We have analyzed Galileo and Voyager images of Io to derive global control point networks, involving 402 (Galileo) and 158 (Voyager) surface points as well as combinations of the two data sets. Using block adjustment techniques, three-dimensional coordinates of the control points are derived with an accuracy of 4.7 km (Galileo) and 8.6 km (Voyager). Three-axial ellipsoids are fitted to the Galileo control points, for which we find the following: a = 1831.1 ¿ 0.3 km, b = 1820.4 ¿ 0.4 km, and c = 1816.6 ¿ 0.5 km. Our results on size and shape confirm that Io (in synchronous rotation and exposed to strong tidal and rotational forces) is in hydrostatic equilibrium and has a pronounced mass concentration toward the center, in agreement with suggested Fe- and Fe/S cores of 740 and 1000 km radius, respectively. However, our shape model is unable to resolve long-wavelength topography beyond the ellipsoidal parameters. Studies of synthetic Galileo and Voyager control point data sets are carried out to track error propagation in our control point networks and to compare the geometric performance of the two spacecraft cameras. We find that Voyager Vidicon images suffer from residual geometric errors on the order of 1/3 of a pixel, which adds to the typical image coordinate measurement errors (slightly less than 1/2 pixel) that apply for Galileo or Voyager images alike. We therefore suggest that previous authors have overestimated the accuracy of their Voyager-based control point networks and global topographic maps significantly.