UPDATE: The New York Times has reported that the Boston Fed reduced their estimates (a tiny bit), and have added a footnote acknowledging NerdWallet and Tim for bringing some of these issues to light. FTW!
At issue: The Boston Federal Reserve is making policy recommendations based on a statistical model showing that the “poor” subsidize credit card rewards for the “rich”. While there’s nothing wrong with writing such papers as an academic exercise, you have to be a lot more careful about your assumptions when writing political papers that are meant to influence popular opinion and public officials.
The initial “WTF”: Back-of-the-envelope math doesn’t really work out. The authors assume a 0.75% reward rate on average (p. 15), so in order for the average card user to earn $1,482 they would have to spend $197,600. Even if we assume the entire merchant fee (they assume 2%) is passed on to card users, that’s a $74,100 spend, which is way too high.
Their assumption: Housing and automobile purchases should be included in estimated credit card spending.
Our debate: Their statement to us was, “In the case of mortgages and cars, the best course of action is to obtain data on actual credit card spending by merchant and product. We have not been able to get these data, however.” So rather than exclude them, they arbitrarily decided to include them. Calls to the top three mortgage lenders confirmed they do not accept credit cards, and CNW research shows that only 1 in 19 auto purchases involve any credit card transactions. Taking these out would reduce their estimates by more than $2 trillion, and their transfer estimates by more than 9%.
Their assumption: Cost of handling cash is significantly lower than handling credit cards (p. 2)
Our debate: They assume a cash transaction cost of only 0.50%, and 2% for credit cards. However, they acknowledge that they are not accounting for any fixed costs of handling cash (e.g. the costs of hiring more cashiers, security, fraud, and bounced checks), because they are not easily quantifiable. Meanwhile the fixed costs of handling credit cards are already factored into the merchant fee, which is another reason they are overestimating any wealth transfer by a significant margin.
Their assumption: Merchants pass through the full merchant fee in their retail prices, and all merchants pass through the same fee (p. 16)
Our debate: If you assume that all merchants pass through the same average fee, then you are assuming that high-income card users and low-income cash users shop in the same stores and buy the same products. However, it’s more likely that high-income card users buy higher-margin items from higher-margin merchants on average than low-income cash users, so Saks is more likely to increase prices for credit cards than a dollar store. In this case, any transfer is mitigated.
Conclusion: In Section 7 of this paper, the authors run a scenario analysis where they stress-test their own assumptions, bringing a couple of them more in line with my arguments above. They show that these new assumptions would reduce their wealth transfer estimates by 50% or so. However, if they had done the same with all of the assumptions that I question above, and brought them more in line with reality, the result would be even more drastic and would possibly even eliminate any claims of a transfer.
Read our full write-up at the American Thinker.
Also check out Felix Salmon’s rebuttal of our rebuttal on the Reuters blog.