Today at the colloquium, Day Manoli presented the above-named paper, which can't be posted yet for IRS clearance reasons. But it promises to be of considerable interest when it comes out. It looks at the data around IRS documentation requests to randomly selected taxpayers, within a group that is deemed to present low or intermediate risk scores re. the need for an auditing adjustment, who have claimed the earned income tax credit (EITC).
It would probably be premature at this point to discuss the study's particular findings, but the administrative intervention that Manoli et al are studying involves "correspondence audits" in which the IRS sends a letter requiring documentation - say, of claimed earnings or dependents - and failure to respond with the requested items results in the EITC's being denied.
EITC denial by reason of the correspondence audits appears to have decreased Type II errors (giving the EITC to those who didn't legally qualify). But it also appears to have increased, perhaps quite significantly as a percentage matter, Type I errors (denying the EITC to those who did legally qualify but couldn't handle the extra hurdles - in some cases because the IRS letter wasn't successfully delivered to them).
Some general observations that I would offer here include the following:
1) By Congressional diktat, audit rates are generally higher for EITC claimants than for any other individual taxpayers other than the very rich. This cannot be justified as a matter of generating the maximum bang for the buck from the allocation of IRS audit resources. I don't think there is good justification for it otherwise either. Among its possible motivations are classism, racism, and fiscal language illusion or bias (viewing over-payment of "transfers" as worse than under-payment of "taxes" even though (a) a dollar is a dollar, and (b) we are all net taxpayers on a lifetime basis, hence it's absurd to code some of the sub-transfers differently than others).
2) Relatedly, the relative priority that public debate gives to focusing on Type II errors in EITC provision, relative to Type 1 errors, appears to me to lack adequate justification.
3) Lots of EITC non-compliance is not deliberate. This reflects the EITC rules' excessive complexity. A prominent NYC tax lawyer of my acquaintance once wrote an article urging his colleagues to offer tax advice pro bono to needy clients, rather than working for more billable hours and giving the extra $$ to charity. One can agree or disagree with his view, given billing rates vs. the value of one's services to someone who can't pay for them, but one of the arguments he made was that in fact the EITC can be daunting enough in practice that skilled tax lawyers actually CAN offer their pro bono clients something that is scarce and of value. The EITC oughtn't to have been designed in such a way that high-end legal skills might come in handy towards claiming it properly.
4) Papers on Food Stamps or SNAP, such as that by Tatiana Homonoff which we discussed several week ago at the colloquium, often discuss "targeting efficiency" - e.g., the relative social welfare costs of SNAP benefit loss to worse-off versus better-off individuals among those whose incomes are low enough to qualify for it. The welfare costs under the EITC not only of Type I errors but even of correcting at least innocent Type II errors (which often reflect the rules' needless complexity) likewise ought to be treated as relevant to the analysis. These are generally people in the lower income ranges, often with dependent children, who may have significant unmet material needs but who are punished for lacking political clout.