This dissertation designs a metadata-driven infrastructure for panel data that aims to increase both the quality and the usability of the resulting research data. Data quality determines whether the data appropriately represent a particular aspect of our reality. Usability originates notably from a conceivable documentation, accessibility of the data, and interoperability with tools and other data sources. In a metadata-driven infrastructure, metadata are prepared before the digital objects and process steps that they describe. This enables data providers to utilize metadata for many purposes, including process control and data validation. Furthermore, a metadata-driven design reduces the overall costs of data production and facilitates the reuse of both data and metadata. The main use case is the German Socio-Economic Panel (SOEP), but the results claim to be re-usable for other panel studies. The introduction of the Generic Longitudinal Business Process Model (GLBPM) and a general discussion of digital objects managed by panel studies provide a generic framework for the development of a metadata-driven infrastructure for panel studies. A first theoretical application presents two designs for variable linkage to support record linkage and statistical matching with structured metadata: concepts for omnidirectional relations and process models for unidirectional relations. Furthermore, a reference architecture for a metadata-driven infrastructure is designed and implemented. This provides a proof of concept for the previous discussion and an environment for the development of DDI on Rails. DDI on Rails is a data portal, optimized for the documentation and dissemination of panel data. The design considers the process model of the GLBPM, the generic discussion of digital objects, the design of a metadata-driven infrastructure, and the proposed solutions for variable linkage.