To improve the estimates of MFP further, development of a data base is required. With the resources available at Statistics Canada, we made progress in putting together a comprehensive data base. Yet the empirical results obtained point out the need for further work, particularly in the following areas: consistent industry classification and aggregation between output and inputs and a change in the measures of capital stocks and labour input to reflect the different vintages of capital and their rates of return and the characteristics of different groups of workers. Indirect taxes and subsidies should be treated differently in order to arrive at purchasers' prices on the input side and producers' prices on the output side.

At the theoretical level we have defined two concepts of MFP and called them traditional and new. The difference between these two is in their treatment of produced inputs. The traditional measures account only for changes in the quantities used, whereas the new account for both the changes in quantities and the lower input content necessary to replace all produced inputs under improved technology. Under conditions of advances in technology, the amount of such required inputs is declining over time. Therefore, the resulting new MFP rate should be greater than the traditional measure. We observed this to be the case for most industries in most years. However, there are exceptions, notably in the post-1973 period, when many industries suffered widespread declines in productivity or in the efficiency with which inputs are used to produce outputs.