Over the past several weeks, Arizona 12News reporters Craig Harris and Joe Dana have repeatedly, falsely, claimed that 20% of the use of the state’s Empowerment Scholarship Account (ESA) program consists of improper “unallowable” purchases. Now, as their narrative implodes under public scrutiny, an actual statistical analysis conducted by the Arizona Department of Education (ADE) shows program misspending is less than 2% (with actual “fraud or egregious purchases” totaling just 0.3%).
Yet rather than retract their false claims, 12News simply recruited “a Princeton education researcher” to bail them out—featuring her in a subsequently unveiled news segment to proclaim ADE’s analysis illegitimate: “Quite frankly this would not fly,” she declared of the department’s analysis. “This would be a failing grade if it was a paper.”
12 News’ Craig Harris promptly laundered her claims as established fact, proclaiming, without any ambiguity:
“The ADE released an audit today by one of its employees who says 2 percent of purchases are unallowed. However, the research is largely skewed, has a miniscule sample and underrepresents two key areas… The data was significantly skewed to lower the unallowed/fraud rate.”
Indeed, citing supposed evidence of a major blunder in ADE’s analysis, 12News’ Ivy League academic confidently assured the network’s audience:
“The innocent explanation would be that this is simply an error on the part of the [ADE] analyst. The impolite explanation is that this is an active attempt to mislead the public about the current rate of unallowables.”
Unfortunately for 12News and their ostensible expert (sociology professor Jennifer Jennings), it turns out (as shown in more detail below) that her words described her own mistaken analysis—meaning the claims parroted by the 12News’ reporting duo to discredit ADE’s findings and absolve themselves of the need to retract their prior stories were yet again…demonstrably false.
This is especially significant because Mr. Harris, Mr. Dana, and Prof. Jennings occupy positions of enormous public trust, representing media and academic institutions they did not build and whose reputations and esteem they did not create. Using the eminence of such institutions to advance claims based on anything but a relentless and objective commitment to the truth is a betrayal of that public trust and of these storied institutions themselves.
These individuals thus have a duty to uphold the standards of journalistic and academic integrity and retract the erroneous and deceptive statements they have made, respectively, against the Arizona Empowerment Scholarship Account (ESA) program, the families it serves, and the men and women who implement it on behalf of over 100,000 children.
I understand that journalism and scholarly inquiry are by their nature messy—and that we should welcome, rather than denounce, good-faith efforts to understand and interpret available evidence even when those interpretations ultimately prove incorrect. This may be indeed be the case with Prof. Jennings, whose misinterpretation of the data could very well be the result of honest error.
But regardless of the explanation, if Mr. Harris, Mr. Dana, and Prof. Jennings are to retain any credibility in their reporting and analysis of Arizona’s school choice programs, they must retract and cease repeating the false claims below. Trying to redirect to the supposed misstatements of others or trying to rewrite the obvious history of 12News’ recent claims are not sufficient or acceptable remedies–particularly when these false claims have been shared with tens of thousands of viewers by 12News and are now used to justify and fuel an active political campaign against a program serving over 100,000 children.
Claim #1: 20% of ESA purchases are unallowed
(Mr. Harris): “Empowerment Scholarship supporters say there is minimal fraud in the program: 1% or less. But new department of education records that 12 News obtained show that unallowed purchases like those cell phones are actually around 20%. That’s 1 in 5.”
(Mr. Dana): A” very big story from my colleague Craig Harris reveals an audit of a very large sample of ESA purchases that shows just over 20% are considered unallowable. 20%. That’s higher than any previous official estimates.”
(12News Mark Curtis): Some people claim the number of people gaming the system is tiny, but now [12News] has uncovered records that show the abuse is much more widespread”
Basic journalistic ethics should have compelled 12News’ reporters to admit at the outset that this 20% figure was not based on a representative sample of ESA purchases overall. Instead, it reflected the rate of unallowable purchases among ESA transactions that had been specifically flagged by the department’s audit protocols as higher risk.
Claiming otherwise would be akin to declaring that if 20% of suspects investigated by the police are guilty of a particular crime, that 20% of Americans overall are guilty of that crime.
Claim #2: ADE’s analysis exposing 12News’ deception was defective & not actually based on a random, representative sample.
Harris: “The ADE released an audit today by one of its employees who says 2 percent of purchases are unallowed. However, the research is largely skewed, has a miniscule sample and underrepresents two key areas.”
Harris: “The Arizona Department of Education says it did a random sample of ESA expenditures to show misspending is low… In looking at the data, it does not appear it was truly random… In the simplest terms, a random sample is supposed to work like pulling names from a hat. If the sample is truly random, it should look broadly like the full set of transactions it came from… But the state’s own datafile has a basic problem: the sample does not appear to match the underlying transaction mix. That means the sample was not random…
Harris: “And, it contains far too few Marketplace and reimbursement transactions… “
Harris: “Had ADE accurately used the correct data set, it would have found over 12 percent of transactions and 14 percent of dollars were “unallowed.””
Harris’ claims were based on the analysis of their Princeton consulted expert, Prof. Jennings, which was glowingly shared in another 12News segment compiled by Mr. Dana:
Dana: “There are 4 ways ESA parents buy items. The state has provided 12News access to two of those ways: reimbursement and marketplace data. All that data for fiscal year 2026 shows marketplace made up 75% of all transactions. Reimbursements made up 25%. In other words, 3 marketplace transactions for every 1 reimbursement. But the audit sample described by the Arizona Department of Ed shows 40% of marketplace and 60% reimbursement. “
Jennings: “It turns out to be the case that those ratios [in ADE’s audit] are roughly flipped. And that wouldn’t matter if they didn’t have different rates of unallowed expenses, but in fact, marketplace has a much higher rate, which then in turn produces a different portrait of the overall unallowed rate.”
Unfortunately, this entire claim—which underpins 12News’ and Prof Jennings’ assertion that ADE’s analysis was defective—is simply, factually wrong. And this is more than a mere minor technicality, as it renders 12News’ primary claims completely false, forcing them to again confront the fact that an actual random sample of ESA purchases has entirely discredited their original claims of 20% unallowable spending rates in the program.
As shown below, when looking specifically at the mix of marketplace and reimbursement transactions that 12News/Jennings analyzed, we see that data confirm the validity of ADE’s audit and expose the erroneous analysis of its critics:
Based on the full datasets made available by ADE and looking at the same time period as 12News/Jennings, it turns out the dollar amounts associated with ESA marketplace purchases in FY 2026 (through Dec. 31, 2025) total $51.3 million dollars. The total for reimbursements was $81.3 million. That means marketplace spending made up 39% of the total marketplace/reimbursement spending, while reimbursements made up 61%.
Notice those are almost exactly the same ratios that appear in ADE’s randomized audit sample, which 12News and Prof. Jennings complained about above, in which they said: “But the audit sample described by the Arizona Department of Ed shows 40% of marketplace and 60% reimbursement.”
In short, ADE’s sample correctly reflects the makeup of spending on marketplace and reimbursement orders in terms of actual dollars spent. (Further confirming the actual alignment between the audit and the program-wide ESA purchases, the ratio of marketplace to reimbursement dollars in ADE’s audit is 37% to 63%.)
But of course, even if the dollars are right, we would still expect the relative number of transactions in ADE’s audit to also resemble the broader program data if it is truly a representative, random sample. According to 12News/Jennings, this is where we see evidence that ADE “flipped the ratios.”
Again, according to 12News/Jennings: “All that data [for the full program] for fiscal year 2026 shows marketplace made up 75% of all transactions. Reimbursements made up 25%. In other words, 3 marketplace transactions for every 1 reimbursement. But the audit sample described by the Arizona Department of Ed shows 40% of marketplace and 60% reimbursement.”
So, did ADE’s sample somehow get the dollars right, but the number of purchases, wrong?
No.
As 12News and Prof. Jennings should have acknowledged, the data from ADE for the full program was provided on an item level for marketplace purchases, but on a transaction level for reimbursements. (Families submit an entire receipt for all items in a transaction together when reimbursed, but all items are individually automatically logged when purchased through the marketplace).
This means that each of the 1 million FY 2026 marketplace records provided by ADE were for individual items. But each of 345,000 reimbursement records were for transactions (many of which have many items). And when you compare the number of marketplace items vs the number of reimbursement transactions, you get (as 12News/Jennings did) a 75% marketplace vs 25% reimbursement ratio.
This mashup of two different units of measure unsurprisingly creates the illusion of a discrepancy when compared to the ratio of marketplace (transactions) vs reimbursement (transactions) found in ADE’s audit sample, which (as 12News/Jennings complained), produces a 40% marketplace vs 60% reimbursement ratio.
Yet, as already shown above, this latter 40%-60% ratio of transactions in ADE’s audit also lines up closely with the ratio of dollars. This alone should have tipped 12News/Jennings off to the fact that something was wrong in their preferred analysis.
And, if 12News/Jennings had bothered to look, they would have seen that ADE’s audit sample even yields almost exactly that same ratio that 12News and Jennings expected (75% vs 25%) if measured using the same mismatched process Jennings used for the overall ESA dataset: ADE’s audit sample includes 1,028 reimbursement transactions and 2,663 marketplace items (excluding 7 returns), for a 72% marketplace vs 28% reimbursement split.
In other words, using their mashup of transactions vs items, 12News and Jennings find a 75%/25% split among ESA purchases overall, while the random sample in ADE’s audit yields a similar 72%/28% split when using the same calculation approach.
That’s obviously a lot of numbers and math, but the takeaways are that:
1) ADE’s audit sample almost perfectly mirrors the overall ESA program data for actual dollars spent by expenditure type: roughly a 60%-40% split between reimbursement and marketplace expenditures in both the overall program data and the audit sample.
2) 12News / Jennings bungled their comparison of the relative share of ESA purchases by failing to acknowledge that half the records they were looking at were items, and half were transactions. Inconsistently comparing these two created the alleged discrepancies they claimed existed in ADE’s data.
3) 12News’ / Jennings’ claims against the representativeness of ADE’s audit sample are simply unsupported by the data. Without the fig leaf of being able to allege a discrepancy in ADE’s audit, 12News should immediately retract its false stories and claims.
Perhaps most damning of all, as I have written previously—to no response from Mr. Harris—his complaints that ADE’s audit “contains far too few Marketplace and reimbursement transactions” is not only an inaccurate claim (as shown above), but also a nakedly hypocritical indictment.
The most recent official ESA quarterly report of all transactions (FY 2026, Q2) shows marketplace transactions made up 12.9% of all ESA dollars (including spending on ‘direct pay’, debit cards, etc.). In ADE’s statistical analysis of the random sample of ESA purchases, marketplace transactions made up 13.2% of sampled ESA dollars. In other words, ADE’s random sample once again mirrors the overall program data. In contrast, in the risk-based audit Mr. Harris relied on to peddle the original fictitious 20% “unallowable” spending narrative, marketplace purchases made up a whopping 44.9% of the total records. Yet inexplicably, Mr. Harris was completely unbothered by the unrepresentative nature of the risk-based audit sample he used for his narrative, while incorrectly attacking ADE’s.
Claim #3: 12 News Never Said “Fraud”
Mr. Harris’ recent assertions that “12News never said 20 percent fraud” (emphasis added) is an intellectually bankrupt attempt to hide the obvious implied meaning in Harris’ statements: As mentioned above, Mr. Harris’ declared in the news segment for his original story:
“Empowerment Scholarship supporters say there is minimal fraud in the program: 1% or less. But new department of education records that 12 News obtained show that unallowed purchases like those cell phones are actually around 20%. That’s 1 in 5.”
And this was not an isolated statement. In his written article for that same bombshell story, Harris reiterated:
“Arizona Republican leaders, including state Schools Chief Tom Horne, have long insisted fraud in Arizona’s Empowerment Scholarship Account program is minimal.
‘One percent or less,’ Horne often has said.
But 12News has obtained new public records — from Horne’s Department of Education — that tell a very different story.
Documents uncovered by 12News Investigates show unallowable purchases — spending explicitly banned under ESA program rules — may account for about 20 percent of transactions.”
If, as Harris now claims, he was not seeking to induce casual readers into believing that fraud was 20%, then it was a spectacular breach of journalist standards to declare that the data “tell a very different story” from the claim that “fraud” is “one percent or less.”
On this point, to my knowledge, Mr. Harris is correct that he has carefully avoided explicitly stringing the words together in a way that would count as outright declaring fraud to be 20%. But the legal standards of a courtroom are not the same as the standards of basic journalistic ethics, which he has far more indefensibly breached.
As is clear from the above, the central premise promoted recently by the 12News team—that ESA “abuse” is “widespread”—has been built on falsehoods, errors, and deceptive framing.
Certainly, any individual that knowingly uses ESA funds to purchase unallowable items—particularly egregious non-educational items—should be subject to rigorous penalties and enforcement.
But 12News’ attempt to now rewrite history and suggest that their 20% “unallowable” spending rate is still a legitimate measure (even after it has been dismantled) because they assert that roughly 20% of parents have made at least 1 unallowable purchase at some point in their ESA careers (out of hundreds or thousands of purchases, and typically for something like a lunchbox) is nothing but a continued effort to stretch all bounds of reason and avoid responsibility for their claims.
To that end, if it is I who am in error with respect to any of the claims or analysis above, I will publicly retract them and express gratitude for the correction.
ESA families and Arizona voters expect and deserve the truth from those charged with reporting the news or with educating many of our nation’s next generation. May those individuals act and speak accordingly.
