax returns and third-party information submitted to the IRS. But Hertz’s findings complicated the story. AUR notifications did not appear to affect later reporting behavior.
The lesson is not that enforcement always changes future behavior. It is that different enforcement tools can produce different behavioral effects. To understand varied taxpayer responses, the IRS needs to know which interventions work, for whom, and for how long.
Making third-party data work harder
The fourth session focused on improving the usefulness of third-party data. Rachel Geiger, of the IRS, considered how machine learning can help make administrative data more useful by bridging gaps between sample-level and population-level information.
Lucas Goodman, with the Office of Tax Analysis at the Treasury Department, examined wage misreporting on individual income tax returns using W-2 and 1099 data. Taxpayers who changed jobs or moved may have less access to W-2s and may be more likely to underreport. Misreporting appeared especially relevant in cases involving multiple jobs or small jobs, and the presentation noted that this type of misreporting is rising.
Riddha Basu, of the George Washington University School of Business, returned to a familiar tax administration question: How does the level of information reporting affect tax compliance? The research found that some small businesses strategically change how they route customer payments in response to reporting rules, but also that expanded reporting can improve compliance.
The broader message was clear: Taxpayers are much more likely to report income accurately when the IRS already has relevant information or can easily verify it.
Some research may not translate immediately into IRS operations. But it still helps identify where the agency should focus, especially as data availability, machine learning, and AI reshape tax administration.
Compliance can be designed into the system
Whether through taxpayer assistance, third-party reporting, machine learning, or targeted enforcement, many presenters emphasized reducing opportunities for error before returns are filed.
But better design also means recognizing that enforcement can have unintended consequences. Because taxpayers respond differently to audits and notices, a one-size-fits-all approach may waste resources, burden compliant taxpayers, or fail to change the behavior of noncompliant ones.
Agency capacity is also a real constraint. IRS resources affect both service delivery and compliance outcomes, suggesting that taxpayer assistance networks should be viewed as complementary rather than competing functions.
Better administrative design can help close the tax gap, and better service, better data, and smarter enforcement can make the tax system work better for everyone.
