South Africa is in the process of a political realignment. The prevailing trend in this realignment is the growth of the extremes in our politics at the expense of our more centrist parties. The 2019 election was the first step in this direction. As previously documented, the results were driven by three separate, but related, trends. 1. The FF+ made significant inroads at the DA's expense in the northern white electorate, particularly in voting districts with large concentrations of more conservative Afrikaans voters.2. The EFF made significant inroads at the ANC's expense in the black electorate, in every province, but lagging slightly in the Eastern Cape.3. Turnout reached a historic low for a national election, indicating growing disengagement with the political process and arguably suggesting frustration with the status quo as represented by the centrist parties, but the ANC in particular.These trends have only accelerated since the election in May. Since then, we have had a large number of by-elections across a diverse group of 366 voting districts encompassing 540 000 registered voters (of which 177 000 voters turned out for these by-elections). We therefore have an enormous 177 000-strong sample from which to draw some analytical inferences about evolving voter preferences since the May election.I found three key messages in these numbers.1) The FF+ is growing (by taking from the DA)Isolating only the suburban districts in the non-coastal (or northern) provinces, provides a useful sample to understand the trajectory of FF+ support. For the purposes of this analysis, I have isolated only the DA and FF+ votes in these areas.Sample size: 37 000 registered voters, of which 11 000 turned out for the 2019 by-electionsIf this trend continues it will be devastating to the DA's performance nationally and in particular in the northern provinces in the 2021 local government election. It may also very seriously complicate the DA's math in its attempt to maintain control in both Tshwane and Johannesburg. It should be noted that the DA has remained mostly steady in the suburban by-elections held in the coastal provinces (excluding the ones that were distorted by independent candidates).2) The EFF is growing (by taking from the ANC)Isolating majority-black voting districts within our sample (and ignoring voting districts where results were distorted by the presence of independent candidates) indicates that we have meaningful samples in four provinces: KZN, Mpumalanga, Limpopo and Gauteng. KZN, as always, is a special case and will be dealt with separately below.(Special note: The ANC and EFF are the only parties to consistently contest all by-elections in the sample, so are therefore the only parties included here. Other provinces did not have large enough samples of voting districts that were not distorted by independent candidates.)Sample size: Gauteng: 22 000 registered voters, of which 6 000 turned out for the 2019 by-elections; Limpopo: 12 000 registered voters, of which 5000 turned out for the 2019 by-elections; Mpumalanga: 51 000 registered voters, of which 15 000 turned out for the 2019 by-electionsIn each of these, the ANC is down on its 2016 results and significantly down on its 2019 results. It is important to note that the achieved 54% of the national vote in the 2016 local government election, suggesting that any further underperformance on the 2016 level, especially in this core portion of the ANC electorate, will bring the party dangerously low to the 50% level in the 2021 local government election. It also appears that the "Ramaphosa bump" the ANC experienced in the lead-up and during the 2019 national election, may be receding.3) The ANC decline in KZN is continuingKZN is hyper-diverse electorally, but we can isolate those voting districts that are IFP-ANC contested to create a meaningful sample of that particular portion of the province. These voting districts are from all over the province, but the sample is currently over-weighting northern and urban KZN (and underweighting southern KZN where ANC support is strongest).Sample size: 65 000 registered voters, of which 25 000 turned out for the 2019 by-electionsThe ANC decline within this (large) sample is rapid and could have serious implications for the ANC's national tally in the 2021 local government election given KZN's very significant contribution to overall national vote tally. This could have significant consequences for the balance of power in a large number of marginal KZN municipalities in the 2021 local government election, especially if we consider the significant growth of the EFF across KZN in the 2019 national election. (The EFF were excluded from this analysis as they did not compete in all voting districts under consideration here.)In conclusion, the numbers from these by-elections reaffirm the trends we witnessed in the 2019 national election: The EFF and the FF+ are growing at the expense of the ANC and the DA respectively and ANC support in KZN is declining.These are early warning signs that the voting patterns for 2021 may be unprecedentedly fractured, resulting in more hung municipalities (and metros) and more complex coalition negotiations/arrangements.