Increasing battery voltage in portable appliances is partially driven by practicality and partially by marketing, but in the last decade or so marketing has definitely been the major factor.
A "powerful" battery powered appliance (drills probably being the most common but not the most power intensive) may have a power rating of 100's of Watts.
Take 100 Watts as an example:
At 100 Watts 12V ~= 8A, 16V ~= 6A, 24V ~= 4A, 36V ~= 3A.
Losses in wiring and connections is mainly due to heat loss = I^R.
For the same resistance losses for 12/16/24/36 volts would be in the proportions
64/36/16/9 so a 36V system may notionally have 9/64 ~= 14% of the losses of a 12V system.
So in practice, as current goes down with increasing voltage, you get less losses with the same resistance or can tolerate somewhat more resistance and still be well ahead.
In a 12V 8A system a one ohm circuit resistance will dissipate I^@R = 8^2 x 1 = 64 Watts - so as that's 64% of total power it would be intolerable. Something more like 0.1 Ohm = 6.4% would be better. It is exceedingly easy to add 0.1 Ohms in wiring and connections, so a 100W 12V system gets annoyingly difficult to build. Even an 18V system with 2/3 the current = 4/9 = 44% of the losses is usefully better.
HOWEVER more voltage requires more battery cells and the room required for interconnects, extra loss in connections and the loss of effective available volume due to square-cubed law effects* means that above a certain Voltage the extra losses start to offset the gains. Marketing doesn't care and the engineers and marketers will have had a behind the scenes stoush to arrive at the final result.
A factor that makes higher voltages easier is the use of LiIon cells. These have a nominal voltage of say 3.6V/cell which is about 3 times that of NiCd or NimH so a 10 cell NimH battery will be 12V nominal but a 10 cell LiIon of the same size will be 36V nominal.
Top grade/quality/cost power tools such as De Walt (Black & Decker in disguise) use LiFePO4 (Lithium Ferro Phosphate) cells in some products with a 3.2V nominal voltage per cell. 10 would give 32 V nominal and this will be "almost sensible" in some applications.
An aside: I understand that De Walt use the industry leading A123 LiFePO4 cells. A123 cells are generally "hard to buy" on the retail market and I have heard of electric vehicle makers buying large numbers of De Walt battery packs to get the cells.
Square-cubed law:
Effects caused by changes in the ratio of area to volume as scale changes.
Volumes are proportional to edge^3.
Surface areas are proportional to egde^2.
so the ratio of volume to edge is proportional to edge^3/edge^2 = edge - which means that volume per surface area increases as objects get bigger.
Secondary effects of this are eg it's harder to cool big things by surface radiation.
Conversely, it's harder to keep small things warm when it's cold.
For a given surface thickness big things have less content per volume.
The latter effect affects batteries.
if a battery can be built with ABOUT the same wall thickness across a range of sizes then big batteries will have more active content per volume than small ones.
One only example.
Two cubes with 1mm thick walls and edges of 1cm and 4 cm.
Wall volumes = 6 x edge^3 x 1mm
Cube total volume = edge ^2
Inner cube inside walls volume ~~= (edge- 2 x wall_thickness)^3
1 cm cube inner/outer volume = (10-2)^3/10^3 = 512/1000 mm^2 = 51%
4cm cube inner/outer = (40-2)^3/40^3 = 54872/64000 = 85%. !!!
The 4 x larger edge cube is 85/51 = 1.59 x more effective a user of available volume than the small one.
Conclusion: High voltage battery packs that use NimH or NiCd may be a bad idea for this reason alone. There are others.