In this paper, we present an approach to relate typological measures of language complexity (based on the grammars of different languages) with empirical measures of that complexity (based on actual texts). It is known as the “structure-conduct-performance paradigm”, and we have taken it from the field of industrial economics. Using a sample of 45 languages for which we have the same text, we apply that approach to capture some relationships that go from phylogenetic, geographic, demographic and sociological characteristics of languages (structural variables) towards some typological variables that determine measures of phonological and morphological complexity, and then have an impact on two corpus-based language ratios (phonemes per word and words per clause). Our results, based on correlation and regression analyses, show some important trade-offs between the typological measures, between the language ratios, and between both sets of variables. Those results are also robust to extending the number of languages in the sample to 81 observations, and to dividing that extended sample into sub-samples.