The answer is pretty simple here. If your outcome variable has both outcomes (the event happens or doesn't happen), then you can easily fit this data to a logistic regression with the event as a binary response. If you only have data on people who have crashed, then its impossible to predict because your response is constant.
If the latter is true, you can consider tackling this from another perspective that doesn't involve the crashes themselves but instead the characteristics of the people in the data. For example, you can consider what the average age is of the people sampled, proportion of gender, etc. There may not be much you can do inferentially with this data given it only encompasses those who have crashed (which may have some bias given the non-crashers don't have a "voice" here).
Edit
Per the comments, you can theoretically flip the script here and try to predict whether crashers are male or female based off the other features in your data. For example, it may be the case that age influences the prevalence of crashes in one gender compared to another. You could fit the logistic regression such that:
$$
\text{Gender of Crasher} = \beta_0 + \beta_1 * \text{Age} + \epsilon
$$
If the binary outcome is coded as $0$ being male and $1$ being female, then the intercept $\beta_0$ would represent the conditional log odds of being a female crasher and the slope $\beta_1$ would be the increase in log odds of this event given increases in age.
However, you would need to include some very important caveats in interpretation based on the fact that you only have crashers here, which may not represent the total population. Indeed males and females who don't crash may have systematic differences from crashers that is not observed in the data, and so your assumptions about the data would be very limited. This is particularly the case with observational data. Thus I think you can probably describe in a general sense what you see but not make strong claims about the analysis given this limitation.
Edit 2
The issue with interpretation with these results lies in the fact that there is no real counterfactual involved in the hypothesis (what could have happened if they didn't crash), as shown from the screen shot of this video, which shows a typical randomization process in experimentation:
Naturally, if you report something like "crashers were twice as likely to be male" or "crashers were twice as likely to be male with a one unit increase in age", it heavily implies that males are more prone to crashing in general and may impress on people that crashing is a gender-based phenomenon, when we simply don't have the data to say that with confidence. You can instead say something like "It was found that males were twice as likely to be reported crashers in this dataset" and explain the caveats I mentioned.
The key here is to be very careful about any causal claims. You can only describe what you see in this kind of scenario. I will add that civilstat's answer is probably better given it tries to account for this in some way, but that depends on if you can get that data.
