I'm aware of merchant transaction fees, but they're insufficient as an explanation. Just run the numbers. Say that a hypothetical consumer spends $10K/year. The person who carries that as a balance will pay $2700 at 27% interest, plus $300 from the 3% transaction fee, for a total of $3000, and they might get $200 as cashback. The person who doesn't carry a balance will pay $300 and get $200 as cashback. It's pretty doubtful that all the other services a credit card company provides - fraud detection, chargebacks, billing, recurring payments - cost < $100/year. Maybe if you never submit a chargeback and your card never gets stolen, but at best we're looking at them breaking even on their non-balance-carrying customers and making all their profits off the customers that carry a balance.

When you're providing a service to one group at or below cost and making all your profits off another group or another service, that's the definition of a cross-subsidy, even if you do get a token amount of revenue off the first group.

> I'm aware of merchant transaction fees, but they're insufficient as an explanation. Just run the numbers.

I don't have to run the numbers, because I read the article where they did the actual research with real numbers, not your back of the envelope estimating, and they came to the opposite conclusion:

| This leads to our second hypothesis: High interest rates are necessary to recoup the high cost of rewards. However, our analysis shows this is not the case. Rewards expenses are more than fully covered by banks’ interchange income—fees collected from merchants based on purchase volume. On average, interchange income amounts to 1.82 percent of purchase volume, while rewards expenses are 1.57 percent.

Chargebacks in particular are eaten by the merchant, not the credit card company, unless the merchant puts in the effort to show that the chargeback is incorrect or fraudulent.