This methodological paper shows that using different local job accessibility models (LJAs) leads to significantly different empirical appreciations of job accessibility. Matching several exhaustive micro data sources on the Paris region municipalities, the paper benchmarks a representative set of LJA measurement models used in the recent literature and an original model where job availability is fully estimated according to a set of individual characteristics, job competition is fully modelled on the local labour market and frontier effects are controlled for. We show that the model-induced empirical differences are spatially differentiated across the Paris region municipalities, and that failing to fully estimate job availability may lead to overestimation of the job accessibility levels of underprivileged municipalities.