A dynamic material flow analysis model is developed to quantify aluminum in-use stocks and old scrap recycling and recovery in the United States for the period of 1900 to 2007. The total in-use aluminum stock in 2007 is estimated as 93million metric tons, which represents approximately 34% of the cumulative apparent consumption since 1900. Alternately, since 1900 nearly 40% of the cumulative discarded aluminum has not been recycled for domestic use in the U.S. or for export to foreign consumers. Statistical time series analysis is used to explore the relationship between model results of in-use stocks and gross domestic product (GDP). Unlike most previous studies of material consumption and economic activity, which ignore the statistical properties of time series data to the detriment of model estimation and inference, data stationarity is explicitly evaluated through unit root testing and model specification is adjusted accordingly. The annual percentage change in GDP is found to have a large and significant association with the annual percentage change in net additions to in-use stocks. Model sensitivity and uncertainty are quantified through the application of the Fourier Amplitude Sensitivity Test and alternate specifications of product lifetime probability density functions.