This paper attempts to fill a gap in Canadian economic statistics. Saving by households is the largest component of net domestic saving in Canada, but lack of data has prevented any detailed examination of the saving behaviour of this sector. The periodical surveys of liquid assets provide information on the income distribution of household savings and some of the forms in which they are held, but no such distribution has been derived for household saving. A first step is made here in the analysis of Canadian household saving behaviour with the construction of income distributions of saving based on data collected by the Urban Family Expenditure Survey for 1959. The survey is not an ideal instrument for this purpose; it was designed primarily to derive detailed information on consumer expenditures, and it is not always possible to separate out, without using approximate methods, data required for saving estimates. In addition, the total sample is relatively small with only a limited number of respondents in the upper income and entrepreneurial categories which contribute substantially to total saving. However, there is no other Canadian source available for this type of study. The results obtained are of interest in themselves and their derivation can point up changes in survey methods and tabulations which would make future survey data more useful for the estimation of household saving.
The 1959 survey was restricted to urban centres with a population of 15,000 or over, but there were no restrictions with respect to family size or income. A probability sample of slightly more than 3000 households was selected for the survey from the eighty-one urban areas which qualified. The sampling rate was 1/1500 in Montreal and Toronto, 1/800 in Vancouver, 1/300 in Newfoundland, and 1/600 in all other areas chosen. To obtain national averages the returns for each region were weighted to allow for differences in sampling and response rates. Response rates varied from 45 per cent in Toronto to 78 per cent in the Prairies, with an over-all average of approximately 62 per cent. No adjustment was made for differences in the response of the various family types. It was noted that “the proportion in the weighted sample of one-person units to families of two or more was below that indicated by available check data.” This lack of adjustment, if differences are significant, would bias the survey results and thus our saving estimates, if the responses varied with family type. Tables IV to VI, given below, indicate that substantial differences exist in the saving patterns of the three family sizes for which estimates were obtained.