Steve.....
The VGS for the part is a measurement of voltage right from Gate to Source. When you connect the load in a common source configuration (Ie the voltage regulator is in the source to GND current path then the available drive voltage from the microcontroller to the MOSFET is reduced by whatever drop develops across the source load. In this case a bias state is established wherein the 3.3V drive from the microcontroller is barely turning on the MOSFET. The regulator probably draws little or no current until its input bias rises above 1.2 volts. In your case you are seeing the thing start to pull current just above this level (ie 1.3 volts). This 1.3 volts is subtracted from the 3.3V microcontroller drive signal leaving a 2.0 volt VGS at the MOSFET. At this drive the MOSFET looks like a resistor of quite a number of ohms.

You are correct that a P-Channel MOSFET would be the correct type of device to put in front of a voltage regulator. Then its VGS is relative to the +V supply. However then in order for the 'LPC932 to drive the gate you would still need to use a buffer to translate the 0->3.3V swing for the output to the V+ to GND swing for the P-Channel MOSFET. By the way...the 2N7002 would be a good choice to do this buffer...tie its source to GND and the drain to the gate of the P-Channel part AND a resistor that goes off to V+.

Michael Karas