In this paper, we examine four alternative methodologies for measuring change by size class for variables like wages and employment. The methodologies are base-sizing, mean-sizing, end-sizing, and momentary or dynamic-sizing. For each methodology, we: discuss the concepts and questions it addresses; give examples and the 10-year graphs of employment change broken down by expansion (March 1992 — March 2001) and contraction (March 2001 — March 2003) periods; and the statistical properties (strengths and limitations). We have chosen firm or employer rather than establishment as the unit of analysis because we are trying to answer the question whether most jobs are created by big or small businesses. For the same reasons, our analysis is on net employment change. The data source is the Quarterly Census of Employment and Wages files that have been linked longitudinally for the time period March 1992 to March 2003. The major thrust of our paper pertains to the results based on longitudinal analysis that provides a deeper understanding into some of the major contributors to the employment growth and losses during the expansion period from March 1992 through March 2001 and the contraction period from March 2001 through March 2003. We decompose the contribution of employment changes by: 1) size of businesses; 2) single- versus multi-establishment firms; 3) new businesses; and 4) age of firm.