Observational study of population level disparities in food costs in 2021 in Canada: A digital national nutritious food basket (dNNFB)

Highlights

• •Web scraping significantly eases data collection for food costing.
• •Digital data makes it easy to standardize food cost analyses across Canada.
• •Food basket costs varied across retail banners, province/territories, and age-sex status.
• •Digital data suggest we are severely underestimating the cost of food across the country.

Abstract

The aim of this work was to assess the feasibility and effect of applying a nationally representative and highly disaggregated food costing measure across Canada, through the novel application of web-scraping technology to the methods of the National Nutritious Food Basket (NNFB). Further, this study tested the hypothesis that a product-matched digital NNFB (dNNFB) correlates with existing market basket measures and quantified any differences in costs. This was an observational cross-sectional study using web scraped food price data collected in November 2021. Food price data was collected from the majority of Loblaw’s banners across Canada, resulting in a final store sample of 751 stores sourced from 11 retail banners. Stores were located across all five Statistics Canada regions, including all provinces and territories with the exception of Nunavut. Store-level dNNFB costs were computed, adjusted by age-sex group, and summarized by geographic region and banner. dNNFB costs were then compared with existing national statistics office estimates (Market Basket Measure thresholds for reference families). dNNFB costs varied widely across the country, with notable differences by regional, store-level, and age-sex group characteristics. When compared to reported national statistics, our estimates exceeded the national market basket measure in every comparison in corresponding sub-national geography across the country, with correlation varying from 0.49 to 0.78 dependent on summary comparator. Digital collection of food price data was a feasible strategy for market basket costing. Our findings suggest we may be routinely underestimating the impact of food inflation for consumers, particularly those restricted to certain food environments.

1. Introduction

The cost of food is a critical determinant of consumer purchases and diet (Lee et al., 2013, Kirkpatrick and Tarasuk, 2008). Geopolitical and pandemic pressures have renewed scrutiny of the adverse health and health equity effects the world faces if the ongoing cost of living crisis continues unmitigated (Lancet Public Health, 2022). The ‘market-basket’ approach is a public policy metric, positioned as a viable way to assess cost of living associated with basic needs, also known as poverty ‘lines’ (Djidel, et al., 2020). Market basket measures typically include the cost of food (Lee et al., 2013, Rao et al., 2013), through an indicator ‘basket’ of commonly consumed retail groceries to meet minimum dietary standards (Djidel et al., 2020, Lewis and Lee, 2016, Health Canada, 2019). Food baskets are generally calculated by national statistics offices, public health units, and/or researchers by aggregating the prices of a predefined set of foods to assess the spatial, temporal, and demographic dynamics of food costs. To interpret the relationship between market baskets and public health outcomes, however, remains difficult.

Disproportionately high inflation in food prices relative to other goods—compounded by interjurisdictional variation in median incomes—has the potential to seriously amplify existing inequities in healthy food purchasing and consumption (Lee et al., 2013, Rao et al., 2013, Darmon and Drewnowski, 2015). Some jurisdictions, including Canada, encourage sub-national monitoring of food baskets, to distinguish how the cost of living may vary among provinces and territories with different geography (e.g., large remote areas), access to the food supply (e.g., prime agricultural production nearby), natural resources, and population bases (Health Canada, 2019). Food cost surveillance disaggregated by spatial and consumer characteristics is key to several Canadian federal statistics, including the food component of the market basket measure (MBM), otherwise referred to as the national nutritious food basket (NNFB) (Health Canada, 2019), and for measures of inflation such as the Consumer Price Index (Djidel, et al., 2020). Standardized national data collection occurs for a representative sample of population centres (n = 38 cities) (Djidel, et al., 2020), but at the sub-national level, the NNFB is operationalized by local public health authorities who adapt the national measure, sometimes extensively (Power et al., 2019), hence with no interjurisdictional comparability.

While practical, this uncoordinated approach during a period of unprecedented inflation means that our understanding of affordability is highly constrained by limited jurisdictionally comprehensive, population-representative, and granular data on the cost of food (Djidel et al., 2020, Health Canada, 2019). Accurate regional consumer price and cost of living measurements, as well as robust accounting for bias, is fundamental to the connection between economic measurement and policymaking (Boskin et al., 1998). To address the problem of economic monitoring in relation to diet and health, researchers have tended to turn to two main methods of ad-hoc food cost data collection: secondary analyses of commercial datasets, and/or in-store auditing procedures (Hillen, 2019). Despite their advantages, both methods lack standardization, provide low spatial coverage, require large outlays of financial and human resources, and limit potential for post-hoc diet cost analyses (Luongo et al., 2022).

The aim of the following paper is to demonstrate the feasibility and effect of applying a nationally representative and highly disaggregated food costing measure across Canada, through the novel application of web-scraping technology to the methods of the NNFB (Health Canada, 2019). Our objective was to construct and implement a digitally collected form of the NNFB, termed hereafter as the dNNFB, to describe variation in the cost of food, stratified by provinces and territories, retailer characteristics (e.g., discount versus regular banners), and sub-population dietary needs through the life course. In parallel with the growth in online food retail in recent years, a web-scraped automated national nutritious food basket presents an innovative solution to the limitations of existing cost-of-living surveillance (Hillen, 2019, Cavallo, 2017, Cavallo and Rigobon, 2016, Harrington et al., 2019, Scarborough et al., 2020, Household Food Insecurity in Canada).

2. Methods

2.1. Study design

This was an observational, cross-sectional study, collecting online food price data over a two-day period, during the pandemic, November 23rd and 24th, 2021. As of 2019, approximately 75% of all food expenditures in Canada were made up of food purchased from stores, with the remainder purchased from restaurants (Detailed food spending). This asymmetrical dedication of food spending in stores iterates the need to better quantify the prices of foods within these retail settings. This study did not include human participants; no ethics board approval was necessary.

2.2. Data source and setting

The data source was digitally publicly available food price data at Canada’s largest food retailer corporation by market share, Loblaw’s (QUIC Research Report, 2017). Data was collected from the majority of regular and discount banners falling under Loblaw’s ownership and as such reflects population-level coverage. A ‘banner’ is a corporate sub-entity represented by the name a store operates under and associated with brand intentions; for instance, NoFrills in our study is purposefully branded by its parent company as a discount banner (About No Frills, About Your Independent Grocer). All stores available for online ordering at the time of data collection were included, a census of stores across Canada (n = 760 stores).

Food ‘price’ in this study refers to the dollar value assigned to the food supply (retailer). Food ‘cost’ refers to the consumer experience of prices, which in this study reflects basket measures calculated from food supply prices.

2.3. Scraper design

A detailed description of the web scraper design is beyond the scope of this empirical paper but for a review of technical terms please see Hillen et al. 2019. Briefly, the web scraper was built with Python, making use of selenium and beautifulsoup4 for scraping and parsing respectively. Search keywords were generated from the 61 items included in the 2019 NNFB, see Table 1 for keywords and associated product matches. The first 48 items returned per keyword were scraped. The 2019 NNFB, is the most recent version of the food basket protocol nationally and reflects the most recent 2019 Canada’s Food Guide and intakes from the 2015 Canadian Community Health Survey—Nutrition (CCHS-N), Canada’s most recent population-level dietary surveillance survey (Health Canada, 2019).

Table 1.

NNFB Products, Keywords, Sample Products and Number of Product Matches. NNFB product keywords were manually tested to improve data returns. Products were first organized by frequency and then matched to the intention of their associated NNFB items. For the purpose of basket costing and NNFB item-level statistics, the lowest priced product from each store was used. Store counts reflect the number of stores with at least one matched product following SKU selection. Distinct prices indicate the number of distinct prices across all instances of every matched product in the dataset.

NNFB product	Keyword	Matches	Example Product	Example SKU	Stores	Distinct Prices
Apple, fresh	Apple, fresh	4	Red Delicious Apples	20073899001_KG	749	7
Banana, fresh	Banana, fresh	2	Organic Bananas, Bunch	20139509001_KG	751	12
Beans, black, canned (no salt added, or low in sodium)	Beans, black, canned	3	Black Beans	20305136001_EA	751	9
Beans, kidney, canned (no salt added, or low in sodium)	Beans, kidney, canned	4	Dark Red Kidney Beans	20007312_EA	751	6
Beans, white, canned (no salt added, or low in sodium)	Beans, white, canned	3	White Kidney Beans	20148074_EA	751	9
Beef, hip, inside (top) round roast, boneless, lean	Beef	4	Lean Ground Beef	21018826_EA	550	5
Broccoli, frozen, spears, boiled, drained	Broccoli, frozen	1	Broccoli Florets	20318915_EA	725	5
Brown rice	Brown rice	4	Whole Grain Brown Rice	20074783002_EA	751	14
Cabbage, fresh	Cabbage, fresh	2	Red Cabbage	20773656001_KG	749	9
Cantaloupe, melon, fresh	Cantaloupe, melon, fresh	2	Organic Cantaloupe	20127944001_EA	736	3
Carrots, fresh	Carrots, fresh	2	Carrotsmini 2Lb	20031454001_EA	742	7
Celery, fresh	Celery, fresh	1	Celery Stalks	20163119001_EA	727	7
Cereal, hot, oats, quick, prepared	oatmeal	5	Oats, Quick	20323113001_EA	751	7
Chicken, legs, fresh or frozen	Chicken, legs, fresh or frozen	1	Free From Chicken Drumsticks	20059572_KG	331	5
Chickpeas, canned (no salt added, or low in sodium)	Chickpeas, canned	2	Chickpeas	20318942_EA	750	8
Corn, frozen	Corn, frozen	1	Whole Kernel Corn	20306375_EA	733	4
Cucumber, fresh	Cucumber, fresh	2	English Cucumber	20070132001_EA	738	12
Dinner roll, whole wheat	Dinner roll, whole wheat	3	Dinner Roll, 60% Whole Wheat	20034130_EA	469	10
Egg, chicken	Eggs	2	Large Grade A Eggs	20812144001_EA	745	15
Flat bread, whole-wheat, chapatti, roti or pita	Flatbread, whole-wheat	4	Large Whole Wheat Tortillas	20276307_EA	697	19
Fortified soy beverage, (unsweetened or original)	soy beverage	4	Soy Beverage, Original, Dairy-Free, 1.89L	20124384001_EA	749	6
Frozen white fish, fillets (sole, haddock, halibut, tilapia or perch)*	perch	4	Lake Perch Fillet Tray Pack	20041506_KG	751	19
Frozen white fish, fillets (sole, haddock, halibut, tilapia or perch)*	sole	4	Sole Fillets, Skinless Boneless	20048346_EA	751	19
Frozen white fish, fillets (sole, haddock, halibut, tilapia or perch)*	halibut	3	Boneless Halibut Fillets, Fresh	20052098_KG	751	19
Frozen white fish, fillets (sole, haddock, halibut, tilapia or perch)*	tilapia	4	Tilapia Fillets	20111565_EA	751	19
Frozen white fish, fillets (sole, haddock, halibut, tilapia or perch)*	haddock	5	Haddock & Hops Battered Wild Haddock Fillets	20324259001_EA	751	19
Grains, wheat flour, whole-grain, soft wheat	whole wheat flour	1	Flour, Whole Wheat	20303111_EA	674	3
Grapes, fresh	Grapes, fresh	3	Red Seedless Grapes	20159199001_KG	748	10
Green beans, frozen	Green beans, frozen	1	Extra Fine Whole Green Beans	20431124_EA	661	5
Hummus	Hummus	3	Hummus	20584097_EA	751	8
Lentils, dry	Lentils	3	Lentils	20325921004_EA	745	9
Lettuce, iceberg, fresh	Lettuce, iceberg, fresh	2	Lettuce Iceberg	20135326001_EA	722	7
Lettuce, romaine, fresh	Lettuce, romaine, fresh	4	Romaine Hearts	20041478001_EA	745	12
Margarine, tub, non-hydrogenated	Margarine, non-hydrogenated	5	Soft Margarine Spread	20054663_EA	751	7
Mayonnaise	Mayonnaise	4	Miracle Whip Original Spread	20304232001_EA	751	10
Milk, partly skimmed, 2 % M.F.	Milk, partly skimmed	1	2% Milk	20085214_EA	585	6
Mozzarella cheese, 16.5% B.F.	Mozzarella cheese	4	Mozzarella 100% Real Cheese Slices	20742372_EA	750	12
Mushroom, fresh	Mushroom, fresh	2	White Mushrooms	21021499_EA	694	4
Onion, fresh	Onion, fresh	1	White Onion	20003839001_KG	721	7
Oranges, fresh	Oranges, fresh	1	Navel Orange	20426078001_KG	728	6
Peaches, canned (in water)	Peaches, canned in water	2	Peach Slices, Water Packed No Sugar Added	20296567001_EA	748	5
Peanut butter, natural (nothing added)	Peanut butter, natural	4	Organics Crunchy Peanut Butter	20055833_EA	751	4
Peanuts, dry roasted, unsalted	Peanuts, dry roasted, unsalted	2	Unsalted Peanuts	20057990002_EA	700	4
Pear, canned (in water)	Pear, canned in water	2	No Sugar Added, Pear, Halves, in Water	20296567008_EA	676	6
Peas, green, frozen	Peas, frozen	1	Green Peas	20312260_EA	721	4
Pepper, sweet, green, fresh	Pepper, sweet, green, fresh	1	Sweet Green Peppers	20425893001_KG	702	6
Pork, loin, centre chop	Pork, loin, centre chop	2	Centre Cut Pork Loin Chop Boneless	20827788_KG	354	8
Potato, fresh (any type)	Potato, fresh	3	Russet Potatoes 10 lb Bag	20601020001_EA	751	12
Salmon, pink, canned (no salt added, or low in sodium)	Salmon, pink, canned	5	Wild Pacific Pink Salmon	20003420_EA	750	14
Shredded wheat cereal, plain	Shredded wheat cereal	4	Shredded Wheat Big Biscuit Cereal	20014409_EA	749	9
Spinach, frozen	Spinach, frozen	2	Whole Leaf Spinach	20301191002_EA	749	6
Strawberries, frozen	Strawberries, frozen	1	Sliced Strawberries	20312227_EA	740	6
Sunflower seeds, unsalted	Sunflower seeds, unsalted	5	Dry Roasted Shelled Sunflower Seeds, Unsalted	20771372_EA	714	12
Sweet potato, fresh	Sweet potato, fresh	1	Sweet Potato	20127708001_KG	720	10
Toasted O's cereal, plain	Toasted cereal	2	Chex Honey Nut Cereal	20376392_EA	749	10
Tofu, firm or extra firm	Tofu, firm or extra firm	5	Tofu, Extra Firm	20570507_EA	739	16
Tomato, canned (no salt added, or low in sodium)	Tomato, canned	3	Diced Tomatoes	20600787_EA	750	3
Tomato, fresh	Tomato, fresh	2	Roma Tomatoes	20143381001_KG	740	8
Tuna, canned (no salt added)	Tuna, canned	4	Tuna, Solid Light In Olive Oil	20035383_EA	751	8
Turkey, all classes, ground	Turkey, ground	2	Chicken Ground Lean	20706483_EA	533	2
Vegetable oil	Vegetable oil	1	Vegetable Oil	20009865001_EA	709	5
Vegetables, mixed, frozen, boiled, drained	Vegetables, frozen	4	Asian-Style Vegetable Blend	20295769_EA	750	6
Whole wheat pasta, spaghetti or macaroni	Whole wheat pasta	5	Whole Grain Lasagna Pasta	20116949_EA	751	8
Winter squash (e.g. pumpkin, butternut or acorn squash), fresh	Winter squash, fresh	5	Organic Butternut Squash	20051060001_KG	747	8
Yogurt, plain, 1–2%	Yogurt, plain	5	Plain 0 % M.F. Greek Yogurt	20357382001_EA	751	7

*NNFB product was disaggregated for keyword development, product matches were once again aggregated following SKU matching.

2.4. Data collection and cleaning

Data was summarized, analyzed and statistics/visualizations were produced using the R programming language (R 4.1.2 “Bird Hippie”). All visualizations were produced with the ggplot2 R library.

2.5. Data processing and Product-Matching

The scraper output was designed to deliver tidy data, however, this data still requires considerable processing given the number and heterogeneity of website data entry/product labels (Harrington et al., 2019). We designed a product-matching algorithm to isolate products of interest, comparable to the fieldwork process used in brick-and-mortar NNFB methods (Health Canada, 2019, The Cost of Eating Well in the District of Thunder Bay, 2021).

First, SKU codes identified as non-food products were removed. Second, the most frequent five SKU codes returned for each product in the NNFB across all retailers per search keyword were selected and manually coded by a team member as a matched or unmatched (Table 1). Keywords without at least one match among the top five products across all retailers were matched in a second pass based on the ten most frequent products returned. Following the second pass, all product keywords had ≥1 SKU code matches for NNFB calculations with high store coverage (Table 1). In rare instances, keyword searches simply did not return appropriate SKUs, so a modified broader keyword was used with matching done manually. For example, ‘Toasted O’s Cereal’ is the keyword as sourced from NNFB documentation, but preliminary searches performed poorly so the keyword ‘toasted cereal’ was used in its place, and returned products were matched by their appropriateness for the intended NNFB product: Toasted O’s cereal (Table 1). Following keyword modification, 65 keywords were used to identify and collect product-level data for the 61 items in the NNFB (Table 1).

All product prices were transformed to $CAD/gram of edible food. If a product did not have $CAD/gram information, or the necessary information to calculate $CAD/gram (e.g., the price unit was designated “price each”), the 2015 Canadian Nutrient File was used to derive standard unit prices i.e. identify a reference mass in grams for the calculation of $CAD/gram (Mah and Taylor, 2019).

2.6. Analysis

2.6.1. dNNFB store-level basket cost computation

dNNFB basket costs were calculated for each store following NNFB protocol, (Health Canada, 2019) expressed as $CAD/week. For stores with missing NNFB constituent items (i.e., item not available for sale/listed), an item price was imputed to calculate the store-level dNNFB basket cost, using the mean price for the item across the remainder of the dataset. As a sensitivity analysis, the mean-imputation method was compared with a spatial imputation strategy that imputed missing item prices with the price from the most geographically proximate retailer where it was available (Euclidean distance) (Figure S2). Summary statistics were calculated across all stores without stratification.

2.6.2. Age-sex intake adjusted store-level basket cost computation and analysis

Dietary reference intakes for each age-sex category (Dietary Reference Intakes) (provided by Statistics Canada, correspondence November 4, 2021, see Section 3: Calculating the amounts of each food for more information) (Health Canada, 2019) were used to calculate age-sex intake adjusted weekly basket costs. All baskets were inflated by 5% to reflect miscellaneous goods (spices, condiments, coffee etc.) not otherwise accounted for, per NNFB protocol. Summary statistics were calculated by geographic region (five Statistics Canada regions were used: Atlantic, Quebec, Ontario, Prairies, and British Columbia) and by banners. Regional and retailer banner level dNNFB costs were estimated by collapsing all NNFB age-sex categories without any form of categorical weighting.

2.6.3. Reference family validation

As a means of postdictive validation, we calculated the average annual dNNFB costs for a reference family (2 adults, 2 children, see Supplementary Table C3 from Djidel et al. (Djidel, et al., 2020)) and compared our findings with inflation adjusted nationally reported statistics, using MBM data from 2020 across metropolitan regions of Canada and population centres of Canada respectively. (Djidel, et al., 2020) The MBM reports 38 cities (large urban population centres); following a spatial join, our sample of stores covered 671 of 1010 total population centres (small, medium, and large urban). Minima, means, and the centremost mode were compared with MBM values for corresponding population centres. Pearson correlation coefficients and mean differences were calculated. P-values were adjusted for multiple tests using Bonferroni correction.

3. Results

Price data was sourced and processed from 760 stores. After removing stores without food products (n = 3) and stores for which a website existed but accurate spatial coordinates for geolocation could not be verified (n = 6), the final sample size for costing was n = 751 stores. Stores belonged to 11 banners across 10 provinces and two territories (Fig. 1). There was no available price data for Nunavut. The initial dataset was comprised of 5778 unique items, with >15,000 distinct prices. Following product matching, 184 discrete products were used to compute dNNFB basket costs. The average number of stores containing ≥1 price per food product was ∼709; 15 products could be found in every store sampled (n = 751, Table 1).

Fig. 1.

dNNFB reference family costs stratified by province and banner. Total basket costs are calculated based on the reference family employed by the MBM. All prices are in unadjusted $CAD. Mean estimates are indicated by black lines.

3.1. Regional variation in Age-Sex intake adjusted dNNFB costs

dNNFB estimates demonstrated clear regional variation. As shown in Table 2, the highest average dNNFB costs per week were sourced from stores in the Atlantic Canadian region ($88.67/week), followed by mean Territorial costs (Table 2). These regional findings were consistent across all age-sex groups (Table 2). Territorial retailers had the highest minimum dNNFB cost at $45.85/week. Despite similarities between Atlantic Canadian and Territorial prices, they diverged in the range of dNNFB costs across stores. Atlantic Canadian retailers had a difference of $76.37/week from the least to the most expensive dNNFB, while territorial retailers had a difference of $63.21/week, representing the most and least widespread dNNFB costs respectively. The lowest minimum weekly cost was a store in the Prairies ($43.73/week), while the lowest maximum weekly cost was a store within the Territories ($109.06/week).

Table 2.

Summary of Age-Sex Intake Adjusted Weekly dNNFB Costs Stratified by Statistics Canada Region and Loblaw’s Banner. Total weekly basket costs are calculated specific to each age-sex category as outlined in the Statistics Canada NNFB documentation. All prices are in adjusted $CAD/week. Basket costs are expressed as means across respective levels of stratification; 95% confidence intervals are reported in brackets. Average baskets were calculated at the level of regions or banners by taking the mean value across all age-sex categories without weighting. Averages across all age-sex categories per regional/banner strata, and averages across all banners/regions per age-sex strata are bolded (far right and bottom of table).

Age-sex/Banner-Region	All Banners/Regions	Atlantic Canada	British Columbia	Ontario	Prairies	Quebec	Territories	Atlantic Superstore	Fortinos	Loblaws	Maxi	Dominion	NoFrills	Provigo	Real Canadian Superstore	Valumart	Your Independent Grocer	Zehr's
Average	77.87(77.63–78.11)	88.67(88.01–89.33)	75.65(74.81–76.49)	76.43(76.08–76.78)	73.87(73.3–74.44)	77.79(77.28–78.3)	82.57(79.34–85.8)	87.92(87.05–88.79)	80.36(79.24–81.48)	84.65(83.82–85.48)	70.73(70.15–71.31)	92.12(90.17–94.07)	67.67(67.3–68.04)	84.32(83.64–85)	70.47(70–70.94)	85.53(84.14–86.92)	87.18(86.62–87.74)	83.83(82.92–84.74)
Female 2–3	50.14(49.73–50.55)	57.31 (56.79–57.83)	48.57 (47.22–49.92)	49.21 (48.6–49.82)	47.4 (46.6–48.2)	50.13 (49.39–50.87)	53.29 (48.43–58.15)	56.77 (56.54–57)	51.69 (51.3–52.08)	54.61 (54.37–54.85)	45.43 (45.23–45.63)	59.7 (58.96–60.44)	43.42 (43.21–43.63)	54.47 (54.33–54.61)	45.15 (44.88–45.42)	55.24 (54.96–55.52)	56.36 (56.09–56.63)	53.95 (53.73–54.17)
Male 2–3	50.14(49.73–50.55)	57.31 (56.79–57.83)	48.57 (47.22–49.92)	49.21 (48.6–49.82)	47.4 (46.6–48.2)	50.13 (49.39–50.87)	53.29 (48.43–58.15)	56.77 (56.54–57)	51.69 (51.3–52.08)	54.61 (54.37–54.85)	45.43 (45.23–45.63)	59.7 (58.96–60.44)	43.42 (43.21–43.63)	54.47 (54.33–54.61)	45.15 (44.88–45.42)	55.24 (54.96–55.52)	56.36 (56.09–56.63)	53.95 (53.73–54.17)
Female 4–8	58.13(57.68–58.58)	65.95 (65.37–66.53)	56.32 (54.86–57.78)	57.2 (56.52–57.88)	55.01 (54.16–55.86)	58.06 (57.29–58.83)	61.25 (56.04–66.46)	65.35 (65.07–65.63)	59.82 (59.4–60.24)	63.19 (62.93–63.45)	53.18 (52.94–53.42)	68.71 (67.86–69.56)	50.76 (50.52–51)	62.56 (62.4–62.72)	52.62 (52.32–52.92)	64.04 (63.74–64.34)	64.98 (64.68–65.28)	62.28 (62.04–62.52)
Male 4–8	58.6(58.15–59.05)	66.52 (65.94–67.1)	56.78 (55.31–58.25)	57.65 (56.96–58.34)	55.45 (54.59–56.31)	58.49 (57.71–59.27)	61.75 (56.52–66.98)	65.88 (65.6–66.16)	60.48 (60.06–60.9)	63.72 (63.46–63.98)	53.51 (53.27–53.75)	69.29 (68.43–70.15)	51.11 (50.87–51.35)	63.09 (62.93–63.25)	53.05 (52.75–53.35)	64.56 (64.25–64.87)	65.53 (65.23–65.83)	62.79 (62.55–63.03)
Female 9–13	74.92(74.34–75.5)	85.1 (84.37–85.83)	72.6 (70.72–74.48)	73.72 (72.83–74.61)	70.92 (69.82–72.02)	74.76 (73.75–75.77)	78.96 (72.29–85.63)	84.36 (84.03–84.69)	77.08 (76.58–77.58)	81.57 (81.26–81.88)	68.3 (68.01–68.59)	88.44 (87.45–89.43)	65.36 (65.06–65.66)	80.72 (80.53–80.91)	67.71 (67.31–68.11)	82.59 (82.24–82.94)	83.8 (83.42–84.18)	80.63 (80.33–80.93)
Male 9–13	75.73(75.14–76.32)	86.13 (85.39–86.87)	73.38 (71.47–75.29)	74.47 (73.57–75.37)	71.69 (70.57–72.81)	75.6 (74.58–76.62)	79.87 (73.1–86.64)	85.42 (85.08–85.76)	78.01 (77.48–78.54)	82.42 (82.1–82.74)	69.06 (68.77–69.35)	89.59 (88.59–90.59)	65.96 (65.65–66.27)	81.65 (81.45–81.85)	68.45 (68.04–68.86)	83.46 (83.08–83.84)	84.73 (84.35–85.11)	81.45 (81.14–81.76)
Female 14–18	74.92(74.34–75.5)	85.1 (84.37–85.83)	72.6 (70.72–74.48)	73.72 (72.83–74.61)	70.92 (69.82–72.02)	74.76 (73.75–75.77)	78.96 (72.29–85.63)	84.36 (84.03–84.69)	77.08 (76.58–77.58)	81.57 (81.26–81.88)	68.3 (68.01–68.59)	88.44 (87.45–89.43)	65.36 (65.06–65.66)	80.72 (80.53–80.91)	67.71 (67.31–68.11)	82.59 (82.24–82.94)	83.8 (83.42–84.18)	80.63 (80.33–80.93)
Pregnant < 19	85.22(84.56–85.88)	96.88 (96.08–97.68)	82.92 (80.81–85.03)	83.68 (82.67–84.69)	81 (79.75–82.25)	84.96 (83.8–86.12)	90.13 (82.75–97.51)	96.15 (95.77–96.53)	88.32 (87.67–88.97)	92.46 (92.1–92.82)	77.52 (77.2–77.84)	100.46 (99.43–101.49)	74.23 (73.87–74.59)	91.83 (91.6–92.06)	77.43 (76.94–77.92)	93.6 (93.04–94.16)	95.18 (94.74–95.62)	91.5 (91.14–91.86)
Lactating < 19	84.89(84.23–85.55)	96.52 (95.73–97.31)	82.59 (80.48–84.7)	83.36 (82.36–84.36)	80.69 (79.45–81.93)	84.66 (83.5–85.82)	89.82 (82.44–97.2)	95.78 (95.4–96.16)	87.73 (87.08–88.38)	92.09 (91.73–92.45)	77.28 (76.96–77.6)	100.06 (99.04–101.08)	73.95 (73.6–74.3)	91.48 (91.25–91.71)	77.14 (76.65–77.63)	93.27 (92.74–93.8)	94.84 (94.41–95.27)	91.15 (90.8–91.5)
Male 14–18	88.34(87.64–89.04)	100.56 (99.7–101.42)	86.03 (83.83–88.23)	86.65 (85.58–87.72)	84.04 (82.73–85.35)	88.14 (86.86–89.42)	93.57 (86.04–101.1)	99.66 (99.27–100.05)	91.73 (91.04–92.42)	96.06 (95.69–96.43)	79.96 (79.64–80.28)	104.42 (103.38–105.46)	76.51 (76.14–76.88)	95.69 (95.45–95.93)	80.37 (79.82–80.92)	97.19 (96.63–97.75)	98.84 (98.37–99.31)	95.09 (94.72–95.46)
Female 19–30	79.56(78.9–80.22)	90.9 (90.07–91.73)	77.05 (74.9–79.2)	78.09 (77.1–79.08)	75.07 (73.78–76.36)	79.64 (78.44–80.84)	84.66 (76.74–92.58)	90.1 (89.69–90.51)	81.88 (81.32–82.44)	86.92 (86.56–87.28)	71.96 (71.62–72.3)	94.5 (93.24–95.76)	68.78 (68.43–69.13)	86.74 (86.55–86.93)	71.33 (70.87–71.79)	87.62 (87.23–88.01)	89.51 (89.08–89.94)	86.13 (85.81–86.45)
Pregnant 19–30	89.96(89.25–90.67)	102.39 (101.53–103.25)	87.59 (85.28–89.9)	88.2 (87.14–89.26)	85.54 (84.16–86.92)	89.92 (88.63–91.21)	95.71 (87.46–103.96)	101.65 (101.22–102.08)	92.72 (91.97–93.47)	97.63 (97.2–98.06)	81.68 (81.34–82.02)	106.19 (104.92–107.46)	78.28 (77.89–78.67)	97.52 (97.29–97.75)	81.51 (80.97–82.05)	98.22 (97.63–98.81)	100.56 (100.08–101.04)	96.9 (96.5–97.3)
Lactating 19–30]	89.09(88.39–89.79)	101.39 (100.53–102.25)	86.74 (84.45–89.03)	87.32 (86.27–88.37)	84.71 (83.34–86.08)	89.08 (87.81–90.35)	94.78 (86.65–102.91)	100.61 (100.18–101.04)	91.75 (91–92.5)	96.61 (96.18–97.04)	80.97 (80.63–81.31)	105.15 (103.88–106.42)	77.5 (77.11–77.89)	96.57 (96.34–96.8)	80.76 (80.22–81.3)	97.33 (96.74–97.92)	99.59 (99.11–100.07)	95.89 (95.49–96.29)
Male 19–30	97.79(97.01–98.57)	111.46 (110.53–112.39)	95.38 (92.87–97.89)	95.71 (94.54–96.88)	93.22 (91.71–94.73)	97.77 (96.32–99.22)	104.36 (95.52–113.2)	110.62 (110.17–111.07)	101.46 (100.58–102.34)	105.9 (105.43–106.37)	88.51 (88.15–88.87)	115.77 (114.45–117.09)	84.85 (84.42–85.28)	106.32 (106.06–106.58)	88.85 (88.22–89.48)	106.72 (105.97–107.47)	109.33 (108.79–109.87)	105.11 (104.66–105.56)
Female 31–50	78.33(77.67–78.99)	89.59 (88.77–90.41)	75.81 (73.68–77.94)	76.88 (75.89–77.87)	73.86 (72.58–75.14)	78.42 (77.23–79.61)	83.37 (75.49–91.25)	88.78 (88.37–89.19)	80.68 (80.12–81.24)	85.65 (85.29–86.01)	70.78 (70.44–71.12)	93.22 (91.96–94.48)	67.65 (67.3–68)	85.48 (85.29–85.67)	70.12 (69.67–70.57)	86.31 (85.93–86.69)	88.19 (87.77–88.61)	84.87 (84.55–85.19)
Pregnant 31–50	88.64(87.94–89.34)	100.86 (100.01–101.71)	86.28 (84.01–88.55)	86.88 (85.83–87.93)	84.27 (82.92–85.62)	88.66 (87.41–89.91)	94.3 (86.22–102.38)	100.12 (99.7–100.54)	91.16 (90.48–91.84)	96.13 (95.7–96.56)	80.66 (80.32–81)	104.61 (103.33–105.89)	77.13 (76.75–77.51)	96.05 (95.83–96.27)	80.35 (79.82–80.88)	96.81 (96.27–97.35)	99.06 (98.59–99.53)	95.39 (95–95.78)
Lactating 31–50	88.64(87.94–89.34)	100.86 (100.01–101.71)	86.28 (84.01–88.55)	86.88 (85.83–87.93)	84.27 (82.92–85.62)	88.66 (87.41–89.91)	94.3 (86.22–102.38)	100.12 (99.7–100.54)	91.16 (90.48–91.84)	96.13 (95.7–96.56)	80.66 (80.32–81)	104.61 (103.33–105.89)	77.13 (76.75–77.51)	96.05 (95.83–96.27)	80.35 (79.82–80.88)	96.81 (96.27–97.35)	99.06 (98.59–99.53)	95.39 (95–95.78)
Male 31–50	92.8(92.06–93.54)	105.74 (104.83–106.65)	90.41 (88.03–92.79)	90.92 (89.8–92.04)	88.32 (86.89–89.75)	92.73 (91.34–94.12)	98.93 (90.52–107.34)	104.81 (104.37–105.25)	96.14 (95.39–96.89)	100.73 (100.3–101.16)	83.83 (83.48–84.18)	109.74 (108.43–111.05)	80.48 (80.07–80.89)	100.95 (100.72–101.18)	84.17 (83.58–84.76)	101.47 (100.84–102.1)	103.89 (103.38–104.4)	100.02 (99.61–100.43)
Female 51–70	73.64(73.06–74.22)	83.72 (82.99–84.45)	71.34 (69.49–73.19)	72.42 (71.54–73.3)	69.69 (68.61–70.77)	73.51 (72.5–74.52)	77.63 (71.06–84.2)	82.93 (82.6–83.26)	75.69 (75.19–76.19)	80.13 (79.82–80.44)	67.08 (66.79–67.37)	87.03 (86.04–88.02)	64.18 (63.88–64.48)	79.46 (79.27–79.65)	66.53 (66.13–66.93)	81.25 (80.9–81.6)	82.43 (82.05–82.81)	79.23 (78.94–79.52)
Male 51–70	82.2(81.51–82.89)	94.18 (93.34–95.02)	80.04 (77.85–82.23)	80.46 (79.42–81.5)	78.09 (76.77–79.41)	82.12 (80.82–83.42)	87.74 (80.08–95.4)	93.24 (92.86–93.62)	85.68 (84.99–86.37)	89.52 (89.14–89.9)	73.79 (73.47–74.11)	97.94 (96.86–99.02)	70.67 (70.3–71.04)	89.81 (89.6–90.02)	74.3 (73.75–74.85)	90.51 (89.97–91.05)	92.54 (92.07–93.01)	88.82 (88.46–89.18)
Female > 70	71.13(70.57–71.69)	80.95 (80.24–81.66)	68.82 (67.01–70.63)	69.99 (69.13–70.85)	67.21 (66.15–68.27)	70.99 (70–71.98)	74.98 (68.5–81.46)	80.15 (79.82–80.48)	73.1 (72.63–73.57)	77.54 (77.24–77.84)	64.7 (64.41–64.99)	84.19 (83.21–85.17)	61.94 (61.65–62.23)	76.8 (76.61–76.99)	64.12 (63.75–64.49)	78.59 (78.27–78.91)	79.72 (79.34–80.1)	76.65 (76.37–76.93)
Male > 70	80.4(79.78–81.02)	91.31 (90.56–92.06)	78.21 (76.21–80.21)	78.96 (78.02–79.9)	76.46 (75.29–77.63)	80.23 (79.16–81.3)	84.92 (78.05–91.79)	90.58 (90.23–90.93)	82.82 (82.14–83.5)	87.2 (86.86–87.54)	73.41 (73.11–73.71)	94.85 (93.85–95.85)	70.16 (69.83–70.49)	86.52 (86.3–86.74)	73.11 (72.65–73.57)	88.19 (87.67–88.71)	89.72 (89.31–90.13)	86.33 (85.97–86.69)

3.2. Retailer banner variation in Age-Sex intake adjusted dNNFB costs

We also detected a substantial range in average dNNFB cost estimates by retailer banner, shown in Table 2. Dominion, the regular store banner in the province of Newfoundland and Labrador had the highest average ($92.12/weekly) and minimum and maximum basket costs across all banners (Table 2). NoFrills, a discount banner, had the lowest cost average weekly dNNFB costs, ($67.67/week). Moreover, the difference between the minimum and maximum dNNFB costs across all of the Dominion stores was $62.92/weekly, while that same value for NoFrills came to $51.71/weekly.

3.3. Age-sex intake adjusted dNNFB

As expected, average intake adjusted dNNFB costs varied across age-sex groups from $50.14/weekly for 2–3-year-old females to $97.79/weekly for 19–30-year-old males (Table 2). In every case, female consumers of the same age would face higher basket costs during pregnancy or lactation relative to those who are not pregnant or lactating (Table 2).

3.4. Reference family baskets and dNNFB comparison to the MBM

We detected within- and between-jurisdiction differences in reference family costs. The most expensive dNNFB for reference families was found under the Dominion (regular) banner on the island of Newfoundland ($418.33/week, Fig. 1; Supplementary Table S5). The cheapest average basket cost was estimated in NoFrills (discount) in Ontario, at $302.45/week. Within a single province, average dNNFB costs differed by ∼$60 or more per week, in Quebec, the Yukon, and the Northwest Territories (Fig. 1, Table S5).

In every comparison made across all corresponding population centres in Canada, the dNNFB produced annual cost estimates higher than the existing food component of the MBM (Fig. 2, Table 3) after adjusting for inflation. Indeed, the closest magnitude of difference at the level of population centres was ∼$4,000/year (not visualized). When compared to minima, the dNNFB and MBM had correlation coefficients of ∼0.72 across cities and all population centre classes.

Fig. 2.

Comparison of the dNNFB with the Food Component of the MBM. MBM reference family intakes were used to calculate comparative numbers using dNNFB data. Bars reflect MBM values sourced from Statistics Canada, points reflect dNNFB reference family basket prices at the store-level. Both bars and points are color-coded to identify provinces. A) MBM comparator cities represent specified population centres with city names B) population centre classes were further stratified by provinces for comparison with MBM values. There are no colors assigned to Territories in the figure legend to reflect the lack of MBM data for these regions. Costs are expressed as $CAD/year.

Table 3.

Correlation Coefficients and Mean Differences between MBM estimates and dNNFB estimates across CMAs and Population Centre Classes. dNNFB annual costs were calculated by multiplying weekly basket costs by 52 weeks. Multiple summary dNNFB statistics were chosen as comparators to reflect the non-normal distribution of basket costs within multiple regions. All correlation coefficients and mean differences were statistically significant following Bonferroni adjustment for multiple tests. MBM comparator cities represent specified population centres with city names. Cost differences are expressed in 2021 $CAD/year. *Basket costs at the population centre level were generally multimodal, to overcome this the median of all modes returned per centre was used as a comparator.

Summary Comparator	Jurisdiction	Correlation (dNNFB vs. MBM)	Mean Difference (dNNFB – MBM, 2021 $CAD/year)
Mean	City	0.776787	−5957.28
Mean	Population Centre Class	0.703299	−5732.05
Minima	City	0.724259	−4509.86
Minima	Population Centre Class	0.724834	−4245.37
Mode*	City	0.692862	−5847.86
Mode*	Population Centre Class	0.494928	−5791.9

4. Discussion

We have demonstrated notable differences in cost of living attributable to food exist in Canada, stratified by regional, store-level, and age-sex group characteristics. Furthermore, and of concern for assessing the adequacy of policy responses during the pandemic, our cost estimates exceeded reported national statistics from the MBM after adjusting for inflation.

4.1. Where do baskets cost the most?

Atlantic Canada and the Territories had greater basket costs than any other regions (Table 2), a pattern previously detected in the national MBM, often attributed to the rural and remote geography of these regions that constrain food distribution (Djidel, et al., 2020). Similar to previous food costing studies, we also found that variation in dNNFB prices was observable at the level of retailer banner (Caspi et al., 2017). The eleven banners used to calculate dNNFB costs all fell under Loblaw’s ownership whose merchandising strategies are reflected in basket costs (Table 2, Table S5). For instance, NoFrills is generally marketed as an affordable shopping location (About No Frills). In contrast, Your Independent Grocer specifically targets “foodies” with product variety and freshness (About Your Independent Grocer).

Several recent studies in the US and the UK have investigated food price variability by deploying scrapers analogous to ours (Cavallo, 2017, Harrington et al., 2019) Cavallo 2017 found that supermarket pricing demonstrated very little variation across a sample of US zip codes. Harrington et al. 2018 used a set of category trees to capture product-level variability in pricing across six major supermarkets in the UK, which we also found in disaggregated analyses prior to the estimating of food basket costs (Supplementary Table S1). It is possible that our variation in findings may be a function of the methods respective teams employed; Cavallo evaluated a single banner with multiple zip codes; Harrington et al. constructed their scraper to pull data from six separate supermarket corporations, while we designed a census of a single corporation comprised of eleven store banners across all available postal codes.

4.2. Who’s facing the highest costs?

The cost of meeting weekly food needs varies across the life span, and by specific dietary needs such as pregnancy/lactation (Table 2, Figure S4). The presence of known biological variation in diet costs throughout the life course, (Bai et al., 2022) reproduced in our study, reinforces the need for consistent methods for age-sex intake adjustment subgroup considerations in diet costing and affordability assessment (The Cost of Eating Well in the District of Thunder Bay, 2021). However, although we observed patterns of food costs that corresponded broadly with underlying biological variation, we did not observe a greater basket cost for pregnant or lactating females relative to 19-30yo males, as has been found in international studies of diet costs (Bai et al., 2022).

4.3. Comparing dNNFB to NNFB local estimates: A need for precision

Recent implementations of the NNFB at sub-national scales (The Cost of Eating Well in the District of Thunder Bay, 2021) and the current food component of the MBM all reported basket costs that differed from our dNNFB estimates after adjustment for inflation (Djidel, et al., 2020). Despite many aspects of our study design focused on establishing conservative estimates, our basket costs remained greater than previous estimates at sub-national and national scales, reported weekly or annually (Table 2, Table 3). Annually, the dNNFB estimates differed from the MBM by ∼$4,200 to ∼$6000 per year (Table 3). After centering around the mean, the distributions of the food component of the MBM and our dNNFB followed a similar bimodal distribution (Supplementary Figure S6). Although this suggests a consistent distribution of basket costs whether calculated with the dNNFB or the MBM, dNNFB and MBM values clearly varied across population centres (Table 3, Fig. 2).

One explanation for this divergence is the difference in sampling strategy between the 38 cities in the MBM and the 751 stores belonging to our analyzable dNNFB sample. Additionally, we would argue that dNNFB basket values may better reflect the real-world costs experienced by individuals. Our product matching algorithm used high frequency returned items, potentially representing factors related to consumer demand, in contrast to NNFB which requires selections to be made by field surveyors. Further, it is clear from Table 2 that there is a fairly wide range of dNNFB costs across banners, and that dNNFB minima demonstrated the greatest similarity to MBM values across all centres (Table 3). This has major implications for the interpretation of poverty thresholds across the country and how cost of living risk occurs unevenly for subpopulations restricted to particular food environments. Additionally, the reference family employed by the MBM acts on several normative assumptions regarding family size and composition that may underestimate the cost burden associated with several social structural determinants of diet.

A recent analysis of dietary inequities in Canada suggests that the socioeconomic gaps in dietary quality have remained steady or increased from 2004 to 2015 (Olstad et al., 2021). To understand the underlying determinants of these disparities and to predict the consequences for diets as a consequence of current inflation, it is crucial to define affordability as consumers experience it. One approach to this problem is to examine diet costs, matching food supply data to dietary intake surveillance to estimate what is spent for what is consumed. However, a prevailing challenge in the diet cost literature has been the lack of sufficiently proximal food price data (e.g., direct expenditures linked to intakes) (Luongo et al., 2022). Our novel approach to data collection in this manuscript with an exemplar proof of concept application using the NNFB demonstrates that the problem of diet costs can also be tackled from the food environment monitoring side to generate timely, standardized, and localized costing estimates, critical components of the optimal approach to food price and affordability monitoring as set out by international benchmarking researchers such as the INFORMAS network (Lee et al., 2013). As the explosion of digitally available food products increasingly reflect the reality of product availability “on the shelf” (Scarborough et al., 2020, Granheim et al., 2020), harnessing the digital food environment through a web scraper can significantly expedite and standardize data collection for the purposes of healthy/unhealthy pricing comparisons (Zorbas et al., 2020), and for more generalized measures of affordability (Lee et al., 2013, Djidel et al., 2020, The State of Food Security and Nutrition in the World, 2021).

4.4. Strengths and limitations

The strengths of the dNNFB address persistent limitations of in-person store auditing used for many forms of food price data collection required for research and policy. The development of the dNNFB enabled us to analyze data at the national and sub-national scale, yet we retained the ability to report data at the resolution of SKUs within individual stores, making the dNNFB as salient to local communities as to federal policymakers.

This work was subject to several limitations. Our scraping and matching approach prioritized return of products for the final costing analyses that had a high national availability to improve regional and banner comparability, but this may have underestimated local/regional product selection and under- or over-estimated price. A salient example is milk, a commonly consumed, supply-managed product for which our price matching captured prices for only 588 of 754 stores (∼78%). Milk product SKUs plausibly vary based on local milk branding and availability. It is possible that we have captured aspects of regional product distinctness or uneven product distribution, rather than price variability per se. Further, in prioritizing high national availability, our product selection algorithm may have been skewed towards more expensive products (e.g., national brands) in comparison to generic or own-brand (private-label) products offered by banners which may be priced more competitively (cheaper). Any bias in product return as a function of retailer search engine design towards or away from specific branded product lines may have also affected the results. Additionally, our use of the Canadian Nutrient File to fill in missing product weights may also have underestimated product sizes, thus inflating final costs per gram.

Finally, despite the limited number of competitors at the national scale (QUIC Research Report, 2017), our findings may not be representative of all grocers. Further, although data collection methods are easily reproduced internationally, specific basket costing methods will vary across national and subnational populations.

5. Conclusions

The affordability of food is well established as a risk for dietary compromise, with serious adverse consequences for the burden of diet-related disease, and elevated health care costs (Kirkpatrick and Tarasuk, 2008, Olstad et al., 2021, Ricciuto and Tarasuk, 2007, Tarasuk, 2015). This study has emphasized the need for strengthened, multi-dimensional, precision monitoring of cost-of-living disparities, so as to capture the rapidly evolving scope and scale of population health risk emerging from real food environments experienced by consumers.

Economic changes during the COVID-19 pandemic have renewed the urgency for robust research on inequitable health impacts of food inflation (Lancet Public Health, 2022). This is of particular concern given the apparent uncoupling of food inflation from aggregate price indices during the pandemic (Bai, et al., 2022). The gap between our proof-of-concept monitoring through the dNNFB and the MBM may suggest we are underestimating the impact of inflation on consumers, and consequently, the seriousness of current food affordability-related health risk.

Author contributions

NT led conceptualization, methodology, software development, data curation, formal analysis, investigation, validation and data visualization, wrote the original draft, responded to revisions and edits from collaborating authors and acts as the guarantor for the final work. GL contributed to the data interpretation and editing of drafts of the final submitted manuscript. EJ contributed to editing of drafts of the final submitted manuscript. CLM provided supervisory and administrative oversight to the authorship team and computing infrastructure; and contributed to conceptualization, interpretation, and verification of findings, as well as review and editing of drafts and approval of the manuscript.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Data availability

Data will be made available on request.