The effects of averting behavior can be complicated by the fact that the pollution is non-rival and non-exclusive (it affects everyone), while the averting behavior is typically private (only the person spending the money benefits). Suppose drinking water quality for everyone is improved (for instance, by cleaning up a toxic spill). Are averting expenditures on water filters likely to increase or decrease in response to the pollution cleanup? If the water filter continues to reduce risk by the same amount, will the benefit measured by the averting behavior method be the same as when water quality was worse? If the value changes, does this mean that the measured benefit is a highly variable number (and therefore more likely to be questioned), or is there a reason that we would expect expenditures to change when the risk is different?