How do votes flow? Local and general elections in İstanbul and Ankara

In the same city, the parliamentary and metropolitan elections produce very different patterns of vote movement between parties. And how much of that movement is a real "transfer," and how much is just appearance?

9 June 2026 · Official YSK data · ballot box / neighborhood level

The question

"When a party's vote drops, where does that vote go?" We cannot see this directly from ballot-box results — no one writes down on the ballot whom they voted for. But we can use a proxy: across the units of the same area (neighborhood/district), as party A falls, does party B systematically rise? If it does, there is a vote swing between the two. This is not a definite transfer; but it shows which parties are each other's "rival pool."

How do we measure this? — The seesaw test

We look at whether two parties move like a seesaw: if party A falls while party B rises in a neighborhood, then there is an inverse swing (seesaw) between the two. The more regular this is, the "stronger" the relationship. Important: this does not mean "A's vote went to B" — it only shows that the two swing in opposite directions.

To wipe out misleading regional differences, we made the comparison among the neighborhoods of the same district (so the "this district is conservative, that one is secular" effect is removed).

1. In the parliamentary election, which party does AK Parti seesaw with?

Between the last two parliamentary elections (2018 → 2023), how much did AK Parti's vote swing inversely with which party? The labels show seesaw strength as Strong → Moderate → Weak → None; the number in parentheses is the technical value (correlation; closer to −1 = strongest).

The result changes strikingly from city to city:

• In İstanbul, AK Parti's strongest seesaw is with Yeniden Refah and the HDP family. With MHP it is almost none — both are under the People's Alliance (Cumhur İttifakı) umbrella and move together. With İYİ it is also none.
• In Ankara the picture is different: the strongest seesaw is with MHP (Moderate); the rest (CHP, HDP, Yeniden Refah, İYİ) are spread out at a weak level.

For the comparison to be fair, a single period (2018–2023) was taken; this way all the parties competing in this interval are compared on equal terms.

2. The local election changes the picture

Because in the metropolitan mayoral election the small parties largely dissolve under two blocs, the race narrows to the AK Parti ↔ CHP axis. The result: in the same city, in the same neighborhoods, the AK Parti–CHP swing becomes many times stronger than in the parliamentary election.

In other words: the AK Parti–CHP seesaw, which remains only weak in the parliamentary election, becomes much stronger in the mayoral election. The "who should govern" question and the "who should represent us" question produce different voting behavior.

3. But is this swing a real "transfer"?

A strong seesaw (as we see in the İstanbul metropolitan election) immediately prompts the sentence: "AK Parti voters switched to CHP." But this is a dangerous leap. There are two traps:

• Mechanical constraint: in a two-candidate race the shares necessarily add up to 100% — if one falls, the other rises mathematically.
• Turnout: if in a neighborhood one bloc's voters stay home while the other's go to the polls, the shares swing inversely without any vote "transferring" at all.

To separate this out, we built a transfer estimate (a statistical model): we split registered voters into four groups — AK Parti · CHP · Other · Did not vote — and estimated the flow from the 2019 metropolitan election to the 2024 metropolitan election.

Here are the two different stories behind the same correlation:

In İstanbul, the strong AK Parti–CHP swing is almost no direct transfer at all: CHP held its own base tightly, while AK Parti lost its vote mostly to other options and to not voting. As one shrinks while the other stays constant in the same neighborhood, a share see-saw forms — but the cross-over transfer is faint.

In Ankara, by contrast, the picture is different: there is a real direct AK Parti→CHP transfer (~16%), and on top of that half of the other parties' voters also gathered at CHP. So the swing in Ankara is fed both by demobilization and by real cross-over transfer.

Conclusion

A three-layered picture emerges:

1. In the general (parliamentary) election, AK Parti's seesaw partner changes by city: in Ankara MHP stands out, in İstanbul Yeniden Refah and the HDP family. The AK Parti–CHP seesaw is only at a weak level in both cities.
2. In the local election, as the race narrows to two blocs the AK Parti–CHP swing becomes much stronger — but this is not proof of a direct transfer.
3. The source of the swing changes from city to city: in İstanbul it is mainly base loyalty + turnout; in Ankara, on top of those, real cross-over transfer and opposition consolidation.

Data and method. Source: Supreme Election Council (acikveri.ysk.gov.tr), ballot box / neighborhood level. Elections: parliamentary 2011, 2015 (×2), 2018, 2023; metropolitan 2009, 2014, 2019, 2024. Correlations with a within-district design (≥10 neighborhoods), Pearson. Transfer estimate: weighted non-negative least squares (Goodman/NNLS) ecological regression, four categories based on the registered electorate.

Limits. Ecological inference does not pin down individual behavior with certainty (aggregation bias). Transfer percentages are indicators that can stick to the corners; for interval estimates King EI is needed. The "Other" category includes the other candidates and invalid votes. Alliance umbrella rows are excluded to avoid double counting.

© Sandık Atlası · Impartial, source visible. This page is part of the regular "Analysis of the Week" series.