Two types of factors, cyclical and secular, are generally assumed to affect membership fluctuations. Cyclical variables, for example, include changes in prices, wages, and unemployment rate. While secular factors are related to the variables like shifts in the propensity of employees to join unions, changes in industrial relations legislation. Based largely on cyclical variables, this paper will examine the key factors in determining the variations of American union membership from 1948 to 1990. To construct a testable model for econometric analysis, I first do an intensive literature review on the current studies. A discussion of the theoretical backgrounds of the model and data sources follows. By comparing the different functional forms of lagged union density in regression model, the evidences suggest that the alternative forms make only slight divverences in the summary statistics. However, upon using the proportional change, instead of lagged from, in union density, the new form produces a striking result in terms of the criterion of overall goodness of fit, the signs, magnitudes, and significance.