Hi Ian,
Your question is an interesting one as it would allow tug fans everywhere to correctly replicate, in a dynamically correct way, the actual towing behaviour of the full size vessel. Often we are content to just divide the full size engine power by the cube of the scale factor, however as has been correctly stated we need to allow for the greater losses between motor/propshaft and bollard present at model level due to our inability to scale down water.
By far the best way then is to measure by effect (bollard pull, what we are all after) rather than cause (engine power, which is reduced by varying and often practically unmeasurable drivetrain efficency, apparent and real propeller slip, etc) when setting up our model powerplant and deciding what prop. diameter/blade number/pitch, motor voltage/gear ratio etc. we will experiment with. Calculators at the ready for the next part!
Simply stated without all the symbols etc. found in the naval architecture/marine engineering texts,
scale bollard pull = 1/scale factor
2.5So if we have, say, a 1/100 scale model of a tug with 90 tonnes bollard pull, the model should have a bollard pull of
90 tonnes divided by (100 to the power of 2.5)
which works out as
0.0009 tonnes = 0.9 kg bollard pull from the model.
The formula applies of course to all boat/ship models and you do n't need to have a scientific calculator, just divide your full size bollard pull by the scale factor, then again, then finally by the square root of the scale factor to get your answer...oh and don't forget to convert from tonnes to kg by multiplying by 1000 at the end...or we are gonna see more underpowered models out there!

Hope this helps!