Technical Documentation HouseholdSpend & FoodSpend July 2018

Transcription

1 Technical Documention HouseholdSpend & FoodSpend 2018 July 2018

2 Wh THEY ARE HouseholdSpend provides current estimes of annual expenditures for 480 variables spanning 18 cegories of goods and used by Canadian households everything from fashion apparel and household furnishings to cell phones and charitable donions. While HouseholdSpend includes a general food cegory, FoodSpend provides a more detailed breakdown of food-reled purchases made by Canadian households, with current estimes of annual expenditures for 284 variables of food reled purchases. With FoodSpend, users can gain valuable insights into wh Canadian households are buying from grocery stores and restaurants everything from specific vegetables, me and fish products to meal types like lunches and dinners. Both dabases also include basic variables describing number of households and household income from DemoSts, which are based on Stistics Canada s Survey of Household Spending (SHS). Both wide-ranging dabases let users analyze potential expenditures by average dollars spent per household and total dollars spent for any geographical from all of Canada to small custom-defined trade areas regardless of where the consumer made the purchase. Table 1 shows counts of variables by cegory for HouseholdSpend and FoodSpend. Table 1: HouseholdSpend 2016 and FoodSpend 2016 cegories by count of variables Cegory Count Product Code 1 HouseholdSpend and FoodSpend Basics* 4 HouseholdSpend and FoodSpend Food Purchases (Cegory Summary) 11 FoodSpend Bakery products 15 FoodSpend Cereal grains and cereal products 17 FoodSpend Dairy products and eggs 30 FoodSpend Fish and Seafood 21 FoodSpend Food purchased from restaurants 6 FoodSpend Fruit, fruit preparions and nuts 36 FoodSpend Me 20 FoodSpend Non-alcoholic beverages and other food products 81 FoodSpend Vegetables and vegetable preparions 42 FoodSpend Household Expenditures (Cegory Summary) 26 HouseholdSpend Clothing 43 HouseholdSpend Educion 10 HouseholdSpend Food 8 HouseholdSpend Games of chance 7 HouseholdSpend Gifts of money and contributions 8 HouseholdSpend Goods and by purchase method** 4 HouseholdSpend Health care 23 HouseholdSpend Household furnishings and equipment 31 HouseholdSpend Household operion 37 HouseholdSpend Miscellaneous expenditures 20 HouseholdSpend Personal care 17 HouseholdSpend Personal insurance and pensions 8 HouseholdSpend Reading merials and other printed mter 6 HouseholdSpend Real este 32 HouseholdSpend Environics Analytics

3 Recreion 76 HouseholdSpend Shelter 69 HouseholdSpend Tobacco and alcohol 8 HouseholdSpend Transportion 38 HouseholdSpend *Total expenditure is cegorized as Basics in FoodSpend, but Household Expenditures (Cegory Summary) in HouseholdSpend. **Goods and by purchase method reflects a different hierarchy from other cegories (such as Clothing, Educion, etc.) listed here. It offers purchase methods such as sales over the Internet or other types of direct sales, but does not include detailed cegories. How THEY WERE built - Key da sources The primary da source for HouseholdSpend 2018 and FoodSpend 2018 is the SHS. Since 2010, the SHS has also tracked food purchased from stores in greer detail than before, effectively measuring wh had been captured in the now-discontinued Food Expenditure Survey (FES). While the SHS has generally expanded in size and improved over time, we have modelled relionships using multiple vintages of SHS to capture cegory sub-divisions no longer measured in the current version. A series of models is calibred on the SHS, and these models are used to score a multi-dimensional cross-distribution of household demographics derived from Environics Analytics DemoSts 2018, PRIZM and custom census cross-tabulions from Stistics Canada. Beyond these estimes, EA collects da from Stistics Canada s Nional Economic Accounts (NEA) program. The NEA s consumption estimes and forecasts the provincial and territorial s form a set of authoritive control numbers against which all of our consumption estimes are mched. These control numbers are derived primarily from the Quarterly Survey of Financial Stements (QSFS) and augmented by the SHS when the sector is not previously measured. In practice, the QSFS program produces robust consumption estimes, with its da derived from the financial stements of household serving institutions rher than relying on the memory recall of household members. This approach allows us to correct for response and reporting biases within our SHS-derived models. Additionally, EA incorpores administrive da from the Canada Revenue Agency (CRA) when available. This CRA da, a combinion of standard and custom da from Stistics Canada, are collected the census tract of geography and higher and we use it to guide our estimes for household income taxes paid, employment contributions, Canada/Quebec Pension Plan payments and charitable contributions. These financial-reled outlays account for about 15 percent of total Canadian household expenditures. How THEY WERE built modelling framework The construction of HouseholdSpend 2018 and FoodSpend 2018 involved three distinct phases: the creion of the initial small-area behavioural estimes; the collection and projection of the control totals; and a mhemical reconciliion process th ensures everything adds up. Small-Area Estimes A series of Heckman selection, log-linear and multinomial logit models were calibred using observion- SHS micro-da. These models predicted consumption using a combinion of demographic, locion, PRIZM5, Environics Analytics

4 seasonality and time da as independent parameters. The log-linear models and Heckman selection models generally estime the base- consumption totals like total current consumption or purchases of primary real este. These estimes are then partitioned using a parent-child structure of household spending cegories, for example, breaking down food purchased restaurants into meals and snacks, then subdividing meals into breakfasts, lunches and dinners. Analysts used these models to score a multidimensional cross-distribution of household demographics derived from Environics Analytics DemoSts 2018, PRIZM and custom census cross-tabulions from Stistics Canada. This da cube was produced for every postal code and census disseminion area in Canada. Altogether, the initial development of HouseholdSpend and FoodSpend required over 700 estimes of consumption using over 200 models and 160,000 coefficients for over 1,000,000 small area geographic units in Canada. Provincial, territorial, and small-area control Totals The spending control totals fall into two cegories: NEA-derived provincial control totals and administrive da-derived small-area control totals. Analysts derived the spending control totals for 49 general cegories of current consumption the provincial and territorial from the Provincial and Territorial Expenditure Accounts (PTEA), a derivive of Stistics Canada s NEA program. PTEA is an authoritive source th compiles and integres a large amount of informion from various survey divisions within Stistics Canada, along with other da, to estime provincial and territorial GDPs, as well as their component cegories and underlying sector accounts. While the PTEA control totals are generally used directly, in some cases they are affected by inconsistencies in definitions between PTEA and the SHS. For example, the shelter and miscellaneous expenditure cegories from PTEA cannot be reconciled directly with the SHS. In these cases, the estimed consumption is controlled back to the aggreged spending values from the SHS, with some adjustments. Income taxes and select deductions such as employment contributions, Canada/Quebec Pension Plan payments and charitable contributions all are controlled to da published the small-area (generally, census tract and higher). Table 2 lists the control sources and years available the established time controls. Table 2: Sources of control totals for HouseholdSpend 2018 and FoodSpend 2018 Cegory Description Source Year Product CC002 Child care outside the home PTEA HouseholdSpend 2018 and FoodSpend 2018 CC014 Child care in the home CL001 Clothing CS001-CS010 Communicions excluding postal, courier and other communicion CS010 Postal, courier and other communicion Environics Analytics

5 ED002 Educion FD003-FD806 Food purchased from stores excluding nonalcoholic beverages FD806 Non-alcoholic beverages FD990 Food purchased from restaurants GC001 Games of chance HC001 Healthcare HC Prescribed and nonprescribed medicines and pharmaceutical products HC020 Total health insurance premiums HE002 Household appliances HE010 Other household equipment HF001 Household furnishings PTEA and equipment HouseholdSpend 2018 HF002 Household furnishings HO001 Household operion HO Pet food and purchase of pets and reled pet goods HO006 Veterinarian and other HO Nursery and greenhouse stock (flowers, plants, seeds), and fertilizers, herbicides, insecticides, pesticides, soil and soil conditioners ME001 Miscellaneous SHS HouseholdSpend 2018 expenditures ME018 Discounts and other SHS HouseholdSpend 2018 fees, goods and ME021 Miscellaneous, other SHS HouseholdSpend 2018 goods and PC001 Personal care PC021 Hair grooming Environics Analytics

6 RE001 Recreion excluding recreional vehicles RE004 Outdoor play equipment and accessories RE020 Photographers' and other photographic RE062 Movie theres RE076 Pre-recorded media, downloads and audio/video media RO001 Reading merials and other printed mter RO004 Books and pamphlets (excluding school books) SH001 Shelter SHS HouseholdSpend 2018 TA002 TA005-TA006 TA006 TR001 TR005 TR TR032+RV011 TR033 TR034 Tobacco products and smokers' supplies Alcoholic beverages excluding those served on licensed premises Alcoholic beverages served on licensed premises Transportion including recreional vehicles, but excluding fuels and lubricants Purchases of automobiles Purchases of vans, and mini-vans, trucks and sports utility vehicles Vehicle insurance premiums for automobiles, trucks and vans and recreional vehicles Tires, bteries, and other parts and supplies for vehicles Maintenance and repairs of vehicles Environics Analytics

7 TR036 Gas and other fuels TR037 Other automobile, van and truck operion TR038+RV016 Parking costs work, school, park-ride and parking meters, and recreional vehicle parking, airport fees; bo storage and harbour dues TR051 Bus, subway, street car and commuter train TR052 Transportion by taxi TR054 Transportion by airplane TX001 Income taxes HouseholdSpend 2018 EP002 Employment insurance and Quebec parental insurance premiums HouseholdSpend 2018 EP004 EP005 ME013 MG006 Canada and Quebec Pension Plan payments Retirement and pension fund payments Dues to unions and professional associions Charitable contributions HouseholdSpend HouseholdSpend HouseholdSpend HouseholdSpend Environics Analytics

8 Due to slight differences in definitions and timeframes in the control totals, both the PTEA-based and administrive da-based control totals needed to be adjusted and projected. As an example of the definitional corrections, PTEA control adjustments were made to realloce net expenditures abroad to each cegory, using da from Stistics Canada s Nional Tourism Indicors (NTI) and Travel Survey of Residents of Canada (TSRC). This year s HouseholdSpend and FoodSpend dabases were built using the most up-to-de PTEA and SHS stistics; however, the provincial/territorial estimes were available up until 2016 and therefore needed to be projected to the current year (2018). The projection of PTEA and SHS control totals for 2017 were completed using the 2017 NEA da by allocing to each province/territory using shares from Furthermore, the 2018 estimes were initially derived by fitting a linear regression with year values as independent variable, for each cegory and for each province/territory. The final estimes were then controlled to the seasonallyadjusted, year-over-year growth re of NEA da between the first quarter of 2017 and the first quarter of A similar approach was undertaken for the small-area administrive da control totals. Missing and suppressed values in census tract- administrive da control totals did occur in the raw da from Stistics Canada. The values were imputed using other years values, if possible, or corresponding shares relive to income from the census metropolitan area (CMA). Except for charitable donions, the share of census tract- control over total income from 2016 to 2018 was projected using a logit model with decay factors applied against the log odd difference for each projected year. The control values for 2016 to 2018 were then derived by multiplying the projected shares to the household income known for each year from DemoSts. The 2017 and 2018 shares of census tract charitable contributions over total income were calculed as the sum of charitable contributions from 2012 to 2016 divided by the sum of total income for the same period. Control values for 2017 and 2018 charitable contributions were then derived by multiplying the projected shares to the household income known for each year from DemoSts. This approach was chosen because the charitable contributions da have a relively large degree of temporal stochasticity and the aggrege charitable donions can be grely affected by large, one-time donions. Reconciliion Process For each year, the control totals have to be reconciled with the initial small-area estimes the postal code and disseminion area. This reconciliion was achieved using a set of non-linear mhemical optimizions th adjust the initial small-area estimes to agree with higher geographic control totals. The controlled cegory totals the small-area were then alloced using the initial estime shares of the cegory components. This reconciliion process results in estimes th mch the control totals different s of geography while deviing as little as possible from the estimes derived from the raw da. Special Note about Comparability with WealthScapes Environics Analytics

9 Starting with the 2016 release, WealthScapes (and its varitions), HouseholdSpend and FoodSpend all use identical definitions for disposable income and discretionary income. In fact, WealthScapes disposable and discretionary income da are produced directly from the HouseholdSpend and FoodSpend dabases. The only difference between the disposable and discretionary income figures for each dabase is the year of reference. All of the stistics in HouseholdSpend 2018 and FoodSpend 2018 are for the current year (2018), and are identical to those found in DemoSts But WealthScapes 2018 s populion and income stistics come from 2017 da, while WealthScapes 2018 s historical year stistics refer to Reference Documents HouseholdSpend and FoodSpend Release Notes: /en-ca/release-notes HouseholdSpend and FoodSpend Variables Lists: /en-ca/variables HouseholdSpend and FoodSpend Metada: /en-ca/metada Environics Analytics