We propose to evaluate the sensitivity of the concentration activities measures to the geographical scale used. Even if the consideration of geography in economic work has become a subject of interest, deficiencies in applied methods persist. In this thesis, we highlight the limits of several commonly applied tools, and more particularly the way in which the Modifiable Areal Unit Problem (MAUP) affects them. In the first chapter we focus our attention on the indices of geographic concentration that serve to apprehend the location strategy of firms. We find that the commonly calculated indices have a bias (positive or negative depending on the index) when the data are geographically aggregated. In the second chapter, we highlight that indices that do not discretize space, and therefore are not sensitive to MAUP, are nevertheless sensitive to other variables that make it difficult to compare results, for example when the size of each sector are not equal. The third chapter is devoted to methods for estimating location factors. The contribution of this work is the comparison of the results with a localization model that we simulate according to several parameters (location strategy, importance of the factors, etc). We find that the variety of location strategy alters the quality of empirical estimates. In view of the work done so far, the last chapter proposes a tool to objectively determine the area of relevant economic activities, a tool that is not sensitive to MAUP. This work uses home-work flows to identify geographical perimeters in relation to a reference site. This operational tool is used to segment visitors from “local” or “non-local” in economic impact studies.