Cosmological models have granted dwarf galaxies a key role: their properties constrain the distribution of dark matter and the physical evolution of their hosts. There is increasing evidence that objects with masses of dwarf galaxies form in the tidal tails of colliding galaxies and speculations that they could become satellite-like galaxies around their progenitors and thus be cosmologically important. Yet, whether these "Tidal Dwarf Galaxies" (TDGs) candidates are really long-lived and not only present in young interacting systems is still an open question to which numerical simulations may give answers. We present a set of 96 simulations of colliding galaxies with various mass ratios and encounter geometries, and statistically study the evolution of their TDG candidates. Among the 593 substructures initially identified in tidal tails, about 75% fall back onto their progenitor or are disrupted in a few 10^8 years. The remaining 25% become long-lived bound objects that typically survive more than 2Gyr with masses above 10^8 M_sun. These long-lived, satellite-like objects, are found to form in the outer most regions of the tidal tails. We infer several basic properties that dwarf galaxies should meet to have a tidal origin and apply these criteria to the Local Group dwarfs. We also find that the presence of TDGs would foster the anisotropy observed in the distribution of satellite galaxies around their host. Identifying the conditions required for interacting systems to form long-lived tidal dwarfs, we roughly estimate their contribution to the overall population of dwarfs. We conclude that a small but significant fraction of them - typically a few percent, and possibly more in dense environments or around early-type galaxies - could be of tidal origin.

Accepted for publication in A&A. Shortened abstract below, full version in the paper. Figs 1, 6, 7 at end fo manuscript