There are not an infinite amount of asteroids, and in large time frames you can reduce their numbers. Therefore, we have a large hit once a century. Large, but largely inconsequential (The Earth is empty. Try this: Pick random coordinates and look them up on a map. Repeat. How long before you end up w/ 10 km of an inhabited area?)
Theories are infinite, funds are not. This is an ineffective solution (we can only deflect small and well predicted orbits) to an unlikely problem: mass extinction from asteroids hasn't happened in a very long time (millions of years) -> Bayesian inference -> very low probability of occurring soon. It will happen. But in a millions years
Compare that with the plethora of reasons we might die out in the next 10, 20, or 100 years.