Tax Compliance and Enforcement: An Overview of New Research and Its Policy Implications

Key Points

  • The government loses almost 14.5 percent of revenue due to noncompliance, enough money to substantially narrow or even eliminate the federal deficit.
  • Third-party reporting of income is effective at improving reporting of income. However, increased reporting of income from third parties does not necessarily lead to increased tax revenue. In addition, most studies indicate that an appeal to moral duty is not effective at improving reporting of income to tax authorities.
  • Instead, increasing the chances of audit is effective at reducing tax evasion.

Tax Compliance and Enforcement: An Overview of New Research and Its Policy Implications

Editor’s Note: This article is part of a series of tax-related articles sponsored by the Penn Wharton Budget Model and the Robert D. Burch Center at Berkeley. All of the articles in this series are forthcoming in a book by Oxford University Press, co-edited by Alan Auerbach and Kent Smetters.

Tax Compliance and Enforcement: An Overview of New Research and Its Policy Implications reviews the leading research on methods to increase tax compliance. Joel Slemrod (2016) finds that, in general, increasing the chance of being audited or financial penalty deters tax evasion. In recent years, research has been helped by increased access to administrative data and the willingness of tax authorities to engage in randomized field experiments.

Tax evasion reduces tax revenue. One of the biggest sources of tax noncompliance stems from self-employed individuals. Based on 2006 tax returns, the Internal Revenue Service (IRS) estimates that tax evasion totaled $385 billion in unreceived revenue, or 14.5 percent of the total tax liability. The IRS receives little information from third parties about income from self-employment. One study, reviewed by Slemrod (2016), estimates how true income differs from reported income and finds a self-employment noncompliance rate of 35 percent. This value is lower than the noncompliance rate of 56 percent estimated by the IRS in Figure 1 because the estimates are produced using difference methods and definitions.

Figure 1: IRS Estimates of Tax Noncompliance Rates

Source: Slemrod (2016)

Third Parties

Noncompliance rates are substantially lower when the IRS receives more information from third parties and when taxes are withheld by third parties. Employers are an example of a third-party tax reporter who also withhold taxes. Income from wages and salaries is reported to the IRS by employers, and taxes on wages and salaries are also withheld by employers and sent to the IRS. Investment management companies are an example of a third-party reporter that does not directly withhold taxes: investment income is reported to the IRS, but the taxpayer is required to make the actual tax payments.

Figure 1 indicates that noncompliance rates are as high as 56 percent when there is little reporting from third parties. However, noncompliance rates fall to around 10 percent when there is a substantial amount of third-party reporting. Noncompliance rates fall to as low as 1 percent when the third party not only reports income to the IRS, but also withholds the taxes.

However, third-party reporting may be more effective at improving tax compliance of some sources of income than others. For example, requiring credit card companies to report payments that they make to businesses increases reported business income, which increases tax revenues. However, this same reporting also encourages businesses to report associated expenses, which decreases tax revenues. On net, therefore, the increase in revenue is smaller than would otherwise be estimated using only estimates of noncompliant business income.

More importantly, increasing third-party reporting of income from self-employment is challenging. For example, it would be extremely difficult to ask households to report payments that they made to electricians and plumbers to the IRS. It would be even more daunting to ask households to additionally withhold taxes on behalf of these service workers. Since income from self-employment has the highest noncompliance rate, much of the literature, therefore, has turned to other forms of increasing compliance, including moral duty and threats of audit.

Moral Duty

A letter from a tax authority threatening an audit can induce people to pay taxes, and a letter from a tax authority about tax benefits can induce people to take the benefit. Can a letter that does not threaten financial damages but appeals to civic duty also be effective at inducing people to pay taxes? Studies show that a letter that describes the benefits of tax-funded projects, indicates that most other people pay taxes, or names public services the person benefits from are not effective at reducing tax evasion. Slemrod (2016) lists six studies as examples that indicate that appeals to moral duty do not increase reporting of income.

However, a recent study from the United Kingdom does show that when filers who owe an overdue tax payment receive a letter that indicates that most people have already paid, the filers pay faster than filers who did not receive a letter. Depending on the text of the letter, filers who received a letter were 1.2 percent to 4.9 percent more likely to pay their overdue taxes.

Moreover, new research shows promise that public disclosure of tax information or shaming by publishing tax evaders may reduce tax evasion. In Norway, when public disclosure of tax information became widely available on the internet, reported income increased by 3 percent. More than 20 U.S. states try to encourage filers with overdue tax bills by posting information to the internet. One study looked at delinquent tax filers in three states and found that, for filers who received a letter containing a threat of shaming, the probability of payment increased by 2.1 percentage points for those who owed small amounts. Nonetheless, while some forms of moral persuasion or shaming might encourage compliance, a majority of research finds little effect.

Chance of Being Audited

Increasing the perceived chance of being audited appears to be the most effective mechanism at deterring tax evasion. When tax authorities send a randomly selected group of tax-filers letters stating that their income taxes will be closely examined, tax compliance increases. Filers who receive a letter report more income that those who do not receive a letter, especially among low- and middle-income filers.

For example, one study that focused on tax filers in Minnesota found that filers with incomes from $10,000 to $100,000 who received a letter increased their reported tax liabilities by 12.1 percent from the year before, compared to tax filers who didn’t receive a letter. In addition, after being audited, reported income, especially from self-employment, increases in the years after the audit. One study, reviewed by Slemrod (2016), found that reported income from self-employment increased by 7.5 percent over three years after the audit.

Some filers evade paying taxes; on the opposite end of the spectrum, some filers do not claim all of the credits and tax benefits they are eligible for. Can methods that are used to increase tax compliance also increase take-up rates of credits and other tax benefits? Studies show that a letter from a tax authority can increase take-up rates of credits and other tax benefits. When filers who are eligible for the earned income tax credit (EITC) receive a letter about the benefit, the take-up rate of EITC increases by 25 percent in the year the letter is received.


Slemrod (2016) concludes that, to deter tax evasion, it is most efficient for the IRS to focus on increasing the threat of audit and financial penalties. Other changes, such as increased reporting and withholding by third parties where possible, simplifying the filing process, and increasing the IRS budget could also help. Finally, if the IRS made data available on both pre- and post-audit returns, then outside researchers could help the IRS learn how to better target tax evaders.

A discussion of this paper is provided by David Weisbach.