The scarcity of data is a widespread problem in applied regional and metropolitan economic research. This note discusses three pragmatic methods to estimate employment from the size-distribution of establishments: the class-interval mid-point method (MP method), lognormal smoothing, and log-logistic smoothing. The methods are compared using data from two sources: the 1996 Recensement des établissements et de l’emploi de Montréal (Banque de Données et d’Information urbaine, INRS-Urbanisation and Ville de Montréal), and Statistics Canada two-digit Standard Industrial Classification (SIC) manufacturing industry data for Québec in 1995. Results show that the smoothing models are clearly superior to the simple MP method, partly because of that method's inherent upward bias.