Income, Wealth, and Household Demand for Deposits
检验货币需求理论中规模变量的选择，发现财富比收入或交易量更能解释存款需求，支持资产组合模型优于交易模型。
Theories of the demand for money typically predict that the quantity of money demanded will increase when: (a) the rate of return paid on money (if any) increases, (b) the rates of return paid on alternative assets decrease, or (c) the of operations (as measured by income, wealth, spending, or permanent income) of the relevant economic actor (firm, household, or nation) increases. Regarding (c), the choice of the most useful measure of involves the guidance of two competing but overlapping concepts of money's role. Transactions models (going back at least to Irving Fisher [1911]) are based on the idea of money as an inventory that bridges the gap created by nonsynchronous receipts and expenditures. For economic actors of a given the optimal inventory should be determined by the opportunity cost of money-holding, the pattern of receipts and expenditures, and the technology of money management. Given these factors, that inventory should increase as the scale, as measured by the rate of receipts or expenditures, rises. Portfolio models (going back at least to John M. Keynes [1930] and popularized by Milton Friedman [1956]) are based on the analogy between asset-holding and budgetary decisions by consumers. The portfolio manager faces a menu of alternative assets with various risk and return characteristics and a given net worth. The optimal holding of money (or any asset) depends upon this wealth constraint, the characteristics of the menu items, and his or her preferences. An increase in wealth should, in general, lead to an increase in money-holding, the magnitude of which depends upon the relative of money as a portfolio component. Such theories suggest that a measure of wealth is the single most useful variable, although the rate of transactions may enter indirectly as a determinant of the otherwise unmeasured marginal return to money. Estimated demand schedules should presumably be based on the most useful scale variable, and some policy issues depend on the choice.1 This choice has been addressed a number of times by studies exploiting time-series data. Table 1 summarizes the results of six regressions from three studies which have approached the question in a nearly uniform way. In equation (i) Alan Meltzer (1963) uses wealth as the variable for aggregate data with great success. The addition of an income measure in (ii) yields no appreciable increase in explanatory power. An equation with income only (not shown) is very similar to (i). The fit is very slightly poorer. In (iii) Charles Lieberman (1977) uses debits to demand deposits as a measure of transactions. The addition of a wealth measure in (iv) yields little additional explanatory power. In (v) Robert Rasche (1987) uses GNP velocity in a first-difference regression. Adding wealth in (vi) yields no significant improvement. Meltzer interprets his results as evidence of the superiority of a portfolio approach. Lieberman interprets his results as evidence of the superiority of the transactions approach produced by a more useful measure of transactions. Rasche favors the use of transactions velocity.