Andrew Cornwall
Home Page

In 1999 I carried out a pilot study to compare the prices of a sample of identical products that are commonly available in supermarkets. In my travels I had noticed that consumable consumer goods were generally higher priced, after accounting for the exchange rate, in the United States than where I live in Nova Scotia, Canada. My experience in this regard has been frequently affirmed by the remarks of travelers to the United States from other regions of Canada. My 1999 Recreation Vehicle (RV) trip across the United States gave me an opportunity to test these impressions by collecting regional price information on a controlled sample of products.

There was a larger purpose in conducting the pilot study. I have a professional background in economics, and am curious about how markets function. Any market analysis requires an unambiguous and accurate measure of price structures. Also, when examining how market factors work to set local prices, it is necessary that there be at least some resulting regional price differentiation. The results of the pilot study indicate that it is possible to collect comparable price information in different regions, and that there are significant regional differences in these prices that warrant more extensive study. An added bonus is the prospect of studying price differences in regions of two free-trading countries.

There is a danger that readers of this study will jump to the conclusion that a particular supermarket in a community is unjustifiably out of line with the prices of its counterparts elsewhere. I have not named the supermarkets surveyed in order avoid such attribution. I believe that the information in this report is accurate, but "to err is human" and I am a member of the species. I did not obtain the approval of management of the supermarkets to gather price information, and they did not have an opportunity to verify the accuracy of the data. I seldom purchased the sampled items at the time price data was being collected. Thus, I am presuming that the prices shown at the shelves were the ones being charged.



How the Sample Was Chosen

Table 1 (Click for Table) describes the 32 products surveyed and reported on in this document. Items included in the sample do not reflect any other agenda than to obtain a cross section of products that should be reasonably standardized by virtue of their brand names and availability across the United States and in Canada. To select the sample I enlisted the expert assistance of my wife, a shopper who has traversed many miles of supermarket isles. My wife also contributed to the study in the capacities of data-gathering quality control supervisor and editorial advisor.

An initial sample of 41 products was compiled during a shopping / data gathering excursion at a supermarket in Phoenix, Arizona. My wife and I selected a few product types from each isle. We did not want to include too many products in order to restrict the amount of effort needed to collect price information. Within a product type we choose a popular brand that we recognized was also available in Canada. Except for dog biscuits, they were the types of products that we might buy for our own use. Of the 41 branded products 9 were subsequently deleted from the analysis of the pilot study because they were not available in one or more regions. The deleted products were (generically) margarine, canned pineapple, tuna fish, peanut butter, baby food, a breakfast cereal, chocolate syrup, orange juice, and instant mashed potatoes.

As part of the process of selecting the sample we also identified a reference package size for each product. Where there was a range of package sizes we generally selected a middle size, forgoing the jumbo pack and smallest container.

We avoided the temptation of extending the sample to include fresh produce, meats (fresh and processed), bakery and dairy products. Many of these products do not have a brand name, or have a local brand. There are also regional differences in how the products are prepared for sale. We also avoided 'house' and 'no-name' products of any variety, because their quality may not be uniform between regions and/or supermarkets. Although the prices of all of these excluded products are worthy of examination, they involve other factors that are too complex to account for in a pilot study.

To some extent, the variations in prices of the sampled products should track the relative regional pricing of the excluded kinds of products. Indeed, this has been our general experience with the exception of fresh dairy and poultry products, where prices in Canada are higher due to government price controls instituted to assist farmers.



Price Data Collection

The prices were collected from mid-September through mid-October in four distinctly separate communities in the United States and one in Canada. In the United States communities we identified the store to be surveyed by asking one or two local residents where to shop to obtain the lowest prices. They invariably referred to a supermarket that was a member of a multi-store chain. The supermarket surveyed in Nova Scotia, also part of a chain, offers the lowest prices in the Halifax area. All of the supermarkets in the survey were members of different parent organizations. The communities and supermarket affiliations are shown below:

- Phoenix, Arizona (Southwestern states regional chain)

- Abilene, Kansas (Kansas and Nebraska chain)

- Frankfort, Kentucky (Mid-West regional chain)

- Concord, New Hampshire (North American chain)

- Halifax, Nova Scotia (national chain)

The rules we used to collect prices were as follows:

1. All prices were recorded before any sales taxes, and are 'cash and carry'. (In any case, none of the stores charged different prices for payment by cash versus credit card or debit card.) Special pricing for damaged or (nearly) outdated product was not considered.

2. If the exact brand and reference size of a product was available then record the lowest price in the store. In many instances this would be a 'sale' price.

3. If the reference size of the brand of product was not available, then record the lowest price of the available size closest to the reference; also record the size for later pro- rata price adjustment based on the proportion of the available size to reference size.

4. Where a product was priced on multiple-unit basis (e.g. 2 for $5.00), then calculate the price on a per-item basis (e.g. $2.50 each)

5. Where a special price for store club members was prominently displayed near the shelved product, then record the lower members' price. Ignore promotional discounts shown only on the cash register receipt.

6. Where a special coupon price applies and a coupon is available nearby the shelved product, then record the after-coupon, discount price.

7. General discounts, promotional points, and cash back schemes based on the overall value of purchases were ignored.

These rules worked well to avoid ambiguities in collecting price information, and in practice were not difficult to apply. In no instance was I left with the impression that the price information was significantly distorted by application of the rules.

The time required to obtain prices in a supermarket for the initial 41 products was approximately one hour. (It was not until after all stores were surveyed that the sample was reduced to 32 products.) Concentrating on particular brand name products of a specific reference size, greatly simplified the process. Judgement was required only when the reference size package was not available and an alternative closest size had to be chosen. Diligence was necessary, however, to make sure that prices (and sizes) were correctly identified - sometimes more difficult when an item was on sale - and accurately recorded.



Results of the Pilot Study

The comparative prices are shown in the Table 2 (Click for Table). Sale prices are noted with an asterisk. Prices that were pro rata adjusted to conform to the reference size are in italics. Canadian prices were converted to United States currency at the prevailing exchange rate of $1 CDN equating to $0.67 U.S.

At least with regard to the sampled products, exchange rate-adjusted prices in Nova Scotia tend to be significantly lower than in three of the four surveyed communities in the United States. For all products combined, prices in Nova Scotia were only 85% of the average of the regions in the United States. Concord, New Hampshire was closest in pricing, which itself was 87% of the United States' regional average.

The lesson for RV travelers from Nova Scotia, is that they should load up on food and related supermarket products at home before traveling in the United States. Conversely, RVers from the United States should come to Nova Scotia with bare cupboards to take advantage of the lower prices here.

More notable results pertain to the significant diversity of prices between the four surveyed regions of the United States. On average for all of the products, prices in Frankfort were highest by 6.6% (an index of 106.6), Phoenix was 4.8% higher (104.8), and Abilene was 2.3% higher (102.3); but Concord was markedly 13.1% lower (an index of 86.9). The difference in prices between Frankfort and Concord was nearly 23%.

Table 3 (Click for Table) examines the ratios of lowest to highest prices by product, between the United States communities sorted by magnitude of proportional price differences. The differences between the highest and lowest prices for the array of products ranged from 10% (ratio of 1.10) for Carnation Instant Breakfast to 102% (ratio of 2.02) for Comet Cleanser. For all products in the sample, the range between the lowest and highest respective prices averaged 47% (ratio of 1.47). None of the regions had uniformly highest nor lowest prices. Including two ties for lowest price amongst the regions (as shown in Table 2 (Click for Table) ), Phoenix had the lowest price for 2 products, Abilene 6, Frankfort 4, and Concord 21. For the most part, the price differences were not due to products being temporarily on sale in one region versus the other regions. Only 3 of the uniquely lowest prices were products 'on-sale'.

The pilot study suggests that there clear differences in the regional prices of many products sold in supermarkets at least in the United States. The same market dynamics may also exist in Canada, but with only one observation for the country it is difficult to say. The information begs to be analyzed; how are regional prices affected by local income levels, employment level, population size, supermarket chain affiliation, etc.? Why is there a greater diversity of pricing for some kinds of products than for others? What is the role of the product manufacturers in these regards? Do these factors act the same between countries? Indeed, the number of possible questions swamps the limited information from the 4, or 5 including Nova Scotia, communities surveyed in the pilot study.

A methodological result of the pilot study is that collecting comparable price information for brand products sold in supermarkets is feasible, and in fact relatively easy. It would be possible to carry out a more extensive study by involving a number of people surveying local supermarket prices in many communities.