Context. The presence of two stellar populations in the Milky Way bulge has been reported recently, based on observations of giant and dwarf stars in the inner and intermediate bulge.Aims. We aim at studying the abundances and kinematics of stars in the outer Galactic bulge, thereby providing additional constraints on formation models of the bulge. Methods. Spectra of 401 red giant stars in a field at (l, b) = (0 • , −10 • ) were obtained with the FLAMES-GIRAFFE spectrograph at the VLT. Stars of luminosities down to below the two bulge red clumps are included in the data set. From these spectra we measured general metallicities, abundances of iron and the α-elements, and radial velocities of the stars. The abundances were derived from an interpolation and fitting procedure within a grid of COMARCS model atmospheres and spectra. These measurements as well as photometric data were compared to simulations with the Besançon and TRILEGAL models of the Galaxy. Results. We confirm the presence of two populations among our sample stars: i) a metal-rich one at [M/H] ∼ +0.3, comprising about 30% of the sample, with low velocity dispersion and low α-abundance, and ii) a metal-poor population at [M/H] ∼ −0.6 with high velocity dispersion and high α-abundance. The metallicity difference between the two populations, a systematically and statistically robust figure, is Δ[M/H] = 0.87 ± 0.03. The metal-rich population could be connected to the Galactic bar. We identify this population as the carrier of the double red clump feature. We do not find a significant difference in metallicity or radial velocity between the two red clumps, a small difference in metallicity being probably due to a selection effect and contamination by the metal-poor population. The velocity dispersion agrees well with predictions of the Besançon Galaxy model, but the metallicity of the "thick bulge" model component should be shifted to lower metallicity by 0.2 to 0.3 dex to well reproduce the observations. We present evidence that the metallicity distribution function depends on the evolutionary state of the sample stars, suggesting that enhanced mass loss preferentially removes metal-rich stars. We also confirm the decrease of α-element over-abundance with increasing metallicity. Conclusions. Our sample is consistent with the existence of two populations, one being a metal-rich bar, the second one being more like a metal-poor classical bulge with larger velocity dispersion.