Measuring the Hubble Constant: The Tension

The Hubble parameter, H(t), describes the fractional growth rate of the cosmic scale factor: H(t) = ȧ/a. The subscript in H₀ means its value today. Despite the word “constant,” H generally changes over cosmic history. Comparing H₀ estimates is therefore different from comparing the expansion rate at two different times.[1]

Individual galaxies also move under the gravity of their surroundings. These peculiar velocities complicate the conversion from redshift to expansion, especially nearby. Cosmologists account for them when selecting and analyzing the galaxies used to estimate H₀.[3]

The nearby route begins with distances that can be calibrated geometrically, then connects them to brighter objects farther away. Several foundations help check the scale:

• Milky Way parallaxes: the apparent annual motion of nearby stars provides distances. Cepheid brightnesses can then be calibrated, with instrumental parallax offsets included.[4]
• Eclipsing binaries in the Large Magellanic Cloud: orbital measurements constrain stellar sizes; a calibrated surface-brightness relation supplies angular sizes. Comparing physical and angular size gives a distance.[5]
• Water masers in NGC 4258: radio measurements track gas orbiting a central black hole. The disk’s positions, velocities, and accelerations constrain the galaxy’s distance.[6]

Cepheid variables connect their pulsation periods with luminosity. Once calibrated, Cepheids in galaxies that hosted Type Ia supernovae help establish those supernovae’s absolute brightnesses. Metallicity and dust corrections are part of this process.[7]

Type Ia supernovae are standardizable candles: their brightnesses are corrected using light-curve shape, color, and other measured properties. The calibrated relation supplies distances to a more distant sample, where smooth expansion is easier to separate from local motions. Uncalibrated supernova brightnesses alone give relative distances, not an absolute H₀.[8]

The cosmic microwave background contains acoustic patterns from the early universe. Their spacing and relative strengths constrain the primordial plasma and a characteristic sound-travel scale. A cosmological model connects that physical scale to its observed angle and the distance to last scattering. The fitted expansion history then gives H(z), including H₀ at z = 0.[9]

The usual reference is flat ΛCDM: general relativity with ordinary matter, cold dark matter, radiation, and a cosmological constant, in a spatially flat background. Planck’s commonly quoted H₀ result is an inference within that model, using temperature, polarization, and CMB lensing measurements.[10]

The investigation extends beyond one satellite. ACT’s 2025 DR6 analysis compared temperature and polarization measurements across instruments and angular scales. Its combined ACT–Planck fit, including CMB lensing and DESI DR2 baryon acoustic oscillations, gave 68.43 ± 0.27 km/s/Mpc within flat ΛCDM. This is a joint result, not an ACT-only measurement or a model-free answer.[11]

A useful benchmark compares Planck’s 2018 results with the SH0ES team’s 2022 Cepheid–supernova analysis. These dated results make the central issue visible; they are not a complete inventory of current methods.

For these two values, treating the reported errors as independent and approximately Gaussian gives a separation of about 4.9 standard deviations. That calculation explains the familiar “roughly five sigma” description. It is conditional on the datasets, uncertainty estimates, and model being compared.

The revised 2025 Chicago–Carnegie Hubble Program report obtained H₀ = 70.39 km/s/Mpc from its adopted HST-plus-JWST TRGB analysis, with separate uncertainties of ±1.22 statistical, ±1.33 systematic, and ±0.70 from supernova calibration. With this broader error budget, the result is statistically compatible with both benchmark determinations. This result belongs to a particular analysis, rather than being the universal answer from red giants.[13]

Carbon-rich giant stars provide the JAGB method, another developing distance indicator. Analyses do not all return a low H₀: the choice of stellar fields and the way their brightness distributions are summarized can alter the calibration. Comparing the same galaxies with different indicators is especially useful for finding such differences.[14]

A Local Distance Network report released in 2025 combined several indicators at the level of their connected measurements and obtained 73.50 ± 0.81 km/s/Mpc. It accounted for shared calibration uncertainties, rather than simply averaging published H₀ values. Because it incorporates existing observations, its result is not an additional independent vote beside those observations.[15]

Baryon acoustic oscillations preserve a characteristic scale in the distribution of matter. Measuring that scale across redshift constrains distance and expansion relative to the ruler’s physical length, usually written r_d. This is the sound horizon at the baryon-drag epoch, slightly later than the last-scattering epoch used for the main CMB acoustic scale.[16]

The inverse distance ladder combines a calibrated early ruler with BAO and relative supernova distances to work toward the present. BAO alone leave an important degeneracy between H₀ and r_d. A physical calibration—through a cosmological model and CMB information, or through primordial-element abundances with standard early-universe assumptions—is needed to infer an absolute H₀.[17]

The strongest cross-checks change the source of the absolute distance scale. Their usefulness depends on what they measure independently and what uncertainties they retain.

Modern TDCOSMO lensing analyses explicitly explore flexible mass profiles. Their inferred H₀ changes with the additional information used. For example, a supernova constraint on the shape of the expansion history can be included without importing the Cepheid-based absolute supernova calibration. Independence is therefore something to describe precisely.[22]

Standard sirens offer another promising route. When no unique host is identified, galaxy catalogs and source-population models can provide statistical redshift information. The LVK collaboration’s GWTC-4 cosmology analysis modeled these ingredients and selection effects, but its H₀ uncertainty remained broad enough to encompass the main competing values. It did not settle the disagreement.[23]

A remaining measurement or calibration issue

Dust, stellar populations, instrument calibration, supernova selection, and local motions all enter the distance inference. Several methods can move together if they share an anchor or calibration. Comparing indicators in the same galaxies and propagating their shared uncertainties helps distinguish a physical discrepancy from one inherited through the measurement network.[15]

Local structure is also testable. In standard ΛCDM simulations that reproduce uneven supernova sampling, the expected variation in locally inferred H₀ is much smaller than the benchmark gap. That makes ordinary local fluctuations an insufficient explanation in those models, while more extreme void proposals must face independent observational checks.[3]

New physics before recombination

Early dark energy is a proposed temporary energy component that increases the early expansion rate and reduces the sound horizon. A smaller ruler can allow a higher inferred H₀ while retaining the measured acoustic angles. But the same change also affects other CMB features and the growth of matter structure, which must remain consistent with observations.[16]

The verdict depends on the model and analysis. A 2025 study combining ACT DR6 and DESI DR2 found early dark energy remained a viable candidate; its preference depended on dataset choices, statistical treatment, and whether SH0ES information was included. A fit that uses the high local value as an input is not an independent prediction of that value.[24]

A broader change to the expansion model

Modified radiation content, neutrino interactions, recombination physics, or late-time dark energy can change cosmological inferences. Moving H₀ alone is not enough: a successful explanation must also reproduce the measured acoustic structure and distance–redshift relation. Late-time changes face the additional requirement of remaining consistent with the calibrated BAO ruler.[9]

A persuasive resolution would do more than make two central values overlap. It would explain why the original analyses differed, survive tests with other observations, and give stable results when reasonable calibration and modeling choices are varied.

The most useful progress comes from tighter connections between different observations: stellar distances checked in common hosts, sirens with better redshift information, lensing with better mass constraints, and acoustic measurements tested across instruments. Each reduces a different route by which an apparent discrepancy could arise.

Research checked: 8 September 2026. Numerical examples are labeled by their analysis and publication period; they should not be averaged as independent measurements.

Cosmology and the Universe’s Large-Scale Structure

1. Cosmic Inflation: Theory and Evidence
2. The Cosmic Web: Filaments, Voids, and Superclusters
3. The Cosmic Microwave Background’s Detailed Structure
4. Baryon Acoustic Oscillations
5. Redshift Surveys and Mapping the Universe
6. Gravitational Lensing: A Natural Cosmic Telescope
7. Measuring the Hubble Constant: The Tension · You are here
8. Dark Energy Surveys
9. Anisotropies and Inhomogeneities
10. Current Debates and Outstanding Questions