Personally I use the simple method of making a note of how far I've travelled since the last full tank, and then note how much has to be put into the tank to replace what's been used.
The car's reported average mpg can be considerably off in either direction (up to 20%). I suspect this is something to do with the profile of journeys, eg cold starts, short or long trips, temperature, humidity, etc.
This only really works to illustrate the unreliability of what the car reports. It's not so useful in measuring the 20 vs 30 fuel consumption, because I'd have to drive quite a long way at 20mph to use enough fuel to remain statistically accurate.
However, with respect to pollution, the 20mph vs 30mph argument about fuel consumption is missing the point. As various people here have pointed out, the bulk of per-car pollution comes when engines are cold, i.e. in the first 5 miles. Using pollution reduction as an argument in favour of 20mph is just rearranging the deckchairs on the Titanic. Cold-engine pollution will continue to hugely outweigh the 20-or-30 difference.
(this ignores the predicted traffic-smoothing effect of 20mph limits which may have a separate beneficial effect on overall fuel consumption)