So, worth saying a few things about this tweet from Tara Roos (x.com/tarathinks/status/2064…) as it is deeply problematic for political reporter to misunderstand the relationship between NPEs and LGEs in this way. It is a mistake she has made often, on multiple platforms. NPEs are not a close proxy for LGEs, they are fundamentally different. They are also not our “closest proxies” either. Those would be by-election results (trends more than percentages, with a few exceptions) and market research. Why are they not close? Well, 1. They have fundamentally different turnout levels. The 2021 LGE, for example, saw 11.7m voters vote on the PR ballot. The 2024 NPE saw 16m voters vote on the provincial ballot. That is a difference of 4.3m voters or 27%: two completely different universes. 2. Obviously, people vote for provincial governments, not local governments, on a provincial ballot – your town, city or metro is simply not on the table. 3. The voters roll itself changes, although generally not a profound influence it most definitely is one. It is true, in general terms, there is a proportional “relationship” in terms of trends between NPEs and LGEs, but to describe the one as a close proxy for the other, is like saying a motorbike is a close proxy for a bicycle. It is also just silly. No serious electoral analyst would ever say or use the two ballots in this way. It is an amateur error, that leads to all sorts of wrongheadedness. For example, Roos argues elsewhere, “Statistically speaking, right, the best data set for us to look at, is actually the 2024 national elections, on the provincial ballot, in the city of Joburg. The DA got 353,000 votes. If everyone who voted for the DA provincially in 2024, voted for the DA locally, she [Zille] would only have 24.9%”. This is called a tautology. And nonsense. What Roos is saying is, “if the provincial election was a local election, the DA would have got a provincial result.” That is a nothing observation. In order to say what I think Roos is trying to say, you need to change the turnout: If the DA gets 353,000 votes in JHB on the PR ballot in 2026, and turnout is the same as 2021, the party would then get 38.1%. That is a far more credible interpretation, but even that is of no real help, because lower turnout impacts different parties to different degrees. You can be very precise about all this, and run different turnout models (scenarios), which tell you exactly what percentage a vote total would transform into at which turnout level, it’s a complicated business (see here: inside-politics.org/2026/02/…) but really that is the best you can do. Unless you're using market research, or a nuanced and comprehensive comparative read of by-election results, it’s just pointless. It is deeply disturbing that this is the Business Day’s current parliamentary political reporter. These are profound mistakes, and the sign of someone who does not understand elections, or election data, or how to read the two things.

When you say ‘growth of PA and ActionSA’ what data do you base that claim on? Specifically in CPT. Does the outcomes of bi-elections (in CPT and surrounding municipalities) support that claim? Especially in instances where all 3 those parties were on the ballot.