In this paper, an Error Correction Mechanism model of U.S. clothing expenditures for the period 1929–1987 is estimated using recent developments in modeling nonstationary variables. Using clothing expenditures as an example, the pitfalls of conventional modeling of nonstationary variables and the advantages of a new modeling procedure that takes into account the properties of data for valid inference about population parameters are pointed out. The basic findings obtained by estimating an Error Correction Mechanism model of clothing expenditures are (1) the demand for clothing is income inelastic both in the short run and in the long run; (2) the price elasticity of demand is unitary in the long run but greater than unity in the short run; (3) an increase in the unemployment rate reduces U.S. clothing expenditures both in the short run and the long run; and (4) an increase in the number of elderly (above the age of 65) increases clothing expenditures in the short run and reduces expenditures in the long run. However, the shortrun impact of an increase in the elderly population on clothing expenditures is statistically insignificant.