This 'finding' is remarkable, even if the constant was found experimentally. But I have problems with the fact that the in-chain transaction volume in USD does include transactions between the same person. This invalidate the whole point in my opinion, making this perhaps a case of 'data overfitting'.
It is possible to include only transactions where a fee was paid. This may be a better metric. If someone wants to send himself money, as long as a fee is paid to miners it's still a real transaction.
