Danger pay is a tricky thing. Lots of jobs are dangerous.
My chosen career has me under 4000 pound hunks of metal held up by 4 posts that were positioned in a hurry for a big part of the day. And often dealing with electrical equipment and open flames in oil soaked clothing. And guiding barely-driveable vehicles around enclosed spaces... and test-driving cars that may or may not be about to fall apart... and and and, you get the idea. But there's no suggestion of auto technicians getting danger pay. Garbage truck operators (at least the ones on the back of the truck) are far more likely to die on the job than firemen, but most of us balk at the idea of a 100,000 dollar garbageman. Of course expertise comes in to that, but it's still clearly not a big part of the formula. Soldiers, of course, stand out as being massively underpaid, and really often live a similar lifestyle and workload to firefighters (lots of downtime but huge risks once in a while).
I think, in the case of firefighters, the question that needs to be studied is how far you could lower wages for a given reduction in effectiveness. If you could project for instance, that a 15% reduction in average salary would mean a 15% increase in 'preventable' deaths or injuries involving firefighter action or lack thereof, that's probably not worth it to most people. Where is that drop off? Is the drop off at a point in terms of savings where it's just not worth the hassle? What would that reduction in quality of service do to the related costs to the government? If 10% cheaper firemen use up 1000 dollars more equipment a year and response time is increased by 15 seconds on average, is that worth it?
Firefighters are really a weird case in that the individual firefighter, and probably even full fire stations, don't really 'do enough stuff that you could statistically analyze. You can check out a cop's conviction rate, who he arrests, how much mileage he puts on his car, etc. But if a firefighter only gets hands-on with 3-4 fires and 10 injuries a year, that's probably not something you can draw real conclusions from.
