Phone number is not about identity, but about reducing spam (they're also not at all tied to your identity, at least not unless you buy data from service providers). The point is that you need something that is very easy for your normal user to do, but very hard for bots - effectively, it's just a glorified captcha.
Often you're not applying these policies to everyone, but only to small sub-populations with a high density of abuse. E.g. if 1% of your accounts are created via VPNs + VOIP numbers, and 99% of those accounts are deemed to be abusive post facto, and you restrict VOIP numbers for just signups for VPNs, the absolute maximum reduction in legit signups is 0.01%. You don't even need to know how many of the legit users would find another way to create the account; you already have an upper bound that's within the guardrails.
Or you can look at what actually happens after you roll out the policy. If it really is watertight, abusers will move away quickly, and all you're left with are the legit users. If that number is low enough, you might even keep the rules in place indefinitely even though all the traffic that is remaining is expected to be FPs.
If those approaches sound dodgy, the default option is to run an A/B experiment. Ban VOIP numbers for the experiment group and allow them for the control group. Then simply look the number of newly created abusive accounts vs. accounts with legit interactions in the two groups.
(A lot of requires you to have a way of distinguishing between abusive and legit users post facto, but if you don't, it's too early to add this kind of restriction in the first place.)