Using a Scan Gauge II, one can accurately measure instantaneous fuel economy. Since the late 1980's, every single fuel injected vehicle on the planet has had the ability to monitor its instantaneous fuel economy but almost every automobile manufacturer chooses not to include this "planet saving" feature in their vehicles. Why? I don't know. 

While driving my wife's 2002 Chevy Venture, I found a less-busy section of interstate where I could setup and run the test. With the aide of my mom who was traveling with me, I recruited her to assist me in recording the data. 47 years ago, she typed my dad's Masters Thesis. Mom was more than capable of writing down a few readings while I focused on keeping the van traveling between the lines. 

After I accelerated above the desired speed, I let off the gas until the van began coasting. This allowed the van to settle out and decelerate to the correct speed. Note: All freeway speed measurements were taken on a flat, strait stretch of road during a high wind storm. This unusual weather provided a nearly constant 25 mph tail-wind which allowed the vehicle to decelerate more slowly. This aided me in taking more accurate, high speed readings.

Initially, the data was inconclusive. The first 2 sets of data showed that coasting in neutral got better gas mileage. The second set of data points revealed that keeping the van in gear was the more fuel efficient choice. 

I decided to try this experiment again later that night, this time running at slower speeds in an undeveloped residential neighborhood. By this time, the wind had died down to only a few mph. With the aid of my 9 year old daughter, we ran 2 sets of “coast tests” at 30 mph and 2 sets at 25 mph. Each set was driven in an opposite direction to eliminate the wind as a factor. But once again, the first 2 tests revealed that Neutral is the superior choice and 2nd set of tests demonstrated that coasting in gear was better. What is going on here? 

Glancing down at the Scan Gauge II display, it hit me. Engine temperature!!!

When an engine first starts up, it is running at temperatures that are out of the operating range of some of its sensors. To combat this, the vehicle's control system runs in an "open loop" (does not receive feedback from its sensors) so it plugs in default values for the sensors. 

As the engine warms up, the engine's control system switches over to “closed loop” where it makes decisions based on the actual readings from the sensors. 

While running with a cold engine, coasting in gear gets lower fuel efficiency and coasting in neutral achieves higher efficiency. After the engine warms up, the opposite is true.  

The effect I was seeing is not related to the closed loop/open loop. After more tests, the Chevy Venture goes into closed loop within a few seconds of starting up the engine. Whatever the phenomenon, the increased in-gear efficiency only happens after about 10 minutes of driving and after the engine coolant temperature reaches about 185 °F. I assume it is when the thermostat opens. Whatever the effect, it is re-producible, predictable and measurable.