Wavelet transforms of a large quantity of light curves of numerous CVs in different photometric states were performed in order to quantify the behaviour of the flickering in a statistically significant sample of systems. The scalegram is used as the appropriate tool to describe the wavelet coefficients of stochastically variable data as a function of the time scale. The (logarithmic) scalegram being largely linear for all light curves shows that flickering is a self-similar process and permits a parametrization in terms of its inclination {alpha} and its value {SIGMA} (flickering strength) at a reference time scale. For a given system, {alpha} and {SIGMA} are stable over many years but can vary over shorter periods and are then loosely correlated. On average flickering on short time scales is somewhat bluer than on longer scales. CVs of different types (and photometric states) occupy distinct regions in the {alpha}-{SIGMA}-plane. This behaviour is particularly clear cut for novalike variables where UX UMa stars overlap only slightly with VY Scl stars, and magnetic CVs populate a small range well separated from the other systems. The intrinsic flickering amplitudes of most dwarf novae vary around the outburst cycle with the square root of the system brightness. In dwarf novae with a strong orbital hump the inclination of the scalegram steepens during the outburst. Due probably to complex functional dependences between observable quantities, the physical origins of the flickering, and dynamical system parameters, no clear correlation (only some trends) between flickering characteristics and dynamical or geometrical properties of the CVs can be seen.