Analytical Studies: Methods and References
Purchasing Power Parities for Consumption and Household Income Across the Canadian Provinces and Territories

Acknowledgments

This project would not have been possible without the support of the Consumer Prices Division at Statistics Canada. The authors would like to thank Michael Henderson and Chris Li for assistance with the early stages of this project, and Kyle de March, Elspeth Hazell, Walid Ezzaouali and Clément Yélou for feedback on the paper. This project was also supported by the National Economic Accounts Division, but particularly Nazrul Kazi, who aided at various stages of the project and provided helpful feedback on the paper. The authors would also like to thank Eric Figueroa and his team from the Regional Directorate of the Bureau of Economic Analysis for their thoughtful and detailed comments.

Abstract

Regional economic disparities have been increasing in many high-income countries, leading to increased interest in the measurement of economic differences between regions. This paper develops regional purchasing power parities (PPPs) for Canada, which can be used to adjust for price levels while making interregional comparisons. Then, it uses these PPPs to create two price-adjusted measures of household disposable income per capita, analyzing differences in real income across the provinces and territories. Differences between the non-price-adjusted measures and the price-adjusted measures suggest that price adjustment is necessary to create an accurate picture of purchasing power across regions. For example, while Alberta has some of the highest price levels in the country, the province ranks highly in terms of PPP-adjusted household per capita, indicating that nominal income is high enough to compensate for high price levels. In contrast, British Columbia and Ontario fall in the rankings after PPP adjustment, indicating that nominal income is not high enough to compensate for high price levels in these provinces.

1 Introduction

After experiencing decreasing regional economic inequality during the 20th century, many high-income countries are now experiencing a period of increasing regional inequality, sometimes referred to as “the great inflection” (Storper, 2018). Inequality between regions can be measured in several ways, and Canadian research focuses on the increasing differences between provinces in terms of economic development, employment precariousness and income per capita (Breau et al., 2020; Brown & Macdonald, 2015; Marchand et al., 2020). Across high-income countries, the great inflection has been linked to increased trade flows (Rodríguez‐Pose, 2012), and, in the Canadian context, it coincides with Canada’s “resource boom” (Brown & Macdonald, 2015). Researchers speculate that this increasing regional inequality is leading many people to feel “left behind” relative to people from other regions, resulting in increased discontentment and political polarization (Marchand et al., 2020; Storper, 2018). This paper provides new evidence on economic differences between the Canadian provinces and territories by developing regional purchasing power parities (PPPs) for Canada, which can be used to adjust for price levels while making interregional comparisons. Then, it uses these PPPs to make price adjustments to two measures of household disposable income per capita, analyzing differences in real income across the provinces and territories.

These results have significance to researchers and policy makers seeking to make interregional comparisons across Canada. Research to date has often assumed that a dollar purchases the same quantity of goods and services in one province or territory as in another.^Note  While this may be roughly true for provinces in proximity, such as the Atlantic provinces or Saskatchewan and Manitoba, it is unlikely to be true for provinces that are farther apart or when making comparisons with the territories. Therefore, the methodology developed in this paper can be used to produce higher-quality comparisons, particularly when examining areas that are geographically dispersed.

More specifically, the paper begins by estimating a PPP for the consumption component of gross domestic product (GDP) based on data from the Consumer Price Index (CPI), the Labour Force Survey (LFS) and the Canadian System of Macroeconomic Accounts (CSMA). It uses the same types of data sources as the regional price parities (RPPs) calculated by the Bureau of Economic Analysis (BEA, 2022) and builds on the inter-city spatial price indexes published by Statistics Canada (Statistics Canada, 2020b). However, rather than using a regression approach (as the RPPs do), the methodology used here is very similar to the multilateral PPP methodology employed by the Organisation for Economic Co-operation and Development (OECD, 2007). The OECD methodology calculates Fisher PPPs between each pair of provinces or territories for each reported commodity grouping. It then applies the Elteto–Koves–Szulc (EKS) formula to enforce transitivity across the provinces and territories. Unlike some other indexes (e.g., the inter-city price index), the calculated PPPs reflect public consumption (e.g., consumption of publicly provided education), as well as private consumption. The paper focuses on results for the year 2021, though results for 2019 are discussed in the appendix. In addition to the overall PPP, the paper provides regional PPPs for several categories of consumption (e.g., shelter).

Across provinces, the results show that the highest price levels occur in British Columbia, Ontario and Alberta. Despite high price levels, Alberta maintains a relatively high PPP-adjusted household disposable income per capita because of high nominal income in the province. British Columbia and Ontario, by contrast, do not have high enough nominal household disposable income per capita to counteract high price levels, putting their PPP-adjusted household disposable income in the bottom half of provinces. Within the territories, price levels are highest in the Northwest Territories and Nunavut, but the Northwest Territories has relatively high PPP-adjusted household disposable income, while Nunavut has the lowest of any region. The results change further when accounting for social transfers in kind—the value of the consumption of publicly provided goods, such as education and health care. When adjusting for social transfers in kind in addition to price levels, Nunavut fares relatively better, and Ontario and British Columbia have the lowest adjusted household income.

The remainder of the paper is structured as follows. Section 2 presents a discussion of PPP methodologies and explains the choice to follow the OECD approach. Section 3 describes the results. Section 4 concludes. The appendix contains a validation against inter-city price indexes produced by Statistics Canada (Statistics Canada, 2020b). The appendix uses 2019 PPPs because the inter-city price indexes are only produced through 2019.

2 Methodology to calculate regional purchasing power parities

Two common approaches exist for calculating PPPs: the country-product-dummy (CPD) approach, which is regression-based, and an approach based on index number calculations.

The CPD approach first uses regressions to predict representative prices. Then, based on price relatives (e.g., price ratios) formed from predicted values in the first stage, the CPD approach estimates the aggregate levels through a second regression. Separate regressions are employed for each sub-aggregate of interest and for the overall level. The first stage of the CPD approach can also be used in combination with another approach. To calculate regional PPPs for the United States, the BEA uses the CPD approach to prepare price inputs, then uses these inputs in the Geary multilateral index formula (BEA, 2022).

The index number approach uses an index number formula to aggregate price relatives to the sub-aggregates of interest and the overall level. This is the approach used by the OECD in its multilateral PPP program (OECD, 2007), by the World Bank in the International Comparison Program (World Bank, 2021), and by Statistics Canada in its Canada–U.S. bilateral PPP calculations (Kazi & Barber-Dueck, 2019) and its inter-city spatial price indexes (Statistics Canada, 2020b). Several index number formulas can be employed, including Laspeyres, Paasche and Fisher spatial indexes. These provide bilateral PPPs, which are not transitive. Multilateral PPP programs, therefore, apply an additional step to enforce transitivity. In the case of the OECD program, the EKS formula is employed. In the case of the inter-city spatial price indexes, transitivity is enforced by linking cities rather than by creating binary pairs that are subsequently adjusted.

In practice, both approaches have been found to produce similar results (OECD, 2007). This study proceeds with the index number approach, which allows for validation against results from Statistics Canada’s inter-city price index program. The index number approach requires data on prices and expenditures. Therefore, this section first describes the measurement of price relatives, and then the measurement of expenditures and the aggregation of the data into regional PPPs.

2.1 Measurement of price relatives

2.1.1 Household consumption

Prices for household consumption are collected from Statistics Canada’s CPI program (Statistics Canada, 2023a). Price collectors for the CPI program are asked to collect prices for representative products based on technical specifications. The data used are from the year 2021 and contain about 1.7 million data points that cover over 600 representative products. Ideally, these products would align exactly across geographies. However, regional differences arise for several reasons. For example, if price collectors across regions are asked to collect prices of women’s winter coats, they collect prices based on the stores and brands that are available in their location. This creates a situation without a one-to-one correspondence between the products that are collected under the representative product headings.

To interpret the regional PPPs, it is assumed that price collectors accurately represent the mix of products as they are sold by different stores within each region. So, if price collectors select winter coats in Vancouver that are lighter but more water resistant than winter coats in Winnipeg, it is assumed that this represents the true state of these markets and that an aggregation of the prices within each market can be used to form a price relative. In other words, because consumers in Vancouver and Winnipeg are buying the best-suited winter coats for their respective locations, the fact that the products differ is not a significant limitation to interpreting the regional PPPs, even if the differences in coat characteristics cause the prices to differ. As noted by the OECD, consumption patterns can vary between regions because of taste, culture, climate and price (OECD, 2007), and this causes expected variation in the prices of products. Therefore, it is best not to judge how expensive or inexpensive a region is based on individual products; rather, aggregate PPPs should be examined because they incorporate consumption patterns via expenditure data and provide a more accurate view of how expensive or inexpensive a region is, based on typical purchases in that region.

The CPI price collectors compile data for representative products for several different outlets within different sub-regions of each province or territory, so the procedure to calculate the price relatives seeks to use this variation to generate distributions of price relatives from which the values for the PPP can be drawn. This is analogous to bootstrapping a price relative distribution.

The first step to calculate the price relative distributions is to take averages for each representative product at the outlet level, for the entire year. Retail chains like Walmart have different outlet identifiers for different geographic areas within a province or territory, so this produces multiple observations at the outlet level. Then, cycling through pairs of regions, Kronecker products of the ratios of these outlet-based average prices are taken, again at the representative product level. For example, if the sampled prices for a representative product in Manitoba are $4 and $7 and the sampled prices for the same representative product in Nunavut are $6 and $8, the Manitoba-to-Nunavut price relative distribution would be 4/6, 4/8, 7/6, 7/8.

This process produces a distribution of ratios for each region pair at the representative product level. Within each region pair, the values are bottom and top coded at the 5th and 95th percentiles. The top and bottom coding accounts for potential aberrant observations within the dataset. By bottom and top coding, the range of the price ratio distributions is controlled across all representative products in a consistent manner, without greatly influencing the central tendency of the overall series.

After bottom and top coding, the representative products are grouped together based on the finest commodity code grouping for which weights are available, where commodity codes are based on Canada’s modified version of the international Harmonized Commodity Description and Coding System (Statistics Canada, 2021a). In many cases, the grouping occurs at the eight-digit commodity code level. For example, there are expenditure weights for the eight-digit commodity codes associated with various types of dairy products, so there are specific groups for products such as fresh milk, cheese and butter. In other cases, a larger grouping is needed. For example, all representative products for various types of fish are grouped under the six-digit commodity code for fish.

Within these commodity code groupings, all the price relatives for the corresponding representative products are used to calculate the median and the mean for each grouping and each region pair.

2.1.2 Government consumption

Government-provided services are paid for through taxation, so no market prices exist for them. Consequently, PPP estimation often follows a methodology that uses input prices as a proxy measure for output prices. Here, average hourly wages from the LFS (Statistics Canada, 2020a) are used as an estimate of the price of government services.

The LFS data for average hourly wages are classified into occupations corresponding to the National Occupational Classification (NOC) 2021 (Statistics Canada, 2023b). To facilitate their use in estimating PPPs, a concordance between NOC 2021 four-digit occupations and four-digit expenditure categories from the Classification of the Functions of Government (COFOG) (Statistics Canada, 2022) is created. The concordance is then used to allocate LFS wage information to the COFOG expenditure categories to calculate price relatives. The concordance between the two classification schemas is not unique.

For example, if the description of a NOC category is senior managers - health, education, social and community services and membership organizations, it can be associated with multiple COFOG categories—in this case, health, education and social protection. Because the goal is to estimate the average price of labour for the associated government functions, the concordance allocates the NOC category to all three COFOG categories.

This does not constitute a problematic double counting because the price relative of government services is the desired measure, rather than a level. There is, however, a measurement problem: managers in health may earn more, on average, than managers in education or social protection.

2.1.3 Housing and rent

The CPI methodology is designed to produce estimates that are comparable over time, rather than by region, so the methodology used for the territories sometimes differs from that used for the provinces if it allows for more consistent temporal indexes. This is the case for CPI rent data. Since PPPs require data that are comparable between regions, rent data are obtained from the 2021 Census of Population (Statistics Canada, 2021b) instead of from CPI data.

The CPI uses the replacement cost of housing (from insurers) as a component because it is correlated with the amount a homeowner would have to spend to maintain the value of their home. However, CPI microdata do not include replacement values for the territories. Therefore, this component was replaced with the average self-reported home value from the 2021 Census.

For electricity, natural gas, water and fuel oil, there are few providers within regions, so the cross-product approach resulted in extreme values, particularly for natural gas and electricity. Because of these extreme values and because natural gas is imputed for the territories, natural gas and electricity data are not used. For water and fuel oil, means and medians are used instead of the cross-product approach.

2.2 Measurement of expenditure values

Consumption values are based on those reported in the CSMA (Statistics Canada, 2018). Consumption data sources are split between those sourced from household consumption variables and those sourced from government consumption. The former begin with detailed data on consumption by households compiled for the provincial and territorial economic accounts. The latter employ consumption estimates compiled to track government consumption based on function (COFOG).

2.2.1 Household consumption

Expenditure weights are based on expenditures reported in each region at various commodity code levels in the CSMA. These are not the same as the CPI basket weights used for the CPI, though there is overlap in the data sources used to derive the two sets of weights.

For a handful of representative products, microdata are not available in certain regions. Therefore, a benchmarking strategy is employed, weighting up the weights at lower levels so that totals at the 1-digit level are enforced (Table 1). For example, in Newfoundland and Labrador, 17.2% of household consumption is represented in the one-digit commodity code for food and non-alcoholic beverages. Suppose that 0.3% of household consumption is spent on food items that are missing from the CPI microdata. If the weights attached to CPI microdata are totalled, they will equal only 16.9% (17.2% minus 0.3%). For the benchmarking strategy, weights for all items listed under food and non-alcoholic beverages would be multiplied by 17.2/16.9. Then, when the weights are totalled for that category, they will add up to 17.2%.

Some components are missing in several regions. Shelter is missing multiple components because the CPI microdata are missing tenant expenses, and natural gas and electricity are dropped. Therefore, the remaining components (rent, home values, homeowner maintenance and mortgage costs) are combined, and a composite weight is used for the aggregated component. Multiple regions are missing alcohol purchased in stores, so the weight for that component is distributed across the other components in tobacco, cannabis and alcohol. Note that these components are missing from the CPI microdata used for this paper but are included in CPI figures published by Statistics Canada. This is because the CPI figures published online use supplementary data sources in addition to direct price collection, allowing for indexes that reflect all components.

Once the weights are adjusted, they are multiplied by the household final consumption expenditure for each denominator province or territory, converting them into dollar values (in millions).

2.2.2 Government consumption

Government consumption is based on aggregates compiled by functional category. These COFOG data are employed at the four-digit level as the base level for aggregation. Aggregates are reported at the two-digit level (Table 2). This leads to variation at finer levels but produces estimates across the provinces and territories that correspond to commonly examined aggregates such as education; health; public order and safety; and other public services, including public administration.

2.3 Calculation of the provincial and territorial purchasing power parities

Aggregation for the provincial and territorial PPPs follows the procedures used by the OECD in its multilateral PPP program. The OECD procedure employs the EKS method to produce transitive multilateral PPPs. Chapter 7 of the Eurostat-OECD Methodological Manual on Purchasing Power Parities, “Calculation and aggregation of PPPs,” provides a detailed description of the aggregation procedure (OECD, 2007). An annotated description is provided below.

The process for aggregating and for generating transitivity follows two steps. First, the Fisher PPP for each pair of provinces and territories for each desired level of commodity classification is calculated. In the second step, the EKS method is used to enforce transitivity on the pairwise Fisher PPPs.

2.3.1 Fisher purchasing power parities

The Fisher spatial index for a given commodity aggregation is defined as the geometric mean of the spatial Laspeyres index and the spatial Paasche index. These are analogous to the inter-temporal indexes, but rather than using period $t$  and $t-1$ , the indexes use province or territory $i$  and province or territory $j$ .

The Laspeyres PPP index holds values in province or territory $j$  constant and allows prices to change. Its formula is the following:

(1)

$$PP{P}_{i,j}^L=\frac{{\displaystyle \sum _m^{}{P}_m^i{V}_m^j}}{{\displaystyle \sum _m^{}{P}_m^j{V}_m^j}}$$

where $P_m^i$  and $P_m^j$  are the prices of commodity $m$  in province or territory $i$  and province or territory $j$ , respectively, and $V_m^j$  is the volume of commodity $m$  in province or territory $j$ .  The product $P_m^j{V}_m^j$  is the nominal expenditure on commodity $m$  in province or territory $j$ .  Because the data used to estimate the PPP are reported as prices and nominal expenditures, it is useful to rewrite equation (1) as follows:

(2)

$$PP{P}_{i,j}^L=\frac{{\displaystyle \sum _m^{}(\frac{P_m^i}{P_m^j})P_m^j{V}_m^j}}{{\displaystyle \sum _m^{}{P}_m^j{V}_m^j}}$$

Equation (2) illustrates that the Laspeyres PPP is the weighted sum for $m$  price relatives between province or territory $i$  and province or territory $j$ , where the weights come from province or territory $j$ .

The Paasche PPP price index holds values in province or territory $i$  constant and allows prices to change. Its formula is the following:

(3)

$$PP{P}_{i,j}^P=\frac{{\displaystyle \sum _m^{}{P}_m^j{V}_m^i}}{{\displaystyle \sum _m^{}{P}_m^j{V}_m^i}}$$

where $P_m^i$  and $P_m^j$  are the prices of commodity $m$  in province or territory $i$  and province or territory $j$ , respectively, and $V_m^i$  is the volume of commodity $m$  in province or territory $i$ . The product $P_m^i{V}_m^i$  is the nominal expenditure on commodity $m$ in province or territory $i$ . Like the Laspeyres index, the Paasche PPP can be rewritten so that its formula is based on prices and nominal expenditure values as follows:

(4)

$$PP{P}_{i,j}^P=\frac{{\displaystyle \sum _m^{}{P}_m^j{V}_m^i}}{{\displaystyle \sum _m^{}(\frac{P_m^j}{P_m^i})P_m^j{V}_m^i}}$$

The Fisher PPP is defined as the geometric mean of the Laspeyres PPP and Paasche PPP:

(5)

$$PP{P}_{i,j}^F={(PP{P}_{i,j}^L\times PP{P}_{i,j}^P)}^{1/2}$$

2.3.2 Enforcing transitivity

To enforce transitivity across PPP estimates, the EKS transformation is applied to the binary Fisher PPPs. The EKS formula uses the geometric mean of the direct and indirect PPPs for each pair of provinces and territories as follows:

(6)

$$PP{P}_{i,j}^{EKS}={({(PP{P}_{i,j}^F)}^2\times {\displaystyle \prod _{k\ne i,j}^N(\frac{PP{P}_{i,k}^F}{PP{P}_{j,k}^F})}{)}^{1/N}}^{MathType@MTEF@5@5@+=\; feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn\; hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr\; 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9\; vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x\; fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamiuaiaadc\; facaWGbbWaa0baaSqaaiaadMgacaGGSaGaamOAaaqaaiaadweacaWG\; lbGaam4uaaaakiabg2da9iaacIcacaGGOaGaamiuaiaadcfacaWGbb\; Waa0baaSqaaiaadMgacaGGSaGaamOAaaqaaiaadAeaaaGccaGGPaWa\; aWbaaSqabeaacaaIYaaaaOaeaaaaaaaaa8qacqGHxdaTdaqeWbqaai\; aacIcadaWcaaqaa8aacaWGqbGaamiuaiaadgeadaqhaaWcbaGaamyA\; aiaacYcacaWGRbaabaGaamOraaaaaOWdbeaapaGaamiuaiaadcfaca\; WGbbWaa0baaSqaaiaadQgacaGGSaGaam4AaaqaaiaadAeaaaaaaOGa\; aiykaaWcpeqaaiaadUgacqGHGjsUcaWGPbGaaiilaiaadQgaaeaaca\; WGobaaniabg+GivdGcpaGaaiykamaaCaaaleqabaGaaGymaiaac+ca\; caWGobaaaaaa@62F4@}$$

where the direct PPP is the Fisher PPP from equation (5) and the indirect PPPs are ratios of the Fisher PPPs, such that for each province and territory pair an estimate of their PPP is inferred indirectly from their PPPs with third-party provinces and territories.

2.4 Level of reported aggregation

The PPP values are calculated for different levels of detail to allow for a greater understanding of price relative structures and the analysis of different aggregates in the national accounts. These aggregations differ from the one-digit consumption commodity codes and COFOG codes used to create the weights for aggregation. Instead, detail is divided partly based on the way products are provisioned and partly based on the broad types of products examined (Figure 1).

The overall PPP is divided into public and private components. This division corresponds to whether the good or service in question is purchased by households (private) or governments (public). Private consumption prices are based on the data collected for the CPI and are measured directly. Public services are priced based on the wages of employees in education, health care and the provision of public services. The private component is further disaggregated among goods (durables, semi-durables, non-durables) and services (shelter, other) to highlight price relative differences among broad groups of products with similar types of life-use characteristics.

Description for Figure 1

The figure shows how the overall purchasing power parity (PPP) index is broken down into smaller categories. “Overall PPP” breaks down into “Private PPP” and “Public PPP.” Private PPP breaks down further into “Goods” and “Services.” “Goods” breaks down further into “Durables,” “Semi-durables” and “Non-durables,” and “Services” breaks down further into “Shelter” and “Other services.”

3 Results

3.1 Provincial purchasing power parities for consumption

Table 3 contains PPPs based on mean prices, while Table 4 presents PPPs based on median prices. Both tables use Ontario as the numeraire province. This means that a PPP of less than 1 indicates that a region is less expensive than Ontario, while a PPP of more than 1 indicates that a region is more expensive than Ontario. For example, the PPP exchange rate from Ontario to New Brunswick is 0.87 (Table 3). In other words, a dollar spent in Ontario buys the equivalent amount of goods and services as 87 cents spent in New Brunswick. Overall, Ontario, British Columbia and the territories (Yukon, the Northwest Territories and Nunavut) are the most expensive regions.

Table 3 also presents PPPs for specific categories of consumption. When New Brunswick and Ontario are used as an example, the table shows that durable goods (e.g., furniture, appliances) and non-durable goods (e.g., food, medicine) are more expensive in New Brunswick, but services and semi-durable goods (e.g., clothing, reading material) are less expensive. Public consumption (e.g., public health, public education) is less costly in New Brunswick relative to Ontario, reflecting lower salaries and wages for workers in those industries. Overall, private consumption is the most expensive in Ontario, Alberta, British Columbia and Nunavut. Public consumption is the most expensive in Ontario, Alberta, and the territories. Results are similar when calculating the PPPs based on median prices, rather than mean prices (Table 4). The range in the private PPP is somewhat narrower than the range of PPPs calculated by the BEA for the United States, which varies from 0.86 to 1.13 in 2021 (BEA, 2024).

This paper produces novel estimates of regional PPPs for Canada, so there is no directly comparable research to validate these estimates against. The closest estimates are Statistics Canada’s inter-city price indexes, but these are only produced through 2019 (Statistics Canada, 2020b). The appendix contains the results of validation using data from the year 2019. The PPP estimates produced in this paper are generally consistent with results from the inter-city price indexes, but there are some differences in terms of the shelter component, indicating that shelter should be an area of focus for future work.

3.2 Real income comparisons

The PPPs estimated here correspond to the consumption component of GDP. They are therefore appropriate for making price-adjusted comparisons of real income across the country. This paper uses two types of income to make comparisons between regions—household disposable income and adjusted household disposable income—and both are based on publicly available data from Statistics Canada (Statistics Canada, 2024c).

The first type of income examined is household disposable income per capita. Disposable income is preferable to household total income because it accounts for taxes and transfers. Household disposable income per capita represents the maximum expenditure level that households can, on average, undertake within a province or territory without having to sell assets or take on debt for current consumption. It is therefore a natural measure for making level-based comparisons of material well-being across jurisdictions. Household disposable income per capita is price-adjusted using the private PPP.

The second type of income examined is sometimes referred to as adjusted household disposable income per capita (Statistics Canada, 2024c). This measure begins with household disposable income, to which an imputation for social transfers in kind is added. Social transfers in kind are designed to adjust for goods and services purchased for households by other sectors in the economy. In Canada, this is predominantly health care and education services purchased for households by governments. This second type of income is price-adjusted using the overall PPP.

Beginning with household disposable income per capita, Yukon, the Northwest Territories and Alberta have the highest incomes per capita without price adjustment, while Nova Scotia, Manitoba and Nunavut have the lowest incomes per capita (Chart 1). PPP-adjusted household disposable income per capita is highest in Yukon, the Northwest Territories, and Newfoundland and Labrador, with Alberta falling to fourth place. It is lowest in Quebec, Manitoba and Nunavut. Ontario and British Columbia—two regions with high price levels—have large changes in rankings, being 4th and 5th, respectively, in terms of nominal household disposable income per capita and 8th and 9th, respectively, in terms of PPP-adjusted household disposable income per capita. Nunavut is heavily affected in absolute terms. In terms of nominal household disposable income per capita, the territory has similar average household disposable income per capita as several provinces, with average values that are $2,283 lower than in Manitoba. In terms of PPP-adjusted household disposable income per capita, Nunavut is the lowest region by far, with average values that are $7,338 below those in Manitoba. It is important to consider that the territories represent locations where data are harder to compile and where economic structures differ from the southern portion of Canada, as do demographic profiles. As a result, results should be viewed with caution for these locations because comparability is less direct.

Description and data table for Chart 1

The chart shows household disposable income per capita, both nominal and purchasing power parity (PPP)-adjusted, in dollar terms. The values are presented in a vertical double bar chart, where the double bars are grouped by province or territory. Within each province or territory, the two bars show the two measures of income: nominal household disposable income per capita and PPP-adjusted household disposable income per capita.

Data table for chart 1
Table summary
This table displays the results of Data table for chart 1 Household disposable income per capita, PPP-adjusted and Household disposable income per capita, nominal, calculated using dollars units of measure (appearing as column headers).

	Household disposable income per capita, PPP-adjusted	Household disposable income per capita, nominal
	dollars
Notes: Ontario is used as the base. Price adjustment is performed using the private purchasing power parity (PPP). Source: Statistics Canada, authors’ calculations.
Yukon	52,840	50,441
Northwest Territories	52,592	51,100
Newfoundland and Labrador	43,373	37,626
Alberta	42,462	42,184
Saskatchewan	40,995	36,459
Prince Edward Island	40,310	35,162
New Brunswick	39,104	33,886
Ontario	37,653	37,653
British Columbia	36,479	41,113
Nova Scotia	36,448	33,836
Quebec	36,247	34,284
Manitoba	35,949	32,967
Nunavut	28,611	30,684

Continuing with household disposable income per capita plus the imputation for social transfers in kind, some results are considerably different (Chart 2). The Northwest Territories and Yukon still have the highest incomes without price adjustment, but Nunavut is now third, rather than Alberta. British Columbia and Ontario are again heavily affected by price adjustment, with British Columbia dropping from 5th to 12th and Ontario dropping from 8th to 13th. As with the first measure of income, price adjustment reduces the average income in Nunavut substantially; however, in this case, the PPP-adjusted measure is in line with the provinces. More specifically, Nunavut and Saskatchewan have similar PPP-adjusted household disposable income, adjusted for social transfers in kind, per capita.

Comparing the two measures of household income (with and without social transfers in kind), the addition of social transfers in kind—which represent the consumption of public-provided services such as health care—changes the ranking of the regions. Nunavut fares better when social transfers in kind are considered, and Ontario and British Columbia fare worse.

Description and data table for Chart 2

The chart shows household disposable income per capita plus imputation for social transfers in kind, both nominal and purchasing power parity (PPP)-adjusted, in dollar terms. The values are presented in a vertical double bar chart, where the double bars are grouped by province or territory. Within each province or territory, the two bars show the two measures of income: nominal household disposable income per capita plus imputation for social transfers in kind and PPP-adjusted household disposable income per capita plus imputation for social transfers in kind.

Data table for chart 2
Table summary
This table displays the results of Data table for chart 2 Household disposable income per capita plus imputation for social transfers in kind, PPP-adjusted and Household disposable income per capita plus imputation for social transfers in kind, nominal, calculated using dollars units of measure (appearing as column headers).

	Household disposable income per capita plus imputation for social transfers in kind, PPP-adjusted	Household disposable income per capita plus imputation for social transfers in kind, nominal
	dollars
Notes: Ontario is used as the base. Price adjustment is performed using the overall purchasing power parity (PPP). Source: Statistics Canada, authors’ calculations.
Northwest Territories	75,007	82,057
Yukon	66,245	68,737
Newfoundland and Labrador	54,600	48,496
Nunavut	53,503	63,139
Saskatchewan	53,365	48,291
Alberta	52,307	52,208
Prince Edward Island	51,225	45,025
Manitoba	50,624	45,546
New Brunswick	50,393	43,590
Nova Scotia	50,225	45,139
Quebec	48,580	44,941
Ontario	47,906	47,906
British Columbia	47,492	50,875

4 Conclusion

This purpose of this paper is to provide new evidence on the economic differences between the Canadian provinces and territories by developing regional PPPs for Canada. The paper follows the multilateral framework proposed by the OECD (OECD, 2007) and then validates the results by making comparisons with Statistics Canada’s inter-city price indexes (Statistics Canada, 2020b). Then, the PPPs are used to create price-adjusted measures of household disposable income per capita. This adjustment changes the rankings of some regions in terms of income, suggesting that price adjustment is an important part of measuring economic differences between regions. Adjusting income for social transfers in kind—a measure of the consumption of publicly provided services—also changes the ranking of regions in terms of household income. These measures shed light on differences in purchasing power between regions—analyzing the amount of money required to purchase an equivalent amount of goods and services in each province and territory.

The paper revealed some issues with the quality of the shelter data, and some inconsistencies between the calculated PPPs and the inter-city price indexes. Future work could focus on finding alternate sources of shelter data that would allow for higher quality estimates. Another limitation of this study is that it focuses only on the cost of consumption, both private and public. Without the inclusion of the cost of private production, the PPP has limited usefulness for businesses that may seek to know which regions are cheaper to produce in. Another possible extension could involve measuring prices for gross fixed capital formation, which would allow the creation of a PPP that could be used to adjust gross domestic income. Lastly, further research may explore the creation of sub-provincial PPPs between selected cities, towns and metropolitan areas where there are sufficient price data. This would also allow more direct comparison with inter-city price indexes.

5 Appendix: Results from 2019 and comparison with inter-city price indexes

Statistics Canada’s CPI data are employed to estimate a set of price indexes that can be used to make inter-city price level comparisons (Statistics Canada, 2020b). Like the PPPs produced in this paper, the inter-city price indexes are calculated using an index number approach and use CPI data. Since the inter-city price indexes focus on private consumption, they are compared with the private PPPs as a validation exercise.

The inter-city price indexes are produced only through 2019, so PPPs based on 2019 must be produced for comparison (Table A.1, Table A.2). PPPs based on 2019 data are like those based on 2021 data, with the largest difference being that British Columbia is relatively less expensive in 2019. Note that rent and house price values are still based on 2021 because they are collected from the census, which occurs only every five years.

While inter-city price indexes are the most comparable indexes to the PPPs calculated here, there is a major difference in that the inter-city price indexes include only the most populated cities in each region. This means that differences between the indexes are expected. This comparison attempts to determine whether the PPPs follow the same pattern as the inter-city price indexes. For Alberta and Ontario, the inter-city price indexes provide estimates for Edmonton and Calgary or Ottawa and Toronto, respectively. To create a single value, the estimates for Edmonton and Calgary are averaged to produce a value for Alberta, while a weighted average for Ottawa and Toronto is employed, where Toronto is given a greater weight. The resulting single-province and single-territory values are then rescaled so that Ontario is the numeraire.

An inter-city price index is not available for Nunavut. Outside Nunavut, when compared with the mean- and median-based PPPs, the inter-city price indexes have much lower variation, with most values closer to 1 than the equivalent PPP (Table A.3). This suggests that there is lower variation between major cities (used for the inter-city price indexes) then there is between entire regions. Overall, the measures follow roughly the same pattern, though the PPP identifies British Columbia as the most expensive province, while the inter-city price indexes identify Ontario as the most expensive province. The difference may occur because most of Ontario’s residents reside in the cities used for the inter-city price index (Ottawa and Toronto), while the inter-city price index for British Columbia includes only Vancouver and does not capture most residents of British Columbia. The standing of Saskatchewan also changes. With the PPP, it is comparable to Manitoba, but with the inter-city price indexes, it is identified as more expensive than Manitoba. The inter-city price index for Manitoba captures most of the population of the province, while the inter-city price index for Saskatchewan does not.

In Yukon and the Northwest Territories, the CPI data used for the PPP are based on data from the capitals, so in theory, the PPP and the inter-city price indexes should align closely for these areas. Despite this, for the Northwest Territories, the inter-city price index is considerably higher than the PPP.

One of the larger differences between the data used for the PPPs and the data used for the inter-city price indexes is in terms of shelter because the PPPs use dwelling values and rent values from the 2021 Census of Population. By comparing price relatives for dwelling values and rent values used for the PPP with the inter-city price indexes, it is apparent that the differences between the PPPs are driven by dwelling values. Dwelling values reflect the price that a home would sell for, while the replacement values in the CPI are based on the cost of replacing a home given current building prices, e.g., building materials. While the use of dwelling values allowed consistent values across provinces and territories (as replacement values for the territories are not included in the CPI microdata used for this paper), a drawback is that the PPPs are less consistent with other measures such as the inter-city price indexes. British Columbia is ranked as more expensive than Ontario using the PPP, and the Northwest Territories is ranked as less expensive. Exploring alternatives for shelter data is a priority for future work; however, the difficulty lies in the fact that few data sources cover the provinces and territories in a consistent way.

References

Ardeni, P.G. (1989). Does the Law of One Price really hold for commodity prices? American Journal of Agricultural Economics, 71(3).

BEA (Bureau of Economic Analysis). (2022). Methodology for Regional Price Parities, Real Personal Consumption Expenditures, and Real Personal Income April 2022.

BEA (Bureau of Economic Analysis). (2024). Interactive Data Application: Regional GDP and Personal Income.

Breau, S., Burkhart, N., Shin, M.E., Marchand, Y., & Sauer, J. (2020). Is it time to start worrying more about growing regional inequalities in Canada? Canadian Geographer / Le Géographe canadien, 65(4): 542-50.

Brown, W.M., & Macdonald, R. (2015). Provincial Convergence and Divergence in Canada, 1926 to 2011. Economic Analysis Research Paper Series, 11F0027M, no. 96.

Ceglowski, J. (2004). The law of one price: Intranational evidence for Canada. Canadian Journal of Economics, 36(2).

Kazi, N., & Barber-Dueck, C. (2019). Purchasing Power Parities between the United States and Canada: Concepts and methods (Income and Expenditure Accounts Technical Series).

Marchand, Y., Dubé, J., & Breau, S. (2020). Exploring the causes and consequences of regional income inequality in Canada. Economic Geography, 96(2): 83-107.

OECD (Organisation for Economic Co-operation and Development). (2007). Eurostat-OECD Methodological Manual on Purchasing Power Parities.

Rodríguez‐Pose, A. (2012). Trade and regional inequality. Economic Geography, 88(2): 109-136.

Statistics Canada. (2018). User Guide: Canadian System of Macroeconomic Accounts.

Statistics Canada. (2020a). Guide to the Labour Force Survey.

Statistics Canada. (2020b). Inter-city indexes of price differentials, of consumer goods and services 2020.

Statistics Canada. (2021a). Canadian International Merchandise Trade Web Application: Commodity Classification.

Statistics Canada. (2021b). Guide to the Census of Population, 2021.

Statistics Canada (2022). Canadian Classification of the Functions of Government (CCOFOG) Methodology.

Statistics Canada. (2023a). The Canadian Consumer Price Index Reference Paper.

Statistics Canada. (2023b). National Occupational Classification (NOC) 2021 Version 1.0.

Statistics Canada. (2024a). Consumer Price Index basket contents organized according to goods and services.

Statistics Canada. (2024b). Table 10-10-0005-01 Canadian Classification of Functions of Government (CCOFOG) by consolidated government component (x 1,000,000) [Data table].

Statistics Canada. (2024c). Table 36-10-0612-01 Adjusted household disposable income, Canada, provinces and territories, annual (x 1,000,000) [Data table].

Storper, M. (2018). Separate worlds? Explaining the current wave of regional economic polarization. Journal of Economic Geography, 18(2): 247-270.

World Bank. (2021). A Guide to the Integration of Consumer Price Index (CPI) and International Comparison Program (ICP) Production Activities.