Although there is consensus that metal-rich stars in the Milky Way bulge are formed via secular evolution of the thin disc, the origin of their metal-poor counterparts is still under debate. Two different origins have been invoked for metal-poor stars: they might be classical bulge stars or stars formed via internal evolution of a massive thick disc. We use N-body simulations to calculate the kinematic signature given by the difference in the mean Galactocentric radial velocity ($\Delta\rm V_{\rm GC}$) between metal-rich stars ([Fe/H] $\ge$ 0) and moderately metal-poor stars (-1.0 $\le$ [Fe/H] $<$ 0) in two models, one containing a thin disc and a small classical bulge (B/D=0.1), and the other containing a thin disc and a massive centrally concentrated thick disc. We reasonably assume that thin-disk stars in each model may be considered as a proxy of metal-rich stars. Similarly, bulge stars and thick-disc stars may be considered as a proxy of metal-poor stars. We calculate $\Delta\rm V_{\rm GC}$ at different latitudes ($b=0^\circ$, $-2^\circ$, $-4^\circ$,$-6^\circ$, $-8^\circ$ and $-10^\circ$) and longitudes ($l=0^\circ$, $\pm5^\circ$, $\pm10^\circ$ and $\pm15^\circ$) and show that the $\Delta\rm V_{\rm GC}$ trends predicted by the two models are different. We compare the predicted results with ARGOS data and APOGEE DR13 data and show that moderately metal-poor stars are well reproduced with the co-spatial stellar discs model, which has a massive thick disc. Our results give more evidence against the scenario that most of the metal-poor stars are classical bulge stars. If classical bulge stars exists, most of them probably have metallicities [Fe/H] $<$ -1 dex, and their contribution to the mass of the bulge should be a small percentage of the total bulge mass.
Comment: 7 pages, 4 figures, accepted for publication in A&A