Knowledge that Transforms

Michael Manove's recent contribution in this Review is fundamental in at least two respects. For the first time he has shown precisely how the formulation of each year's plan could be facilitated by taking advantage of data derived in the course of constructing the plans of previous years; and he has demonstrated the relationship between the changes in supply and demand generated in the present year to the imbalances of previous years. Although the logic of Manove's models is unasssailable, I do have some reservations on the way he interprets, or seemas to interpret, what is going on in some of these iterative processes he describes. It clearly emerges from his paper, in particular, that a procedure that calls for more will necessarily produce absolutely smaller imbalances in the long run, for given yearly final demands and increments in output. In other words, the absolute value of the elements of the vector of supply-demand imbalances in year T will be smaller under the centralized procedure that includes retrospective iterations (Manove's equation 21), than under the procedure that does not (equation 20), provided that the input-output matrix A is productive. In symbols, ETI< IE'j,where

In a recent article in this Review, Martin Weitzman argued that the observable slowdown in the of output (gy) of the Soviet economy in the 1960's need not be associated with a fall in the of total factor productivity (ga), as is usually suggested, but rather can be better shown to be a manifestation of diminishing returns to capital. By directly estimating a Constant Elasticity of Substitution (CES) production function' for the two decades following World War II, he found an elasticity of substitution of capital for labor (o-) significantly less than one. From this he concluded that the slowdown in the of that economy could largely be explained in terms of the diminishing returns to capital which resulted from the small substitutability between capital and labor and rapidly increasing overall capital deepening in the economy. Weitzman concluded that Instead of capital, labor and technical change will have to be increasingly relied upon as alternative sources of future economic growth (p. 685); and [that due to demographic trends] This rests the spotlight finally on technical change .. the way of raising g, is now to increase ga because gL iS more or less fixed .. . (p. 686). We should like to advance the proposition that the record of of the Soviet economy during the 1950's and 1960's (as presented in Weitzman's Table 1, p. 677) points to aspects of the underlying Soviet macro-production process other than the small elasticity of substitution as possibly the kev culprits effecting the noted slowdown in g,. Furthermore it is suggested that perhaps the most appealing way of raising g, may after all be not through the overall productivity relationship A (or ga), but rather through the term slighted by Weitzmanthe rate of the labor force gL. We fit the data in Weitzman's Table 1 to a maximum likelihood, non-linear regression program,2 similar to that used by Weitzman. A more general model was employed which imposed neither a geometric time trend, nor unitary returns to scale on the data. The specification used was:

In his comment, Alex Cukierman argues that to obtain better estimates of price, or money-illusion, effects in an aggregate consumption function one should disaggregate the consumer price index (CPI) into its components and include these separate price components in the equation, rather than just including the CPI as we did. He then estimates a consumption function for our sample period, 1955 1-1965 IV, using, our data for real per capita consumption, net labor income, and wealth, and disaggregated data on five individual price series the CPI components for food, housing, apparel, transportation, and health and recreation.' In his representative equation, the lag on income is shortened from seven quarters to four quarters while the lags on the individtual prices vary from one to three quarters, as compared with our original seven-quarter lag on the CPI,2 As is clear from Cukierman's Table 1, the coefficient sums of Cukierman's best equation (his 1-3) are fairly similar to those of our final equation (his 1-1). The sums of his income coefficients and price coefficients are a bit smaller than ours, and his wealth coefficient is a bit larger. The main difference between equations 1-1 and 1-3 in Cukierman's Table 1 is that the sum of his price coefficients (in 1-3) is only 2.85 times its standard error, while ours (in 1-1) is 11.6 times its standard error. From this result, Cukierman concludes that the money-illusion coefficient is smnaller and less sigynificant when the consumiier price index is disaggregated. Nevertheless the results still seemto indicate some deg,ree of monev illusion. The procedure Cukier-nian uses raises two questions that are best handled sequentially,,. First, to what extent are his estimi-ates the resuilt of changing the lag lengths in the estimated equation, and to what extent are they duie to disagg(regration of the price variable? Second, if disaggwreg-ation. is the imip ortant cause of the divergence between his results and ours, what is the best wav to interpr-et his resuilts? The first two sections below consider these two questions, and the third section concluides with somie further comnments.