Abstract: I study the location patterns and the effect of agglomeration on plant-level productivity of manufacturing industries in the Russian Empire. The analysis is based on a novel micro-geographical dataset on the population of Russian manufacturing plants in 1908, augmented with geo-coded data on Imperial railroads and province-level indicators. Using continuous distance measures of geographic concentration, I show that most industries in the Russian Empire were localized at distances about 100--200 km. Also, I reveal coagglomerated industry pairs forming input--output linkages. I estimate impacts of location characteristics of each plant on plant-level productivity. I find that a firm-level TFP is bell-shaped with respect to the number of plants of the same industry in the same geographic area. I reveal a strong positive impact of closeness to railroads on firm-level productivity. These findings reflect non-trivial interplay between agglomeration and dispersion forces to which my estimation results impart empirical content.