This paper presents a performance evaluation of space-time block coding (STBC) employing a realistic mobile radio channel model in macrocellular and urban environments. The bit error rate (BER) is computed by Monte-Carlo simulations in the down-link to evaluate its sensitivity to channel correlation. We consider a horizontal uniform linear array at the base station (BS) formed by up to four antenna elements, and one and two uncorrelated antenna elements at the mobile station (MS). The channel model includes the probability density function (pdf) of the azimuth and delay of the impinging waves and their expected power conditioned on the azimuth and delay. The statistical properties of the model are extracted from macrocellular measurements made in urban environments. Simulation results show that the use of STBC can provide significant gains with acceptable sensitivity to the channel correlation under realistic conditions.