A Welfare Analysis of Tax Audits Across the Income Distribution
Bibliographic record
Abstract
Abstract We estimate the returns to IRS audits of taxpayers across the income distribution. We find an additional $1 spent auditing taxpayers above the 90th income percentile yields more than $12 in revenue, while audits of below-median income taxpayers yield $5. We construct our estimates by drawing from comprehensive internal accounting information and audit-level enforcement logs. We begin by estimating the average initial return to all audits of U.S. taxpayers filing in tax years 2010–2014. On average, $1 in audit spending initially raises $2.17 in revenue. Audits of high-income taxpayers are more costly, but the additional revenue raised more than offsets the costs. Audits of the 99–99.9th percentile have a 3.2:1 initial return; audits of the top 0.1% return 6.3:1. We then exploit the 40% audit reduction between tax years 2010 and 2014 to examine the returns to marginal audits. We find they exceed the returns to average audits. Revenues remain relatively unchanged, but marginal costs fall below average costs due to economies of scale. Next, we use randomly selected audits to examine the effect of an initial audit on future revenue. This individual deterrence effect produces at least three times more revenue than the initial audit. Deterrence effects are relatively consistent across the income distribution. This results in the 12:1 return above the 90th percentile. We conclude by estimating the welfare consequences of audits using the MVPF framework and comparing audits to other revenue-raising policies. We find that audits raise revenue at lower welfare cost.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.009 | 0.030 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.003 | 0.003 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".