Misspending Rates in Arizona’s ESA Program Have Been Wildly Overstated

As of this school year, over 100,000 students have enrolled in Arizona’s Empowerment Scholarship Account (ESA) program. Their families are spending more than $900 million per year on a variety of educational environments for their children. When the number gets that big, there is bound to be scrutiny. Taxpayers want to know their dollars are being spent as intended.

That is why it made a splash when the Arizona Republic published an article under the headline, “Over 20% of Arizona ESA purchases were ‘unallowable,’ audit finds.” It was not the only media outlet pushing this idea. The television station 12 News has broadcast multiple pieces popularizing the idea that “misspending could be 20% of all purchases in the $1 billion voucher program.”

This 20% number appears to come from a risk-based audit of questionable purchases in the program conducted by the state department of education, obtained via a freedom of information request. That sampling method will be important to remember.

On March 12, the Arizona Department of Education (ADE) released their own analysis of ESA spending, which included a broader and more scientific sampling approach than the prior analysis provided. The Department found that just 2% of all ESA spending was for items that are not allowed. To be clear, that doesn’t mean that parents were trying to game the system. The ADE contends that many parents didn’t realize that some items, like water bottles, can’t be purchased with ESA funds. Dollars from fraudulent or egregious purchases account for 0.3% of ESA spending.

Unlike the sample used by 12 News, which only looked at transactions that had been flagged as questionable, the sample in ADE’s analysis was randomly drawn, which allows for a more representative illustration of all program transactions

We were granted access to data on all ESA transactions for the school year 2024-25 (the year prior to the one ADE studied). After our review, the ADE’s numbers seem plausible. Not only that, but unallowable expenses seem to be almost entirely concentrated in one source, which should make improvements simpler.

To understand any of these numbers, it is important to go over some context. There are three ways to spend money in the program. (There is a fourth way that involves a Bank of America debit card, but it is being phased out.)

First, families can pay for something and ask for reimbursement. In the school year 2024-25, there were about 156,000 transactions totaling about $350 million, or about $2,000 per transaction on average. There are about 175 vendors, mostly curriculum companies, and everything seems on the up and up.

Second, parents can pick an approved vendor, often a private school but also sometimes a supplemental service like tutoring, and ask Class Wallet to send a payment. This is called the Direct Pay system. Class Wallet manages the system and there are gates up to prevent money being sent to non-approved vendors or for non-approved categories. In school year 2024-25, there were about 341,000 Direct Pay transactions totaling $455 million, or about $1,350 per transaction. There don’t appear to be issues there either.

Now we have accounted for more than $800 million out of approximately $900 million spent in FY 2024-25, or 89% of the total. Prior to the Arizona Department of Education’s randomized audit published March 11, none of these transactions were taken into account in error rates being reported.

The final way parents can spend their ESA dollars is in the Class Wallet Marketplace. This is similar to having an employer-provided card that is automatically paid that you are allowed to use for appropriate expenses. This appears to be where the problems arose.

In 2024-25, there were nearly 1.4 million transactions in the Marketplace, totaling about $80 million, or $55 per transaction. Amazon is, by far, the most-used vendor in the Marketplace. ESA dollars can be relatively easily spent while you’re logged into your own Amazon account, albeit with an extra step to select this payment type. Amazon alone had over 1.1 million Marketplace transactions in 2024-25.

After scanning through the 1.4 million Marketplace transactions in 2024-25 (as thoroughly as can be done in a couple of days), essentially all the potentially controversial purchases are found in Amazon transactions.

How many are there? It’s hard to say. A simple content analysis finds that there are some clearly unallowable transactions, and we are working on a more thorough and large-scale analysis to quantify exactly how many purchases are egregious.

But even if every Amazon transaction was eyebrow-raising, we would only be questioning around 7% of the total spend on the program. And not every Amazon transaction was eyebrow-raising. Far from it. Examining the records shows thousands of computers, thousands of books, and thousands of school supplies purchased this way. All perfectly allowable.

If we use the Amazon dollars ADE identified as unallowable in their study as a guide, more than 80% of Marketplace dollars spent on Amazon are perfectly fine, so the actual total amount of questionable purchases should be around 1 or 2%.

To be clear, there should be protocols to minimize unallowable purchases, as errors with taxpayer dollars are bad.

But even if there is room for improvement around Amazon, based on our preliminary research and the Arizona Department of Education’s study, the ESA withstands more scrutiny than many public programs. It’s easy to make headlines about any social service—unemployment insurance, SNAP, public schools—if you look only at a list of all transactions flagged as questionable.

When it comes to governance, however, you have to look at the entire picture. And so far, the ESA is turning out fairly photogenic, and its one notable blemish should be easily fixed.

---

For more on Arizona, check out:

Is Arizona’s ESA Plagued by Unallowable Expenses?
What 12News Missed (and Misrepresented) in EdChoice’s ESA Report