Notes. Column (1) – The four bins used in the analysis are all objects, and bins (1), (2) and (3), see Sect. 2, Table 1, and Fig. 1 for details. Within each bin, – brightest refers to stacks made using only the brightest 18-pixels area (which roughly corresponds to the area of one seeing disk) of Hα emission per object, while – fainter refers to stacks made using only the remaining, less intense Hα emission pixels per object. Column (2) – Hα luminosity of the stack in units of 10⁴¹ erg s^-1; Col. (3) – The logarithm of the average Hα surface brightnesses of each stack in units of erg s^-1 kpc^-2.  ⟨ Σ_Hα ⟩  is calculated by dividing the total Hα luminosity of the stack by the total isophotal area of the galaxies making up the stack. The total isophotal area is estimated including all pixels that are 3 times the RMS of the background of each data set (see Lehnert et al. 2009; and Le Tiran et al. 2011, for details). Column (4) – FWHM of the Hα line obtained without weighting the widths of individual objects by their Hα flux (in km s^-1); Col. (5) – HαFWHM obtained when weighting the widths of individual objects by their Hα flux (in km s^-1); In the next Cols. 5−8, we provide the characteristics of our fits of a single broad component underlying the Hα and [Nii] emission. However, we emphasize that we do not think of these fits as realistic (see text for details) but are given for comparison with Shapiro et al. (2009). The parameters are only provided for fits that have high significance. Note the considerable uncertainties in the fits, suggesting they are not very constraining, which is the reason we adopted a Bayesian approach; Col. (6) – Hα luminosity of the single broad line component fitted to the spectrum (in units of 10⁴¹ erg s^-1); Col. (7) – FWHM of the single broad component fitted to the spectrum (in km s^-1); Col. (8) – velocity offset between the best-fit broad component and the narrow Hα line component (in km s^-1), where negative numbers indicate a blue-shifted broad component; Col. (9) – flux ratio of the broad to narrow Hα line components; Col. (10) – Bayesian likelihood, ln(B), of the model with a broad line compared to the model without, where values above 5 indicate a significant enhancement of the fit quality by including a broad line and negative values rule out a broad component; Col. (11) – [Sii]λ6716/[Sii]λ6731 line flux ratios; Col. (12) – electron densities (in cm^-3) derived from the [Sii] ratios given in Col. (11), with their 1σ uncertainties. Throughout our analysis, we adopt the cosmology H₀ = 70 km s^-1 Mpc^-3, Ω_Λ = 0.7 and Ω_M = 0.3.