ABSTRACT We present an empirical analysis of the properties of dust-continuum emission in a sample of 17 galaxies in the early Universe (4 < z < 8) with well-sampled far-infrared spectral energy distributions (SEDs) compiled from the literature. We place our results into context by self-consistently comparing to samples of nearby star-forming galaxies, luminous infrared galaxies (LIRGs), and quasars. With the exception of two sources, we find no significant evolution in the dust emissivity index across cosmic time, measuring a consistent value of βIR = 1.8 ± 0.3 at z > 4, suggesting that the effective dust properties do not change dramatically for most galaxies. Despite having comparable stellar masses, we find the high-redshift galaxies to be similar to, or even more extreme than, LIRGs in the Herschel (U)LIRG Survey, where (U)LIRG refers to (ultra-)LIRG, sample in terms of dust temperature ($T_\text{dust} \gt 40 \, \mathrm{K}$) and infrared (IR) luminosity ($L_\text{IR} \gt 10^{11} \, \mathrm{L_\odot }$). We find that the dust temperature evolves mildly towards high redshift, though the LIRGs and quasars exhibit elevated temperatures indicating a more efficient and/or additional heating mechanism. Where available, we compare stellar mass estimates to our inferred dust masses, whose degeneracy with dust temperature can only be mitigated with a well-constrained SED. In merely half of the cases, the dust yield may be explained by supernovae alone, with four sources ($44{{\ \rm per\ cent}}$) significantly exceeding a highly optimistic yield where Mdust ≈ 0.01M*. We discuss possible explanations for this apparent inconsistency and potential observational biases in the measurements of the dust properties of high-redshift galaxies, including in the current IR-bright sample.