Science · 7 min read
The Hubble Tension: Measurements, Explanations, and Open Tests
A map of the measurements behind the Hubble tension and the observations that could distinguish error from incomplete cosmology.

The Hubble Tension: Measurements, Explanations, and Open Tests
The Hubble constant, written H₀, describes the present relationship between the distance to a galaxy and the speed at which cosmic expansion carries it away. Astronomers can estimate it by building a local distance ladder from nearby geometric measurements to variable stars and supernovae. They can also infer it from the early universe by fitting a cosmological model to the cosmic microwave background and other observations. These routes have repeatedly produced different values: roughly 73 kilometers per second per megaparsec for a prominent local ladder and about 67.4 for the Planck satellite's inference under the standard ΛCDM model.
That disagreement is called the Hubble tension. The measurements are not two direct readings of the same instrument. They rely on different objects, calibrations, epochs, and assumptions. The question is therefore not simply which number is correct. Researchers are testing whether an unrecognized systematic error affects one route, whether the standard cosmological model is incomplete, or whether several smaller effects combine. A 2026 community synthesis called the Local Distance Network combined correlated local indicators and reported 73.50 ± 0.81 km/s/Mpc. The discrepancy has persisted, but its cause remains open.
The early-universe route is a model-based inference
The European Space Agency's Planck mission mapped temperature and polarization variations in the cosmic microwave background, radiation released when the universe became transparent about 380,000 years after the Big Bang. The angular pattern of those variations encodes the universe's composition and geometry. Under a six-parameter ΛCDM model, the Planck collaboration's 2018 analysis inferred H₀ near 67.4 km/s/Mpc with a sub-one-percent quoted uncertainty.
Calling this an “early-universe measurement of H₀” is convenient but incomplete. Planck observed the microwave sky. The value of H₀ emerges from fitting that sky with a model containing cold dark matter, a cosmological constant, ordinary matter, and assumptions about early physics. Baryon acoustic oscillation data can reinforce the low value by measuring a standard scale in the later distribution of galaxies, but that standard ruler also depends on the sound horizon established before recombination.
This route is powerful because the model fits a wide range of observations with few parameters. Its precision is conditional on that model and the treatment of instrumental and astrophysical effects. A small error bar does not cover the possibility that ΛCDM omits a relevant component. Conversely, disagreement with a local ladder does not by itself identify new physics.
The local ladder joins several measured rungs
The SH0ES collaboration's distance ladder begins with anchors whose distances can be found geometrically. Examples include Milky Way parallaxes, eclipsing binaries in the Large Magellanic Cloud, and water masers orbiting the black hole in NGC 4258. Astronomers use those anchors to calibrate the relation between the pulsation periods and luminosities of Cepheid variable stars. Cepheids in galaxies that have hosted Type Ia supernovae then calibrate those brighter explosions. Supernovae farther into the smooth Hubble flow connect distance with redshift.
In a 2022 Astrophysical Journal Letters analysis, Adam Riess and the SH0ES team reported H₀ around 73.0 km/s/Mpc with about 1 km/s/Mpc uncertainty. The result differs from Planck's ΛCDM inference by far more than the quoted statistical errors. Yet the ladder contains possible systematic vulnerabilities: dust, stellar metallicity, photometric zero points, crowding of Cepheids by unresolved neighboring stars, supernova selection, and the treatment of peculiar galaxy velocities.
Researchers have attacked these vulnerabilities by changing instruments and tracers. The James Webb Space Telescope has finer infrared resolution than Hubble. Comparisons of Cepheid photometry in the same supernova-host galaxies found close agreement between the telescopes, weakening the specific proposal that Hubble crowding alone creates the high local value. The size and calibration of JWST samples still matter, and teams have differed over how much existing Webb data reduce the statistical significance.
Other local indicators include the tip of the red-giant branch, carbon-rich J-region asymptotic giant branch stars, masers, surface-brightness fluctuations, and gravitational-wave standard sirens. Wendy Freedman's Carnegie-Chicago program has obtained values between the best-known Planck and SH0ES estimates in some analyses. Such results may indicate residual calibration differences, or they may reflect smaller samples. They do not yet supply a single independent verdict.
The H₀ Distance Network collaboration's 2026 Astronomy & Astrophysics report treated these indicators as a connected graph rather than a collection of isolated ladders. Its covariance-weighted baseline changed little when Cepheids or red-giant-tip measurements were removed, supporting internal consistency among its selected local routes. That is a stronger cross-check than one ladder alone, but it remains a published analysis with choices about data inclusion, correlations, and calibration. A separate 2026 reanalysis using different priors reported a lower local value, illustrating why model and prior sensitivity remain active subjects rather than closed bookkeeping.
A tension can survive while its interpretation changes
The reported number of “sigma” separating results changes with datasets and covariance assumptions. Sigma summarizes how unusual a difference would be if specified statistical models and error estimates were correct. It is not a probability that ΛCDM is false. Unknown systematics are especially difficult to represent in a single significance value.
Several cross-checks nevertheless make a one-line error explanation less plausible. Reanalyses of Cepheid photometry, additional geometric anchors, Webb comparisons, and alternative supernova calibrations have not revealed a correction large enough to bring the canonical SH0ES value to 67.4. At the same time, several local methods do not cluster at 73 with comparable precision. The evidence supports a persistent inconsistency between particular high-precision inference chains, not a unanimous split across observations of the early and late universe.
Large-scale-structure measurements add another dimension. The Dark Energy Spectroscopic Instrument maps baryon acoustic oscillations across cosmic time. DESI's 2024 and 2025 releases strengthened discussion of whether dark energy might evolve when BAO data are combined with certain supernova and microwave-background datasets. That issue is connected to the broader adequacy of ΛCDM, but changing late-time dark energy alone does not automatically resolve H₀. A successful model must fit the microwave background, BAO, supernovae, structure growth, and local distances together.
New physics proposals must pass several tests
One large family of proposals changes physics before recombination. “Early dark energy” models briefly add energy density before fading, reducing the inferred sound horizon so a higher H₀ can fit the observed angular scale. Other proposals alter the number or interactions of relativistic particles, recombination, neutrino properties, gravity, or primordial magnetic fields. Late-time proposals change dark energy, gravity, or the local expansion environment.
These models face a common constraint: the standard model already fits much of the data. Moving H₀ can worsen the fit to the microwave-background damping tail, galaxy clustering, gravitational lensing, or element abundances. Reviews by Eleonora Di Valentino and colleagues and later model-comparison studies show that some extensions reduce the numerical tension, but added parameters and conflicts with other observations often limit the improvement. No alternative has secured field-wide agreement as the replacement for ΛCDM.
A local cosmic underdensity could make nearby expansion look faster, but surveys constrain how large and unusual such a void could be. Cepheid physics could still contain subtle population effects, yet the required shift must remain compatible with multiple anchors and infrared observations. The plausible search space has narrowed; it has not closed.
Observations that can discriminate
Better precision matters only when it attacks different assumptions. Larger JWST samples can compare Cepheids, red-giant tips, and J-region stars in the same hosts. More megamaser distances can add geometric anchors. Gravitational-wave standard sirens infer distance from waveform amplitude and can provide a ladder-independent H₀ when host redshifts are known or statistically modeled. Time delays between images of strongly lensed quasars and supernovae supply another route, although lens mass models introduce their own systematics.
On the early side, improved microwave-background polarization, small-scale anisotropy, and large-scale-structure surveys can test changes to the sound horizon and structure growth. A proposed solution becomes persuasive only if it explains the discrepant H₀ chains while predicting patterns that new data then observe.
The established finding is a durable mismatch between leading Planck-ΛCDM and Cepheid-supernova-ladder estimates. It remains contested how other indicators should be weighted and whether remaining systematics are large enough. The unknown is the cause. More decimal places on either route will not answer that alone; the decisive evidence will come from methods whose errors and model dependencies overlap as little as possible.
Sources
7 sources, none of them with a link, so treat them as reading, not proof.
- Planck Collaboration, “Planck 2018 Results VI: Cosmological Parameters,” Astronomy & Astrophysics, 2020.
- Adam G. Riess and the SH0ES collaboration, “A Comprehensive Measurement of the Local Value of the Hubble Constant,” Astrophysical Journal Letters, 2022.
- Wendy L. Freedman and collaborators, Carnegie-Chicago Hubble Program measurements using the tip of the red-giant branch.
- Adam G. Riess and collaborators, JWST–HST Cepheid photometry comparisons, 2024–2025.
- Eleonora Di Valentino and collaborators, 2021 review of proposed Hubble-tension solutions, Classical and Quantum Gravity.
- DESI Collaboration, first- and second-data-release baryon acoustic oscillation results, 2024–2025.
- H₀ Distance Network Collaboration, “The Local Distance Network: A Community Consensus Report on the Measurement of the Hubble Constant at ∼1% Precision,” Astronomy & Astrophysics, 2026.
