Similarly, as the percentage of folks who use heroin increases, the relative rate of overdoses declines. But that doesn't mean that heroin itself has become any more or less safe.
The best way I can think to reject the risk-aversion hypothesis (without doing a RCT) would be to see a time-lag. If accident rates drop after ridership increases, that would suggest it's not risk-adverse people coming in later, whereas if ridership increases lead the rate drop, that would indicate that it's not a risk-tolerance thing.
On the other hand, ridership increase proceeding accident rate dropping might need to account for newer riders being less experienced.
I've read/read about other studies that found the same thing as this one, but I'm having trouble finding them right now. I found [0], but the methodology doesn't look like it would address the risk-aversion hypothesis.
Personally, I think we should encourage cycling even if it doesn't lower accident rates, but evidence of improved safety makes it a more rational choice for urban planners.
[0] https://www.washingtonpost.com/news/wonk/wp/2014/04/17/actua...
Take it to the logical extreme, imagining legalization of heroin and subsequent widespread use and sale in convenience stores, and it'd be way safer, because you'd always be getting a standardized product.
That definitely was true for a while, not sure about in the last few years. A less controversial analogy might be with coffee, where the rate of people overdosing is probably lower now that it's no longer mainly used by Sufi mystics in enormous doses for entheogenic purposes.
Most people would agree that driving has become a safer activity if you could show them that the number of injuries or fatalities per driver or distance travelled has been falling. The same is true for cycling.
For heroin, this is probably true, for cycling not so much.
I'm really not. Read what I actually wrote instead of projecting onto it.