Substorm‐injected particles are time dispersed with pitch angle such that low pitch angle particles arrive earliest on the dayside at geosynchronous orbit, opposite the sense predicted by drift in a dipole field. In contrast to energy dispersion which is well understood, pitch angle dispersion of particle injections has not been quantitatively explained. We present a statistical analysis of energetic ion injections observed by the Active Magnetospheric Particle Tracer Explorers/Charge Composition Explorer (AMPTE/CCE) satellite, and we compare the results with model calculations of single‐particle drift in realistic model magnetospheric electric and magnetic fields. The data analysis is based on observations made from August 1984 to February 1986, which cover all local times, radial distances between 5.5 and 8.8 RE, and dipole latitudes between −16° and 16° and includes over 400 events mostly beyond 8 RE. Using events observed between L = 8 and L = 9, where the effects of the satellite radial motion are minimal, we study the local time dependence of the injection delay time between different energies and pitch angles. We find that high pitch angle injection is delayed from low pitch angle injection and that the delay increases westward from the evening sector reaching ∼2 min (67° versus 23° pitch angles) at 1200 magnetic local time (MLT) at 120–220 keV. The observations are compared to proton drift calculations in static magnetic (T89c) and electric (Volland‐Stern) field models. The calculations reproduce the observed drift times, pitch angle dispersion, and local time dependences, showing for the first time that sufficiently realistic magnetospheric magnetic field models can account for reverse pitch angle dispersion. Detailed examination of the curvature and gradient drift terms shows that the pitch angle dispersion is due principally to compression of the dayside magnetosphere, which preferentially suppresses the gradient and curvature drifts of near equatorially mirroring particles.