Something bugs me about Italian opinion polls. There's not enough variance: FdI has been at 27-and-a-bit in 12 of the last 13 polls. My suspicion was herding, so I went down a statistical rabbit hole, which involved inventing a new measure of herding (which may have picked up a different issue!). 🧵

Bots/LLMs might also ne an issue, no?

So this is the rolling-average ISS for Italy (in blue). For 15 years the blue line is at ‘sensible’ values (~1000: roughly the preference-expressing respondents you'd expect in an Italian poll). From 2020 it EXPLODES. Polls now behave as if samples were 3x their reported size (orange/green lines).

Ever wondered what attributes are relevant in your conjoint experiment? And what assumptions you need to estimate it? Asya and I got you covered. In a new working paper we propose the APNS: the probability that an attribute's level change is necessary and sufficient for the selection of a profile.

Thank you for the praise ;)

Unsere Band heißt einstürzende Brandmauern und das nächste Stück heißt: Das BSW hat nur fünf Mandate.

The CDU is crucial to the stability of German democracy, but this doesn’t look stable at all.

Noch einmal zur Erinnerung: In Berlin gibt es seit eh und je eine stets knappe Mehrheit der Stimmen für Parteien links der Mitte. Was sich verschoben hat, sind die Kräfteverhältnisse innerhalb des linken Blocks.

We are hiring at LMU Munich: two PhD positions and one postdoc on democracy, political psychology and computational social science. Start: January 2027 or as soon as possible. We review applications on a rolling basis until October 20. Details below. Reposts very welcome! 1/7

Unterschiede gibt es allerdings in der Struktur der Agrarwirtschaft und dortigen Besitzverhältnissen. Das preußische Vorpommern ist stärker von Großbetrieben geprägt, häufig in der Hand adliger Junkerschichten. Die Landwirtschaft in Mecklenburg eher von kleinbäuerlichen Familienbetrieben.

Someone is clearly tenured...