1. Census data and source

• Basic summary tabulations and profiles released by Statistics Canada for the 1996 census:
• Source for socioeconomic characteristics
• Contain comprehensive data from 2A and 2B census forms by large and small levels of geography.
• Include Canada, the provinces and territories, census sub-divisions, census tracts, enumeration areas, federal electoral districts and forward sortation areas.

• Enumeration level data for Manitoba:
• Used to derive the 1996 SEFI
• Available through the ‘public use census files’ via the data liberation initiative (DLI) at the University of Manitoba libraries.
Note: In some cases, it may be better to use census subdivision data for rural areas

• Winnipeg Neighbourhood Resource Network  (NRN) areas are based on the centroid of each enumeration area in Winnipeg and the location of the centroids within the larger areas (NRN).

• Data from the Community Data Network, which is aggregated at Winnipeg specific levels, could not be used since it does not contain some of the information from the basic summary tables in the public use census data.

• Variables Used (Compressed DLI file name, Beyond 20/20 file name, Census description of file):

• Age dependency ratio. Census profile (95F0185XDB96001a, pcan10, Prairie and Pacific Region. Area profiles contain data from the 100% database as well as the 20% sample database.  The suppression rules for the 100% database differ from those used for the 20% sample database.  For this reason, some geographic areas will show 100% data but the 20% sample data will be suppressed).  The Manitoba DLI files include individual profiles for Manitoba and Winnipeg at the enumeration area level. This variable expresses the ratio of the population aged 65 or older in a region to the population aged 15-64.

• Single Parent households. Basic Summary table (95F0196, B06EA, Census Families in Private Households by Age Groups of Never-married Sons and/or Daughters at Home (15), Showing Family Structure (7), for Canada, Provinces, Territories, Federal Electoral Districts (1987 Representation Order) and Enumeration Areas, 1996 Census (20% Sample Data)). Percent of single parent households among households with children aged 0-14.

• Female single parent households (same as Single Parent households). Percent of single female parent households among households with children aged 0-14.

• Labour force participation female (Census profile). Women working or seeking work on census day.  Denominator for this is based on the count of all women over the age of 15.

• Unemployment 15-24, 25-34, 35-44, 45-54 Basic Summary table (95F0230, b24ea, Population 15 Years and Over by Age Groups (17) and Marital Status (6), Showing Labour Force Activity (8) and Sex (3), for Canada, Provinces, Territories, Federal Electoral Districts (1987 Representation Order) and Enumeration Areas, 1996 Census (20% Sample Data)). The unemployed include persons during the week prior to the census that were without work, had looked for work in the previous four weeks and were available for work in the week of the census.  The denominator for each age group was the count of the total labour force in that age group.

• Education 25-34, 35-44, 45-54 Basic Summary table (95F0226, b34ea, Population 15 Years and Over by School Attendance (4), Age Groups (11A) and Sex (3), Showing Highest Level of Schooling (8), for Canada, Provinces, Territories, Federal Electoral Districts (1987 Representation Order) and Enumeration Areas, 1996 Census (20% Sample Data). Education rates were based on the count of the number of residents on census day reporting attaining a minimum of a high school diploma.  The age-specific rates were computed by dividing the count by the total number of residents in the age group.

2. Aggregating Census Data

• All of the census data were assigned a municipal code. Generally census subdivisions (CSD) match the municipal areas used by Manitoba Health; some modifications were made where multiple municipal areas linked to a single CSD.
• CSDs that did not link were assigned a municipal code with the 1996 PCCF file and the Manitoba Health Postal Code Conversion file for postal codes to municipal codes .
• The few remaining enumeration areas that did not link using the above methods were assigned by hand based on the observed geographic location of the enumeration area.
• All of those areas falling inside the Winnipeg Regional Health Authority, as defined by municipal code, were assigned a NRN based on the location of the geographic centroid of the enumeration area.
• Enumeration area centroids were calculated within the GIS package ArcView; a spatial join was used to attach NRNs.

• The weighted mean value of each of the census variables was calculated for each municipal area (outside Winnipeg) and each NRN within Winnipeg.
• The weighting for the mean value is based on the appropriate population for the variable (e.g. Unemployment aged 25-34 is weighted by the labour force population aged 25-34).

• Characteristics that were missing at the municipality code level were imputed a value according to the following strategy:

• Area suppression - adopted by Statistics Canada to protect the confidentiality of individual responses
• The deletion of all characteristic data for geographic areas below a specified size.
• Income distributions and related statistics are suppressed if the total non-institutional population in the area from either the 100% or 20% databases is less than 250.
• Other characteristics are suppressed if the total non-institutional population in the area from either the 100% or 20% databases is less than 40.

• The municipality code level value of each of the socioeconomic characteristics was calculated by the weighted mean of the enumeration area values (where the weight is dependent upon the specific characteristic).
• If all of the enumeration area values for a specific characteristic were suppressed within the municipality, the municipality code level value was missing.
• To calculate the imputation value, each municipality code was defined as either a first nation community or not.  Then weighted mean values were calculated for each characteristic by physician service area and first nation community status.
• Each missing municipality code within a physician service area was imputed the appropriate first nation community average value or non-first nation community value.

3. Defining the SEFI score

• The SEFI score was defined for each municipal/NRN area using the first principal component factors from principal components analysis on the standardized (0 mean/1variance) value of each of the census variables.
• The education and unemployment variables were reduced to a single education and unemployment factor using principal component analysis with standardized values.  These two factors were used in place of the seven education/unemployment variables.

• The first principal component factor was used as the Socio-Economic Factor Index score.  This provided a SEFI score for each of the municipal/NRN areas.

4. Calculation of SEFI at various geographic levels.

SEFI for larger geographic areas were calculated using mean population-weighted SEFI.

• Obtain municipality code level SEFI values and population values.  We used the first principal component factor from the SAS PRINCOMP procedure (standardized to have mean 0 and variance 1) and population values from the MCHP population databases.

• Calculate the weighted provincial mean SEFI value using the appropriate population as the weight.

• Calculate the weighted geographic level mean values of the municipality code level SEFI values weighted by population.

• Centre the geographic level values by subtracting the weighted provincial mean.

Notes/Issues:

• Larger areas must be aggregations of municipal/NRN areas.

• SEFI scores are dependent on the group of geographic areas that were used to derive the variables.  These scores should not be compared to SEFI scores built on a different set of underlying areas.

• Researchers may be tempted to apply the component scores (factor loadings) for each variable to the standardized values of each census variable. Since the SEFI and the scores are standardized to the group of geographic areas in which they were defined, applying the scores to a different set of areas may not be appropriate.

5. SAS programs used to calculate SEFI scores.

• SAS program information available at: SEFI_SAS.html (internal access only)

MORE INFORMATION:

• Frohlich,N. & Mustard,C.A.(1998) Socio-Economic Characteristics Population Health Information System 1991/92 - 1986 Census Version Manitoba Centre for Health Policy