Is the cosmic dipole anomaly a real motion, nearby structure, or a survey artifact?

Why it matters: The cosmic microwave background (CMB) is slightly warmer in one direction, read as our motion through the universe at about 370 km/s. If so, distant quasars and radio galaxies should bunch up toward the same point, by an amount that speed sets. They do lean the same way, but two or more times too strongly: 4.9 sigma in CatWISE2020 quasars (Secrest et al. 2021), about 6.4 sigma once radio and quasar results are combined (a 2025 review). If it holds, matter and the universe’s oldest light don’t share one rest frame, an assumption standard cosmology is built on.

What’s known: Answers disagree. In Gaia’s Quaia quasars, a Bayesian analysis finds agreement with the CMB once the Galactic plane is cut (Mittal et al. 2024), while a direct count finds three to four times the expected size (Singal 2024). A joint fit with Planck finds the tension strongest for infrared quasars and weaker in radio (Land-Strykowski, Lewis and Murphy 2025), and others find it swings with sample and fitting choices. Each study uses its own catalog, sky mask, estimator, and treatment of how the survey picked its sources, so it’s unclear what drives the spread.

What could settle it: The three explanations leave different fingerprints. Our motion predicts the same ratio of measured to expected dipole, the expectation set by each sample’s own brightness counts and spectra, in every survey, wavelength, depth, and redshift. Nearby structure predicts an excess that fades with redshift. A survey artifact predicts one that tracks scan depth, dust, Galactic latitude, or sources near the faint limit, and differs between instruments. Public catalogs from independent instruments exist: CatWISE2020 (infrared), Quaia (optical, with redshifts), and NVSS, RACS, VLASS, and LoTSS (radio). The question is whether that ratio holds steady across all of them when each is measured the same way, with significance judged against simulated skies that carry each survey’s own selection and clustering.