Once you separate completed sales from withdrawals, the usable sample may be too small for a broad Tokyo conclusion. That is still better than treating every vanished listing as a buyer found within 26 days.
I would keep the analysis narrow: use consistent neighbourhood boundaries, then record...