Re-weighting and historical revision of Survey of Household Spending
public use microdata files 

This SHS PUMDF uses new survey weights which take into account new
population projections from the 2001 Census. In order to make estimates
comparable over time, the weights for all SHS PUMDFs have been revised
using the new projections. 

The weights and new calibration strategy were implemented for SHS for
the years 1997 and onward resulting in revised estimates of household
spending for each year. To ensure comparability, users of SHS data
should take care that comparisons are made between years that use the
revised weights. 

See the Excel file "weights-reweighting-yyyy" for a table showing the
principal statistics of the revised weights. The estimation of
population characteristics from a sample survey is based on the idea
that each sampled household represents a certain number of other
households in addition to itself. 

These numbers are called the survey weights of the sample. To improve
the representativity of the sample, the weights are adjusted so that the
estimates from the sample are in line with population totals, or
benchmarks, from other independent sources of information that are
considered reliable. 

This is called weight calibration. 

SHS uses two sources for calibration. The first source is the Census of
Population which provides demographic benchmarks. From 1997 to 2003, SHS
used benchmarks derived from the 1996 Census. Since the Census is
conducted once every five years, Statistics Canada projects the Census
results for later years (up to the present), and then revises those
estimates when the next Census data become available. 

The projections use a variety of secondary information, including
administrative data on births, deaths and migration. The second source
used for adjusting the survey weights for SHS are T4 data from Canada
Revenue Agency, which ensures that the estimated distribution of earners
in the survey matches the one in the Canadian population. 

It was decided to take advantage of this historical revision to also
introduce an improved calibration strategy for the SHS weights.
Improvements to the calibration strategy were deemed necessary to put
emphasis on SHS needs (such as the age groups used for calibration) and
to take into account the quality of the benchmarks. 

It was also felt that there were too many benchmarks leading to too many
constraints on the weights, and that this produced undesirable results,
such as negative weights, which were not acceptable. 

The changes made were as follows: 

+ Age-sex 

- At the provincial level, controls for sex have been eliminated. There
are 8 age groups (0-6, 7-17, 18-24, 25-34, 35-44, 45-54, 55-64, 65+)
instead of the 18 agesex groups used previously. 

- No change at the CMA level: 2 age groups (0-17, 18+) 

+ Controls for the type of household (lone parent and couples with
children) were removed. 

The following calibrations were unchanged: + No change to the controls
for size of household (1 person, 2 persons, 3+) 

+ No change to the T4 adjustments to the weights of the population for
income from 

wages and salaries (0-25th percentile, 25th-50th, 50th-65th, 65th-75th,
75th-95th, 95th- 100th) 

Due to their smaller population, only two age groups are used for the
three northern territories: number of persons under 18 and number of
persons 18 and older. The weights are also calibrated to the totals for
one-person households, two person households, and households with three
or more persons. 

Before the historical reweighting the calibration strategy varied
slightly between the territories and between survey years. The northern
calibration is now consistent across all three territories and over
time. 

=========================================================================