This article looks into the processes and outcomes of setting up and maintaining a probability-based longitudinal online survey, which is recruited face-to-face and representative of both the online and the offline population aged 16-75 in Germany. This German Internet Panel studies political and economic attitudes and reform preferences through bimonthly longitudinal online interviews of individuals. The results presented here demonstrate that a carefully designed and implemented online panel can produce high-quality data at lower marginal costs than existing panels that operate solely in a face-to-face mode. Analyses into the representativeness of the online sample showed no major coverage or nonresponse biases. Finally, including offline households in the panel is important as it improves the representation of the older and female segments of the population.