Emmanuel Saez, et al, out of Berkeley modeled we would see 10% leave. He is one of the world's most respected economists. Here's a quote from the paper used to model this:

"Won’t billionaires leave California to avoid the tax?

Billionaires who are California residents, as defined by law, as of January 1, 2026 will have to pay the one-time tax in full; leaving after January 1, 2026 will not allow billionaires to avoid the tax. Note that the law for determining residence is not a simple binary under California law; it is a standard that looks to both subjective and objective factors. Therefore, the taxpayer’s residency is already largely set and moving does not allow billionaires to avoid the tax."

So, the premise is essentially "we'll catch them by surprise, so they won't leave." The whole model is built around this element of surprise. From a modeling standpoint, that feels risky. And more than that: zero variance. Exactly 10% is the impact.

We already have data suggesting reality is a bit messier. If you look at just recent high-profile departures, six billionaires publicly left the state: Larry Page ($269B), Sergey Brin ($245B), Peter Thiel ($28B), Don Hankey ($8.2B), Steven Spielberg ($7.1B), and David Sacks ($2B). Plus, the paper is based off of Forbes, and there is a lot of older, quieter money that Forbes isn't going to capture.

The problem is once somebody leaves, they take away some amount of their "normal" ongoing tax revenue. Then, because the precedent is set, you risk making anyone sensitive to this,like budding billionaires, htink twice about staying. I still think the state gets a nice one-time payment, but it might not be as much as projected, and it could create a hole in the broader revenue system. If you run a Time Value series, the ROI is marginal or may even turn out to be negative.

The paper then seems to assume that after this surprise tax, the next crop of billionaires will just shrug it off and continue to be created in California at the same rate. This is a bold statement, unbacked by data, with no potential whiskers on the bar. As an engineer, no process has zero variation, but that's what their paper implies. By the way, I think European data clearly shows that without surprise, it obviously cannot work. And so surprise is great, but it has a tendency only to work one time.

The final twist is we'll likely see a massive ad campaign by these individuals to market against the bill. The biggest issue is now if the proposal fails at the ballot, the state might have damaged trust in its revenue system, didn't get the one-time boost, and an already big problem potentially gets worse.

I'm not saying that you can model everything. However, I think some modelling up front, with an error bar, would really be a good place to start. If the 10% leaving is wrong, we need to reset the model. The latest version of the paper, updated in Feb 25, doesn't do this.