The Canadian Consumer Price Index Reference Paper
Chapter 7 – Quality Change and Adjustment

7.1 The Canadian Consumer Price Index (CPI) aims to measure pure price change, that is excluding that portion of price changes that is due to differences in the quality of products purchased by consumers. It achieves this mainly by employing the matched-model method which tracks identical products each month in the same outlets.

7.2 The universe of products bought and sold in the marketplace evolves over time. Updating the sample of items for any given elementary aggregate is inevitable in order to maintain its representativeness. As products in the market change (for example, an old model might be discontinued), observed product offers (PO) may change. It may be important to determine techniques for a valid comparison of prices for the new and the old products. This means that the matched-model framework at times does not hold, and therefore price changes could reflect a mixture of price and quality differences. In order to measure pure price change, quality adjustments are performed.

7.3 There are multiple techniques, both implicit (indirect) and explicit (direct), available to account for quality differences between exiting and entering POs. This chapter will present the different methods used in the CPI.

7.4 It is not always necessary or possible to adjust for quality change when a PO must be replaced in the CPI sample. There are various reasons why adjusting for quality change may not be required and a direct price comparison between entering and exiting POs is the best option. Direct price comparison, an implicit method of quality adjustment, is the simplest approach used in the CPI.

7.5 The CPI employs the direct price comparison method when there is no perceived difference in quality between entering and exiting POs. This method assumes that POs are equivalent in terms of quality.  

7.6 The use of the direct price comparison method for these elementary indices is not likely to lead to any systematic bias in the CPI because the majority of these indices fall under one of the following categories.

7.7 The use of overlap pricing can also eliminate or significantly reduce the need to make explicit quality adjustments. This implicit method allows for the reduction of unexpected disappearances of sampled POs and ensures that new representative products (RP) can be introduced into the sample before the replaced ones disappear from the market or become unrepresentative. The overlap pricing method is most commonly used in conjunction with the profiles method, enabling the collection of a replacement profile before the obsolescence of an existing one. This method can be used for certain high technology goods (video game consoles, televisions, other consumer electronics), where both new and old models are available on the market during the same months before the disappearance of the old models.

7.8 Overall mean imputation is another implicit method used in the CPI to make quality adjustments between the prices of POs entering and exiting the sample. With this method, the price movement applied to entering POs is based on the observed average price movement of all other POs for the same representative product. Overall mean imputation relies on the assumption that the donor POs are comparable to the PO being imputed.

7.9 The link-to-show-no-change method for quality adjustment, another indirect method, involves forcing a price relative of unity (equals no price change) when replacement POs enter the sample. Currently, this practice is being reduced across the CPI because it introduces a degree of undue price stability in the index.Note 

7.10 Quantity adjustment entails accounting for changes in the quantity (e.g. package size, number of tissue ply, etc.) of observed POs. This is another implicit method of quality adjustment because it is assumed that the quality per standardized unit is the same over time.

7.11 Quantity adjustment is the default treatment for nearly all of the POs in the food major aggregate as well as some of the products in the household operations, and personal care supplies and equipment aggregates.

7.12 For the majority of elementary indices, not covered by the implicit methods described above, it is necessary to make explicit quality adjustments when POs enter or exit the sample.

7.13 To make the appropriate quality comparison, Statistics Canada is usually guided by market valuations of the two POs. Where possible, the two POs are compared in terms of the quality features they offer to consumers. A PO is thought to provide a range of features to the consumer which, grouped together, determine the market price.Note  This general framework is the basis for many of the explicit quality adjustment methods described below.

7.14 The CPI relies on the hedonic quality adjustment technique for certain elementary aggregates, notably in the case of high-technology goods or services. Currently, the CPI uses hedonic quality adjustment for computer equipment, software and supplies, Internet access services, rent, used cars and cellular services from telecommunication service retailers that provide Statistics Canada with their administrative data. The hedonic method of quality adjustment is most appropriate for products whose markets are competitive and experience rapid turnover, and where the characteristics of these products change quickly but are readily and consistently observable.

7.15 The hedonic method is applied in the case of forced replacements. This approach assumes that a relationship exists between the price of a PO and its characteristics. Under the hedonic imputation variant, hedonic specifications have to be defined using standard regression techniques and a previous period price is imputed for the replacement product. In period t (when a previously observed PO is no longer available) a regression is used to estimate the unobserved price for the entering PO in period t-1. The estimation of the t-1 price is based on quality differences between the entering and exiting POs, as well as the t-1 price of the exiting PO.

7.16 For the computer equipment, software and supplies index, a hedonic adjustment method is used to impute monthly prices of incoming and outgoing items for desktops and laptops. For monitors and printers, which tend to change less frequently, a matched-model price approach is used.

7.17 The log of monthly prices for laptops and desktops are modelled as a function of a set of explanatory variables using a random forestNote  algorithm. While each product has a separate model, the explanatory variables used are mostly the same. The variation in log price is thus explained by characteristics such as storage space, storage type, total Random Access Memory (RAM), type of RAM, display size (for desktops this variable is set to zero if the desktop in question is not an all-in-one desktop), number of Central Processing Unit (CPU) cores, CPU speed, CPU brand, Graphics Processing Unit (GPU) brand, product weight, the presence of a touch screen (laptops only), item manufacturer, and item retailer. For categorical variables, categories with low counts or observations with unknown values are grouped together as ‘other’. A model is estimated every month using the latest available data.

7.18 For the Internet access services (IAS) elementary index, a pure matched-sample cannot properly account for the rapid technological change and marketing practices that characterize the Internet access industry in Canada. Therefore, a symmetric hedonic method is used to adjust the prices of both entering and exiting Internet access services plans.Note  For the regression model specification, characteristics are transformed as appropriate. Like the method for the computer equipment, software and supplies index, coefficients are estimated at every index calculation period from sets of plans that are used to calculate the index and Internet plan weights.

7.19  For IAS, rather than estimating a single multiple regression, three separate simple regressions are estimated at every index calculation period. In each of these regressions, the dependent variable, log price, is regressed on the intercept term and a single explanatory variable consisting of either log download speed, log upload speed or log usage cap.

7.20 The least squares method is used to solve for the B MathType@MTEF@5@5@+= feaagKart1ev2aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8 qacaWGcbaaaa@36DD@ vector of parameters in the following formulation:

ln p i t =B X i t + ε i MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8 qaciGGSbGaaiOBaiaadchapaWaa0baaSqaa8qacaWGPbaapaqaa8qa caWG0baaaOGaeyypa0JaamOqaiabgkci3kaadIfapaWaa0baaSqaa8 qacaWGPbaapaqaa8qacaWG0baaaOGaey4kaSIaeqyTdu2damaaBaaa leaapeGaamyAaaWdaeqaaaaa@45A8@

where

p i t MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8 qacaWGWbWdamaaDaaaleaapeGaamyAaaWdaeaapeGaamiDaaaaaaa@395E@  is the price of plan i MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamyAaaaa@36E5@ from period t MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamiDaaaa@36F0@  ,

ε i MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8 qacqaH1oqzpaWaaSbaaSqaa8qacaWGPbaapaqabaaaaa@3906@  is a random error term with an expected value of zero, and

X i MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamiwamaaBa aaleaacaWGPbaabeaaaaa@37EE@  is plan i MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamyAaaaa@36E5@  ’s characteristic (either log download speed, log upload speed or log usage cap).Note 

7.21 Once all three regressions have been estimated, results from each of the regressions are used to predict a price for each plan, leading to three predicted prices. A weighted average of these three predicted prices is calculated as a single predicted price; the weights are defined such that a regression with a higher value of the coefficient of determination R-squared will have more weight. The missing prices of entering and exiting plans are imputed. The missing price of plan i MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamyAaaaa@36E5@ in period t MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamiDaaaa@36F0@   from Internet Service Provider (ISP) h MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamiAaaaa@36E4@   is calculated as p ih t = p ^ ih t × A ih t MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8 qacaWGWbWdamaaDaaaleaapeGaamyAaiaadIgaa8aabaWdbiaadsha aaGccqGH9aqpceWGWbWdayaajaWaa0baaSqaa8qacaWGPbGaamiAaa WdaeaapeGaamiDaaaakiabgEna0kaadgeapaWaa0baaSqaa8qacaWG PbGaamiAaaWdaeaapeGaamiDaaaaaaa@45C5@ . Here A ih t MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8 qacaWGbbWdamaaDaaaleaapeGaamyAaiaadIgaa8aabaWdbiaadsha aaaaaa@3A1C@  is an adjustment factor calculated from the plans available in period t MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamiDaaaa@36F0@  from ISP h MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaamiAaaaa@36E4@   while p ^ ih t MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8 qaceWGWbWdayaajaWaa0baaSqaa8qacaWGPbGaamiAaaWdaeaapeGa amiDaaaaaaa@3A5B@  is the imputed average predicted price.Note  For further details on the adjustment factor, refer to the paper “Internet Access Services Index Methodology in the Consumer Price Index” available online.

7.22 For the used vehicles price index, a hedonic regression modelling approach is used to account for quality change and depreciation over time. In order to control for quality change and to estimate pure price change, a time dummy hedonic method is employed along a rolling five-month estimation window. Dummy variables for the four latest monthsNote  of the estimation window, as well as car model fixed effects, are included in the regression. The coefficients of the time dummy variables measure the average price changes of a typical used vehicle (characterized by a given car model, a fixed odometer and a fixed age) in these four months compared to the fifth month of the window. Consequently, the change in the coefficients of dummy variables relating to two consecutive months within an estimation window measures the month over month change in used vehicles’ prices – in practice, the CPI uses the differences in the last two dummy coefficients as a measure of monthly price change for a given class of used vehicle. For details about the hedonic model used in the estimation of the used vehicles price index, refer to the paper “Measuring price change for used vehicles in the Canadian Consumer Price Index” available on the Statistics Canada website.

7.23 For the rent index, a hedonic model is estimated using monthly cross-sections of the Labour Force Survey (LFS) data at the national level. The lowest geographical level indices are constructed using average characteristics as quantities and estimated coefficients as prices, while the higher level indices use weighted averages of lower level estimated expenditures.

7.24 The hedonic model for the rent index is a log-linear regression in which the explanatory variables include observed unit characteristics, such as the number of bedrooms, as well as locational characteristics captured by postal codes. The regression specification is as follows:

y * = β 0 + β 1 services+ β 2 age+ β 3 bedrooms+ β 4 dwelling+ β 5 FSA+ϵ MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8 qacaWG5bWdamaaCaaaleqabaWdbiaabQcaaaGccqGH9aqpcqaHYoGy paWaaSbaaSqaa8qacaaIWaaapaqabaGcpeGaey4kaSIaeqOSdi2dam aaBaaaleaapeGaaGymaaWdaeqaaOWdbiaadohacaWGLbGaamOCaiaa dAhacaWGPbGaam4yaiaadwgacaWGZbGaey4kaSIaeqOSdi2damaaBa aaleaapeGaaGOmaaWdaeqaaOWdbiaadggacaWGNbGaamyzaiabgUca Riabek7aI9aadaWgaaWcbaWdbiaaiodaa8aabeaak8qacaWGIbGaam yzaiaadsgacaWGYbGaam4Baiaad+gacaWGTbGaam4CaiabgUcaRiab ek7aI9aadaWgaaWcbaWdbiaaisdaa8aabeaak8qacaWGKbGaam4Dai aadwgacaWGSbGaamiBaiaadMgacaWGUbGaam4zaiabgUcaRiabek7a I9aadaWgaaWcbaWdbiaaiwdaa8aabeaak8qacaWGgbGaam4uaiaadg eacqGHRaWktuuDJXwAK1uy0HwmaeHbfv3ySLgzG0uy0Hgip5wzaGqb ciab=v=aYdaa@7700@

where  y * MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8 qacaWG5bWdamaaCaaaleqabaWdbiaabQcaaaaaaa@380E@  is the log of observed rent, services MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8 qacaWGZbGaamyzaiaadkhacaWG2bGaamyAaiaadogacaWGLbGaam4C aaaa@3DA3@  represents whether the rent cost includes furniture, a washing machine, refrigerator, cable, or heat, age MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8 qacaWGHbGaam4zaiaadwgaaaa@38D3@  represents the age of building, bedrooms MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8 qacaWGIbGaamyzaiaadsgacaWGYbGaam4Baiaad+gacaWGTbGaam4C aaaa@3D9A@  represents the number of bedrooms, dwelling MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8 qacaWGKbGaam4DaiaadwgacaWGSbGaamiBaiaadMgacaWGUbGaam4z aaaa@3D95@   represents the type of the building, and FSA MathType@MTEF@5@5@+= feaagKart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9 vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=x fr=xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaaeaaaaaaaaa8 qacaWGgbGaam4uaiaadgeaaaa@3880@  is a vector of dummies defined from the first three digits of the postal code that corresponds to a neighborhood (in urban areas) or a region (in rural areas).

7.25 The option cost method is another explicit approach for making quality adjustments to entering POs in the CPI sample. This technique relies on having data about the specific costs for adding options or quality characteristics to a product. In this explicit method, an adjustment to the last observed price of the exiting PO is made so that it can be compared with the observed price of the entering PO. The option cost method is most commonly used for products where the manufacturer or retailer provides pricing details for the available product characteristics. The CPI has used the option cost method for some time in the elementary aggregates corresponding to the purchase of passenger vehicles index.

7.26 Expert judgment has, in the past, been a predominant practice for explicit quality adjustment in the CPI. This relies upon an employee with expertise in a particular product market to assess and give a valuation to differences in quality between exiting and entering POs. However, the practice of quality adjustment by expert judgment is not arbitraryNote  and follows procedural guidelines for choosing the most plausible quality ratio between exiting and entering POs. The expert judgment method is primarily used for elementary indices under the clothing and footwear major aggregate.

7.27 The option cost and expert judgment explicit approaches to quality adjustment are used in the CPI for cases where a complex decision has to be made, and where it is not appropriate to apply an implicit method such as overall mean imputation.


Date modified: