THE discussion concerning Professor Alilton Friedman's permanent income hypothesis led some investigators to conclude that observable quantitative differences between so-called within-group consumption elasticities and between-group consumption elasticities support the Friedman hypothesis. I am in qualified agreement with this conclusion-qualified in that I consider the reasoning that leads to the conclusion incomplete. The explanation of my position will start with emphasis on agreement, but it will end with the emphasis shifted to the qualifications. Take for some year a cross-section of households at large and compute the consumption-income relation for this universe. This measured relationship will not depend solely on differences between the "permanent" incomes of various households and on differences between their "permanent" (or regular) consumption. The relationship will also be affected by differences between "transitory" components of income, which may be positive or negative, and by differences between "transitory" components of consumption, which are in the nature of random additions to the regular consumption flow of the household or of random subtractions from it. If we accept the permanent income hypothesis, according to which the true relationship between income and consumption is a relationship between "permanent" income and "permanent" consumption and according to which the transitory components of income do not influence either constituent of consumption (or do not influence them appreciably), then we are led to expect that the apparent consumption-income relationship, computed from a cross-section of households at large, will understate the "true" consumption elasticities with respect to income. The simplest way of expressing the essential reason for this understatement of consumption elasticities may be to point out that in a universe of households at large the spread between measured incomes must be greater than the spread between permanent incomes and that statistical measurements attribute the observable differences in consumption to the greater differences between measured incomes rather than to the smaller differences between the consumption-governing permanent incomes. Assume, for example, that, out of a permanent income of 100 units, 80 units are consumed, while out of a permanent income of 200 units, 160 units are consumed. If these were the only two existing sizes of permanent income, statistical measurements would attribute the observable 80-unit differences in consumption per household, not to the 100unit differences in permanent income, but to greater differences. Hence consumption elasticities would be understated. This is because, as a consequence of negative transitory incomes, some measured incomes would be smaller than 100 units; and, as a consequence of positive transitory incomes, some measured incomes would exceed 200 units. Instead of considering cross-sections at large, let us now classify the universe of households into groups whose permanent income-earning capacities are typically different. Occupationally or locationally defined groups are of this character, and so are groups defined by age. Within each of these groups the household-to-household comparisons will lead to an underestimate of the "true" consumption elasticities, for reasons that are identical with those explained for cross-section measurements at large. But we may now also compute consumption