Levels of parental risk aversion can remain fixed while behaviors change dramatically

If the function at play is consistent incremental risk preference. If the odds of all cause mortality for a 5-10 year old are 60 in 100,000 in 1950 and you allow or deny your child to undertake activity X which increases or decreases their risk by 6 in 100,000 then you have adjusted the risk of catastrophe by 'only' 10% of total risk. If the odds in 2020 are 10 in 100,000 then the same refusal improves your odds by over 50%. If there is an ongoing risk reduction over time (driven by societal factors or broader health improvement or whatever) and each parent responds to the prevailing rate, then specific actions become 'riskier' and the reward for further constraints becomes 'greater'. This doesn't require any change in relative risk preference across that period, and the effect could be observed entirely via improvement in very basic health and safety (e.g., vaccines reducing death from disease, seatbelts reducing death in car accidents).

In the above scenario the important claim is that the perceived risk of X has not changed. This is not a Weber's Law scenario in which people 'notice' risks they would not have before due to lower background risk rates. People in 2020 do not incorrectly believe that these activities are more dangerous now than they were for their predecessors, and predecessors did not ignore the risk or believe the risk levels to be lower than they actually were. Nonetheless, since in all periods the function at play is incremental risk mitigation the reasonable decision to make regarding X can shift.