The
datasheet is fairly useless, but like most CMOS logic, I'd expect the supply current to be proportional to the clock frequency. As its internal oscillator jumps from 16KHz low power mode to 512MHz Fast mode (a factor of 32) for up to twelve seconds after a touch is detected, I would expect its current consumption to vary by at least an order of magnitude, depending on what proportion of its standby current is due to the 1MHz sensor oscillator.
A low current potential divider is therefore unsuitable to power it - you'd need up to 0.5mA standing current in the divider to get a reasonably stable (+/-10%) output voltage.
The best option if no lower voltage is available would be to use a low quiescent current 5V linear regulator, as shown in the datasheet application circuit. Make sure the regulator has an adequate input voltage rating - less than 18V could be problematic due to connection transients, and use an ordinary aluminum electrolytic of x10 the value in parallel with any ceramic input capacitor to damp connection transients.
OTOH its quiescent current is low enough that it could be supplied from a tap on the battery pack e.g via the balance connector on a RC LiPO pack, from the bottom cell for a 3.0V to 4.2V supply without significantly unbalancing the pack. If you are using Alkalines, I'd suggest tapping across the bottom three cells of the pack for a 2.4V to 4.7V supply.