Aims. Although the variability in the ultraviolet and optical domain is one of the major characteristics of quasars, the dominant underlying mechanisms are still poorly understood. There is a broad consensus on the relationship between the strength of the variability and such quantities as time-lag, wavelength, luminosity, and redshift. However, evidence on a dependence on the fundamental parameters of the accretion process is still inconclusive. This paper is focused on the correlation between the ultraviolet quasar long-term variability and the accretion rate. Methods. We compiled a catalogue of about 4000 quasars including individual estimators for the variability strength derived from the multi-epoch photometry in the SDSS Stripe 82, virial black hole masses M derived from the Mg line, and mass accretion ratesṀ from the Davis-Laor scaling relation. Several statistical tests were applied to evaluate the correlations of the variability with luminosity, mass, Eddington ratio, and accretion rate. Results. We confirm the existence of significant anti-correlations between the variability estimator V and the accretion rateṀ, the Eddington ratio ε, and the bolometric luminosity L bol , respectively. The Eddington ratio is tightly correlated withṀ. A weak, statistically not significant positive trend is indicated for the dependence of V on M. As a side product, we find a strong correlation of the radiative efficiency η with M in our sample. We show via numerical simulations that this trend is most likely produced by selection effects in combination with the mass errors and the use of the scaling relation forṀ. The anti-correlations of V withṀ, ε, and L bol cannot be explained in such a way. The strongest anti-correlation is found between V andṀ. However, it is difficult to decide which of the quantities L, ε, andṀ is intrinsically correlated with V and which of the observed correlations of V are produced by the L − ε −Ṁ relation. A V −Ṁ anti-correlation is qualitatively expected for the strongly inhomogeneous accretion disks. We argue that the observed amplitudes of the variability at far-UV wavelengths, the stochastic nature of variability, and the variability time-scales are not adequately explained by the simple multi-temperature black-body model of a standard disk and suggest to check whether the strongly inhomogeneous disk model is capable of reproducing these observations better.