Here is my Don Quixote Moment:
Lotta money being wasted by some of you people. Don't you know that air filter efficiency improves as the filter loads up? You are actually doing more harm replacing a slightly dirty filter for that reason, plus when you open the air box up, you are letting dirt in by accident. On top of that, as you will read below, fuel economy isn't effected until the filter gets so plugged the car basically won't run. Power may be effected but since the average filter has 20-40 percent more airflow capacity than the engine can use, you have that percentage to play with before power is effected. There's where a restriction gauge plays a part (read below).
I have access to a flow bench (Superflow SF600) and tested the stock air filter assembly on my F150. It flowed an average of 621.58 cfm @ 28" H2O. When I calculate the airflow needs of the 5.4L engine at it's rated maximum power (300 hp @ 5000 rpm) it's 480 cfm figuring 100% VE, so I have 23 percent to play with (it's less really because 100% VE isn't often achievable... usually 0.85 is more real world and that gives me nearer my 40 percent)
So, my cars now have air filter restriction gauges and the air fitlers aren't touched until restriction reached 2.5kPa or, if I have the manufacturers spec, I'd use that. Actually, 2.5kPa (about 10" H20 may be conservative)... the generic upper limit I've found is 5kPa (20" H20). Going on about 5 years with our Honda and three on the F150 and neither is anywhere near 2.5 kPa.
LookHere:
Air Filters & Fuel Economy http://www.fueleconomy.gov/feg/pdfs/Air_Filter_Effects_02_26_2009.pdf
This 2009 report tests the effect of dirty air filter on fuel economy using standard EPA dyno routines. They used later model EFI cars (2003-2007)and one 1975 carbureted car. The conclusion what that the fuel injection could compensate for the restriction up to the point where the car's performance had deteriorated to barely running and fuel economy decreased only by a percent or so. Performance degraded, of course. The carbureted car was more effected but fuel economy wasn't lowered by nearly as much as "common knowledge" dictates... only 2.5 percent.
In every case lower end performance was unaffected until the restriction reached the level at which insufficient air was being inhaled. Performance dropped mightily at that point but the EFI could still trim fuel to keep the mixture about right. They found that in "normal" driving situation, the car was almost unaffected by a clogged filter because the airflow needs were so low. In a few cases, the filters were so clogged that the engine sucked them into the intake tube when they tried to make full power tests.