We carried out the largest (> 3.5 × 10 5 Mpc 3 , 26 deg 2 ) Hα narrow band survey to date at z ∼ 0.2 in the SA22, W2 and XMMLSS extragalactic fields. Our survey covers a large enough volume to overcome cosmic variance and to sample bright and rare Hα emitters up to an observed luminosity of ∼ 10 42.4 erg s −1 , equivalent to ∼ 11M yr −1 . Using our sample of 220 sources brighter than > 10 41.4 erg s −1 (> 1M yr −1 ), we derive Hα luminosity functions, which are well described by a Schechter function with φ * = 10 −2.85±0.03 Mpc −3 and L * Hα = 10 41.71±0.02 erg s −1 (with a fixed faint end slope α = −1.35). We find that surveys probing smaller volumes (∼ 3 × 10 4 Mpc 3 ) are heavily affected by cosmic variance, which can lead to errors of over 100 per cent in the characteristic density and luminosity of the Hα luminosity function. We derive a star formation rate density of ρ SFRD = 0.0094 ± 0.0008 M yr −1 , in agreement with the redshift-dependent Hα parametrisation from Sobral et al. (2013). The two-point correlation function is described by a single power law ω(θ ) = (0.159 ± 0.012)θ (−0.75±0.05) , corresponding to a clustering length of r 0 = 3.3 ± 0.8 Mpc/h. We find that the most luminous Hα emitters at z ∼ 0.2 are more strongly clustered than the relatively fainter ones. The L * Hα Hα emitters at z ∼ 0.2 in our sample reside in ∼ 10 12.5−13.5 M dark matter haloes. This implies that the most star forming galaxies always reside in relatively massive haloes or group-like environments and that the typical host halo mass of star-forming galaxies is independent of redshift if scaled by L Hα /L * Hα (z), as proposed by Sobral et al. (2010).