For more than a half a decade the fact that expenditure on durables can be well approximated by a random walk has remained a hidden puzzle, challenging almost any theory in which agents smooth the use of their wealth. This paper shows that once a nonparsimonious approach is used, or lower frequencies of the data are examined, the fact itself disappears; changes in expenditures on durables reveal a degree of reversion consistent with the permanent income hypothesis (PIH), although this reversion occurs at a rate significantly slower than what is suggested by a frictionless PIH model. © 1990, by the President and Fellows of Harvard College and The Massachusetts Institute of Technology.