O N How to Calculate Permanent Income

Some recent studies on the Permanent Incomes Hypothesis (PIH) seem to view permanent income as some sort of a moving average of measured inc0me.l It is our contention that this approach is essentially misleading. There are at least three reasons for this contention. First, there appears to be some misunderstanding about the meaning of "horizon" on the part of some authors using the moving average method. In some studies a three year moving average has been used. This may be partly because Friedman in his calculations of the consumption function for the United States estimated the consumer units horizon as approximating three years. Tobin appears to have been misled by this, for in a review of Friedman's A Theory of Consumption Function, he asserts, ". . . permanent income is something like a moving average of observed income over a period of two to five year^."^ Friedman's very vague and somewhat incorrect usage of "horizon" in his book may have contributed to this confusion. In his later studies however, Friedman corrected his earlier vagueness. In Friedman's own words: "I have referred to the consumer units 'horizon' and estimated it from empirical data as approximating three years in length. But 'horizon' here means something very different from what it usually does. In particular the concept is definitely not to be interpreted as 'a moving average of observed income for a three year period.' As it has generally been used, 'horizon' implies a cutting-off point-that a consumer does not look beyond three years, if that is the length of his horizon. As I use it, it means that it is the dividing line between the effects he considers transitory and those he considers permanent; that he will not adjust his consumption pattern to actual or possible changes in circumstances, except as these can be expected to affect his consumption possibilities for a period longer than three years; and hence that he does not look beyond three years. His permanent income for a three year period may differ from his average measured income for those three years just as the expected