This article presents the construction of a new dataset on respect for civil liberty and provides in-depth discussions of the methodological issues related to the reliability, validity, and dimensionality of the scores. The dataset covers 28 (post-) communist and 20 Latin American countries on an annual basis from the end of the 1970s till 2003. It is theoretically well-grounded in liberal theory and consists of five indicators: 1) independence of courts; 2) freedom of opinion and expression; 3) freedom of assembly and association; 4) freedom of thought, conscience, and religion; and 5) freedom of movement and residence. Statistical tests show that the data is characterized by a fairly high degree of inter-coder reliability and that the indicators reflect a common latent dimension. Apart from proposing a new dataset and an index on civil liberty, this study offers a meticulous guideline for the creation of subjective measures that addresses the choices concerning focus, scope, conceptualization, measurement, and aggregation.