The number of unique daily addresses used does measure a degree of economic activity, making it a fundamental measurement rather than a technical trading measurement. The amount of new money coming in is unfortunately proprietary information kept by exchanges. The importance of the two fundamental indicators of adjusted transaction quantity and daily unique addresses, as calculated by Blockchain.info [ ... ]
However, since both new addresses and transactions are free, users and programmers have no reason to use them sparingly. A substantial and variable fraction of those counts could be "fake" transactions, between addresses owned by the same person -- such as tumbling, hotwallet/coldwallet motion, software testing, etc.. Bitcoin deposits and withdrawals at the exchanges should also be counted as "fake" in this sense. The blockchain traffic may also include testing of new altcoins or services built on top of bitcoin.
Unfortunately there is practically no reliable and meaningful data on the bitcoin economy.
To correct a few assumptions: there is a small transaction cost as well as a management cost in handling many private keys and an escalating risk in dealing with more keys than one can comfortably secure. (If this is happening and I'm sure it is on a small scale, it is a practical case of big wallets subdividing, to represent growth that has already happened)
Tumbling addresses as I understand are not part of the final count as they end up empty. And deposits at an exchange are most often pooled often in identified address, and wouldn't add a significant count. (I've done a few purchase lately with bitpay and found they even reuse address that end up empty so that doesn't add either. I think exchange withdrawals are a legitimate addition to the total as those are the final unique address holding coins that are counted.
While the measurements may not be accurate I feel confident they are magnitudes more accurate than the tools to manage existing monetary policies.
As things stand this metric is waiting validation, and manipulation doesn't prove anything or give anyone a quantitative competitive advantage just yet.